the syntax and semantics of relativization and quantification

THE SYNTAX AND SEMANTICS OF RELATIVIZATION
AND QUANTIFICATION: THE CASE OF QUECHUA
A Dissertation
Presented to the Faculty of the Graduate School
of Cornell University
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
by
Rachel Elizabeth Hastings
May 2004

c○ Rachel Elizabeth Hastings 2004
ALL RIGHTS RESERVED

THE SYNTAX AND SEMANTICS OF RELATIVIZATION AND
QUANTIFICATION: THE CASE OF QUECHUA
Rachel Elizabeth Hastings, Ph.D.
Cornell University 2004
This dissertation examines the structure of relative clauses, the semantics of
quantifiers and the interactions between them in two dialects of Quechua. It is
shown that the meaning of Quechua relative clauses with quantified heads is unexpected
in light of previous work on relative clause structure. In particular, the head
position of an internally-headed relative is not limited to indefinites, as predicted by
analyses in which head-raising is followed by intersective modification of the head
by the embedded clause. Although languages like Japanese have also been shown
to violate the indefiniteness restriction on the head, the interpretation of apparently
similar structures in the two languages is not the same. In Cuzco Quechua
even universally quantified internal heads take wide semantic scope over the clause,
whereas Japanese internal heads are interpreted clause-internally. The wide scope
of Cuzco Quechua internal heads is explained via a process of determiner incorporation.
In a framework in which relative clauses are analyzed as CPs selected directly
by D, it is proposed that the determiner of the head incorporates into the external
determiner thus gaining scope over the relative clause. This process is linked to

the fact that the head is not overtly Case-marked. It is proposed that the Case of
the head is checked by a matrix clause element and this process is also licensed via
determiner incorporation.
Imbabura Quechua is then shown to differ from Cuzco Quechua both in quantifier
scope and in Case-marking on relative clause heads. These differences lead to the
proposal that head-raising is not mandatory in Imbabura as it is in Cuzco, and
hence Imbabura relative clause heads behave semantically more similarly to their
Japanese counterparts.
Further issues in DP syntax and quantifier semantics are then pursued through
the study of extraction from noun phrases in Cuzco Quechua. It is argued that
the possessor of a noun phrase in existential position may be extracted from its
containing DP, yielding a possessive sentence interpretation. Semantic differences
between continuous and discontinuous noun phrases are studied, and an analysis is
proposed in which an element extracted from a DP must be interpreted externally
to the definiteness head of that phrase.

Biographical Sketch
Rachel Hastings grew up near Buffalo, New York and attended Williamsville public
schools until entering Harvard University in 1987. She graduated with an A.B. in
Physics and Mathematics in 1991, and spent the next two years as a Peace Corps
Volunteer in Niger, West Africa, where she taught secondary school mathematics in
Tanout and Diffa. In 1993 she entered the Ph.D. program in Applied Mathematics
at Cornell University. She earned an M.A. in this field in 1996 and a Ph.D. in 1998.
After one semester as a Visiting Assistant Professor in the Cornell mathematics
department, she entered the Ph.D. program in Linguistics at Cornell in January of
1999. She received an M.A. in Linguistics in January, 2002.
iii

To my family, broadly construed
iv

Acknowledgements
When I entered the field of Linguistics at Cornell in 1998 I thought I knew, probably
better than most, what I was getting myself into. I decided that the opportunity
to engage in the fascinating study of language was worth going through yet more
years in the often ill-fitting role of a graduate student. What I could never have
anticipated is the level of support, camaraderie and friendship that would buoy me
along a course that at the time seemed like a twisted form of masochism. I wish that
every doctoral student could be granted such a remarkable collection of colleagues
and friends.
Although there are only four official members of my graduate committee many
other Cornell linguists deserve an honorary membership. I cannot really do justice
here to all the help I have received from students and faculty alike. My coadvisors,
Chris Collins and Molly Diesing, have gone above and beyond the call
of advisor-hood in their generous donations of time and thought to the content of
the dissertation, the balance between data and theory, and the negotiation of travel
and fieldwork throughout the process. They have also helped me tremendously by
providing brilliant and inspiring models of how linguistics can be fit into a life.
v

Molly’s holistic approach to advising recognizes that the big picture always (and
necessarily) involves more than just linguistics. So just as she has helped steer my
research, strategize on funding, and negotiate conferences and teaching, she has also
been available on a moment’s notice to discuss turtles, vicuñas and pigs, and to
administer chocolate whenever necessary (or as a preventative measure).
Chris has put countless hours into reading, thinking about and commenting on
both the theoretical and the practical issues that I have struggled with in writing
this dissertation. He has provided a compelling example of how fieldwork-based
syntactic research can both remain true to the language being studied, providing
important insights into the way in which that particular language is organized, while
at the same time speaking to deep theoretical issues that are of relevance to the wider
community of linguists.
My other committee members have similarly provided practical linguistic assistance
and broader personal mentorship in approximately equal measures. Sally
McConnell-Ginet not only initially helped me to believe that a wholesale change
from mathematics to linguistics was possible, but has helped to make it possible
by providing valuable feedback on my work and sharing her infectious enthusiasm
for the field. John Whitman has also put a lot of energy into commenting on my
work, pointing to useful directions for research and relevant pieces of the literature,
and has used his finely honed skills as a teacher to help me reformulate my own
rudimentary ideas into actual hypotheses and theories.
Many other members of the Cornell linguistics faculty have shaped my current
understanding of the field in their roles as teachers and colleagues. In particular,
vi

through valuable courses and comments John Bowers, Dorit Abusch, and Mats
Rooth have helped with my work on Quechua syntax and semantics, Michael Weiss
assisted with my project on historical relationships between Khoisan languages, and
I have benefitted greatly from the teaching of Carol Rosen, Wayne Harbert and
Draga Zec. I am also heavily indebted to Junko Shimoyama who, during her year
as a visitor at Cornell and her research into Japanese relative clauses, inspired much
of the work that has led to this dissertation.
Thanks to Abby Cohn for expertly steering the department during her years as
Chair, and to Angie Tinti and Sheila Haddad for setting the tone of the department
at friendly and responsive and for keeping the administrative hurdles of graduate
life as low as possible.
The amorphous ranks of my fellow graduate students are probably largely responsible
for the fact that I enjoy going into work in the morning (or afternoon or
evening, as the case may be). As I think of my incoming class, it is clear that we
started as and remain a very diverse group, and yet from our formative first year I
have enjoyed and appreciated both the diversity and the sense that as a cohort we
have supported one another through this shared experience. So thanks to Jiahong,
Yoshi, Paul, Evelyn, Arthur, and Tanya. Several of you have been not only good
friends but also valuable commentators on my work, or on life or both depending
on your areas of expertise. And special thanks to Tanya for all the understanding,
the fun, and for being such a source of valuable perspective on the whole enterprise.
Thanks to the rest of the ling-grads community for making the basement a fun
place to be, even if only to complain about being in the basement. Listing everyone
vii

who should be thanked is impossible, but among these are: Aggrey, Ana, Andrew,
Changguk, Dan, Devon, Diego, Edith, Ellert, Eunchong, Irene, Johanna, Marisol,
Rebecca, Rina, Rob, Sang Doh, Whitney, Yuping and all the rest of you who make
the group what it is.
I have benefitted from the expertise of many scholars outside of Cornell, whose
feedback at conferences and during my fieldwork has also helped shape my dissertation
and my view of linguistics. I wish particularly to thank Martina Faller,
colleague in Quechua semantics, fellow fieldworker, and intrepid Andean hiker, for
friendship, support and encouragement. Thanks also to other Quechua linguists
whose past work has been so important to my own, and perhaps most importantly
to the dozens of Quechua speakers who have shared their language and their lives
with me during my trips to Peru and Ecuador. Certainly none of this would have
seemed worthwhile without the personal relationships that have been forged through
this work. I cannot say enough about the contributions of my teachers and language
consultants in Peru and Ecuador, whose insights into their language and ability to
communicate to me so many important nuances of meaning have been the cornerstone
of the work represented in this dissertation. Special thanks for contributing
a tremendous number of hours to this research are owed to the following individuals:
Inés Callalli Villafuerte, José Lema Maldonado, Elena Muenala Pineda, Natalia
Pumayalli Pumayalli, and Edith Zevallos Apaza. Thanks also to Hirmenegilda Contreras,
Veronica Fuentes, Marguerita Pumayalli, and Cirilu Pumayalli. Additional
thanks to my Quechua teacher at Cornell, Luis Morató-Peña, for years of patient
instruction and for sharing his own insight and experience as a Quechua scholar.
viii

I have also been the beneficiary of institutional and financial support from the
U.S. Department of Education and the Cornell University Latin American Studies
Program in the form of two academic-year FLAS fellowships for the study of
Quechua. Thanks to Mary Jo Dudley of LASP for her help in this regard. I also
gratefully acknowledge the support of the National Science Foundation through
Dissertation Research Improvement Grant BCS-0132445 which funded most of my
fieldwork in Peru and Ecuador, and provided some research support at home.
During the 2003-2004 academic year I have been a Visiting Instructor at Syracuse
University, and owe thanks to the wonderful students who have made my teaching
at SU such a joy, and to the faculty for their support.
As always, it is friends and family who really determine the hue of an existence,
and I doubt I could have gotten through one PhD, let alone two, without a dedicated
group of people who might not be up on the latest developments in the Minimalist
Program but whose expertise in other parts of life have made my own so much
easier during the last several years. Thanks in particular to Craig and Jenny for
our several years together as a household. You both have not just the patience for
linguistics but a delightful interest in language that kept the fun side of the game
always prominent in my mind. And thanks also to Alec, Julia, Sean, and the many
other friends who have hung in there through two bouts of graduate school and still
haven’t given up on me.
Also in this last category are my parents, whose amazing fortitude has allowed
them to register unwavering delight, encouragement and enthusiasm through what
must have seemed like an endless series of classes, papers, theses and degrees. Withix

out all of your love and support I would never have made it through even one of
these hurdles, and I know I continue to rest on this baseline security as I navigate
my way to the next adventure (and no, it won’t be another degree of any sort).
Finally, thanks to Paul, who has not only survived my struggles through two
PhDs but who has steadily increased his role throughout the process (to the point
where I currently find myself barely able to tie my own shoes without him, much
less submit a dissertation). Unfortunately, where thanks are the most due, words
are the least adequate and all the linguistic theory I can muster won’t help me with
that.
x

Table of Contents
1 Introduction 1
1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Theoretical framework and assumptions . . . . . . . . . . . . . . . . 5
1.3 Background on Quechua . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.1 The Quechua language family . . . . . . . . . . . . . . . . . . 8
1.3.2 Morphosyntax . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 The data and methodology . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Cuzco Quechua Relative Clauses with Quantified Heads ∗ 15
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Some previous analyses of IHRs . . . . . . . . . . . . . . . . . . . . . 19
2.2.1 Williamson 1987 . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.2 Cole 1987 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.3 Culy 1990 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2.4 Basilico 1996 . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.5 Shimoyama 1999, 2001 . . . . . . . . . . . . . . . . . . . . . . 24
2.3 The challenge posed by Cuzco Quechua . . . . . . . . . . . . . . . . . 26
2.3.1 Cuzco Quechua quantifiers . . . . . . . . . . . . . . . . . . . . 27
2.3.2 IHRs with quantified head . . . . . . . . . . . . . . . . . . . . 30
2.4 The structure of Cuzco relative clauses . . . . . . . . . . . . . . . . . 33
2.4.1 Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4.2 Calculating truth conditions . . . . . . . . . . . . . . . . . . . 42
2.5 Further evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.5.1 Scope interactions between the relative clause head and a matrix
clause quantified DP . . . . . . . . . . . . . . . . . . . . . 45
2.5.2 Scope interactions between the head and the subject . . . . . 47
2.5.3 Interaction with distributivity . . . . . . . . . . . . . . . . . . 48
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
xi

3 The Syntax of Head-Raising in Cuzco Quechua 54
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Morphosyntactic facts of Cuzco Quechua relatives . . . . . . . . . . . 57
3.2.1 Agreement and Case . . . . . . . . . . . . . . . . . . . . . . . 57
3.2.2 What can be a head . . . . . . . . . . . . . . . . . . . . . . . 59
3.2.3 Nominalizing morphology . . . . . . . . . . . . . . . . . . . . 62
3.3 Syntax of Cuzco Quechua Relative Clauses . . . . . . . . . . . . . . . 63
3.3.1 Basic framework . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.3.2 Explaining the Case-marking pattern . . . . . . . . . . . . . . 67
3.3.3 Which DP may raise? . . . . . . . . . . . . . . . . . . . . . . 80
3.4 Extending the analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.4.1 Subject-headed RCs, and explaining the nominalizing morphology
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.4.2 Revisiting weakly quantified heads . . . . . . . . . . . . . . . 84
3.4.3 Island constraints on relativization . . . . . . . . . . . . . . . 90
3.4.4 Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4 Comparison of Imbabura and Cuzco Quechua Relative Clauses 105
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.2 Morphosyntactic comparison between the dialects . . . . . . . . . . . 107
4.2.1 Review of similarities between the dialects . . . . . . . . . . . 107
4.2.2 Four morphosyntactic differences between the dialects . . . . . 109
4.2.3 Summary of CQ/IQ differences . . . . . . . . . . . . . . . . . 113
4.3 Syntax of IQ subordinate clauses . . . . . . . . . . . . . . . . . . . . 115
4.3.1 Explaining the morphological differences . . . . . . . . . . . . 115
4.3.2 Structure of IQ subordinate clauses . . . . . . . . . . . . . . . 115
4.4 Semantics of IQ vs. CQ internal heads . . . . . . . . . . . . . . . . . 119
4.4.1 Four differences between CQ and IQ internal heads . . . . . . 119
4.4.2 Structural implementation of the intuition . . . . . . . . . . . 122
4.4.3 How the analysis explains the CQ/IQ semantic differences . . 124
4.5 Do IQ heads ever raise? . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.5.1 Evidence from Islands . . . . . . . . . . . . . . . . . . . . . . 125
4.5.2 Evidence from quantifiers . . . . . . . . . . . . . . . . . . . . 130
4.5.3 Evidence from arguments vs. adjuncts in headless relatives . . 132
4.5.4 A proposal for movement of arguments only . . . . . . . . . . 140
4.5.5 Does the head raise for Case reasons? . . . . . . . . . . . . . . 144
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
xii

5 Existential and Possessive Sentences in Quechua 150
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5.2 The basic facts of Cuzco Quechua existentials . . . . . . . . . . . . . 152
5.3 Structure of CQ existential sentences . . . . . . . . . . . . . . . . . . 155
5.3.1 Quechua adverbial evidence that associate is not a Subject . . 158
5.3.2 Relative clause morphology indicates associate is not a Subject162
5.3.3 The syntax of CQ existential sentences . . . . . . . . . . . . . 166
5.4 Structure of CQ Possessive sentences . . . . . . . . . . . . . . . . . . 170
5.4.1 Some problems: ‘Maria has llamas’, and ‘There are those
mountains’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.4.2 Some past approaches to unifying existentials with possessives 173
5.4.3 A proposal for CQ possessive sentences . . . . . . . . . . . . . 177
5.4.4 Possessive adverbial and relative clauses . . . . . . . . . . . . 189
5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
6 The Semantics of Discontinuous Noun Phrases in Quechua ∗ 194
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
6.2 More data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.2.1 Co-Case marking . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.2.2 Quantifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.3 Interpretation and structure . . . . . . . . . . . . . . . . . . . . . . . 203
6.3.1 Previous work . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6.3.2 Semantic effects of co-Case marking . . . . . . . . . . . . . . . 206
6.3.3 Addressing the indefiniteness of discontinuous noun phrases . 209
6.4 Co-Case marking indicates scope outside the DP . . . . . . . . . . . . 212
6.4.1 The basic interpretive structure . . . . . . . . . . . . . . . . . 212
6.4.2 LF structures . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.4.3 Revisiting the syntactic options . . . . . . . . . . . . . . . . . 216
6.5 Why sapa ‘each’ won’t behave . . . . . . . . . . . . . . . . . . . . . . 222
6.5.1 The problem with sapa ‘each’ . . . . . . . . . . . . . . . . . . 222
6.5.2 Two different lexical items? . . . . . . . . . . . . . . . . . . . 223
6.5.3 Role of the suffix -nka . . . . . . . . . . . . . . . . . . . . . . 224
6.5.4 Proposal for sapa . . . . . . . . . . . . . . . . . . . . . . . . . 225
6.5.5 So why can’t sapa “float”? . . . . . . . . . . . . . . . . . . . . 227
6.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Bibliography 229
xiii

List of Tables
2.1 Cuzco Quechua Quantifiers . . . . . . . . . . . . . . . . . . . . . . . 30
4.1 Summary of CQ/IQ morphosyntactic differences . . . . . . . . . . . 114
4.2 Case-marking on subjects . . . . . . . . . . . . . . . . . . . . . . . . 117
xiv

List of Abbreviations
abl=ablative case
acc=accusative case
augm=augmentative
bi.adv=bipersonal adverbializer
cis=cislocative
conj=conjunction
dat=dative case
delim=delimitive
dem=demonstrative
dimin=diminutive
distr=distributive
euph=euphonic
evid=evidential marker
gen=genitive case
inch=inchoative
nm=nominalizer
nm.sbj=nominalizer, subject-headed relative clause
nm.nonsbj=nominalizer, nonsubject-headed relative clause
pl=plural
prog=progressive
pst.rep=past reportative
Q=interrogative particle
top=topic marker
uni.adv=unipersonal adverbializer
xv

Chapter 1
Introduction
1.1 Overview
This dissertation presents syntactic and semantic analyses of relative clauses and
quantification in Quechua. It uses quantifier scope data to explore relative clause
structure and then goes on to study quantifier behavior and noun phrase structure in
a wider context, including existential constructions and discontinuous noun phrases.
The narrow goal is to study the interaction of quantification and relativization, and
in particular the nature of the dual relationship of the relative clause head to the
matrix and embedded clauses. A broader goal is to further our understanding of
quantification and clausal modifiers cross-linguistically, incorporating descriptive
generalizations uncovered through a body of new Quechua data. In this respect
the dissertation is intended as a contribution to current research on the behavior of
quantifiers cross-linguistically and on the typology of relativization, with emphasis
on the role of syntactic structures as input to semantic interpretation.
1

2
The most immediate puzzle of Quechua relative clauses, which has drawn attention
in the past syntactic literature, is presented by the internally headed relatives.
In this construction, illustrated in (1.1), the noun phrase which is modified by the
embedded clause appears within the clause itself. In (1.1), the head is waka ‘cow’.
(1.1) [Juan-pa waka ranti-sqa-n]-qa yuraq-mi
Juan-gen cow buy-nm-3sg-top white-evid
‘The cow that Juan bought was white.’
ka-ra-n.
be-past-3sg
This construction raises the question of how the head is understood to be the
element which is modified by the relative clause. Syntactic studies which have
addressed this question specifically in Quechua include [Cole, Harbert & Hermon
1982], [Cole 1987a], [Lefebvre & Muysken 1982], and [Lefebvre & Muysken 1988]. A
natural analysis which has been proposed by these authors for Quechua and also in
other work on a variety of languages is that the head must raise out of the clause
at some level of structure in order to achieve a sisterhood relationship with its
modifier. Head-raising analyses of relative clauses date back to Vergnaud’s [1974]
work on French, and have been more recently revitalized by Kayne [1994] as part
of his Antisymmetry framework. Bianchi [1995, 1999, 2002] and others have further
developed a theory of relative clauses in Kayne’s framework.
This dissertation presents a new theory of head-raising in Quechua, based on
both syntactic and semantic evidence. On the semantic end, I show in Chapter 2
that certain quantified internal heads take interpretive scope over the entire relative
clause. I also show that some typologically unusual properties of Quechua relative
clause heads point to the same structural conclusion. In particular, Quechua relative

3
clause heads can violate an indefiniteness restriction on the head of an internally
headed relative first noted for Lakhota by Williamson [1987] and supported in crosslinguistic
studies by Culy [1990] and Basilico [1996]. The analysis of quantifier data
of this sort is motivated by recent work on Japanese internally headed relative clauses
by Hoshi [1995] and Shimoyama [1999, 2001] who study why in that language headraising
is not compatible with the meaning of certain internally headed relatives. In
light of this past work on internally headed relatives, and other recent work regarding
the typology of relative clauses (e.g. [Grosu & Landman, 1999]), I find that the
Quechua data illustrate a surprising clause type: internally headed relatives with
externally interpreted strongly quantified heads. On the syntactic side, I propose in
Chapter 3 that head-raising is associated with the Case-marking of the head by an
element within the matrix clause. I support this theory by providing evidence that
an internal head in Quechua cannot be Case-marked within the relative clause. This
fact also points to a head-raising analysis of internally headed relatives in Quechua,
and I argue that both the syntactic and semantic facts can be jointly explained via
a determiner-incorporation process along the lines proposed for English in [Bianchi
2000, 2002].
The next set of results, presented in Chapter 4, involve a close comparison between
relative clauses in Cuzco and Imbabura Quechua. I show that by assuming
a parametric difference in the feature settings on functional heads in the extended
relative clause construction, it is possible to explain a series of differences in the
morphology, syntax and semantics of relatives in the two languages. These results
provide support for an understanding of linguistic diversity based on the lexicon:

4
the availability of slightly different functional heads in the two dialects yields a host
of related implications for the structural aspects of the language.
Having looked at relative clause structure and quantifier behavior in relative
clauses in detail, I turn more generally to other issues relating to the structure of
noun phrases in Cuzco Quechua. I examine two particular aspects of this question,
both of which build on the basic DP structure developed in the earlier chapters. In
Chapter 5 I show that relative clauses in which the embedded sentence is existential
provide insight into a general problem in Quechua existential sentences: there seem
to be a set of systematic violations of the definiteness effect typically found in this
construction. This is unexpected from a cross-linguistic perspective, and my analysis
suggests that this is yet another case in which an apparent definiteness effect
violation is handled in Quechua by covert extraction. In fact, extraction of a possessor
from a noun phrase in an existential context yields a possessive sentence, making
Quechua surface structure an overt instantiation of the cross-linguistic analysis of
possessive sentences in [Szabolcsi 1994] and [Kayne 1994] in which the subject of a
possessive sentence is taken to be extracted from a noun phrase-internal possessor
position.
Finally, in Chapter 6 I look at the problem of discontinuous noun phrases in
Quechua, which connects apparent relative clause extraposition to other cases of
modifier and quantifier extraction. The main result here is that these cases of discontinuity
have a particular semantic effect: the (in)definiteness of the apparently
extracted element determines the (in)definiteness of the overall noun phrase. I propose
that this is due to an interpretive configuration in which the modifier appears

5
outside of the definiteness head of the noun phrase at LF.
1.2 Theoretical framework and assumptions
In this section I briefly outline the main theoretical tools that I adopt in this dissertation,
and the assumptions that I will make about phrase structure in Quechua.
The basic syntactic model will be a Minimalist framework as in [Chomsky 1995,
2000, 2001a]. Thus I assume that the fundamental syntactic operation by which
two lexical items or syntactic constituents may be combined is Merge. External
Merge is the derivational step by which two independent syntactic units α and β
are combined to form the set {γ, {α, β}}, where γ is the label of α. Internal Merge
is a similar operation with the additional proviso that β is a constituent of the structure
α. I will often refer to Internal Merge informally as “movement” of β. I will
also frequently use a notation in which the point from which Internal Merge takes
place is indicated with a trace (t) in the syntactic tree. When it becomes relevant in
the text I will discuss the Spell-Out of the syntactic structure and the importance
of the Copy Theory of movement, by which the lower “trace” position is filled by
one copy of the Internally Merged element.
In this framework, Internal Merge is triggered by a combination of an Agree
relationship between the head H of α, and β, and an EPP (OCC) feature on H.
Agreement can occur when H contains uninterpretable features which are identical
to interpretable features of β. In this case, H is called a probe, and β is the goal. In
order for the Agree relationship to be established, β must be the nearest matching

6
goal to the probe H. The detailed content of this restriction is known as the Minimal
Link Condition (MLC), which hinges on the notion of C-Command. I adopt the
following definitions of these terms:
(1.2) C-Command: A constituent α C-commands a constituent β in a syntactic
tree if every node of the tree that dominates α also dominates β, but neither
α dominates β nor β dominates α.
(1.3) Minimal Link Condition: An Agree relationship may be established between
head H and category β with matching features as long as there is no
intervening category γ such that H C-commands γ, γ C-commands β and
the same uninterpretable feature set of H also matches features of γ.
Note that under this definition of the MLC, even a matching intervening category
γ which has had its uninterpretable features checked already will still block a match
between H and β. This fact is known as the defective intervention constraint.
With respect to noun phrase structure, I assume that determiners head a functional
projection, the DP, as proposed in [Abney 1987]. Determiners normally bear
an uninterpretable Case feature which is valued in conjunction with an Agreement
relation between a probe and the Determiner. In Quechua, I assume that relative
clauses are CPs, which can be directly selected by a determiner D. In Chapter 3 I
present arguments in favor of this construction. In this regard I follow the work of
[Kayne 1994] and [Bianchi 1999, 2000] who have developed a theory of head-raising
in relative clauses and proposed this to be valid cross-linguistically. Although I follow
this aspect of the Antisymmetry framework quite closely, I do not adopt all of

7
the theoretical consequences of the LCA (Linear Correspondance Axiom) of [Kayne
1994], which among other things entails that all trees are right-branching. In this
dissertation I assume that Quechua is a mixed word order language. I therefore
place each head in its unmarked surface structure position (to the extent that this
can be determined from the data). Thus, DPs are head-initial since determiners
precede nouns, while VPs are head-final since SOV is the unmarked word order of
a basic transitive sentence. Although I believe it would be quite possible to reframe
my conclusions in the more stringent structure required by the LCA, I do not choose
to do so here because I think it would obscure the main ideas I am presenting.
On the semantic side I assume a type-driven truth conditional semantics and
adopt the general approach and framework represented by [Heim & Kratzer 1998].
I assume that interpretation of a syntactic tree proceeds compositionally, through
the processes of functional application or predicate modification. More specifically,
denotations of individual lexical items will be either that of entities (type e), truth
values (type t), or functions built up from these basic types (e.g. type ,
). The basic operations for calculating the denotation of nonterminal
nodes of a binary branching syntactic tree will involve either applying the
denotation of one daughter node (the function) to that of the other (the argument),
or by a modificational relationship in which both daughters denote sets, and the
mother node denotes the intersection of the two sets.

8
1.3 Background on Quechua
This section provides an overview of the Quechua language family and a brief sketch
of Quechua morphosyntax of particular relevance to this thesis.
1.3.1 The Quechua language family
The Quechua languages are spoken in regions of South America encompassing parts
of Ecuador, Colombia, Peru, Bolivia, Argentina and Brazil. The number of speakers
has been estimated at over 8.3 million [Cerrón-Palomino 1987], with the vast
majority of these speakers in Peru, Ecuador and Bolivia. The languages have been
divided into two main sub-families, known as Quechua A and B ([Parker 1969]) or
Quechua I and II ([Torero 1964]). Quechua B/I is spoken in coastal central and
northern Peru, and dialects outside this region are all Quechua A/II. In this thesis
I look at two dialects which are both classified as Quechua A languages but
which are geographically rather distant: Cuzco Quechua is spoken in the highlands
of Southern Peru, while Imbabura Quechua is spoken in northern Ecuador (where
the language is known as Quichua). Although there are many sociolinguistic differences
between these two regions, in both areas Quechua is in close contact with
Spanish and many speakers are bilingual. Rural Quechua-dominant communities
are found in both regions, where particularly the older speakers are often monolingual.
Substantial linguistic work has focused on the sociolinguistic and structural
aspects of this contact, and in particular [Mannheim 1991] offers a detailed look
at the linguistic history of the region since the arrival of the Spanish in the 16th

9
century.
Although the majority of the data in this dissertation come from my own fieldwork
in the regions of Cuzco and Imbabura, my research has also benefitted from a
substantial body of past scholarship on Quechua linguistics in general and Cuzco and
Imbabura dialects in particular (e.g. [Cusihuamán 1976/2000], [Cole 1985], [Cole &
Hermon 1994], [Lefebvre & Muysken 1982, 1988], [Muysken 1989]), narratives and
collections of stories (especially [Valderrama & Escalante 1977]) and relative clauses
in Quechua ([Weber 1978, 1994], [Jake 1985], [Cole 1987a], [Lefebvre & Muysken
1982, 1988]). In particular, I have heavily relied on the specific contributions represented
by [Lefebvre & Muysken 1988], on nominalization in Cuzco Quechua, and
[Cole 1985], a descriptive grammar of Imbabura Quechua. I refer to these works
extensively throughout this dissertation.
Other past work on the Quechua languages include historical work on the relationships
between dialects including [Parker 1969], [Torero 1964], [Cerrón-Palomino
1987], grammatical studies of specific dialects ([Solá 1967], [Cusihuamán 1976/2000],
and many others), dictionaries (e.g. [Cusihuamán 1976/2000] for Cuzco Quechua),
grammars (e.g. [Stark 1973] for Imbabura Quechua), and other theoretical linguistic
contributions such as [Cerrón-Palomino 1987], [Cole 1987a], [Cole & Hermon 1994],
[Cusihuamán 1976/2000], [Lefebvre & Muysken 1979, 1982, 1988], [Jake 1985], [Weber
1978, 1994], [Sánchez 1996], [Faller 2002], and many others.

10
1.3.2 Morphosyntax
In this section I provide a brief introduction to some of the basic facts of Quechua
morphology which will be relevant to examples presented throughout the dissertation.
Many of the details vary substantially among the Quechua languages, and
since here I am unable to provide more than the most relevant general information I
refer the reader to some of the excellent grammars available (notably [Cusihuamán
1976/2000] for Cuzco Quechua, [Cole 1985] for Imbabura Quechua, and to [Cerrón-
Palomino 1987] for an overview and comparison of a number of dialects, as well as
to the other references listed above.
The Quechua languages have a basic Subject-Object-Verb word order but with
a great deal of variation allowed. The unmarked word order in a main transitive
clause is illustrated in (1.4) from Cuzco Quechua (CQ).
(1.4) Juan
Juan
waka-ta
cow-acc
‘Juan bought a cow.’
ranti-rqa-n.
buy-past-3sg
(CQ)
The order of constituents is largely determined by stylistic and pragmatic considerations,
and not by truth-conditional semantic or purely syntactic considerations.
Thus in (1.4), all six of the logically possible word orders are acceptable and convey
the same meaning. In this sense the Quechua languages are highly nonconfigurational.
However, Quechua does seem to obey some restrictions in word order
particularly in subordinate clauses and within noun phrases, making it somewhat
more rigid than the well-studied example of Warlpiri, a classically nonconfigurational
language (e.g. [Hale 1983]). For example, subordinate nominalized clauses in

11
Quechua seem to be mandatorily verb-final. The issue of the configurationality of
the DP constituent will be discussed further in Chapter 6.
Quechua morphology is of the agglutinating type and both nouns and verbs can
support multiple suffixation (indicating such things as tense, direction, person and
number agreement, etc.) Examples of these affixes can be found in the verbs in
(1.5) and (1.6), from [Valderrama & Escalante 1977], the autobiography of Gregorio
Condori Mamani (GCM), a cuzqueño porter.
(1.5) Manan
NOT
mancha-ri-ku-ra-ni-chu.
fear-inch-refl-past-1sg-neg
‘I did not get scared.’ (GCM p.31)
(1.6) Ña
already
uywa-kuna-pas
animal-pl-conj
mikhu-sqa-n-ku-manta
eat-nm-3sg-pl-abl
kuti-y-mu-sha-n-ku-ña... (GCM p.29)
turn-aug-cis-prog-3sg-pl-already
‘The animals (also) were already returning from their grazing.’
Case marking is overt as can be seen on the direct object in (1.4) and the ablative
in (1.6).
A variety of subordinate clause types will be relevant to this dissertation. These
include relative clauses, adverbial clauses and complement clauses. In the case of
relative clauses there are different strategies available but I will focus almost exclusively
on the most common type which involves nominalization of the relative
clause. One piece of evidence for the nominal nature of the clauses is their ability to
be overtly Case-marked. We can observe that the clausal Case-markers are licensed
by the nominalized verb and not simply by the presence of a nominal head by noting

12
that complement clauses exhibit the same Case-markings as well as the same nominalization
morphology (the same as in relative clauses in which a non-subject has
been relativized). Indeed, subordinate nominalized clauses are sometimes ambiguous
between complement clauses and relative clauses. Case-marking of nominalized
clauses, and the relative/complement clause ambiguity are illustrated in Cuzco
Quechua in (1.7).
(1.7) Mariya
Maria
[[nuqa-q
I-gen
wasi
house
‘Maria saw that I made a house.’
ruwa-sqa-y]]-ta
make-nm-1sg-acc
riku-ra-n.
see-past-3sg
‘Maria saw the house that I made.’ (Lefebvre & Muysken 1988)
(CQ)
Syntactically, nominalized clauses are similar to main clauses with the exception
of verb position, as mentioned above: it is argued in [Lefebvre & Muysken 1988] for
CQ and [Cole 1985] for IQ that a nominalized verb must always appear in the rightmost
position within its clause. Other constituents of subordinate clauses exhibit
fairly free word order, as is the case in main clauses. Furthermore, Case-marking
of most categories of subordinate clause constituents is the same in subordinate
and main clauses. Some systematic exceptions to this, found in the Case marking
of an internal head, of direct objects, and of subjects, are dialect-specific and are
discussed in Chapters 3 and 4.
1.4 The data and methodology
The work presented here is based on my own fieldwork with Quechua language
consultants in Peru and Ecuador in various trips between 2000 and 2003. My pri-

13
mary consultants were seven Cuzco speakers (most of whom live in the village of
Chinchero near Cuzco) and three Peguche, Imbabura speakers of Quechua, but many
other friends and acquaintances contributed to my understanding of the intricacies
of Quechua through their conversation, judgments and comments on their language.
All my main consultants are bilingual in Spanish, but to varying extents. Some
were balanced bilinguals while others were more comfortable in Quechua. My consultants
themselves ranged from highly experienced teachers of Quechua to speakers
who had never studied the grammar of their language before. My field sessions with
consultants were conducted in a mix of Spanish and Quechua.
I have also relied heavily on the past work on Quechua mentioned earlier in this
chapter, although the majority of the facts reported in this dissertation come from
my own fieldwork. One reason for this is simply that many of the facts that I found to
be relevant to the problems I study here were not to be found in the past literature on
these dialects. However, in the interest of the internal consistency of the current work
I have verified previous results with my own consultants, since the dialect reflected
in my consultants’ judgments disagreed at certain points with judgments reported
in the past. There are of course a variety of possible explanations for this fact,
ranging from differences in dialect or ideolect, to language change, to methodological
discrepancies among researchers, and I have tried explicitly to indicate important
points of disparity in judgments. I have also cited many examples from past work
which my own consultants agree with, and in these cases I have noted the source in
the text. Examples for which no source is noted come simply from my own work.
This brings me to the question of data collection. The methodology I have

14
used is a combination of traditional elicitation and solicitation of acceptability judgments,
along with searching for examples in spontaneous speech and texts (here I
consulted most extensively the autobiographical narrative of Gregorio Condori Mamani
[Valderrama & Escobar 1977]). However, to be consistent and in keeping with
the considerations discussed in the preceding paragraph, I have also checked spontaneous
examples of these sorts with my consultants to be sure they reflected the
grammatical system I am studying here. Because of the complex semantics of some
of my examples, I made extensive use of pictures and other concrete visual aids to
confirm contexts. Of course, there were also many cases in which my consultants’
opinions varied, but in the end I have tried to use examples which are broadly agreed
upon and which reflect systematic responses, for the core analyses I have given. In
cases in which relevant data evoked mixed responses I have tried to report this explicitly
in the dissertation, and where possible to provide an explanation for these
responses.

Chapter 2
Cuzco Quechua Relative Clauses
with Quantified Heads ∗
2.1 Introduction
In this chapter I look at the interpretation of Cuzco Quechua relative clauses in which
the head of the relative clause is quantified. I consider cases of both internally headed
relatives and externally headed relatives. I use the ultimate scope of the quantifier
as evidenced by the possible meanings of the relative clause to probe the structure
of the relative clause itself.
An internally headed relative clause (IHR) is a subordinate clause which semantically
modifies one of its own constituent nominals. Cross-linguistically this is a
fairly rare construction, but it does show up in such diverse languages as Japanese,
Quechua and certain North American languages, such as Mojave and Lakhota. Syn-
∗ This chapter is based on [Hastings 2001].
15

16
tactically, it poses a challenge for theories of relative clauses in which the modified
nominal, or head of the relative clause is generated externally to the clause and
coindexed with an internal element. Semantically, IHRs raise the question of how
the head is recognized as the element to be modified.
The following pair of sentences illustrates an externally headed relative clause
(EHR) (2.1) and an IHR (2.2) in Cuzco Quechua.
(2.1) [[Juan-pa ranti-sqa-n] waka]-qa yuraq-mi
Juan-gen buy-nm-3sg cow-top white-evid
‘The cow that Juan bought was white.’
ka-rqa-n.
be-past-3sg
(2.2) [[Juan-pa waka ranti-sqa-n]]-qa yuraq-mi
Juan-gen cow buy-nm-3sg-top white-evid
‘The cow that Juan bought was white.’
ka-rqa-n.
be-past-3sg
In (2.1) the head waka ‘cow’ appears outside the subordinate clause, while in
(2.2) it is internal to that clause, reflecting the basic SOV word order of Quechua.
Note that the relative clause is nominalized with the suffix -sqa, which is also marked
for relative tense (main clause vs. subordinate clause). 1
1 There also exist non-nominalized relative clauses in Quechua. These constructions
involve a fully finite clause subordinated by chay ‘that’ (more typically used
as a demonstrative). An example is shown in (i).
(i) Chay Juan yacha-sha-ra-n Antonio-q suwa-sqa-n qulqi-ta,
dem Juan know-prog-past-3sg Antonio-gen steal-nm-3sg money-acc
chay wasi-y-pi ka-sha-n.
dem house-1sg-loc be-prog-3sg
‘The money Juan knows Antonio stole is in my house.’
Here I study only the more common, nominalized relative clauses as shown in the
main text.

17
In a main clause, accusative Case is marked with the suffix -ta, while in a nominalized
clause, the accusative Case marker is typically null (-∅). 2 Hence the lack
of overt Case marking on waka in (2.2) is in accordance with this general pattern.
The analysis of this phenomenon is a major topic of Chapter 3.
English allows only externally headed relative clauses, and hence syntactic and
semantic analyses of relative clauses based on languages like English have typically
assumed structures like that suggested by (2.3).
(2.3) [ DP The cow [ CP that John bought e]]
Here, e indicates the position in which the thematic role of cow within the subordinate
clause would normally be assigned. One popular analysis of English relative
clauses, which I will call the Operator analysis, suggests that the empty category e
is the trace of an empty operator which is coindexed with cow, and which raises to
Spec of the subordinate CP. In this analysis the head of the relative clause is the
noun phrase cow, which is adjacent to the within the external DP. 3,4 The syntactic
structure of the DP in (2.3) would hence be as shown in (2.4) under the operator
analysis.
2 For Lefebvre and Muysken’s [1988] consultants, -ta also could appear on some
nominalized clause direct objects, though it was dispreferred. My consultants rejected
-ta in these cases.
3 I will refer to the DP (or NP) which contains a relative clause as the outer DP
(or NP). In my own analysis I do assume the existence of a Determiner Phrase as
proposed in Abney (1987).
4 Another analysis, the Adjunction analysis of relative clauses, suggests that the
relative clause is adjoined to the DP the cow. Such an analysis raises immediate
problems for compositionality of the DP interpretation since the constituent ‘the
cow’ suggests there is one unique contextually relevant cow (this problem is discussed
by Partee (1976)). A solution to this problem proposed by Bach and Cooper (1978)
is mentioned in connection with Quechua in Section 2.4.

18
(2.4) DP
✟ ✟✟ ✟ ❍❍
❍
❍
D NP
The ✟ ✟✟ ✟ ❍❍
❍
❍
N CP
cow ✟ ✟✟ ✟ ❍❍
❍
❍
i
OP i C
✟ ✟✟ ✟
that
❍❍
❍
❍
IP
✟ ✟✟ ❍
❍
❍
DP VP
✟ ✟ ❍ ❍
John V NP
bought
t OPi
Most early work on IHRs (e.g. [Cole 1987a]) assumed that IHRs and EHRs had
essentially the same semantic distribution. Thus, languages allowing both constructions
were seen as having a built-in optionality in terms of head position. For this
reason, it was attractive to assume that both IHRs and EHRs had the same LF
structure. Since the EHR structure was already consistent with the modificational
meaning of a relative clause, it was naturally hypothesized that the head of an IHR
raises covertly to look like an EHR head at the level of interpretation. This view is
advanced by Cole [1987a] for Ancash and Imbabura Quechua and also by Williamson
[1987] for Lakhota, though with some differences in syntactic detail.
Other studies, however, suggest that it is incorrect to assume that EHRs and
IHRs have essentially identical LF structures. First Basilico [1996] claims that IHRs
are actually quantified NPs. It has further been shown in work on Japanese, that
in fact the meaning of IHRs and EHRs is not always identical, indicating a need for
independent semantic analyses of these constructions. This observation was made
by Kuroda [1974], and recently developed in detail by Hoshi [1995] and Shimoyama
[1999, 2001], who claim that IHRs are interpreted as independent sentences and the
head is identified for its role within the matrix sentence via E-type anaphora.

19
In this chapter I argue that a similar approach is necessary in Cuzco Quechua, but
must be implemented in quite a different way. I limit the discussion to sentences
in which an object 5 of the subordinate sentence is being relativized, as subjectrelativized
clauses are morphologically distinct from nonsubject-relativized clauses
in Cuzco Quechua. 6 The case of subject relativization will be discussed further in
Chapter 3.
In Section 2.2 I provide summaries of the earlier analyses of IHRs mentioned
above. In Section 2.3 I present some of the Cuzco Quechua data which demonstrates
the inadequacy of the previous theories to predict the interpretations of
relative clauses in this language. In Section 2.4 I propose structures for certain
Cuzco Quechua relative clauses. In Section 2.5 I give further evidence for the proposed
analysis involving quantifier scope interactions and the distributive suffix -nka.
Section 2.6 is the conclusion.
2.2 Some previous analyses of IHRs
2.2.1 Williamson 1987
Williamson [1987] argues based on data from Lakhota that the head of an IHR raises
at LF from its clause-internal position to a position which C-commands the relative
5 Although subordinate-clause nominals bearing a variety of grammatical roles
may be relativized, in this chapter the examples I study involve direct object heads.
In Chapter 3 I will look more closely at the identity and the syntactic position of
the head.
6 Subjects and non-subjects are distinguished in Cuzco Quechua in a number of
ways. In the case of relative clauses, the nominalizing morphology is distinct: in
subject-relativized clauses, the suffix -q appears on the verb, which is not inflected
for subject agreement or for tense. Non-subject relativization is signalled by the
nominalizing suffixes -sqa or -na which are selected on the basis of relative tense.
Roughly, they mark that the subordinate event occurs before or after the matrix
event respectively.

20
clause (specifically, she suggests a position adjoined to the S of the relative clause).
The determiner of the relative clause is external to S, being in Spec of the outer NP,
as shown in (2.5) and (2.6).
(2.5) S-Structure ([Williamson 1987, 24a]): NP i
✟ ✟✟ ❍
❍ ❍
S Det
✏ ✏✏✏
NP i
(2.6) LF ([Williamson 1987, 24b]): NP i
✟ ✟✟ ❍
❍ ❍
S’ Det
✟ ✟✟ ❍ ❍
S NP i
✏ ✏✏
t i
Williamson demonstrates that in Lakhota there is an indefiniteness restriction
on the head of an IHR, whereby in (2.5), NP i cannot be definitely marked (with
a definite determiner or universal quantifier). The external Det, of course, has no
such restriction. These facts are illustrated in (2.7) and (2.8). Example (2.7) is a
permissible IHR, with the indefinite head ‘a quilt’. However, (2.8) is bad due to the
definite determiner on the internal head ‘the quilt’.
(2.7) [Mary
Mary
owiža
quilt
wa
a
kaǧe]
make
ki]
the
‘the quilt that Mary made’ (Williamson 1987 (4a))
(2.8) *[[Mary
Mary
owiža
quilt
ki i
the
kaǧe]]
make
ki
the
(‘The quilt that Mary made’) (Williamson 1987 (5))
Williamson explains this definiteness condition by suggesting that insofar as the
relative clause is providing a restriction on the domain of the head NP, such a restriction
is semantically incompatible with an NP which is definite. Referring to
Heim’s [1982] treatment of definite and indefinite NPs as quantifier-free variables,

21
Williamson suggests that an NP marked with the definite determiner also cannot
be the head of an IHR because it represents old information, which is then not compatible
with further restriction. She goes on to predict that all languages allowing
IHRs will also exhibit this definiteness restriction.
2.2.2 Cole 1987
Cole [1987a], like Williamson, suggests that EHRs and IHRs have identical structures
at LF, through raising of the internal head. However, he also argues that an
IHR has an empty external head at S-structure, which is coindexed with the head
noun within the relative clause. This empty head is then replaced at LF by the
lexical head, which leaves a trace within the IHR. Both the empty external head
and the LF-raised lexical head are daughers of the outer NP . Thus, Cole suggests
the following structures of an IHR at S-Structure (2.9) and at LF (2.10).
(2.9) S-Structure ([Cole 1987a, 3]): NP
✟ ✟✟✟ ❍ ❍ ❍
¯S NP
✏ ✏✏✏✏ e
NP i
i
✏ ✏
(lexical)
(2.10) LF ([Cole 1987a, 4]): NP
✟ ✟✟ ❍ ❍
¯S NP i
✏✏
✏ ✏
...t i ... (lexical)
Cole’s work is based largely on data from Imbabura and Ancash Quechua. He
does not address the issue of definiteness in these languages, nor elaborate upon
Williamson’s definiteness restriction (note that his structure creates the same problems
for semantic parsing mechanisms as does the Adjunction analysis of EHRs,
mentioned in footnote 4).

22
2.2.3 Culy 1990
Culy [1990] provides a cross-linguistic survey and analysis of IHRs in nine languages.
He finds that internal heads are incompatible with universal quantification in all
these languages, as predicted by Williamson’s analysis of IHRs.
Culy’s analysis of the syntax and semantics of IHRs is different from Cole’s and
Williamson’s in that it assumes that the head NP is associated with a co-indexed
wh operator that moves to COMP at LF. The S-structure of an IHR is shown in
(2.11).
(2.11) S-Structure of IHR (Culy p.96): NP i
✟ ✟✟ ✟ ❍❍
❍
❍
— N —
S
✟ ✟ ❍ ❍
COMP S
❍
✟ ✟✟✟✟
❍
❍
❍
❍
— NP i —
✟ ✟✟ ❍ ❍ ❍
— N i —
wh i
Culy adopts the interpretation framework proposed by Heim [1982]. Skipping the
details of the steps involved in interpretation, I show in (2.12) a slightly simplified
version of Culy’s LF structure for an IHR.
(2.12) LF of IHR (from Culy p.166):
NP i
✟ ❍❍
✟ ✟✟✟✟ ❍
❍
❍
❍
DET
N i
S
✟ ✟✟✟ ❍ ❍ ❍
COMP S
✟ ✟✟ ❍ ❍
wh i NP i S
✏ ✏
(head) ...t NPi ...

23
For example, Culy states that under his analysis the (simplified) interpretation
of the Donno So sentence (2.13) would be (2.14) (Culy does not provide details of
the parsing).
(2.13) [[aru
[[agreement
kundaa
make-psp
ǫmmǫ]
1pl]
bo]
AUX-DEF]
agalaa
undo-PSP
ti
send-PN-3sg
(Donno So)
‘He has undone the agreement that we had made.’ (Culy p.163)
(2.14) [the agreement(x 1 ) & (we had done x 1 )] (he undid x 1 )
2.2.4 Basilico 1996
Basilico’s [1996] proposal concerning IHRs is motivated by certain word-order facts
in languages such as Diegueño and Mojave, wherein the internal head is disambiguated
through clause-internal movement. That is, some IHRs in which the embedded
sentence contains two objects are ambiguous with regard to the identity
of the head unless the actual head has raised to a higher position, but still within
the embedded clause. An example of an ambiguous clause in Cocopa (a Yuman
language) is given in (2.15). In (2.16) this sentence is disambiguated through IHRinternal
movement (xat ‘dog’ becomes unambiguously recognizable as the head when
it moves to the left of xu:r ‘rock’).
(2.15) [[John
John
xu:r
rock
xat
dog
pa:cu:s-p-t y ]]
hit-DEM-SUBJ
u:n y iL y cis.
black.EMPH
?‘The dog John hit with the rock was black.’
(Cocopa)
‘The rock John hit with the dog was black.’ (Basilico [1996], from Gorbet
[1976]:60)
(2.16) [[John
John
xat
dog
su:r
rock
pa:cu:s-p-t y ]]
hit-DEM-SUBJ
‘The dog John hit with the rock was black.’
(Basilico [1996], from Gorbet [1976]:53)
u:n y iL y cis.
black.EMPH
(Cocopa)

24
To explain this phenomenon, Basilico suggests that IHRs are instances of quantification,
and as such become adjoined to the matrix IP at LF. The determiner
of the IHR binds the variables in the IHR, which are associated to the internal
head as well as to the relative clause restriction, and this determiner also adjoins
to the matrix IP at LF. According to Basilico, the clause-internal movement of the
head illustrated in (2.15) and (2.16) must take place at least in the covert part of
the grammar in order for the quantification to obey Diesing’s Mapping Hypothesis
([Diesing 1992]). This hypothesis implies that indefinites must move out of the VP
at LF in order to avoid existential closure and be bound by the determiner of the
IHR. Basilico gives supporting evidence by showing (like Williamson) that the head
of an IHR can only be an indefinite.
In Cuzco Quechua, however, the contrast noted in (2.15) and (2.16) above is not
found, as seen in (2.17) and (2.18).
(2.17) [[Irqi-q
Child-gen
alqu-man
dog-dat
rumi
rock
ch’anqa-sqa-n]]-qa
throw-nm(nonsbj)-3sg-top
ka-rqa-n.
be-past-3sg
‘The rock that the child threw at the dog was black.’
(2.18) [[Irqi-q
Child-gen
rumi
rock
alqu-man
dog-dat
ch’anqa-sqa-n]]-qa
throw-nm(nonsbj)-3sg-top
ka-rqa-n.
be-past-3sg
‘The rock that the child threw at the dog was black.’
yana
black
yana
black
(CQ)
(CQ)
The method by which the head is identified is different in different dialects of
Quechua, and is a principal topic of Chapters 3 and 4.
2.2.5 Shimoyama 1999, 2001
Shimoyama’s [1999, 2001] theory of Japanese IHRs is significantly different from the
last two in that she does not rely on the presence of an operator to bind the internal

25
head. Rather, she assumes with Hoshi [1995] that the IHR is a closed sentence
and its interpretation involves E-type anaphora. The idea is that the head is never
raised out of its clause (or, indeed, its base position) but rather the role it plays in
the matrix clause is understood through the context of utterance, which determines
the interpretation of a null proform that combines the information present in the
predicative part of the head with the information about the head provided by the
rest of the IHR. The IHR as a whole will adjoin to the matrix IP at LF to give the
proper interpretation.
For example, sentence (2.19) illustrates a Japanese IHR with a universally quantified
head.
(2.19) Taro-wa
Taro-top
[[Hanako-ga
Hanako-nom
dono sinbun-mo
every newspaper
katte kita]-no]-o
buy came-nm-acc
tana-ni
shelf-on
narabeta.
placed
‘Hanako bought (and brought) every newspaper and Taro shelved them.’
([Shimoyama 1999 36a])
(2.20) LF of (2.19) ([Shimoyama 1999, 36b]):
IP
❍ ✟ ✟✟✟✟✟
❍
❍
❍ ❍
CP i
IP
✏ ✏✏✏ ✏
✟ ✟✟ ❍
❍ ❍
Hanako bought VP I
every newspaper ✟ ✟ ❍❍ ❍
Taro V’ past
✟ ✟✟ ❍
❍ ❍
DP V
✟ ✟ ❍ ❍
t i D’ shelve
✟ ✟ ❍ ❍
NP D
N
P
nm
Shimoyama postulates an LF structure as in (2.20). The proform P is a free
variable of type which gets its denotation from the context c. Here, the

26
function g c assigns to the index 3 associated with the proform the property of being
newspapers that Hanako bought.
(2.21) g c :=[3→ λx ∈ D e . x is newspapers that Hanako bought].
Thus the gloss in (2.19) accurately represents the proposed interpretation.
Note that sentence (2.19) appears to be an example of precisely what Williamson,
Culy and Basilico claimed not to exist: a universally quantified internal head. 7 Using
data such as this, Shimoyama brings to light a number of facts concerning the
interpretation of IHRs which were not remarked on by these previous studies. Most
significantly, she points out that IHRs in Japanese do not always have the same
truth-conditions as their externally headed counterparts (both constructions are
permissible in that language). For instance, if ‘Hanako’ is replaced by an existentially
quantified DP in (2.19), then in the interpretation this DP must take scope
over ‘every newspaper’, as would be the case in the corresponding independent sentence.
However, if scrambling had occurred within the IHR, then ’every newspaper’
could take wide scope over the DP. These are exactly the facts of Japanese singleclause
sentences, and they support the claim that IHRs should be interpreted with
no raising of the head. If the head were to raise at LF, then we would expect ‘every
newspaper’ to always take wider scope than the subject DP, regardless of whether
embedded clause scrambling had occurred.
2.3 The challenge posed by Cuzco Quechua
In this section I present the data from Cuzco Quechua which remain unexplained
under existing theories of IHRs. My analysis of these data appears in Section 2.4.
7 In fact, Basilico does report apparently similar S-structures in Mooré and
Navajo, but suggests that in these languages the apparent universal quantifier might
instead be functioning as a verbal operator. He does not provide a detailed account
of this option however.

27
Since the key IHRs all have quantified heads, I start out in Section 2.3.1 by briefly
summarizing the Quechua quantifiers which appear in the IHR data.
2.3.1 Cuzco Quechua quantifiers
The quantifiers relevant to this chapter are D-quantifiers (determiner-like), in the
terminology of [Bach et al. 1995]. They are structurally part of a DP, typically
appearing before the noun and adjective, if there is one. Although Quechua has
no overt definite and indefinite articles, it does have non-quantificational apparent
determiners in the form of the demonstratives kay ‘this’, chay ‘that’, and haqay
‘yonder’. Sentences (2.22) and (2.23) illustrate one property of Quechua quantifiers
which indicates that they are D-quantifiers: when the quantifier does appear in a
non-canonical position within the sentence, it must receive its own Case marking.
This is a general syntactic diagnostic in Cuzco Quechua for apparent movement out
of a DP in an argument position, discussed by Lefebvre and Muysken [1988]. This
phenomenon is investigated in detail in Chapter 6.
(2.22) Llipi-n
All-3
runa-kuna-ta
person-pl-acc
riku-rqa-nki-chu?
see-past-2sg-Q
‘Did you see all the men?’ ([Lefebvre & Muysken 1988, p.142])
(2.23) [e i Runa-kuna-ta]
person-pl-acc
llipi-n-ta i
all-3-acc
riku-rqa-nki-chu?
see-past-2sg-Q
‘Did you see all the men?’ ([Lefebvre & Muysken 1988, p.142])
Note that in (2.22), the quantifier llipin ‘all’ (which, incidentally, appears to
have the same distribution as tukuy ‘all’) is not Case-marked, and appears in its
canonical position before the head noun runakuna ‘people’. By contrast, in (2.23),
llipin appears to the right of the head noun, and the fact that it is no longer part
of the DP is indicated by the accusative Case-marking which now appears on the
moved element.

28
Since Quechua allows a great deal of null anaphora, the noun itself may be
omitted if it is understood, in which case the relevant Case-marking (which appears
on the last element of the NP) will be located on the adjective or on the determiner
or quantifier. 8 This possibility is illustrated in examples (2.24) and (2.25).
(2.24) Juan
Juan
wakin
some
llama-lla-ta
llama-delim-acc
‘Juan saw some (of the) llamas.’
rikhu-ra-n.
see-past-3sg
(2.25) Juan wakin-lla-ta
Juan some-delim-acc
‘Juan saw some of them.’
rikhu-ra-n.
see-past-3sg
As in many other languages, Quechua distinguishes between strong and weak
quantifiers, as defined in [Milsark 1977]. As discussed in that work (and variously
by Barwise and Cooper [1981], deHoop [1995] and others), existential constructions
provide the canonical environment which distinguishes between the two. Thus, in
English we call some a weak quantifier as evidenced by the acceptability of such
sentences as There are some llamas in the field. By contrast, most is a strong
quantifier as indicated by the unacceptability of *There are most llamas in the field.
In Cuzco Quechua, there-sentences are expressed using the third singular form
kan of the verb kay ‘to be.’ (This verb is obligatorily dropped in copula constructions
8 Lefebvre and Muysken [1988], adopting a strong view of the lexical hypothesis
and presenting evidence that certain Case markers are indeed affixes and not clitics,
argue that adjectives and determiners are really nominal in nature. For the purposes
of this chaper, I will begin with the assumption that quantifiers are determiners,
although the identity of the true quantifiers in this sense will be clarified later
in Section 2.4. In [Bittner & Hale 1995] it is argued that Warlpiri D-quantifiers
are themselves nominal (and indeed that Warlpiri makes no use of the category
“determiner”). However, Quechua differs from Warlpiri in that the base structure
of a non-clausal Quechua DP is fairly fixed (movement aside, as discussed above)
while the corresponding structures in Warlpiri are quite free. For these reasons, I
assume that Quechua quantifiers do have access to the structural position D, and
withhold judgment on whether they can also head NPs.

29
with third person singular subjects, so there is no ambiguity.) A syntactic analysis
of this construction is the topic of Chapter 5. Sentences (2.26) and (2.27) show that
ashka ‘a lot’ is compatible with the existential construction while tukuy ‘all’ is not. 9
(2.26) Ashka
many
llama-kuna
llama-pl
chacra-pi
field-loc
ka-n.
be-3sg
‘There are many llamas in the field.’
(2.27) *?Tukuy
all
llama-kuna
llama-pl
chacra-pi
field-loc
‘There are all llamas in the field.’
ka-n.
be-3sg
Similar examples with other quantifiers are shown in (2.28).
(2.28) *?Llipin/*?Wakin/Kinsa/Pisi
every/some/three/a few
llama-kuna
llama-pl
chakra-pi
field-loc
‘There is (are) every/some/three/a few llamas in the field.’
ka-n.
be-3sg
Table 2.1 below lists the Cuzco Quechua quantifiers mentioned in this chapter,
with English gloss and classification as strong or weak according to the above criterion.
10 Note that I have not identified cases in which a determiner may be either
strong or weak depending on context, although such cases certainly occur in other
languages and may also exist in Quechua.
9 Some consultants take the intended reading of (2.27) to involve a copula reading
of the verb, but suggest that the correct verbal form would involve progressive
marking. Thus the same sentence with ka-sha-n (’be-prog-3sg’) means ‘All the
llamas are in the field.’ Existential constructions are studied further in Chapter 5.
10 I note in passing that for some consultants sapa is, in fact, compatible with the
test sentence, but in this environment takes on the meaning ‘only’, and thus can
appear, for instance, before a proper noun as well as before llama. This use of sapa,
however, often triggers person/number agreement of sapa with the following noun.
A study of other non-agreeing uses of sapa can be found in Chapter 6.

30
Table 2.1: Cuzco Quechua Quantifiers
Quantifier English gloss strong/weak
ashka many, a lot weak
huk, iskay, etc. one, two, etc. weak
pisi a few, a little weak
sapa each, every strong
tukuy all strong
wakin some (of) strong
2.3.2 IHRs with quantified head
As discussed in Section 2.2, sentences in which the head of an IHR is a quantified NP
can provide important evidence for or against various structural possibilities. Recall
that Cole’s [1987a] analysis of (Ancash and Imbabura) Quechua assumes that an
IHR and its corresponding EHR have identical meanings and therefore an analysis
in which both structures look identical at LF may be advantageous since the same
parsing strategy may apply to each. By contrast, Hoshi [1995] and Shimoyama
[1999] point out certain important differences in the meaning of EHRs and IHRs in
Japanese, necessitating distinct parsing mechanisms.
In Cuzco Quechua, I find that in general EHR/IHR pairs, even those with quantified
heads, are semantically indistinguishable. (The exceptions I am aware of involve
the distributive suffix -nka on the head as discussed in Section 2.5.3. In these cases
the EHR is strongly preferred.) Thus, sentences (2.29) and (2.30), containing an
IHR and EHR respectively, both have the same interpretation, compatible with an
external head at LF.

31
(2.29) [[Juan-pa tayta-n-pa wakin wasi ruwa-sqa-n]]
Juan-gen father-3sg-gen some house make-nm-3sg
‘Some houses that Juan’s father made are big.’
hatun.
big
(2.30) [[Juan-pa tayta-n-pa ruwa-sqa-n] wakin
Juan-gen father-3sg-gen make-nm-3sg some
‘Some houses that Juan’s father made are big.’
wasi]
house
hatun.
big
Consultant comment on both sentences: There are other houses that he
made that we don’t know the size of.
The translations suggest an interpretation compatible with an LF-raising analysis,
along the lines of Cole’s [1987a] or Williamson’s [1987] proposals. However,
the fact that the LF head-raising analysis does not fully capture the facts of Cuzco
Quechua relative clauses is revealed when we consider data in which the head is
quantified by the weak quantifier, pisi. The following IHR/EHR pair of sentences
((2.31) and (2.32)) shows that this quantifier does not allow its NP to take scope
over the relative clause, as might be expected under Cole’s or Williamson’s analysis.
(2.31) [[Asunta-q
Asunta-gen
apa-ra-ni
bring-past-1sg
pisi
a little
aqha
cornbeer
aqha-sqa-n]]-ta
make corn beer-nm-3sg-acc
‘Asunta made a little corn beer and I brought it (the little corn beer that she
made).’
(2.32) [Asunta-q
Asunta-gen
apa-ra-ni
bring-past-1sg
aqha-sqa-n]
make corn beer-nm-3sg
pisi
a little
aqha]-ta
corn beer-acc
‘Asunta made a little corn beer and I brought it (the little corn beer that she
made).’
Consultant comment on both sentences: I brought all the cornbeer that she
made.
In this case, both the IHR version (2.31) and the EHR version (2.32) have, again,
the same meaning, but this time with an internal-scope reading of the quantified

32
head. Furthermore, these sentences clearly state that Asunta only made a little
cornbeer and I brought that entire quantity. 11
My consultants uniformly reject the possibility of continuing with “...but I left
the rest behind,” or any other suggestion that Asunta could have made more than
a little cornbeer. That I brought all that she made is entailed by both sentences.
To express that Asunta made a little cornbeer and I only brought a little of that,
two consultants suggested (2.33).
(2.33) [[Asunta-q pisi aqha aqha-sqa-n]]-manta (aswan)
Asunta-gen a little cornbeer make cornbeer-nm-3sg-abl (more)
pisi-lla-ta apa-ra-ni.
little-delim-acc bring-past-1sg
‘Asunta made a little cornbeer and I brought only a little of it.’
To further complicate this picture, when the head of the relative clause is quantified
by the strong quantifier tukuy ‘all’, the head is interpreted with mandatory
wide scope over the relative clause. This is illustrated in sentences (2.34), which
contains an IHR, and (2.35), with an EHR.
11 Srivastav [1991 p.103], in her dissertation on correlative clauses, mentions that
the following sentence from Ancash Quechua (which is quite distantly related to the
Cuzco dialect) exhibits an interpretation pattern in line with the interpretations of
the pisi-headed sentences above.
(i) nuna ishkay bestya-ta ranti-shqa-n alli bestya-m ka-rqo-n
man two horse-acc buy-perf-3 good horse-Valid. be-past-3
‘The two horses that the man bought were good horses.’
Srivastav notes that this sentence includes the information that the man bought
(only) two horses. She further states that this is not the case for the externally
headed version. The same maximalizing property of Quechua IHRs is reiterated
also in [Grosu & Landman 1998] and [Grosu 2002]. If these facts are correct for
Ancash Quechua, then they indicate a pattern more similar to Japanese than to
Cuzco Quechua, but still problematic for such previous treatments as [Cole 1987a].
To my knowledge these facts have not been further investigated in Ancash Quechua.
However, in Chapter 4 I will show that Imbabura Quechua exhibits a semantic
pattern of the same sort.

33
(2.34) Asunta-(qa)
Asunta-top
[[Mayta-q
Mayta-gen
plaza-pi
plaza-loc
tukuy
all
planta
plant
p’iti-ra-n.
prune-past-3sg
‘Asunta pruned all the plants that Mayta planted in the plaza.’
(2.35) Asunta-(qa)
Asunta-top
[[Mayta-q
Mayta-gen
plaza-pi
plaza-loc
planta-sqa-n]
plant-nm-3sg
planta-sqa-n]]-ta
plant-nm-3sg-acc
tukuy
all
p’iti-ra-n.
prune-past-3sg
‘Asunta pruned all the plants that Mayta planted in the plaza.’
planta]-ta
plant-acc
Both (2.34) and (2.35) can be followed up with a statement like “...but she did
not touch the rest of the plants in the plaza (the ones that Mayta did not plant).”
Consultants are clear that these sentences do not make the statement that Mayta
planted all the plants in the plaza.
Recall that according to the generalizations and predictions of Williamson [1987]
and Culy [1990], universally quantified heads should not be allowed in an IHR, as
we have in (2.34) 12 . Now, this generalization is also contradicted by Japanese, but
in Japanese the expected translation of the equivalent sentence would be ‘Mayta
planted all the plants in the plaza, and Asunta pruned them.’ Thus, sentence (2.34)
is surprising under any previous analysis of IHRs.
2.4 The structure of Cuzco relative clauses
2.4.1 Structures
In this section I propose LF structures for two types of Cuzco Quechua relative
clause. I first look at relative clauses with pisi-quantified heads, then those with
tukuy-quantified heads, representing weak and strong quantifiers respectively. The
12 The fact that Quechua internal heads do not appear to be restricted to indefinites
is also mentioned in Footnote 12 of [Cole & Hermon 1994].

34
basic syntactic analysis given here will be developed in more detail in the next
chapter.
2.4.1.1 Heads quantified by pisi ‘a little’
First, to accommodate the pisi-headed data in (2.31) and (2.32), in which we find
that the content of the subordinate clause is implied by the matrix sentence, I
propose that an E-type anaphora interpretation is appropriate for both IHRs and
EHRs in Cuzco Quechua. That is, I propose that the EHR construction is not in
fact one in which the head takes interpretive scope over the relative clause. Either
the external head is reconstructed to a clause-internal position at LF, or perhaps the
apparent external head is best considered the result of clause-internal scrambling. 13
Either way, the subordinate sentence is interpreted as a proposition, and the head
is recognized by E-type anaphora.
Concretely, I propose the following LF Structures for the sentences in (2.31) and
(2.32). This proposal will be modified and refined in Chapter 3 to explain some
differences between Imbabura and Cuzco Quechua in terms of how the head gets
identified.
13 The exact nature of the clause-final head position deserves further research.
Lefebvre and Muysken [1988] argue that all nominalized clauses must be verb-final,
citing evidence that this is the case in nominalized complement clauses. For relative
clauses the situation is more complex, however. In fact, Lefebvre and Muysken
identify a “COMP-like Case position” which is rightmost in an IHR, but still not
external to the clause.
Because of various data differences discussed later in section 3.4.4, although Lefebvre
& Muysken distinguish a clause-internal and clause-external position for a rightmost
head (based on Case-marking and distributional phenomena), this distinction
is not evident in the dialect I am studying here. I look more closely at the syntax of
head-raising in the next chapter but for the moment simply note that if scrambling
is not a viable explanation for the apparent head-final structures, then reconstruction
to base position at LF still seems possible. The main point is that this position
empirically does not give interpretive scope over the relative clause in the case of a
pisi-quantified head, nor even is it a syntactically viable scope position in the case
of strong quantifiers, as will be discussed below.

35
(2.36) LF of (2.31), ((2.32) similar):
IP
✟ ❍
❍
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟✟ ❍
❍
❍
❍
CP i
IP
✟ ✟✟ ❍
❍ ❍
VP I
✏ ✏✏✏✏✏✏✏✏✏✏✏
✟ ❍❍
Asunta-q pisi aqha aqha-sqa-n-ta
✟ ✟✟✟ ❍
❍
❍
Asunta-gen little cornbeer make-nm-3sg-acc NP V’
Nuqa
I
✟ ✟✟✟ ❍ ❍ ❍
DP V
✟ ✟ ❍ ❍
D NP aparani
✟ ✟ ❍❍ ❍ brought
t i N
R
R is a variable of type which receives its denotation from the context of
utterance via the following assignment:
(2.37) g c := [5 → λx ∈ D e . x is cornbeer that Asunta made]
Here I have basically adopted Shimoyama’s [1999] structure of IHR sentences
in Japanese, with a couple of changes. First, I place CP as the underlying sister
to N rather than in [Spec,DP] to allow a greater parallel with other relative clause
constructions in Cuzco Quechua, and in keeping with syntactic arguments made by
Hoshi [1995:4.2.1.1.2] in support of NP-internal relative clauses. Secondly, note that
the nominalizing morphology in Quechua (a verbal suffix) does not lend itself to a
Determiner analysis, as Shimoyama suggests for -no in Japanese. Thus in (2.36),
D is null, though apparently definite in accordance with the maximality effect of
E-type anaphora (as in Sells [1986]).
Another possibility, which I reject, is that pisi-quantified heads do raise at LF
to an interpretive position with scope over the relative clause, but that there is

36
something about pisi which induces a definite interpretation. However, in simple
matrix clauses, there is no tendency to interpret pisi (or its associated determiner)
as a definite, as seen in (2.38), where there is no implication that there was only a
small amount of cornbeer available.
(2.38) Nuqa
I
pisi-(lla)
a little-(delim)
aqha-ta
cornbeer-acc
‘I drank a little cornbeer.’
ukya-ra-ni.
drink-past-1sg
Recall that in Williamson’s analysis of Lakhota (a language with overt definite and
indefinite articles), the definiteness restriction held for an internal head, but there
was no restriction on the definiteness of the external determiner, of the outer DP.
That is, either a definite or an indefinite article could appear in that position. Thus,
it would be unexpected if Cuzco Quechua, under the head-raising analysis of IHRs,
had a null article in this position which had a forced definite interpretation.
It is worth mentioning at this point one aspect of pisi-headed relative clauses
which is perhaps unexpected under the E-type analysis adopted here. I am referring
to the unambiguity of the head of the relative clause. This is partly due to the fact
that the nominalizing morphology differs depending on whether the subject of a
non-subject is being relativized. However, we might nonetheless expect that two
non-subjects could compete as the head of a relative clause. For example, why
can’t (2.39) (repeated from (2.17)) mean ‘the dog that Juan threw the rock at was
black’? A purely pragmatic construal of the head would predict this alternative
meaning should be possible, and yet it is unavailable. I leave aside this problem
for the moment but bring it up again in the context of the syntactic analysis of
head-raising given in Chapter 3, which aims to address this very problem.

37
(2.39) [[Irqi-q
Child-gen
alqu-man
dog-dat
rumi
rock
ch’anqa-sqa-n]]-qa
throw-nm(nonsbj)-3sg-top
ka-rqa-n.
be-past-3sg
‘The rock that the child threw at the dog was black.’
yana
black
2.4.1.2 tukuy-quantified heads
The E-type anaphora analysis accounts for the data in (2.31) and (2.32), but still
does not explain why other quantifiers behave quite differently from pisi. In particular,
in (2.34) and (2.35), we saw that tukuy-quantified heads receive mandatory
wide scope over the relative clause. Furthermore, we saw that the LF-raising of a
universally quantified head appears to violate syntactic and semantic restrictions
which prohibit this configuration in other IHR languages.
To begin to answer the question of how exactly the tukuy-quantified RCs are to
be interpreted, I note first that (2.34) and (2.35) have a paraphrase in which the
quantifier appears explicitly in front of the relative clause, shown in (2.40).
(2.40) Asunta
Asunta
[tukuy [Mayta-q plaza-pi planta
all Mayta-gen plaza-loc plant
p’iti-ra-n.
prune-past-3sg
‘Asunta pruned all the plants which Mayta planted in the plaza.’
planta-sqa-n]]-ta
plant-nm-3sg-acc
In (2.40), the general proposal of Williamson [1987] for Lakhota neatly supplies
a structure in which tukuy appears in the Determiner position of the outer DP. In
this position this quantifier has scope over the relative clause, in keeping with the
gloss. The LF structure of (2.40) would then be as in (2.41).

38
(2.41) LF head-raising in (2.40):
DP
❍
❍
❍
✟ ✟✟✟✟✟✟
❍
❍
❍ ❍
D
CP
❍ ❍
❍
tukuy
all ✟ ✟✟✟✟✟✟
❍
❍
❍ ❍
CP
DP j
✏
✏ ✏✏✏✏✏✏✏✏✏✏ Mayta-q plaza-pi t j planta-sqa-n-ta
Mayta-gen plaza-loc t j plant-nm-3sg-acc
✏ ✏
planta
plant
Returning to (2.34) and (2.35), in which the quantifier appears adjacent to its
associated nominal, there are two related issues to be addressed by any proposal
regarding their LF structure and mechanism of interpretation. First, the quantifier
is apparently interpreted outside of its S-structure clause, in violation of broad
cross-linguistic evidence for the clause-boundedness of strong quantifiers. Second,
the associate of the quantifier at S-structure (apparently just the nominal ‘plant’)
seems to be different from the associate of the quantifier at the interpretive level
(apparently the entire phrase ‘plant that Mayta planted’). I mention here briefly
some possible approaches to each of these issues. A detailed study of this problem
is the topic of the next chapter.
Basilico [1996] addresses the second of these points (the LF disassociation of the
apparent Determiner from its apparent sister nominal) by suggesting that a universal
quantifier associated with an internal head may not be a Determiner at all, but
rather a verbal operator. This he proposes in response to the existence in Mooré
[p.524] and Navajo [p.529] of certain examples in which an internal head appears
to be associated with a universal quantifier (contra his generalization that this configuration
is impossible). These examples look very much like (2.34). Basilico’s
verbal operator theory is compatible with his theory of IHRs because it identifies

39
the head as simply an unquantified nominal, which then provides a variable which
can be bound by an external determiner. Even if it turns out that tukuy is functioning
in sentences such as (2.34) as an A-quantifier (an adverbial-like quantifier,
as opposed to a D-quantifier), however, the clause-escaping ability of that quantifier
still requires explanation. Basilico does not address this problem in his discussion
of Mooré and Navajo.
Another proposal for relative clause structures which takes care of the apparent
disassociation problem but not the island-escaping problem is that of Bach & Cooper
[1978]. This proposal attempts to outfit the adjunction analysis of relative clauses
with an appropriate interpretation scheme. Briefly, Bach and Cooper propose that a
DP can contain a free property variable, and an adjoined relative clause can function
to supply this property. For example, in the English DP in (2.3) (‘the cow that John
bought’), the cow would contain a free property variable, supplied by the property
expressed by the relative clause: being bought by John. This proposal could be
applied directly to the structure of the EHR (2.35) and extended to the IHR (2.34)
if LF head-raising is applied to the entire inner DP. Like Basilico’s verbal operator
proposal, this solution still would not explain how the strong quantifier has escaped
its clause.
An entirely different approach which immediately gives an appropriate structure
for (2.35) is that of Kayne [1994]. Kayne suggests that even externally headed relatives
are derived via (pre-Spellout) head-raising from an initially internally headed
syntactic structure. An S-structure like (2.35) is generated by moving the embedded
IP to adjoin to DP. The relevant S-structure is shown in (2.42).

40
(2.42) Kayne’s head-raising applied to (2.35):
DP
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟
❍
❍
❍
❍
IP i
DP
✏
✟ ✟ ✟ ❍
❍
❍
✏ ✏✏✏✏✏✏✏✏✏✏ D CP
✟
Mayta-q plaza-pi t j planta-sqa-n-ta
✟ ❍ ❍
tukuy NP j CP
Mayta-gen plaza-loc t j plant-nm-3sg-acc all ✏ ✏
✟❍
planta C t i
plant
head.
However, this explanation for (2.35) does not address (2.34), with its internal
Finally, a possibility more in keeping with the intuition that (2.34) and (2.35) are
interpreted in the same way, is that these sentences are related to their paraphrase
(2.40) by movement. This would suggest that the structure in (2.41) is essentially
the correct LF structure also for (2.34) and (2.35). This is the most straight-forward
structure schematically, but the question of when a universally quantified DP may
escape its clause in Cuzco Quechua would need to be resolved, 14 as would the question
of the nature of the trace of the moved quantifier. It is this basic analysis,
however, that I adopt and develop in Chapter 3, building on recent work by Bianchi
[2000] on relative clauses in the framework of [Kayne 1994].
If the LF structures of the tukuy- and pisi- headed clauses are as I have suggested
in (2.36) and (2.41), the question remains as to what induces the different
14 It is clear that this clause escape is specifically associated with a relative clause
head and not, say, any embedded universally quantified phrase. An example of a
clause-bound universally quantified non-head is shown in (i) from the narrative of
Gregorio Condori Mamani [Valderrama & Escalante 1977 p.33]
(i) ...tapu-ra-nku [[papel lliw movilizable-man qo-sqa-nku]]-manta
ask-past-3pl [[paper all mobilized soldier-dat give-nm-3pl]]-abl
‘They asked for the paper that they gave to all “mobilized soldiers”.’
Here, the universally quantified DP lliw movilizable ‘all mobilized soldiers’, not being
the head of the RC, is clearly clause-bound.

41
interpretive positions of these quantifiers. A possible explanation for this difference
lies in syntactic analyses of weak and strong quantifiers which suggest that
these two quantifier types actually occupy different positions within the DP. Here I
adopt the general analysis of DP structure in Bowers [1990]. Bowers suggests that
strong quantifiers are true determiners which occupy the head position of a DP,
whereas weak quantifiers are actually adjectival adjuncts to a Number Phrase (#P)
appearing between DP and NP.
Thus, taking tukuy and pisi to be classic strong and weak quantifiers respectively,
(2.43a) and (2.43b) show the expected structure of two quantified nominals.
(2.43) (a) tukuy llama
all llama
‘all llamas’
DP
✟ ✟ ❍ ❍
D NP
✏✏
tukuy llama
all llama
(b) pisi llama
a few llama
‘a few llamas’
DP
✟ ✟✟ ❍
❍ ❍
D #P
✟ ✟ ❍❍ ❍
AP # ′
✏✏
✟ ❍
pisi # NP
a few ✏✏
llama
llama
If these structures are correct, then clearly pisi is not eligible to occupy the head
position of the outer DP, as was proposed for tukuy in (2.41).
In fact, some immediate evidence for this syntactic analysis comes from the fact
that the position available to tukuy in front of the relative clause is indeed unavailable
to pisi. Thus (2.40) contrasts with the ungrammatical (2.44).
(2.44) *?Pisi
little
[[Asunta-q aqha
Asunta-gen cornbeer
apa-ra-ni.
bring-past-1sg
‘I brought a little cornbeer that Asunta made.’
aqha-sqa-n]]-ta
make.cornbeer-nm-3sg-acc

42
In general, other weak and strong quantifiers in Cuzco Quechua follow the basic
patterns of pisi and tukuy studied here. In particular, universal quantifiers always
take wide scope over the clause regardless of their internal or external surface positions.
Judgments on weak quantifiers other than pisi can be variable, however,
and this issue will be looked at further in Chapter 3. Although the internal reading
seems to be always available, sometimes an external reading is possible too.
Consider, for instance, examples (2.29) and (2.30). Since wakin ‘some’ is strong
in Cuzco Quechua, these sentences mean something like “Some of the houses that
Juan’s father built are big,” and not “Juan’s father built some of the houses and
they are big.” On the other hand, the same sentences with the number kinsa ‘three’,
a weak quantifier, in place of wakin mean “Juan’s father made three houses and
they are big.” In this case, he made exactly three, and they are all big. These
interpretations are in line with the strong and weak quantifier positions suggested
by pisi and tukuy.
2.4.2 Calculating truth conditions
In this section I spell out the technical details of the parsing mechanism for the RC
structures I have proposed.
I begin with the pisi-headed relative clause in sentence (2.31). This sentence is
repeated here as (2.45), and I now gloss its (slightly simplified) structure with the
denotations of each node, in (2.46).
(2.45) [[Asunta-q pisi
Asunta-ge little
apa-ra-ni
bring-past-1sg
aqha
cornbeer
aqha-sqa-n]]-ta
make corn beer-nm-3sg-acc
‘Asunta made a little corn beer and I brought it (the little corn beer that she
made).’

43
(2.46) LF:
IP
true ⇐⇒ Asunta made a little
cornbeer and I brought
the cornbeer that Asunta made
✟
✟ ✟✟✟✟✟✟✟
CP i
true ⇐⇒
Asunta made
a little cornbeer
✏
✏ ✏✏✏✏✏ Asunta-q pisi aqha
Asunta-ge little cornbeer
aqha-sqa-n-ta
make-nm-3sg-acc
❍❍
❍
❍
❍
❍
❍
❍
❍
IP
true ⇐⇒ I brought the
cornbeer that Asunta made
✟ ❍
✟ ✟✟✟
NP
Nuqa
I
❍
❍
❍
❍
VP
[ λx ∈ D e . x brought the
cornbeer that Asunta made]
✟ ❍❍
✟ ✟✟✟✟✟✟
DP
[The cornbeer that
Asunta made]
✟ ❍
✟ ✟✟✟✟✟✟
D
[ λf ∈ D . the maximum entity
y ∈ D e such that f(y) = t]
❍
❍
❍
❍
❍
❍
❍
❍
❍
❍
❍
❍
❍
NP
[ λx ∈ D e . x is cornbeer
that Asunta made]
✟ ❍❍
✟ ✟✟✟
t i
V
[ λx ∈ D e .[ λy ∈ D e .
y brought x ]]
❍
❍
❍
aparani
brought
N
[ λx ∈ D e . x is cornbeer
that Asunta made]
R
[5→ λx ∈ D e . x is cornbeer
that Asunta made]
Note that the truth value of the root IP node is determined by the truth values of
the nodes it dominates in a conjunctive fashion. In this I have followed Shimoyama,
and I also leave open the details of the mechanism which yields the conjunctive
interpretation.
I now turn to the case of the tukuy-headed relative clause, using sentence (2.34)

44
as an example. This sentence is repeated here as (2.47), and as above a slightly
simplified structure is presented here with labels indicating the denotation of each
node.
(2.47) Asunta
Asunta
[[Mayta-q plaza-pi tukuy planta
Mayta-ge plaza-loc all plant
p’iti-ra-n.
prune-past-3sg
‘Asunta pruned all the plants that Mayta planted in the plaza.’
planta-sqa-n]]-ta
plant-nm-3sg-acc
(2.48)
IP
true ⇐⇒ ∀x such that x is a plant
that Mayta planted, Asunta pruned x
❍ ❍
❍
❍
✟ ✟✟✟✟✟✟✟
❍
❍
❍ ❍
DP [λx ∈ D e . Asunta pruned x]
[λf ∈ D . ∀x such that
❍ x is a plant that ✟ ✟✟✟✟ ❍
❍ ❍
Mayta planted x, f(x) = 1] 1 IP
❍
true ⇐⇒ Asunta pruned [[t
✟ ✟✟✟✟
❍
1 ]]
❍
❍
❍
❍
D
CP
✟ ✟✟✟✟
❍
❍
❍
❍
[λx ∈ D e . x is a plant NP
VP
tukuy and Mayta planted
[λx ∈ D 3 .[x pruned [[t 1 ]]]]
all
Asunta
x in the plaza]
✟ ✟✟ ✟ ❍❍
❍
❍
❍ ❍
t 1
✟ ✟✟✟✟✟
IP
true ⇐⇒ Mayta planted
[[t j ]]in ✏ the plaza
✏ ✏✏✏✏ Mayta-q plaza-pi t j
Mayta-ge plaza-loc t j
planta-sqa-n-ta
plant-nm-3sg-acc
❍
❍
❍ ❍
DP j
✏ ✏
planta
plant
V
[ λx ∈ D e .[ λy ∈ D e .
y pruned x ]]
p’itiran
pruned

45
2.5 Further evidence
In this section I will offer further evidence for the generalizations and analyses
presented in the previous sections, and argue against some alternative analyses.
2.5.1 Scope interactions between the relative clause head
and a matrix clause quantified DP
I have found that quantifiers within a clause can engage in scope interactions similar
to those observed in English. One such example is illustrated in (2.49).
Here, the subject is quantified by sapa ‘each’ 15 and the object is quantified by
kinsa ‘three’. Note that the distributive nature of sapa(nka) results in the “three
apples per child” reading of this sentence.
(2.49) Sapa(nka)
each
irqi
child
kinsa
three
manzana-ta
apple-acc
‘Each child took three apples.’
apa-ra-n.
take-past-3sg
It is interesting, then, to note how the scope possibilities are affected by the
presence of a relative clause modifying either the subject or the object. Consider
the sentence (2.50), in which sapanka irqi has become the internal head of a relative
clause.
(2.50) [[Nuqa-q
I-gen
apa-ra-n.
take-past-3sg
sapa(nka)
each
irqi
child
riqsi-sqa-y]]
know-nm-1sg
kinsa
three
‘Each child that I know took three apples.’ (three each)
manzana-ta
apple-acc
15 Some consultants will not accept sapa with this reading without the suffix -nka.
This use of -nka is discussed further in Chapter 6. The current facts remain the
same for these consultants, with sapanka in place of sapa.

46
Note that the matrix clause distribution is still possible, indicating that the
apparently embedded quantifier sapanka in fact is available for scope interactions
within the matrix clause. The situation changes when kinsa manzana ’three apples’
from (2.49) is made the head of a relative clause as in (2.51) and (2.52) where it
is an internal and external head respectively. Some consultants find (2.51) to be
preferable to (2.52), but still feel that in (2.52) a total of only three apples was
bought. 16
16 Similar examples can be found using tukuy ’all’, as shown in (i), (ii) and (iii)
(comparable to (2.49), (2.50) and (2.51)). However, for some consultants the narrow
scope reading of huk ‘one’ with respect to tukuy ‘all’ is hard to get without a
distributive suffix -nka on the object (cf. [Faller 2001]). I discuss data involving the
distributive suffix in Section 2.5.3.
(i) Tukuy llama-(kuna) huk platanu-ta mikhu-ra-nku.
all llama-(pl) one banana-acc eat-past-3pl
‘All the llamas ate one banana.’ (one each or one total)
(ii) [[Juan-pa tukuy llama ranti-sqa-n]] huk platanu-ta mikhu-ra-nku.
Juan-gen all llama buy-nm-3sg one banana-acc eat-past-3sg
‘All the llamas that Juan bought ate one banana (each).’
‘All the llamas that Juan bought ate one banana (together).’
(iii) Tukuy llama [[nuqa-q huk platanu ranti-sqa-y]]-ta mikhu-ra-nku.
All llama I-gen one banana buy-nm-1sg-acc eat-past-3pl
‘I bought one banana and all the llamas ate it.’ (All the llamas ate the one
banana that I bought (together)).
*‘All the llamas ate one banana that I bought (each).’
In (ii), the same basic sentence as (i) is repeated but this time the subject tukuy
llama is the internal head of a relative clause. Consultants report that both translations
given in the glosses are appropriate here, too, indicating that the ability of the
head to interact with another matrix clause quantifier is not affected by the relative
clause. This is consistent with an analysis in which the universally-quantified head
is external to the relative clause at LF. By contrast in (iii), in which the object huk
platanu is the head of an IHR, consultants report that the sentence must mean that
one banana total was bought, effectively indicating that huk platanu ‘one banana’ is
no longer participating in scope interactions within the matrix clause. This is consistent
with the E-type anaphora analysis of relative clauses with weakly quantified
heads.

47
(2.51) Sapa(nka)
each
irqi
child
[[nuqa-q
I-gen
kinsa
three
manzana
apple
mikhu-ra-n.
eat-past-3sg
‘I took three apples and each child ate them.’
apa-sqa-y]]-ta
take-nm-1sg-acc
(2.52) Sapa(nka)
each
irqi
child
[[nuqa-q
I-gen
apa-sqa-y]
take-nm-1sg
mikhu-ra-n.
eat-past-3sg
‘I took three apples and each child ate them.’
kinsa
three
manzana]-ta
apple-acc
Consultant comment on both sentences: A total of three were bought. The
children may have shared them.’
2.5.2 Scope interactions between the head and the subject
Further evidence for the E-type anaphora analysis of relative clauses with weakly
quantified heads, and against an analysis in which a raised head sometimes has
a forced definite interpretation comes from sentences in which the subject of the
relative clause is quantified. An example is given in (2.53). Here, the internal head
is quantified by huk, ‘one’, which appears to induce a narrow scope interpretation
as I have proposed for pisi.
(2.53) Asunta mikhu-ra-n [[sapa irqi-q huk t’anta
Asunta eat-past-3sg each/every child-gen one bread
ranti-sqa-n]]-ta
buy-nm-3sg-acc
‘Each child bought one roll and Asunta ate them (one roll per child).’
Note that (2.53) does not mean ‘Asunta ate one roll that each child bought’ (even
on the reading where ‘each child’ seems to escape the relative clause to gain scope
over ‘one roll’ 17 ) because it contains the information that each child bought exactly
one roll. Furthermore, (2.53) raised no problems for my Quechua consultants, while
17 Abusch [1994] notes that English each does seem to have this clause-escaping
property.

48
some English consultants find the sentence “Asunta ate the one roll that each child
bought,” to be understandable with the meaning indicated above, but to be semantically
odd and/or difficult to parse. 18 Other consultants find it understandable only
with the unlikely reading that there is only one roll which was bought individually
by each child. In the Quechua sentence, if we were to imagine that the head huk
t’anta undergoes LF raising, and then for some reason is obligatorily associated with
a null definite determiner (this is the analysis which I argue against in Section 2.4)
we would expect (2.53) to run into the same problem as is found in the English
version. On the other hand, an E-type anaphora analysis allows us to interpret first
the relative clause ‘each child bought one roll’, and then the matrix clause: Asunta
ate the maximal entity which is rolls bought by the children.
2.5.3 Interaction with distributivity
Further evidence for the close relationship between the head of a relative clause and
the embedded verb suggested by the E-type anaphora analysis is provided by data
involving the distributive 19 suffix -nka (studied in detail in [Faller 2001]). This suffix
18 Sharvit [1996] explores acceptable English sentences with (externally headed)
relative clauses resembling the IHR in (2.53). She advocates a functional analysis
of the so-called “multiple individual reading” of sentences such as (i).
(i) The woman every man invited to the party was his mother. [Sharvit 1996, p.3]
(The relevant readings are those in which with the woman varies with the man.)
Rejecting analyses in which every man escapes its clause to take matrix scope,
Sharvit argues that this effect is achieved because the operator trace in the relative
clause is interpreted as a function variable (of type ). Because Sharvit’s
analysis relies on the presence of operator movement and not head-raising, it is not
easily applicable to the case of Quechua IHRs. Furthermore, since there is no definite
marking on huk, a functional analysis would not explain the unavailability of the
reading in which each child bought several rolls and Asunta ate one roll from each
child’s stash. These considerations lead me to conclude that the E-type anaphora
analysis is the correct one for Cuzco Quechua.
19 This is the traditional description of this suffix, but Faller [2001] gives evidence
that in fact it serves a more complicated function than simple distributivity, and at
least in some uses takes on a group-forming function.

49
typically appears on either the noun or the quantifier of certain quantified DPs (its
acceptability depending on the quantifier). In this section I will first summarize the
relevant uses and distribution of -nka, then show how -nka interacts with relative
clauses.
Examples (2.54) to (2.58) illustrate the compatibility and interpretation of -nka
with various DP types.
First, -nka is compatible with the weak quantifiers pisi ‘few/a little’, ashka
‘many/a lot’ and huk ‘one’ (and other numbers). In its distributive use, it marks the
DP which is being distributed. 20 It indicates that the marked DP will be distributed
in units of the size specified by the quantifier. Examples with ashka ‘a lot’ (2.54),
(2.55) and iskay ‘two’ (2.56) are shown below. The suffix -nka can optionally surface
on the quantifier or on its sister noun, as seen in (2.54) and (2.55). When on a direct
object noun, the accusative marker becomes optional, which I take to be a purely
phonological effect.
(2.54) Runa-kuna-man-qa
person-plural-dat-top
qu-yku-nki.
give-intensifier-2sg(fut.)
ashka-nka
much-nka
coca-cha-ta
coca-dimin-acc
‘You will give a lot of coca (leaves) to each person.’
(2.55) Runa-kuna-man-qa
person-plural-dat-top
qu-yku-nki.
give-intensifier-2sg(fut.)
ashka
much
coca-cha-nka-(ta)
coca-dimin-nka-(acc)
‘You will give a lot of coca (leaves) to each person.’
(2.56) Pisi
a few
runa
person
iskay
two
wik’uña-nka
vicuña-nka
riku-ra-nku.
see-past-pl
‘A few people saw two vicuñas each.’
20 This DP is called the Distributive Share in the terminology of [Choe 1987].

50
However, -nka may not appear on nominals quantified with the strong quantifiers
tukuy or wakin (as in (2.57)), or on the bare noun cocacha ‘coca leaves’ in (2.58). 21
(2.57) Runa-kuna-man-qa wakin coca-cha-(*nka)-ta qu-yku-nki.
person-pl-dat-top some coca-dimin-nka-(acc) give-intensifier-2sg.fut
‘You will give some of the coca leaves to each person.’
(2.58) *Runa-kuna-man-qa coca-cha-nka-(ta)
person-pl-dat-top coca-dimin-nka-(acc.)
‘You will give coca leaves to each person.’
qu-yku-nki.
give-intensifier-2sg.fut
Now we may ask whether the head of a relative clause can be marked with
distributive -nka. Note that the incompatibility of this use of -nka with strong
quantifiers makes the question only relevant to weakly quantified heads.
I find that -nka may only appear on the head of an IHR when the subordinate
verb supports a distributive interpretation with respect to that head. For example,
sentences (2.59) and (2.60) illustrate IHRs in which the head is marked by the
distributive use of -nka.
(2.59) [[Huk
one
platanu-nka
banana-nka
ranti-sqa-y]]-ta
buy-nm-1sg-acc
tukuy
all
llama
llama
mikhu-ra-nku.
eat-past-pl
‘I bought (the llamas) one banana each and all the llamas ate them.’
Consultant comment: I bought one each, but did they actually eat one each,
or perhaps some had half a banana and others had more?
21 It is not the case that -nka is always incompatible with strong quantifiers, however.
In the following example, -nka appears twice, and in sapa-nka irqi ‘each-nka
child’, it is associated with the strong quantifier sapa. Here, the role of -nka can be
seen as group-forming, while the second use marks the Distributive Share.
Sapa-nka irqi pisqa t’anta-nka mikhu-nqa-ku.
Each-nka child five bread-nka eat-fut-pl
‘Each child will eat five rolls.’
For current purposes, however, I concentrate on the simple distributive use of this
suffix.

51
(2.60) [[Juan-pa
Juan-gen
iskay
two
wik’uña-(*nka)
vicuña-nka
muna-sqa-n]]-ta
like-nm-3sg-acc
pisi
a few
runa
person
riku-ra-nku.
see-past-pl
‘Juan likes two vicuñas and a few people saw them.’
(One consultant points out that (still without -nka) this can also mean: ‘A
few people saw that Juan likes two llamas.’)
Note that in (2.59), ‘buy’ is compatible with a distributive interpretation, and
in fact speakers report that the sentence conveys that I had (deliberately) bought
enough bananas so there would be one per llama. The embedded verb ‘like’ in (2.60)
is incompatible with such an interpretation. 22 Therefore this sentence is acceptable
only if the suffix -nka is omitted.
Note that these results are surprising under any analysis in which the head of
the IHR raises at LF to a position in which it is external to the relative clause.
For example, in the English sentence ‘A few people saw two vicuñas that Juan
likes’ there is no problem distributing pairs of vicuñas over people. However, an
analysis such as the one I have outlined in the preceding sections in which an IHR
is interpreted as a sentence would predict that the distributive nature of -nka must
be compatible with this subordinate clause.
There remains one mystery, which I will leave to further research. It appears
that there is some level of incompatibility between an external head and the suffix
-nka. For example, the EHR version of (2.59) is unacceptable, as seen in (2.61).
(2.61) *[[Ranti-sqa-y]
buy-nm-1sg
huk
one
platanu-nka](-ta)
banana-nka-acc
tukuy
all
llama
llama
mikhu-ra-nku.
eat-past-pl
‘I bought one banana each (per llama) and all the llamas ate them.’
Sentence (2.61) is simply not understandable to my Quechua consultants, and in
fact the distinction between the acceptable IHR sentence (2.59) and the unacceptable
22 One consultant did suggest that (2.60) could possibly be acceptable with -nka
under a reading in which Juan likes vicuñas in pairs. This is consistent with the
group-forming uses of -nka discussed in Faller [2001].

52
EHR sentence (2.61) is the strongest acceptability contrast I have found between
EHR and IHR pairs with direct object heads.
2.6 Conclusion
In this chapter I have shown that the meaning of Cuzco Quechua relative clauses is
not captured by previously proposed relative clause interpretation schemes. Specifically,
I have shown that the position in which the head of a relative clause is interpreted
depends on the quantifier associated with that head. I have drawn contrasts
with Japanese, in which the surface position of the head determines its interpretive
position. These facts have led me to propose that while relative clauses with head
marked by pisi ‘a little’ are amenable to an E-type anaphora analysis, those whose
head is marked by tukuy ‘all’ appear to involve head-raising. In this second class of
relative clause, the universal quantifier is interpreted outside of its clause at LF.
These findings are in general support of the idea expressed in [Basilico 1996],
[Hoshi 1995], [Shimoyama 1999, 2001] and others that relativization is not achieved
cross-linguistically by a single syntactic structure. They also show that, to adopt
the vocabulary of the relative clause typology in [Grosu 2002] 23 , both restrictive
(head-raising) and “maximalizing” relative clauses can appear in a single language.
Although my proposal contradicts predictions of [Williamson 1987] and [Culy 1990]
since I have found universally quantified internal heads in Cuzco Quechua, the data
nonetheless supports the more general point made by Williamson and Culy, that
strong and weak DPs behave differently as the head of a relative clause. In Cuzco
23 [Grosu 2002], building on [Grosu & Landman 1999] and much other work, looks
at a broad range of languages and relative clause types. Although here I am unable
to examine fully the significance of Quechua within the resultant typology, I hope
that the present focused study, revealing some of the diversity of Quechua both
within and between dialects, will lead to future work of this sort.

53
Quechua, both types of internal heads are allowed, but the semantic patterns exhibited
by the two are quite distinct. In the next chapter I turn to the question of the
exact syntactic structure of the two relative clause types proposed here, and offer
an explanation for the apparently typologically unusual behavior of Cuzco Quechua
with regard to relative clauses, based on Case-marking patterns and features on
functional heads of the clause.

Chapter 3
The Syntax of Head-Raising in
Cuzco Quechua
3.1 Introduction
We have seen in Chapter 2 that in Cuzco Quechua relative clauses, both internal
and external positions are available for the head. Furthermore, we have seen that at
least in the case in which the head is quantified with a universal quantifier, the head
takes scope over the entire relative clause, regardless of its surface position. This
chapter examines the syntax of this construction in detail. In particular, I address
the question of what the detailed relative clause structure must look like in Cuzco
Quechua, and how the quantifier can gain scope over its associate as well as over
the relative clause. I develop a theory by which the scope effect is achieved when
the quantifier is incorporated into the external determiner head. In this view, the
failure of weak quantifiers to achieve wide scope is due to the fact that they are not
determiners and hence cannot undergo determiner incorporation.
Although I have focused on relative clauses with quantified heads in order to
detect semantically the interpretive scope relationships, in fact raising analyses have
54

55
been proposed for non-quantified relative clauses also. Consider the internally and
externally headed relative clause pair in (3.1) and (3.2).
(3.1) [[Juan-pa waka ranti-sqa-n]]-qa yuraq-mi
Juan-gen cow buy-nm-3sg-top white-aff
‘The cow that Juan bought was white.’
(3.2) [[Juan-pa ranti-sqa-n] waka]-qa yuraq-mi
Juan-gen buy-nm-3sg cow-top white-aff
‘The cow that Juan bought was white.’
ka-rqa-n.
be-past-3sg
ka-rqa-n.
be-past-3sg
In [Cole 1987a], it is argued that Quechua 1 sentences such as (3.1) and (3.2) have
identical LF structures, related by LF head-raising. That is, Cole proposes that the
head waka of (3.1) raises at LF to the position to the right of the clause, allowing
both (3.1) and (3.2) to be interpreted in the same way. This idea is supported,
with some modifications, by [Kayne 1994], and in my analysis too there is a raising
relationship between the heads in these two examples.
In addition to object heads as in the above examples, subjects can also appear
as internal or external relative clause heads. In this case the nominalizer -q appears
on the verb. The syntax of these relative clause types will also be discussed in
this chapter. Examples are shown in (3.3) and (3.4), illustrating an EHR and IHR
respectively.
(3.3) [Chay [waka ranti-q] runa]-qa hatun-mi.
dem cow buy-nm man-top tall-evid
‘That man who bought the cow is tall.’
(3.4) [Chay
dem
[runa
man
waka
cow
ranti-q]]-qa
buy-nm-top
‘That man who bought the cow is tall.’ 2
hatun-mi.
tall-evid
1 Cole’s work focuses on Ancash and Imbabura Quechua.

56
Note that for a head-raising analysis to be viable, it requires that both internally
headed relative clauses (IHRs) and their externally headed relative clause (EHR)
counterparts have the same truth conditions. It also requires that the head of the
relative clause have interpretive scope over the remainder of the clause. In Chapter 2,
I showed that the first of these conditions is true in general of Cuzco Quechua relative
clauses by examining the truth conditions of IHRs and EHRs with quantified heads.
However, I argued that only in some cases do Quechua relative clause heads take
interpretive scope over the clause. Relevant examples with universally quantified
heads are given in (3.5) and (3.6).
(3.5) Asunta
Asunta
[[Mayta-q plaza-pi planta-sqa-n] tukuy
Mayta-ge plaza-loc plant-nm-3sg all
p’iti-ra-n.
prune-past-3sg
‘Asunta pruned all the plants that Mayta planted in the plaza.’
(3.6) Asunta
Asunta
planta]-ta
plant-acc
[[Mayta-q plaza-pi tukuy planta
Mayta-ge plaza-loc all plant
p’iti-ra-n.
prune-past-3sg
‘Asunta pruned all the plants that Mayta planted in the plaza.’
planta-sqa-n]]-ta
plant-nm-3sg-acc
In sentences like those in (3.3) and (3.4) in which the head is overtly a bare
nominal, there is no immediate semantic criterion for deciding the scope of the
head. The example sentences in this chapter contain relative clauses with bare noun
phrase heads as well as some with quantified heads. To account for the possibility of
the head taking scope over the whole clause, I adopt a head-raising analysis based on
2 My consultants are often more hesitant to accept subject internal heads than
they are to accept object internal heads. The additional cue of the demonstrative
chay seems to help these consultants recognize that a relative clause is intended. I
do not at present have an explanation for this subject/object asymmetry. Past work
including [Lefebvre & Muysken 1988] suggests that subject IHRs are unproblematic
for many speakers.

57
that in [Kayne 1994] as modified in [Bianchi 2000]. However I argue that in Cuzco
Quechua this raising is explicitly connected to Case-marking and Case-checking.
Evidence comes from the fact that an internal head cannot be Case-marked. In
transitive main clauses I assume that the Case-checking of internal arguments is
achieved in the presence of functional heads possessing uninterpretable Case features
and associated with Transitivity (following ideas in [Koizumi 1995], [Collins 2001,
2003] and [Bowers 2002]).
The organization of this chapter is as follows. In the next section I present
some more background on Quechua morphosyntax. In section 3.3 I propose explicit
syntactic structures for Cuzco Quechua relative clauses. In section 3.4 I broaden
the analysis by showing how subject relativization and weakly quantified heads fit
into the proposed structures. Section 3.5 is the conclusion.
3.2 Morphosyntactic facts of Cuzco Quechua relatives
In this section I expand on the morphosyntactic background given in Chapter 1 to
provide more details relevant to the relative clause structures studied in this chapter.
3.2.1 Agreement and Case
In Cuzco Quechua, verbs nominalized with -sqa or -na agree overtly with their subjects.
Subject-nominalized relatives, with the -q nominalizer, do not. This adheres
to a broader generalization in Cuzco Quechua: possessed nouns agree with their
possessors, as illustrated in (3.7) and (3.8).

58
(3.7) Juan-pa wasi-n
Juan-gen house-3sg
‘Juan’s house’
(3.8) Nuqa-q wasi-y
I-gen house-1sg
‘My house’
This generalization is relevant because subjects are genitive-marked in nonsubject
relativized clauses 3 and in complement clauses as illustrated in the relative
clause in (3.9). Thus, genitive marking of a subject in both cases is associated
with a (nominal) agreement pattern on a lower noun or nominalized verb. These
cases consequently both contrast with subject RC heads as in (3.10), which are not
genitive-marked and do not agree with the nominalized verb. Note that in (3.9) but
not (3.10) the verb agrees with the subject ‘boy’.
(3.9) [wayna-q
boy-gen
waka
cow
ranti-sqa-n]
buy-nm(past/nonsubj)-3sg
‘the cow which the boy bought’
(3.10) [wayna
boy
waka
cow
ranti-q]
buy-nm(subj)
‘the boy who bought the cow’
I now turn to the question of Case-marking of internal arguments of the nominalized
verb. According to the judgments of my consultants, in Cuzco Quechua
3 Lefebvre & Muysken [1988] found that the genitive marking is optional in some
cases. This is part of the basis for a distinction they make between NP and VP
projections of subordinate nominalized verbs (and this distinction is in turn a core
aspect of a more general theory of “mixed categories”). In the case of a null-marked
subject, which they take to show that the nominalized verb heads a VP projection,
they found that the accusative Case marker -ta was optional (though dispreferred)
on a direct object. These differing judgments will be discussed further in section
3.4.2. It is important to note at this stage, however, that my consultants and those
of Lefebvre and Muysken all agree that in the case where the direct object is the
internal head of a relative clause, it cannot be overtly Case-marked.

59
the direct object of a nominalized clause (for instance, waka ‘cow’ in (3.9)) has no
overt Case-marker. 4 Furthermore, other subordinate arguments of the verb 5 have
no overt Case-marking when they are the internal head of a relative clause as illustrated
in (3.11). Note that in a Main clause these Case-markers are not optional,
as illustrated in (3.12). A similar pair of examples is shown in (3.13) and (3.14).
(3.11) [Nuqa-q llaqta-*pi/*-ta/-∅
I-gen town-loc/-acc/-∅
‘The town I live in is big.’
tiya-sqa-y]]-qa
live-nm(past/nonsubj)-1sg-top
hatun.
big
(3.12) Nuqa
I
kay
this
llaqta-pi/*ta/*∅
town-loc/-acc/-∅
‘I live in this town.’
tiya-ni.
live-1sg
(3.13) [[Kama-*pi/*ta/∅] puñu-sqa-y]-qa
bed-loc/sleep-nm-1sg-top big-evid
‘The bed that I slept in is big.’
hatun-mi.
(3.14) Nuqa
I
puñu-ni
sleep-1sg
huk
one
‘I sleep in a big bed.’
hatun
big
kama-pi/*ta/*∅.
bed-loc/-acc-∅
It is important to emphasize at this point that while the -ta marker may be
missing on any direct object of a nominalized clause, other markers like -pi above
are only dropped on the head of a relative clause.
3.2.2 What can be a head
In general, and in keeping with the preceding discussion, only DPs with no overt
Case marker can be internal heads. This fact is illustrated in the following examples,
4 Again see Section 3.4.2 for discussion of data, reflecting different judgments, in
which the -ta marker is allowed under some circumstances.
5 I use the expression “argument” to refer to DPs which have a semantically close
relationship to the verb, and which can appear as the head of an IHR in Cuzco
Quechua.

60
which show a subject internal head (3.15), a direct object internal head (3.16), a
locative object internal head (3.17) and an ablative internal head (3.18).
(3.15) [[Runa-(*q) hamu-q]]-ta riqsi-ni.
person-gen come-nm-acc know-1sg
‘I know the person who is coming.’
(3.16) [[Juan-pa runa-(*ta) riqsi-sqa-n]]-wan
Juan-gen person-acc know-nm-3sg-com
‘I spoke with the person who Juan knows.’
rima-rqa-ni.
speak-past-1sg
(3.17) [[Wasi-(*pi) tiya-sqa-y]]-ta riqsi-nki-chu?
House-loc live-nm-1sg-acc know-2sg-Q
‘Do you know the house I live in’
(3.18) ?[[Runa-(*?manta) rima-sqa-y]]-qa
person-abl speak-nm-1sg-top
‘The person who I talked about came.’
hamu-rqa-n.
come-past-3sg
The acceptability of the last example is unclear since consultants’ judgments
vary. In fact, internal adjunct heads seem to be unacceptable regardless of the
presence or absence of a Case marker. Examples are shown in (3.19) and (3.20).
(3.19) *[[Lapiz-(wan) qilqa-sqa-y]]-qa
pen-com write-nm-1sg-top
‘The pen that I wrote with is blue.’
azul-mi.
blue-evide
(3.20) [[Rumi
stone
ñan-(pi)
road-loc
riku-sqa-y]]
see-nm-1sg
yana.
black
*’The road in which I saw the stone is black.’
‘The stone that I saw in the road is black.’ (with -pi)
Thus the acceptability of (3.18) seems to be a result of the status of the ablative
element as an argument of the verb for some consultants. 6
6 Unlike examples (3.15)–(3.17), some consultants do find (3.18) to be somewhat
improved with the Case-marker than without it, which is a violation of the general-

61
It thus follows from the discussion so far that direct objects and other nullmarked
internal arguments can be Cuzco Quechua internal heads. Furthermore,
subjects may be internal heads as long as the subject-head nominalizer -q appears
on the verb, since in these cases subjects have no overt Case marking. (Recall
that the nonsubject-head nominalizers -sqa and -na appear with genitive-marked
subjects.) In fact, this is the totality of the internal heads that I have found in
Cuzco Quechua.
External heads vary much more widely. Not only do all of the examples (3.15)–
(3.18) have externally headed versions, but the adjunct relativization cases do too
((3.19) and (3.20)). For completeness I include these EHRs below, in (3.21)–(3.26).
(3.21) [[Hamu-q] runa]-ta riqsi-ni.
come-nm person-acc know-1sg
‘I know the person who is coming.’
(3.22) [[Juan-pa riqsi-sqa-n] runa]-wan rima-rqa-ni.
Juan-gen know-nm-3sg person-com speak-past-1sg
‘I spoke with the person who Juan knows.’
(3.23) [[Tiya-sqa-y]
live-nm-1sg
wasi]-ta
House-acc
riqsi-nki-chu?
acc know-1sg-Q
‘Do you know the house I live in’
(3.24) [[Rima-sqa-y] runa]-qa hamu-rqa-n.
speak-nm-1sg person-top come-past-3sg
‘The person who I talked about came.’
(3.25) [[Qilqa-sqa-y]
write-nm-1sg
lapiz]-qa
pen-top
azul-mi.
blue-evide
‘The pen that I wrote with is blue.’
ization I have made above regarding the prohibition on Case-markers on an internal
head. These consultants find the externally headed version to be the truly correct
form, however. I hypothesize that for these consultants, the presence of the Casemarker
may be a way of “saving” the adjunct internal head by simply allowing for
the embedded clause to be interpreted as a sentence. This mechanism is highly
marked within Cuzco Quechua, however.

62
(3.26) [[Rumi
stone
riku-sqa-y]
see-nm-1sg
ñan]
road
hatun.
black
‘The road in which I saw the stone is black.’
3.2.3 Nominalizing morphology
As has been previously mentioned, Cuzco Quechua can distinguish by means of its
nominalization morphology on the verb of a relative clause whether it is the subject
or a non-subject of the clause that is the head. The examples given above illustrate
this. Notice that in (3.15) and (3.21) the nominalizer is -q while in (3.16)–(3.18)
and (3.22)–(3.26) the nominalizer is -sqa. The same generalization holds for both
internally headed and externally headed relatives.
Nominalizing suffixes can also contain information about relative tense (main
clause vs. subordinate clause). Although the subject nominalizer -q is neutral
for tense as illustrated in (3.27), the non-subject nominalizers -sqa and -na reflect
that the subordinate clause action is prior to or later than the main clause action,
respectively. Examples are given in (3.28) and (3.29).
(3.27) [[Ripu-q]
leave-nm
runa]
person
‘The man who left/is leaving/will leave’
(3.28) [[Riku-sqa-y] runa]
see-nm-1sg person
‘The man who I saw’
(3.29) [[Riku-na-y] runa]
see-nm-1sg person
‘The man who I will see’
Also: ’The man who I have to see’

63
3.3 Syntax of Cuzco Quechua Relative Clauses
In this section I look at the mechanism of head-raising in Cuzco Quechua relative
clauses. I will focus for the present on the core case of internal argument raising.
In the next section I look at subject heads and also at relative clauses with weakly
quantified heads. Here I will suggest an implementation of head-raising in Cuzco
Quechua which explains the patterns of Case-marking, etc. seen in the previous
sections.
3.3.1 Basic framework
Since I am not aware of any evidence in Quechua head-raising constructions for a
category (say, NP) between the DP which dominates the relative clause and the
clause itself, I will assume with Kayne [1994] that CP is the sister of the determiner
D. There are, of course, two basic possibilities for how the head-raising proceeds.
The head could raise to the right, as proposed in [Cole 1987a] and the clause itself
remains insitu. Alternatively, the head could raise to the left and the clause could
raise beyond it to achieve the EHR S-structure word order. These possibilities are
illustrated in (3.30) and (3.31) respectively.
(3.30) DP
✟ ✟✟ ❍
❍
❍
D CP
✟ ✟ ❍ ❍
CP DP i
✏ ✏
...t i ...
(3.31) DP
✟ ✟✟ ❍
❍
❍
CP j DP
✏ ✏ ✟ ✟ ❍ ❍
...t i ... D CP
✟ ❍
DP i
t CPj

64
The tree in (3.30) shows the S-structure of an EHR and the LF of either an EHR
or an IHR. Similarly, (3.31) gives the S-structure of an EHR. According to Kayne
[1994], the S-structure of an IHR under a left-movement analysis is achieved under
the copy theory of movement ([Chomsky 1995]) by simply spelling out the trace t i
and not the head DP i . In (3.31), of course, we must assume reconstruction of CP j
to its base position at LF in order for interpretive scope to be given to the head
DP. Given that I am not aware of Quechua-internal evidence in favor of this more
complicated derivation, I will currently assume (3.30). In fact, one piece of evidence
explicitly in favor of (3.30) is that the determiner can overtly appear in either the
external D position as in (3.32) or (3.33) or in the head of DP i position.
(3.32) [Tukuy [Juan-pa waka ranti-sqa-n]] yuraq-mi.
all Juan-gen cow buy-nm-3sg white-evid
‘All the cows that Juan bought are white.’
(3.33) [Tukuy [Juan-pa ranti-sqa-n] waka]
all Juan-gen cow buy-nm-3sg
‘All the cows that Juan bought are white.’
yuraq-mi.
white-evid
Under the analysis in (3.31), we would have to say that in (3.32) there is either
iteration of the DP projection (so tukuy is not in fact the D head of the DP represented
in the tree above), or that in this case, exceptionally, the relative clause has
not raised to Spec,DP but remains in the t CP position at Spell-Out. In fact, I believe
that either of these possibilities could be implemented along with the analysis I give
in the following sections.
There is one non-standard aspect common to both trees which needs to be explained.
Note that in both cases, the head nominal is the determiner phrase DP i .
Under standard assumptions like those in [Kayne 1994] the head is an indefinite of
category NP. However, here I follow the discussion in [Borsley 1997], picked up in
[Bianchi 2000], where it is argued that the head is a full DP. Borsley points out that

65
since the head noun phrase occupies an argument position within the relative clause,
we would expect it to be generated as a DP (presumably the category licensed in
argument positions in a main clause). He presents a variety of evidence using binding,
island and parasitic gap tests showing that the gap within the relative clause
behaves like a DP-trace. In fact, in Quechua there is independent evidence for this
assumption: the head can be universally quantified.
Bianchi [2000] accepts these arguments and further suggests that in English
head-raising structures, the null determiner D on the raised head is licensed by
incorporation into the determiner of the dominating DP. Thus she proposes that
the structure of the English relative clause in (3.34) is as in (3.35).
(3.34) the picture that Bill liked
(3.35) DP
❍ ❍
✟ ✟✟✟✟ ❍
❍ ❍
D Rel +the CP
✟ ❍
✟ ✟✟ ❍
❍
❍
DP i CP
✟ ❍
✟
t picture
✟✟ ❍ ❍
that IP
✏ ✏✏
Bill liked t i
(Bianchi 2000 (4))
A few aspects of Bianchi’s analysis need to be clarified at this point. First of all,
this analysis would seem to have a problem similar to the original NP head analysis
of Kayne’s: just as a bare NP cannot be generated in argument position, how is it
that a DP with a null determiner can be generated here. That is, although (3.34)
is a grammatical expression, the sentence version of its relative clause is not:
(3.36) *Bill liked [∅ picture].

66
Bianchi suggests that it is the incorporation which licenses the null determiner
in English. Therefore, (3.36) is bad because the null determiner never incorporates
into another determiner.
There is also a question of how exactly the determiner incorporation is licensed.
Bianchi proposes that the feature structures (as in the work of Pollard and Sag
[1994]) of the two determiners must be compatible in the sense that for any given
feature, the two morphemes are either identically specified for that feature, or one
of them is underspecified for that feature. In our case of two determiners, the null
determiner of the relative clause head is underspecified for definiteness. Thus, the
definiteness feature of the external head (+ or - definite) determines the definiteness
of the resulting “unified” D head. Furthermore, both determiners share the
φ features of the NP part of the head of the relative clause. Bianchi notes that
incorporation is plausible since once the head has raised to Spec,CP the two determiners
are in a strictly local configuration: the outer D immediately C-commands
the lower one. She argues that under these circumstances, the actual incorporation
step is licensed by an economy principle:
(3.37) Economy of Representation (Bianchi 2000 (8))
Incorporate a functional head to a host whose feature structure is consistent
with its own.
I adopt a similar analysis for Quechua, but assume that the reverse licensing
can also take place, in which case it is the dominating DP’s head D which is null
while the raised DP is overt. 7 Details of this analysis are provided in the following
sections.
7 A variation on the null determiner analysis pointed out to me by Chris Collins
(p.c.) is that the null determiner is actually a null wh-operator, so (3.36) would be
bad for the same reason that “*Bill liked which picture.” is bad. Under this analysis,
Determiner incorporation would not be a necessary operation in English (and indeed
would be prohibited by a feature class between ‘the’ and the wh-operator). Even if

67
3.3.2 Explaining the Case-marking pattern
Now let us turn to some of the details of Cuzco Quechua. Consider the curious
pattern of Case-marking within subordinate clauses as opposed to main clauses.
Subordinate clause null-marking has traditionally been seen as simply a direct object
Case marker (e.g. [Lefebvre & Muysken 1988]), parallel to -ta in main clauses.
However, I noted in Section 3 that in (3.11) and (3.12), repeated here as (3.38)
and (3.39), we see that a non-direct object argument can also be null-marked in a
nominalized clause but cannot be -ta marked in a main clause.
(3.38) [Nuqa-q llaqta-*pi/*-ta/-∅
I-gen town-loc/-acc/-∅
‘The town I live in is big.’
tiya-sqa-y]]-qa
live-nm(past/nonsubj)-1sg-top
hatun.
big
(3.39) Nuqa
I
kay
this
llaqta-pi/*-ta/*-∅
town-loc/-acc/-∅
‘I live in this town.’
tiya-ni.
live-1sg
On the other hand, a non-head argument (other than the direct object) must be
marked with its regular Case-marker, as shown in (3.40).
(3.40) Juan
Juan
[[nuqa-q
I-gen
kay
this
llaqta-*(pi)
town-loc
tiya-sqa-y]]-ta
live-nm(past)-1sg-acc
‘Juan knows that I lived in this town.’
yacha-n.
know-3sg
How can we explain these different Case paradigms in Main and Subordinate
clauses? It could be that nominalized clauses offer a special option which is not
available in Main clauses: the null-marked Case. Or, it could be that the Main
clause offers a special option not available in nominalized clauses: the -ta-marked
Case. One version of the first possibility is that the accusative DP can incorporate
into the subordinate verb, and that this process is licensed by the nominalization.
this is correct for English, in Quechua we need determiner incorporation to explain
the ultimate scope of the strong quantifiers.

68
This is the analysis given in [Cole 1985] for Imbabura Quechua. However, at least
in the case of Cuzco Quechua there are some problems with this idea (the case of
Imbabura Quechua will be discussed in Chapter 4). The null-marked DP does not
need to be a bare noun but can have a specifier and complement. An example of
this is shown in (3.41).
(3.41) [Chay [hatun wawa-kuna riqsi-sqa-y]]-qa futbalista-kuna
dem tall child-pl know-nm-1sg-top football player-pl
‘Those tall children that I know are football players.’
ka-nku.
be-3pl
Furthermore, even when scrambling occurs in the subordinate clause, the direct
object is null-marked, making it unlikely that it was ever incorporated into the
nominalized verb. An example of a scrambled internal direct object head is given
in (3.42).
(3.42) [[Juan-pa regalo Mariya-man qu-sqa-n]]-ta
Juan-gen gift Maria-dat give-nm-3sg
‘I like the gift that Juan gave Maria.’
muna-ni.
like-1sg
Let us instead pursue the second possibility: that an option available in the Main
clause is unavailable in the subordinate clause. This option has the immediate
advantage of adhering to Ross’s [1973] “Penthouse Principle” which states that
any Subordinate clause structural possibilities are also available in Main clauses.
Furthermore, recent work on transitivity by [Koizumi 1995], [Collins 2001, 2003] and
[Bowers 2002] suggests that transitivity can be associated to a clausal projection
of a transitivity head. Under this view, we can say that the Transitivity Phrase
responsible for assigning -ta-marked Case is unavailable in subordinate clauses.
The TrP I am referring to here is that suggested in [Bowers 2002]. The transitivity
phrase discussed in [Collins 2003] is somewhat different (associated to the
presence of an additional argument) and in fact I will argue below that this phrase,

69
too, is needed in Quechua. Thus I will effectively be arguing that all structural Case
is associated to clausal projections with functional Case-assigning heads.
3.3.2.1 Case Marking in a Main Clause: the TrP projection
Let me begin by introducing TrP in a Main transitive clause. The role of Tr is to
check Accusative Case (the idea of a functional head to check accusative Case is
developed in [Koizumi 1995]) and its presence is overtly manifested in the presence
of -ta on the direct object. That is, a noun phrase which merges with this functional
head must be Case-marked with -ta. I assume that the verb in a main clause raises to
a v head (as is standard based on the work in [Chomsky 1995, 2000, 2001a] (called Pr
in [Bowers 1993, 2002]) and that the subject is generated in the Specifier position
of this head. In the case of nominalized clauses with genitive subjects, I posit
instead the presence of n as in [Chomsky 1995], which I assume to be manifested
by the nominalizing suffixes. However, a null version of n can be found also in noun
phrases containing possessors, and here it is the possessor which Merges with n.
As stated in Chapter 1, contra the Antisymmetry framework of [Kayne 1994] I do
not limit myself to a right-branching tree. Instead, I place the head in the right or
left branch according to the unmarked word order of each projection type. Thus, I
assume D heads to occupy a left branch while N and V heads occupy right branches.
Furthermore, somewhat arbitrarily I place n, v and Tr to the right and C and T
to the left. I also place all Specifiers to the left. I should note at this point that
my main aim here is to provide an analysis of the basic constituent structure, and
deriving word-order under stricter structural guidelines (such as the LCA), while
quite possible, would mask the basic line of argument that I will be making.
The structure of the simple transitive clause in (3.43) is shown in (3.44).

70
(3.43) Mariya
Maria
Juan-ta
Juan-acc
‘Maria saw Juan.’
riku-ra-n.
see-past-3sg
(CQ)
(3.44) TP
✟ ✟✟✟
❍ ❍ ❍
Maria i T’
❍
✟ ✟✟✟ ❍
❍
❍
T vP
❍
❍
✟ ✟✟✟✟✟
❍
❍
❍
❍
t i v’
❍
✟ ✟✟✟✟
TrP
✟ ✟✟ ✟ ❍❍
❍
❍
Juan k -ta Tr’
✟ ❍
VP Tr
✟❍
t k V t j
❍
❍
❍
❍
v
see j -past-3sg
t j
3.3.2.2 Case-marking in a Complement clause: the LkP projection
Now let us ask what happens in subordinate nominalized clauses. I separate these
into two cases: complement clauses, and relative clauses. In the case of complement
clauses, we have all the same argument possibilities which exist in main clauses,
but with -∅ Case marking in place of -ta. I relate this to the case of Main clauses
with more than one argument. In both cases I suggest that the extra arguments
(those either not taken care of by Tr, or those which occur in clauses in which Tr
is disallowed) are accommodated by another functional projection. I follow [Collins
2003] in calling this functional head Lk (for ‘linker’). Note, however, that the head
itself is not overtly expressed in Quechua as it is in languages like |=Hoan. As in
[Collins 2001, 2003], the functional head Lk is associated with additional arguments
of the verb. I suggest that more than one such projection can occur in a single

71
clause. The Lk head checks structural Case, but it is not associated with a particular
grammatical or thematic role. A null-marked direct object has its Case checked by
Lk.
We must now explain the impossibility of Tr in a nominalized clause. Recall
that whereas Main clauses contain a v head, to which I have attributed the Main
tense morphology above, Subordinate clauses contain a n head, to which I attribute
nominalization (including reduced tense information) morphology. I propose that n,
unlike v, cannot select TrP, but must select LkP or VP directly. Furthermore, I will
assume for the moment that the T which selects n checks genitive Case, while the
Main clause T checks nominative Case. 8 This position will be refined in Chapter 4.
The following trees (3.46) and (3.49) give structures for the ditransitive Main
and Complement clauses in (3.45) and (3.48) respectively to illustrate the role of
LkP. 9
8 This proposal is quite similar in spirit to one in [Lefebvre & Muysken 1988]. In
that study, the authors argue that the nominalized verb has the feature set [+N,+V]
and may head either an NP or a VP projection. In this way they explain that for
their consultants the genitive marking on the subject of an embedded clause was
optional, as was the accusative marker on the direct object in a nominalized clause.
Further, they found that the presence of the accusative marker was specifically
prohibited just in the case where the subject did receive genitive marking. They
thus associate the VP node with a nominative subject and the optionality of -ta,
while the NP node is associated with genitive marking and impossibility of -ta. My
consultants were not able to confirm the range of sentences analyzed by Lefebvre
and Muysken (and, I should note that Lefebvre and Muysken’s consultants also
found the presence of -ta in a nominalized clause to be only marginally acceptable).
However, I nonetheless believe the analysis I present here to be very similar to that of
Lefebvre and Muysken, with the difference that for my consultants the nominalized
verb simply cannot serve as the head of a VP node, under the analysis of Lefebvre
and Muysken. Although in the current framework I attribute the nominal nature
of this node to a n head in place of v head, I believe the spirit of the analysis to be
very much in the same vein. This point is discussed further in 3.4.2
9 Here I am assuming that ‘give’ is indeed ditransitive in Quechua and that both
objects merge with V. A simpler analysis would involve just a single LkP, and assume
that the dative noun phrase is an adjunct. Since my main use of LkP will be for
direct objects of nominalized verbs, this issue is not crucial to my discussion of
relative clauses below. As the trees stand, however, the apparent MLC violation as

72
First (3.45) and (3.46) illustrate a main clause ditransitive.
(3.45) Mariya
Mariya
Juan-man
Juan-dat
qulqi-ta
money-acc
‘Maria gave money to Juan.’
qu-ra-n
give-past-3sg
(3.46)
TP
❍ ❍
✟ ✟✟✟✟ ❍
❍ ❍
Maria i T’
✟ ❍
❍
✟ ✟✟✟✟ ❍
❍
❍
❍
T
vP
✟ ❍❍
❍
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟✟✟ ❍
❍
❍
❍
t i v’
✟ ✟✟✟✟✟✟✟✟
TrP
❍ ❍
❍
✟ ✟✟✟✟✟✟✟
❍
❍
❍ ❍
money l -ta
Tr’
❍ ✟ ✟✟✟✟
LkP
✟ ❍
✟ ✟✟ ❍
❍
❍
Juan k -dat Lk’
✟ ✟ ❍❍ ❍
VP Lk
✟ ❍
t k V’ t j
✟❍
t l V
❍
❍
❍
❍
❍
❍
❍
❍ ❍
❍
❍ ❍
Tr
t j
v
give j -past-3sg
Note that to achieve the surface word order we must posit further movement of
the embedded constituent TrP, as shown in (3.47). The scrambling of the indirect
the indirect object moves over the direct object would have to be resolved either by
a principle of equidistance of the objects, or by a (different) feature-match between
the Probe and the Goal in each case.
t j

73
object over the direct object is represented here by the movement of LkP to the
Spec of a functional XP phrase containing TrP.
(3.47) TP
❍
❍
❍
✟ ✟✟✟✟✟✟
❍
❍
❍
❍
Maria T’
❍
❍
❍
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟✟✟✟
❍
❍
❍ ❍
T
vP
❍ ❍
❍
❍
❍ ❍
✟ ✟✟✟✟✟✟
XP i
❍ ❍
✟ ✟✟✟✟✟✟
❍
❍
❍ ❍
LkP j
X’
✟ ❍❍
✟ ✟✟✟ ❍
❍
❍ ✟ ✟✟✟
❍
❍
❍ ❍
Juan k -dat Lk’ X TrP
✟ ✟✟ ❍ ❍
✟ ✟✟ ❍
VP Lk money l -acc
✟ ❍
t k V’ t m
✟❍
t l V
v
gave m
❍ ❍
Tr’
✟❍
t j Tr
t m
Now (3.48) and (3.49) illustrate a complement clause ditransitive.
t m
(3.48) Nuqa [[Mariya-q Juan-man qulqi
I [[Maria-gen Juan-dat money
qu-sqa-n]]-ta
yacha-ni
give-nm(past/nonsubj)-3sg]]-acc know-1sg
‘I know that Maria gave money to Juan.’

74
(3.49)
CP
(3.47).
TP
❍
✟ ✟✟✟✟
❍
❍
❍
❍
Maria-gen i T’
❍ ✟ ✟✟✟✟✟
❍
❍
❍ ❍
T
nP
✟ ❍
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟ ❍
❍
❍
❍
t i n’
❍
❍
❍
✟ ✟✟✟✟✟✟
❍
❍
❍
❍
LkP
n
❍ ❍
❍
✟ ✟✟✟✟✟✟✟
❍ give j -nm-3sg
❍
❍ ❍
money l -∅
Lk’
❍ ✟ ✟✟✟✟ ❍
❍ ❍
LkP Lk
✟ ❍
✟ ✟✟ ❍
❍
❍ t j
Juan k -dat Lk’
✟ ✟ ❍❍ ❍
VP Lk
✟ ❍
t k V’
✟❍
t l t j
Final word order will be achieved by movement of the dative as shown above in
3.3.2.3 Case-marking in a Relative clause
In this section I will limit myself to relative clauses with non-subject heads. Subject
heads will be discussed in Section 3.4.4. In other words, the heads will be internal
arguments of the subordinate verb, and the -sqa (or -na) nominalizer will appear on
the nominalized verbs.

75
How can we explain the fact that null-marked internal arguments may raise to a
clause-external position, either post- or pre-Spellout? Note that it is not sufficient
to say that the element which appears in the first Spec,LkP position can raise, since
this does not explain the fact that non-direct object internal arguments (such as the
locative in (3.38)) are marked differently depending on whether they are the head
or not.
To answer this question I return to an important aspect of the determinerincorporation
analysis of [Bianchi 2000], a version of which is adopted here for
Quechua as discussed in Section 4.1. Under the DP analysis, it is natural to assume
that Case is assigned to the determiner of a DP. Bianchi assumes that Case is
assigned to the relative DP within the subordinate clause and to the outer D within
the matrix clause. However, I would like to make a different proposal for Cuzco
Quechua. I propose that the head does not receive Case within the subordinate
clause. This explains why heads have no Case marking in the subordinate clause.
I therefore distinguish between the null-Case accusative found in some subordinate
clauses, and the head of a relative clause which I take to be unmarked for Case.
Note that such an analysis is possible, and easily formulable, in a framework in
which structural Case assignment is attributed to functional heads and not to the
verb itself.
How, then, does the head receive Case? To be concrete, I propose that the
(non-subject) head of a relative clause is generated simply in the Spec of VP, which
under the current analysis is not a Case position at all. The head then raises either
overtly or at LF to the Spec of the dominating CP. At this point I can see two
possibilities. One possibility is that the [Spec,CP] position is itself a Case-checking
position; that the relative complementizer C has Case features itself. 10
A second
10 Although this is an unorthodox proposal, in fact Bianchi [2000] suggests in a
footnote that it is possible that the external D of a relative clause is associated

76
possibility is that the Case of the head is not checked until after D-incorporation
has occured, and in fact the Case of the two merged D’s is checked simultaneously
within the matrix clause. Although either option can be implemented within the
current framework, the second one has the advantage of linking Case-marking with
determiner incorporation. Let us consider this option in more detail.
Recall that in Chapter 2 we found that Quechua defies cross-linguistic generalizations
in that it allows a strong quantifier to appear on an internal relative clause
head, but ultimately to have wide scope over the relative clause. Furthermore, in
the present chapter we have been considering the lack of Case-marking of the relative
clause head. I am not aware of other languages (with overt Case morphology)
which have either of these two properties. For each in turn let us examine which
aspect of Cuzco Quechua syntax may be responsible for it. The analysis I present
will attribute them both to the special nature of the determiner which selects a CP
complement in Cuzco Quechua.
The unique feature of Quechua here seems to lie in the fact that Quechua allows
determiner incorporation to happen even when the surface structure spells out the
quantified head in its relative clause-internal position. Other IHR languages do not
have this option, since they disallow quantified noun phrases in the internal position.
Furthermore, if Bianchi’s analysis is correct (at least for English and the Romance
and Slavic languages she discusses), then (some) EHR languages also do not have
this option because their version of determiner incorporation involves S-structure
immediate C-command of the two determiner positions.
We can capture these facts within the current framework and in the spirit of
Bianchi’s solution to the challenge presented by Borsley by positing that Quechua
has a null determiner (heading the main clause DP dominating the relative clause)
with an Agr o phrase which dominates the relative CP and that the head moves to
[Spec,Agr o ] after moving to [Spec,CP].

77
without its own Case feature. This deficient determiner must have its features completed
at some point in the derivation via incorporation with a complete determiner
(that of the relative clause head): i.e., one with a Case feature. Recall that in
Bianchi’s analysis (shown in (3.35)) a null determiner in English was licensed by
incorporation of an overt determiner. Here, I am suggesting that a Caseless determiner
is licensed by incorporation with an overt determiner. With this assumption
the structure of a Quechua relative clause such as that in (3.50), is shown in (3.51).
(3.50) [[Mayta-q tukuy planta planta-sqa-n]]-ta
Mayta-ge all plant plant-nm-3sg-acc
‘all the plants that Mayta planted’
(3.51)
✟ ✟✟✟✟✟✟
D+tukuy
DP
❍ ❍
❍
❍
❍
❍ ❍
- - - - k CP
✟ ❍❍
✟ ✟✟✟✟
❍
❍
❍
❍
C’
DP i
✟ ✟✟✟
❍
❍
❍ ✟ ❍
❍ t k -planta
- - -
C TP j
❍
✟ ✟✟✟ ❍
❍
❍
Mayta-q l T’
❍
✟ ✟✟✟ ❍
❍
❍
T nP
✟ ❍
✟ ✟✟✟✟
❍
❍
❍
❍
❍
t l n’
✟ ✟✟✟ ✟
VP
✟ ✟✟ ❍
❍
❍
DP i
✟ ✟ ❍ ❍
tukuy planta
t m
❍❍
❍
❍
❍
n
planta-sqa-n-ta m
In this tree the crossed-out versions of tukuy and planta represent their LF po-

78
sitions. The derivation of (3.51) proceeds in the following steps (ignoring verb
movement):
1. Merge(tukuy,planta)
2. Merge(tukuy planta,plantasqan)
3. Merge(VP,n)
4. Merge(Mayta,nP)
5. Merge(nP,T)
6. Agree(T,Mayta)
7. Merge(Mayta,TP)
8. Merge(TP,C)
9. Agree(C,DP i )
10. Merge(DP i ,CP)
11. Merge(CP,D)
12. Match(D,tukuy)
13. Merge(D,tukuy)
It is worth addressing explicitly the issue of when the Case of the relative clause
head is checked. Under the analysis I have given, it appears that the Case of an
internal head is not checked until after it raises, at which point its Case is checked
within the Main clause. There are two options for the precise timing on this, which
I will simply point out without attempting to decide between them. If we adopt
the Copy theory of movement as in [Chomsky 1993] (as adopted for IHRs in [Kayne

79
1994]) then we may assume that raising has taken place and Case-checking has
occurred prior to Spell-out. On the other hand if we assume that the upper copy
of any chain formed before Spell-out must iself be spelled out then we are forced
to allow Determiner incorporation and Case checking on the resulting DP in the
matrix clause to be covert operations. Within the framework in [Chomsky 1995]
this latter would not be allowed.
The question of where exactly the Case of the head is checked is closely related to
another important issue: why does the head raise at all? Bianchi [2000] suggests that
the head raises to satisfy a selectional [+N] feature of the Determiner which merges
with the CP. However, I find this analysis to be unsatisfactory because I expect a
selectional feature to be just that: a prerequisite for selection. Thus I would not
expect this feature to be satisfied by movement after Merge. On the other hand,
if we take this feature to be one checked by a probe-goal relationship, then we run
into the problem that the movement which occurs after the Agree relationship has
been established (between the outer D and the head DP) is not to [Spec,outer DP]
but to [Spec,CP]. Therefore, contra [Bianchi 2000] I propose that the head raises
to [Spec,CP] because of an EPP feature in C and in conjunction with a matching
feature of C itself. The nature of this feature will be discussed further below.
There is also a rather parallel analysis in which it is the lower determiner (tukuy)
which is Caseless, and the clause-selecting D has a Case feature. I reject this analysis
in favor of the one given for two reasons. First, under this alternative analysis we
would have to posit two versions of each strong quantifier that fits this paradigm,
as opposed to a single null D. And second, it seems reasonable to assume that it is
the Caseless D that also has the property of selecting the relative clause CP headed
by a [+D] C. This selectional restriction explains the fact that within embedded
nominalized complement clauses, all arguments are Case-marked clause-internally.

80
3.3.2.4 Why no -∅-marking in a Main clause?
The previous analysis might lead one to falsely predict that -∅ marking could occur in
a Matrix clause, in a situation in which Tr happens not to be part of the Numeration.
Under these circumstances we might expect that a direct object could simply be
handled by a Lk head, as proposed for subordinate clauses such as (3.49). Note
that this fact cannot be explained by simply saying that v, as opposed to n must
select a TrP, since in the case of simple intransitive sentences and sentences with
only (say) a locative argument we do not detect the presence of TrP (there being no
-ta suffix). Instead, it seems that the Lk head associated with a direct object cannot
be selected by v. Only n can select this Lk directly. This assumption predicts that
the only valid transitive matrix sentences will be achieved via the presence of TrP,
and hence will exhibit -ta marking on the direct object.
3.3.3 Which DP may raise?
What constrains the choice of the DP which may raise? In particular, why can’t a
non-argument be an internal head? Also, why can’t a Case-marked argument be an
internal head? (Again, I am leaving aside Subject-headed clauses for the moment,
though these will end up fitting into essentially the same analysis as described here.)
These questions are quite quickly answered based on the preceding discussion.
A non-argument, and an overtly Case-marked argument, will receive Case within
the subordinate clause (I assume that the Case-markers or postpositions associated
with non-arguments assign Case to the DPs they attach to). Therefore, the Case
feature requirement of the external D will not be met by the incorporation of a
determiner which has already had its Case feature deleted.
The question of how an internal object (instead of the subject) can raise at all
must also be addressed. For example, in (3.51) why is the movement of tukuy planta

81
over Mayta not a violation of the Minimal Link Condition? This question leads
immediately to the question of what forces the head to move at all? In short, what
triggers the movement of the relative clause head and not of any other, possibly
intervening, DP?
Within a Minimalist framework, the answer to these questions must be that
there is a Probe-Goal relationship established by a matching pair of features, and
that the intervening DP (the subject) does not have this same feature. Note that
it is not enough to say that the subject has previously valued φ-features because
under the framework in [Chomsky 2001a p.129 ] this would still be a violation of the
defective intervention constraint, defined in Chapter 1. Thus, in the construction
under discussion it cannot be simply the set of interpretable φ features on the head
DP which is the goal to an uninterpretable φ-feature probe.
Above I proposed a partial answer to this puzzle: a feature of C serves as a probe
which matches the DP head as its goal. However, we now see that this feature cannot
be purely a +D feature or a set of φ-features for the reasons just mentioned. One
possible answer is simply to say that the head and the C have [+wh] features, and
this would make Quechua fit into the relative pronoun model of English. Although
I am not aware of language-internal evidence for the presence of a [+wh] element in
Quechua relative clauses, I will assume that this is the relevant feature.
Under this analysis, the entire picture would look as follows: a Caseless D head
selects a C with an uninterpretable [+wh] ([uwh]) feature. In the Spec of this CP
is the relative clause head which was Merged into this position after its [wh] feature
Matched the wh feature of C. After Merge with D, the D-incorporation takes place,
completing the feature set of the defective outer determiner. Like Bianchi, I assume
that a locality condition is in place for this sort of head-to-head incorporation,
satisfied by the immediate C-command relationship between the two. Furthermore,

82
the feature sets of each D are complementary in the sense that there is no feature
clash between them.
Under this analysis, what distinguishes Quechua from a language like Lakhota
which obeys the indefiniteness restriction on an internal head is the presence in
Quechua of a Caseless determiner which allows incorporation of an overt determiner
from an internal head after head-raising has taken place. What distinguishes
Quechua from Japanese is the probe C which allows for head-raising in the first
place, as well as the Caseless determiner. And what distinguishes Quechua from
English is (at least) the option of post-Spellout head-raising, which perhaps can be
attributed to a strong/weak feature distinction on C.
3.4 Extending the analysis
3.4.1 Subject-headed RCs, and explaining the nominalizing
morphology
Recall that in Cuzco Quechua, the nominalization pattern is quite straight-forward:
the nominalizer -q appears when the subject is the (internal or external) head of a
relative clause. The nominalizers -sqa/-na tell us that the subordinate clause action
occured before/after the main clause action. These last two are used in both relative
clauses and complement clauses. (Simultaneous subordinate/main action seems also
to be compatible with -sqa.)
Recall further that while -sqa and -na are accompanied by overt subject agreement
morphology, -q-marked subject relativizations do not show subject/verb agreement.
Lefebvre and Muysken [1988] propose that while -sqa and -na are parts of the
Cuzco Quechua system of Tense, -q is an agreement morpheme. Contrary to this
analysis, under the analysis I am developing here, the lack of overt agreement with

83
the subject of a subject-headed relative clause would be directly related to the fact
that this head is not Case-marked within the embedded clause but rather within
the matrix clause. (Similar facts in Turkish externally headed relative clauses are
given an analysis along these same lines in [Kornfilt 2002].) I propose that the T
which selects the -q head does not have the Case and Agreement features which the
standard T head is presumed to have. (This is compatible with the fact that -q is
not inflected for person, nor does it indicate tense.) Therefore, this T is not a probe,
and the Subject is free to move to CP before or after Spellout to become the head
of the relative clause. Note that a derivation in which an internal argument jumps
over the Subject to the [Spec,CP] position would, in the presence of -q, leave the
Subject with unchecked Case and the derivation would crash.
As an example, I show the LF structure of (3.52) in (3.54). The internally headed
version of (3.52), (3.53), would have essentially the same structure.
(3.52) [[Juan
Juan
riku-q]
see-nm(subj)
runa]
person
‘the person who saw/sees/will see Juan’
(3.53) [(Chay)
dem
[runa
person
Juan
Juan
riku-q]]
see-nm(subj)
‘the person who saw/sees/will see Juan’

84
(3.54) DP
✟ ❍❍
✟ ✟✟✟✟ ❍
❍
❍
❍
D
CP
✟ ❍❍
✟ ✟✟✟✟ ❍
❍
❍
❍
CP
person i
✟ ✟✟ ✟ ❍❍
❍
❍
C TP l
❍ ❍
✟ ✟✟✟✟
T
❍
❍ ❍
nP
❍ ❍
✟ ✟✟✟✟✟
❍
❍
❍ ❍
person - - - - i n’
❍
✟ ✟✟✟
LkP m
✟ ✟✟ ❍
❍ ❍
Juan j Lk’
✟ ✟❍
❍
VP Lk
✟❍
t j t k
❍
❍
❍
see k -q
Note that in order to achieve the observed word order it is not necessary (though
possible) to assume that T has some +D feature which can match the subject, and
hence the subject need not Merge into Spec,TP on its way to the final head position.
I am not aware of evidence that would decide between these possibilities.
3.4.2 Revisiting weakly quantified heads
Having gone through the syntactic analysis of head-raising associated with the semantics
of strongly quantified internal and external heads as discussed in Chapter 2,
we are left with the problem of how weakly quantified heads should be incorporated
into our paradigm. Before going into the details, let us consider a pair of conclusions
from Chapter 2: strong quantifiers but not weak quantifiers are determiners; and,
strong quantifiers but not weak determiners are interpreted externally to the clause.
The discussion in the previous sections brings us to an immediate association of

85
these two facts: since weak quantifiers are not determiners, they cannot undergo determiner
incorporation. Thus, we do not expect them to take scope over the relative
clause. I will now proceed to flesh out this basic idea.
Recall from Chapter 2 that based on data of the sort shown in (3.55) and (3.56),
in which a strong but not a weak determiner can appear in the pre-clausal position
in a relative clause construction, I adopted the familiar notion that only strong
quantifiers are in fact determiners.
(3.55) Asunta
Asunta
[tukuy [Mayta-q plaza-pi planta
all Mayta-gen plaza-loc plant
p’iti-ra-n.
prune-past-3sg
‘Asunta pruned all the plants which Mayta planted in the plaza.’
(3.56) *?Pisi
little
[[Asunta-q aqha
Asunta-gen cornbeer
apa-ra-ni.
bring-past-1sg
‘I brought a little cornbeer that Asunta made.’
aqha-sqa-n]]-ta
make.cornbeer-nm-3sg-acc
planta-sqa-n]]-ta
plant-nm-3sg-acc
In Chapter 2 I argued that pisi-headed relative clauses are not interpreted via
a head-raised structure but rather lend themselves to the E-type anaphora analysis
of relative clauses developed for Japanese in [Hoshi 1995], [Shimoyama 1999, 2001]
and others. The fact that these relative clause heads can nonetheless appear in
clause-external position meant that reconstruction to base position was necessary
at interpretation. Thus, a null proform was posited which played the role of the
head in the matrix clause. A variety of evidence pointing to the presence of this
null pronoun was shown. However, there is a point which needs to be clarified in
light of the structure of head-raised relative clauses discussed here. Namely, what
is the role of head-raising (at least in the case of externally headed relative clauses)
for weakly quantified relatives?

86
Given the discussion thus far the answer is clearly again related to Case. A
weakly quantified relative clause head is no more Case-marked within its clause than
a strongly quantified head. I will return shortly to exactly how the Case-checking
is achieved in this case. However, the head-raising must serve another purpose: it
must provide the identity of the head.
One point not addressed in the discussion thus far of weakly quantified heads
is the unexpected lack of ambiguity of the head. That is, in Cuzco Quechua I
have not found instances of ambiguous head identity. Part of this can be explained
by the different nominalizing morphology in the cases of subject and non-subject
relativization. But even if we limit ourselves to non-subject relativization we have a
rather unexpected situation in which the un-Casemarked DP must be the head. The
pragmatic construal analysis offers no explanation for this lack of ambiguity. On
the other hand we cannot return to a simple head-raising analysis here due to the
binding and scope effects discussed in Chapter 2. A crucial example is given here
in (3.57) (repeated from (2.53)), with the similar (3.58) provided for comparison.
(3.57) Asunta mikhu-ra-n
Asunta eat-past-3sg
ranti-sqa-n]]-ta
buy-nm-3sg-acc
[[sapa(nka)
each/every-(nka)
irqi-q
child-gen
huk
one
t’anta
bread
‘Each child bought one roll and Asunta ate them (one roll per child).’
(3.58) Sapa(nka) irqi-qa [[pro huk t’anta
each-nka child-top pro one bread
‘Each child, he bought one roll and ate it.’
ranti-sqa-n]]-ta
buy-nm-3sg-acc
mikhu-ra-n.
eat-past-3sg
To make the conundrum clear, I list in (3.59) the main arguments in favor of
positing an E-type pronoun within the matrix clause in the case of sentences like
(3.57). I also list the ways in in which this pronoun does not exactly follow the
expected behavior of an E-type pronoun.

87
(3.59) • E-type behavior:
– Exact referent not necessarily a constituent of antecedent sentence,
as in (3.57)
– Consistent with narrow scope of weak quantifiers
• Non-E-type behavior:
– Head is not ambiguous
– Head may be raised, and hence structurally not pragmatically
determined
What we need is a mechanism by which the proform gets assigned its denotation.
The mechanism seems to involve the local relationship between the proform and the
head of the relative clause. This local relationship is achieved once the head has
raised to the Spec of the relative CP. What triggers this raising? Note that under
the structural analysis suggested in Chapter 2, which exactly paralleled the analysis
for IHRs in Japanese, incorporation of a determiner is impossible in this case due to
the lack of locality between the CP and the DP projection: there is an intervening
NP. (Recall that under Bianchi’s analysis of determiner-incorporation, the fact that
there was a relationship of immediate C-command between the D which selects CP
and the C head itself was taken to be a licensing condition for incorporation to be
viable.) The problematic tree suggested in Chapter 2 as (2.36) is repeated here in
(3.60).

88
(3.60) IP
❍
❍
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟✟✟
❍
❍
❍
❍
CP i
IP
✟ ✟✟ ❍
❍ ❍
VP I
✏ ✏✏✏✏✏✏✏✏✏✏✏
✟ ❍❍
Asunta-q pisi aqha aqha-sqa-n-ta
✟ ✟✟✟ ❍
❍
❍
Asunta-gen little cornbeer make-nm-3sg-acc NP V’
Nuqa
I
✟ ✟✟✟ ❍ ❍ ❍
DP V
✟ ✟ ❍ ❍
D NP aparani
✟ ✟ ❍❍ ❍ brought
t i N
R
At this point I would like to suggest a revision to (3.60) which departs from
the analysis of Japanese in [Shimoyama 1999, 2001] in ways suggested by the nonprototypical
behavior of the E-type pronoun in Quechua. Furthermore, the revision
addresses a rather unexpected aspect of the tree in (3.60) given the discussion of
head-raising which has formed the bulk of this chapter. Ideally we would like to
posit parallel syntactic structures for weakly and strongly-headed relative clauses
and I now proceed to outline such an analysis.
I start by assuming that in fact the motivation for head-raising in all nominalized
relative clauses in Quechua is the same: the uninterpretable [+wh] feature of the
complementizer C acts as a probe which finds the head as its goal. This is all
as discussed in Section 3.3.2.3. Furthermore, in no case is there an intervening
NP but rather the Caseless determiner discussed above always selects a CP and
requires the incorporation of a determiner with a Case feature. This now radically
alters our picture in (2.31). We are forced to posit the presence of a null article
even in the presence of weakly-quantified heads, and this determiner incorporates
into the external D. Where is the “E-type” pronoun in this case? I posit that in

89
the case in which a null article has incorporated into it, the external determiner
itself is interpreted as the E-type pronoun within the matrix clause. 11 On the other
hand when an overt strong quantifier has incorporated into the external determiner,
the denotation of this strong quantifier becomes the denotation of the complex
determiner after incorporation.
(3.61).
The revised version of (3.60) that we arrive at under this analysis is shown in
(3.61) IP
✟ ✟✟✟✟✟✟✟✟ ✟
CP i
✏ ✏✏✏✏✏✏✏✏✏✏✏
Asunta-q [t j pisi aqha] aqha-sqa-n-ta
Asunta-gen little cornbeer make-nm-3sg-acc
❍
❍
❍
❍
❍
❍
❍
❍
❍
❍
IP
✟ ✟✟ ❍
❍ ❍
VP I
❍ ✟ ✟✟✟✟ ❍
❍ ❍
NP V’
Nuqa ✟ ✟✟✟ ❍ ❍ ❍
DP V
I
✟ ❍
D+D j t i
aparani
brought
For this analysis to be viable we must assume that the determiner trace t j is
not interpreted and the resulting bare NP is interpreted as an indefinite DP would
be (either by QR or existential closure). Note that the determiner incorporation
provides an explanation for why the resulting pronoun, which I am positing in place
of the prototypical E-type pronoun which seems necessary for Japanese does not
have the property of ambiguity with respect to the head. Rather, the index (j) on
the incorporated determiner forces the pronoun to refer back, in this case, to the
cornbeer. On the other hand, in an example with a quantified subject of the relative
clause such as (3.58), we do not have to insist that the E-type pronoun is an exact
11 See [Elbourne 2003] for another analysis of E-type pronouns as definite articles.

90
reconstruction of the head (in that case, huk t’anta ‘one roll’) but could in fact refer
to all the rolls bought by children. 12
3.4.3 Island constraints on relativization
My analysis that Head-raising in Quechua is associated with Case-checking of the
raised head has some important implications for relativization out of embedded
clauses, and relativization of genitives. First, since the relative clause head must
have its Case unchecked in the embedded clause, and I have captured this by the
assumption that the head merges directly with V rather than with a Case-assigning
head such as Tr, we expect that possessors of embedded DPs would not be possible
heads of relative clauses at all. This is born out in the internally headed relative
12 In fact, as mentioned in Chapter 2 there is some variation in behavior among
different weak quantifiers. It turns out that some uses of weak quantifiers do seem
to be compatible with a true head-raising analysis. For example, some consultants
on different occasions report either reading is possible for (i).
(i) Juan [tayta-n-pa kinsa wasi ruwa-sqa-n]-ta muna-n.
Juan father-3sg-gen three house make-nm-3sg-acc like-3sg
‘His father made three houses and Juan likes them.’
¿Juan likes three houses that his father made.’
Note that these glosses show that kinsa ‘three’ is crucially different from pisi
‘a little’, which shows only the type of behavior reflected in the first gloss. The
second translation is surprising under the pure E-type anaphora analysis. Since
I am essentially proposing a new, highly constrained type of E-type pronoun it
remains to further work to decide whether such a pronoun may be needed in other
constructions and in other languages too.
Under the present analysis we have two options. First, it is possible that certain
weak quantifiers, like numerals, are structurally ambiguous and may incorporate
into the external determiners, as strong quantifiers can do. Another possibility is
that the second reading of (i) can be attributed to the fact that ‘Juan’s father made
three houses.’ does not entail that he made no more than this number. The matrix
sentence ‘Juan likes them.’ would then refer simply to the three houses mentioned
in the first sentence. Under this second analysis the difference between kinsa ‘three’
and pisi ‘a lot’ would be one of entailment: ‘Asunta made a little cornbeer’ in
examples like (2.36), shown in (3.60) entails that she did not make any more than
this amount. I leave a choice between these options to future research.

91
clause example shown in (3.62).
(3.62) [[pro Waka-q
cow-gen
uña-n
baby-3sg
‘the cow’s calf that they stole’
*‘the cow whose calf they stole’
suwa-pu-sqa-nku]]
steal-definitively-nm-3pl
However, consultants do accept the externally headed version shown in (3.63):
(3.63) [[e
e
Uña-n
baby-3sg
suwa-pu-sqa-nku]
steal-definitively-nm-3pl
‘the cow whose calf they stole’
waka]
cow
This example does not violate the generalization that an internal head must be
an argument of the clause. However, it does raise the question of the syntax of
an externally headed relative in cases in which there is no IHR counterpart and the
current head-raising analysis is not viable. In fact, there are other classes of examples
of this general sort. First, there are cases of external heads which correspond to
adjunct positions within the relative clause (and consequently do not have IHR
counterparts). This is illustrated in (3.64) and (3.65).
(3.64) [[e
e
Qilqa-sqa-y]
write-nm-1sg
lapicero]
pen
‘the pen I wrote with’
(3.65) *[[lapicero-(wan) qilqa-sqa-y]]
pen-instr write-nm-1sg
‘the pen I wrote with’
Second, there are cases of external heads which seem not to correspond to any argument
within the subordinate clause, as illustrated in (3.66). Note that jugo ‘juice’
does not appear to be a natural argument of ch’arway ‘squeeze’ since consultants
reject (3.67), and propose instead (3.68) as an appropriate way to relate the roles
of the juice and the oranges in the squeezing event.

92
(3.66) [[naranja
orange
ch’arwa-sqa-y]
squeeze-nm-1sg
jugo]
juice
‘the juice that I made by squeezing oranges’
(3.67) ?*Nuqa
I
jugo-ta
juice-acc
ch’arwa-ra-ni.
squeeze-past-1sg
(3.68) Nuqa-n
I-evid
jugo-ta
juice-acc
ruwa-ra-ni
make-past-1sg
[naranja-ta
orange-acc
‘I made juice (by) squeezing oranges.’
ch’arwa-spa-y]
squeeze-adv-1sg
I believe that all of these examples should be analyzed as complex NPs along the
lines discussed for Japanese in [Murasugi 2000]. This possibility will be discussed
further in Section 4.5.4. Evidence for a modificational type analysis of this sort is
that the external head must have a pragmatic connection to the modifying clause
itself, rather than simply a possible syntactic connection. This is not the case when
a direct object is relativized.
In fact, further evidence for the status of the heads as in (3.64) being basegenerated
externally to the clause can be found in attempts to add weak quantifiers
to these heads. Recall that in cases of internal arguments which are weakly quantified
external heads, an internal-scope reading is still found, as illustrated in (3.69).
(3.69) [[Ana-q aqha-sqa-n] pisi
Ana-gen make cornbeer-nm-3sg a.little
‘Ana made a little cornbeer and I drank it.’
aqha]-ta
cornbeer-acc
ukya-ra-ni.
drink-past-1sg
If the head (pen) in (3.64) had raised from an internal position we might expect
this internal reading to be available. In fact, this is not the case. Compare (3.70)
and (3.71). The context is that I go into a store with the intention of buying a pen,
and try out some pens to decide which one I want.
(3.70) Pisi lapiz-wan qilqa-ra-ni.
a.few pen-instr write-past-1sg
‘I wrote with a few pens.’

93
(3.71) *?[[pro qilqa-sqa-y] pisi lapiz]-qa azul-mi.
write-nm-1sg a.few pen-top blue-evid
Intended reading: ‘I wrote with a few pens and they were blue.’
In fact, (3.71) is judged to be quite odd by consultants. One suggested meaning
was roughly ‘I wrote with a small piece of blue pen’ but this was considered strange.
This is presumably due to the fact that the weakly quantified expression pisi lapiz
cannot be reconstructed clause-internally (since it was never in this position), nor
can it be modified externally by a clause (in the complex NP analysis), since pisi
would not have scope over that clause.
Another pair of examples involving a locative adjunct is shown in (3.72) and
(3.73).
(3.72) Ashka
many
manka-pi
pot-loc
wayk’u-ni.
cook-1sg
‘I cook in many pots.’
(3.73) *[[pro wayk’u-sqa-y] ashka manka] hatun-mi.
cook-nm-1sg a.lot pot big-evid
Intended reading: ’I cooked in many pots and they are big.’
The preceding examples show that the relationship between an external adjunct
head to the relative clause is in fact quite different from that of an argument external
head to its relative clause.
Going back to the genitive construction (3.62), I propose that the agreement
marker on uña ‘calf’ indicates that there is a small pro within the embedded clause
which is construed as co-referential with the head waka, ‘cow’. The structure of this
example would be as follows:

94
(3.74) DP
✟ ❍❍
✟ ✟✟✟✟ ❍
❍
❍
❍
CP
✏
✏ ✏✏✏✏✏✏
pro [pro uña-n] suwa-sqa-nku
(they) stole her calf
D’
✟✟ ❍❍
D NP
waka
cow
Note that unlike [Murasugi 2000] for Japanese, I am not proposing structure
(3.74) for all relative clauses in Cuzco Quechua. 13 Rather, in Quechua I have analyzed
argument relativization as necessarily involving head-raising. On the other
hand the possibility of gapless relatives such as (3.66) does suggest that a structure
such as that in (3.74) must also be available to speakers. I am unable to provide
a full analysis here of when this latter option is available (possibly it is present as
an alternative structure even when head-raising is also available), but leave this
question to future research.
In Japanese, evidence pointing to a no-movement (necessarily) analysis of some
relative clause types has been provided in such work as [Kuno 1973], [Hoji 1985],
[Saito 1985] and others. For example, it is shown in examples like (3.75) that
relativization can occur out of an embedded complex NP. Such examples have been
posited to contain a small pro in the position e i corresponding to ‘gentleman’.
(3.75) [ NP [ IP [ NP [ IP e i e j kiteiru] yoohuku j ]-ga yogoreteiru] [ NP
is-wearing suit-nom is-dirty
sinsi i ]]
gentleman
‘the gentleman who [the suit that he is wearing] is dirty’ [Murasugi (16)]
This is an area in which Quechua departs significantly from Japanese, and indeed
from many other languages which allow relativization at least out of a subordinate
13 In fact, a similarly diverse view of Japanese relative clauses, in which some but
not all such clauses involve head-raising, is also advocated in such work as [Kaplan
& Whitman 1995].

95
complement clause. Examples like (3.76) and (3.77) show that relativization out of
a complex NP is not possible in Quechua.
(3.76) *[[[chura-sqa-n
put.on-nm-3sg
punchu]
poncho
qhelli
dirty
ka-sqa-n]
be-nm-3sg
runa]
person
‘the man who the poncho he is wearing is dirty’
(3.77) *[ NP [ IP [ NP [ IP e i e j monta-sqa-n/monta-q]
ride-nm-3sg/nm(sbj)
ka-sqa-n/ka-q] [ NP runa i ]]
be-nm-3sg/nm/sbj man
‘the man who [the horse that he is riding] is black’
caballo j ]
suit-nom
yana
black
Even more compellingly, neither internal nor external versions of relative clauses
such as (3.78) and (3.79) in which an embedded object is relativized externally
or inernally, or (3.80) and (3.81) in which an embedded subject is relativized are
acceptable to speakers of Cuzco Quechua. In order to express these concepts my
consultants suggested various paraphrases which do not involve relative clauses at
all. 14
(3.78) *[[Juan-pa [Antonio-q [e] suwaku-sqa-n] yacha-sqa-n]
Juan-gen Antonio-gen e steal-nm-3sg know-nm-3sg
‘the money that Juan knows that Antonio stole’
qulqi]
money
(3.79) *[[Juan-pa [Antonio-q qulqi suwaku-sqa-n]
Juan-gen Antonio-gen money steal-nm-3sg
‘the money that Juan know that Antonio stole’
yacha-sqa-n]]
know-nm-3sg
14 Another option is relativization involving the chay ‘that’ subordinator and fully
finite embedded clauses, also mentioned in Footnote 1 of chapter 2. The form given
there, repeated here in (i), was suggested as an alternative to (3.78) and (3.79).
(i) Chay Juan yacha-sha-ra-n Antonio-q suwa-sqa-n qulqi-ta,
dem Juan know-prog-past-3sg Antonio-gen steal-nm-3sg money- sc acc
chay wasi-y-pi ka-sha-n.
dem house-1sg-loc be-prog-3sg
‘The money Juan knows Antonio stole is in my house.’

96
(3.80) *[[pro
pro
[[e]
e
Lima-pi
Lima-loc
runa]
person
‘the person who you know lives in Lima’
tiya-q/sqa-n]
live-nm(subj)/nm(nonsubj)-3sg
yacha-sqa-yki]
know-nm-2sg
(3.81) *[[pro [Runa Lima-pi tiya-q/sqa-n]
pro person Lima-loc live-nm/nm-3sg
‘the person who you know lives in Lima’
yacha-sqa-yki]]
know-nm(nonsubj)-2sg
Given the analysis of IHRs and EHRs presented in this chapter, the unacceptability
of these expressions is easily explained. Recall that the Complementizer C
which is associated with head-raising (i.e. that served as a probe) must be selected
directly by the external D. In the examples here, however, the embedded clause
is in neither of these positions but is in fact in an argument position of the “you
know...” or “Juan knows...” clause. Therefore, a head cannot move to the CP of its
immediately containing clause (since this C is not a probe). Furthermore, it cannot
move to the CP of the relative clause because it is trapped in an embedded clause
island. Note that in order to rule out successive cyclic movement it is crucial that
head-raising is an operation specifically licensed by the probe in the C head selected
by D. In other words the probe which I have called a [uwh] feature on C cannot
be identical to the [uwh] probe on main clause complementizers in languages with
overt wh-movement in English. 15
In fact, this seems intuitively reasonable based
on the discussion so far since movement of the head involves movement of an un-
Casemarked element, while traditional wh-movement always involves movement of
a Case-marked category (A’ movement).
Furthermore, the external heads are ruled out for two reasons. First, the external
head does not bear a direct pragmatic relationship to the less-embedded clause,
15 The availability of wh- movement in Quechua is unclear. Cole & Hermon [1981,
1994], argue for it while [Lefebvre & Muysken 1982, 1988] argue that the observed
phenomenon is simply a generic fronting process available to a wide variety of categories.

97
which appears to be its surface sister. Thus it cannot be base-generated externally
to this clause. Second, the external head could not have raised to this position, for
the same reasons given above for ruling out internal heads in these cases.
Certain data involving the “future” nominalizer -na at first glance seem to violate
the generalization illustrated in (3.78) to (3.81), namely that elements of embedded
clauses cannot head a relative clause. Note that this generalization was just
presented based on examples containing the -sqa nominalizer. In fact we find that
external heads are acceptable from a -na embedded clause. An example is shown in
(3.82).
(3.82) [[ [[e] Ripu-na-n]
e leave-nm(fut)-3sg
‘The man you want to leave’
muna-sqa-yki]
want-nm-2sg
runa]
person
However, there is in fact strong evidence that a subjunctive embedded clause is
not an island to movement in the way that the factive embedded clauses in (3.78)
to (3.4) are. This evidence contradicts the treatment of these two construnctions as
structurally equivalent, as suggested in [Lefebvre & Muysken 1988]. The similarity
between the two construction types can be seen in (3.83) and (3.84), each containing
complement clauses in which the action of the embedded clause takes place after
the action of the main clause and hence the nominalizer -na (as opposed to -sqa) is
selected in each case.
(3.83) Nuqa
I
muna-ni
want-1sg
[chay
dem
runa
man
‘I want that man to leave.’
ripu-na-n]-ta.
leave-nm-3sg-acc
(3.84) Nuqa
I
yacha-ni
know-1sg
[chay
dem
runa
man
ripu-na-n]-ta.
leave-nm-3sg-acc
‘I know that that man will leave.’

98
Despite the similarity in (3.83) and (3.84), however, only in the subjunctive
example in (3.84) can the embedded subject become the head of a relative clause.
This contrast is illustrated in (3.85) and (3.86).
(3.85) [[[e ripu-na-n] muna-sqa-y]
leave-nm-3sg want-nm-1sg
‘the man I want to leave’
runa]
person
(3.86) *[[[e ripu-na-n] yacha-sqa-y]
leave-nm-3sg know-nm-1sg
‘the man I know will leave’
runa]
person
A similar contrast can be seen when the relativization is from the object position
of the embedded clause. Here too it is only in the case of an embedded subjunctive
that relativization is possible, as illustrated in the contrast between (3.87) and (3.88).
(3.87) [[[Juan-pa
Juan-gen
e yanapa-na-n]-ta
help-nm-3sg-acc
ripu-sqa.
leave-past/surprise
‘The man I wanted Juan to help left.’
muna-sqa-y]
want-nm-1sg
runa]
person
(3.88) *[[[Juan-pa
Juan-gen
e yanapa-na-n]
help-nm-3sg
yacha-sqa-y]
know-nm-1sg
‘The man I know Juan will help left.’
runa]
person
ripu-sqa.
leave-past/surprise
I leave to future research the question of the exact structure of subjunctive clauses
in Quechua, but note simply that embedded clauses with the -na nominalizer that
are the true equivalent of the -sqa embedded clauses do in fact behave as expected
under my analysis.
3.4.4 Variation
I have alluded to further dialect variation in head-marking and head identification.
In particular, the “Case-float” phenomenon studied in [Lefebvre & Muysken 1988],

99
and phenomena detected by [Weber 1978] in Huanuco Quechua, should eventually
be placed within the general framework I am establishing here. In this section I
will focus on certain facts reported for Cuzco Quechua by Lefebvre and Muysken
(L&M) and how these data might work into the current analysis. These represent
areas in which my consultants’ judgments differ from those authors’ consultants’,
and so I do not have data from my own fieldwork in these cases. Unfortunately, it
is not possible here to survey the entire field of potential differences between the
idiolects represented, but I will look at and attempt to explain a few key points of
distinction.
First, I have mentioned that L&M’s consultants do allow the use of an overt -ta to
mark the accusative Case in nominalized complement clauses, but this use is limited
to clauses in which the subject does not receive genitive Case. My consultants
always report that the subject should have genitive Case and the -ta marker should
be absent. Thus, the following example is good for L&M:
(3.89) [Kay warmi qusa-n-ta maqa-sqa-n]-ta yacha-ra-nki-chu.
this woman husband-3sg-acc beat-nm-3sg-acc know-past-2sg-Q
‘Do you know that this woman beat her husband?’ (L&M 26c p.119)
L&M do report that this construction is marginal, while my consultants reject it
altogether. For L&M, the alternative construction in which the subject has genitive
Case and no -ta marking appears on the direct object of a nominalized clause is
preferable. Note that this judgment is in line with that which I have reported for
my own consultants. Data such as (3.89) are not difficult to fit into the framework
which I have outlined thus far. We can simply say that for some speakers, v (as well
as n) can in fact appear within a nominalized clause. In these cases, selection of a
TrP by the v is possible, as evidenced by the -ta on ‘husband’ in (3.89).
The situation gets more complex when we look again at relative clauses. Some
of the thorniest data discussed by L&M involve what they call “case float” in RCs.

100
Consider their examples (3.90) to (3.93). All four examples are acceptable to some
of their consultants, with the meaning ‘the girl I saw will come.’ Note in each case
the presence (or absence) and position of the -ta marker.
(3.90) Warma
girl
riku-sqa-y-ta,
see-nm-1sg-acc
hamu-nqa.
come-3fut
‘The girl I saw will come.’ (L&M p.187 (60))
(3.91) Riku-sqa-y
see-nm-1sg
warma-ta,
girl-acc
hamu-nqa.
come-3fut
‘The girl I saw will come.’ (L&M p.187 (61))
(3.92) Riku-sqa-y-ta
see-nm-1sg-acc
warma
girl
hamu-nqa.
come-3fut
‘The girl I saw will come.’ (L&M p.187 (62))
(3.93) Riku-sqa-y
see-nm-1sg
warma
girl
hamu-nqa.
come-3fut
‘The girl I saw will come.’ (L&M p.187 (63))
L&M report that (3.90) and (3.93) are acceptable to all of their consultants. It is
also true that my consultants uniformly accept (3.93). However, and interestingly,
they uniformly reject (3.90). I should add at this point that there is no uncertainty
or inconsistency in this judgment. It is completely out for my consultants under the
intended reading, so we are truly dealing with different ideolects here. As for (3.91)
and (3.92), L&M report that each of these is acceptable to some consultants. These
examples too were uniformly rejected by my consultants under the intended reading.
I should note, however, that (3.90) and (3.92) do have a possible interpretation,
which is ‘The girl will come by/to the place that I know.’, but under this reading
obviously ‘girl’ is in the matrix, not embedded clause. (In Cuzco Quechua, the verb
rikuy ‘to see’ can also mean ‘to know’ when referring to knowing a location.)
The interesting thing about (3.90), (3.91) and (3.92) is that the -ta Case marker,
associated with the accusative Case of the head within the subordinate clause, ap-

101
pears, variously, on the embedded verb (with the head internal or external) and
on an external head. I believe that (3.90) and (3.91) lend themselves to a slightly
different analysis, which I mentioned earlier as an alternative to the Case-checking
via Determiner Incorporation analysis I have advocated. This is that the C head
itself checks the Case of the head. This is essentially the analysis which L&M themselves
propose in the form of a “COMP-like Case poisition.” The -ta morphology
on the external head would be licensed under this Case-checking relationship. The
pair (3.90) and (3.91) again suggest the Copy theory of movement analysis, with
the -ta marker on the verb in (3.91) being associated with the upper copy of the
head. The idea is that the Complementizer C is checking the Case of the head (even
when the lower copy of the head is the one to be actually pronounced) and licensing
the presence of the -ta Case-marker. I leave open the question of whether this Case
is truly Accusative Case in the absence of Tr. In the case of the internally headed
relative clause in (3.90), the appearance of the Case marker on the verb itself is
presumably a result of the null phonology of the external copy of the head. With
regard to (3.92) it appears that here the manifestation -ta of the Case-marker C
may be generated in the Complementizer position itself.
Finally, L&M report that for their consultants, non-subject externally headed
relative clauses were subject to a Case-matching restriction whereby the internal and
external Cases of the head had to be identical (when the head appears in argument
position within a clause). This was not true for my consultants, nor was it true for
Gregorio Condori Mamani in his narrative, as illustrated in (3.94). In (3.95), data
from L&M showing Case matching is shown in (3.95) and (3.96).
(3.94) ...tapu-ra-nku
ask-past-3pl
[[papel
[[paper
lliw
all
movilizable-man
“mobilized soldiers”-dat
qo-sqa-nku]]-manta
give-nm-3pl]]-abl
‘They asked for the paper that they gave to all “mobilized soldiers”.’

102
(3.95) *Santiyagu
Santiago
Xwancha-q
Juan-gen
rima-sqa-n
speak-nm-3sg
warma-wan
girl-with
riku-ra-n.
see-past-3sg
‘Santiago saw the girl with whom Juan spoke.’ (L&M p.195 (76))
(3.96) Warma
girl
rima-sqa-y-wan
speak-nm-1sg-with
puklla-ra-ni.
play-past-1sg
‘I played with the girl that I talked with.’ (L&M p.194 (74))
In fact, for my consultants (3.95) is also bad, but can be fixed by changing the
-wan ‘with’ marker on ‘girl’ to -ta (Accusative). This would violate the Matching
Condition which was present for L&M’s consultants, however. Note that for L&M,
the example in (3.95) is a “headless” relative clause because the Case on the head
reflects the Case of the head within the embedded clause. The modification just
suggested would turn the example into a “headed” relative clause. Recall that the
headless type illustrated in (3.95) is not acceptable to my consultants. In fact, my
consultants have no objection to other instances of Matching Condition violations,
such as those found in the examples (3.97) and (3.98).
(3.97) Nuqa-qa [[Juan-pa riqsi-sqa-n] wawa]-wan
I-top Juan-gen know-nm-3sg child-com
‘I played with the child who Juan knows.’
puklla-ra-ni.
play-past-1sg
(3.98) Placida-qa [[qilqa-sqa-y] lapicero]-ta suwa-ra-n.
Placida-nm write-nm-1sg pen-acc steal-past-3sg
‘Placida stold the pen that I wrote with.’
In (3.97) the head wawa ‘child’ is the direct object of the subordinate clause but
part of an adjunct noun phrase within the main clause. In (3.98) the situation is
reversed: the head lapicero ‘pen’ is an adjunct of the subordinate clause but a direct
object in the matrix clause. Both types of Matching Effect violations were perfectly
acceptable to my consultants.
Within the current analysis, however, the matching effect as illustrated thus far
can be explained (for those speakers for whom this is a part of their grammar) by

103
simply stating that the Case of the external head must be the Case of the head within
the external clause. This is nicely explained within a Determiner-incorporation
analysis in which the head’s D is assigned Case by C, by positing that the Case
assigned to each of the two determiners must be identical. (Even though for these
speakers, the Case assigners themselves are different—C for the head, and a matrix
clause Case assigner for the outer D.) However, there remains the problem that for
L&M’s consultants, it is possible to violate this generalization if the relative clause
and the noun it modifies are in topic position. This is illustrated in (3.99).
(3.99) Xwancha-q
Juan-gen
rima-sqa-n
speak-nm-3sg
warma-wan-(qa),
girl-with-top
(pay-ta-puni)
he-acc-emph
‘He saw the girl with whom Juan spoke.’ (L&M p.194 (75))
riku-ra-n.
see-past-3sg
Examples like (3.99) give no indication of the role of the topic within the matrix
clause. They are bad for my consultants. For the speakers consulted by L&M,
however, we can simply say that clashes between the Case assigned by C (presumably
matching the internal Case) and the Case of the head within the matrix clause are
not tolerated. This is resolved when the entire clause is in topic position (a non-Case
position).
3.5 Conclusion
In this chapter I have analyzed the syntactic structure of Cuzco Quechua relative
clauses. I have argued for the presence of functional heads within Main and nominalized
clauses which are responsible for Case checking of arguments of the verb. I
have furthermore posited a n head which appears in nominalized clauses and possessor
phrases, and parallels v within Main clauses. I have claimed that the lack of
Case-marking on an internal head is due to the fact that the Case of this head is
checked in the matrix clause, via a proceess of determiner incorporation. Thus, the

104
analysis presented here also provides support for Bianchi’s [2000] proposal that the
head of a relative clause starts as a full DP within the subordinate clause, and its
raising out of the clause is followed by determiner incorporation between the inner
and outer D’s. I have further linked this determiner incorporation in Quechua to the
fact that universal quantifiers even on an internal head can take wide interpretive
scope over the relative clause.

Chapter 4
Comparison of Imbabura and
Cuzco Quechua Relative Clauses
4.1 Introduction
This chapter introduces data from a second dialect of Quechua, spoken in the
province of Imbabura in Ecuador. 1
I take the analysis of Cuzco Quechua (CQ)
developed in Chapters 2 and 3 as a point of departure for studying a set of systematic
differences between CQ and Imbabura Quechua (IQ) relativization facts and
suggest that these can be accounted for in terms of basic structural distinctions
between relative clauses in the two dialects. Specifically, I propose that IQ does not
have a Caseless Determiner head which selects a CP as I have proposed for CQ.
This difference means, in particular, that internal heads do not obligatorily raise
in IQ as they do in CQ. A second important difference between the two dialects is
found in the licensing mechanism for genitive Case marking.
1 The judgments in this chapter are only those of consultants from the village of
Peguche in Imbabura. [Cole 1985] discusses some variation within the province of
Imbabura with respect to subordinate clause morphology, and I do not know if the
judgments reflected here extend throughout the province or not.
105

106
Recall that Cuzco Quechua and Imbabura Quechua allow both internally headed
and externally headed relative clauses as illustrated in (4.1) and (4.2).
(4.1) Externally Headed examples
(a) [[Juan-pa
Juan-gen
e i
ranti-sqa-n]
buy-nm-3sg
waka]-qa
cow-top
‘The cow that Juan bought is black.’
yana-n.
black-evid
(CQ)
(b) [[Juan
Juan
e i
ranti-shka]
buy-nm
vaka]-ka
cow-top
‘The cow that Juan bought is black.’
yana-mi.
black-evid
(IQ)
(4.2) Internally Headed examples
(a) [[Juan-pa
Juan-gen
waka
cow
ranti-sqa-n]]-qa
buy-nm-3sg-top
‘The cow that Juan bought is black.’
yana-n.
black-evid
(CQ)
(b) [[Juan
Juan
vaka
cow
ranti-shka]]-ka
buy-nm-top
‘The cow that Juan bought is black.’
yana-mi.
black-evid
(IQ)
Although the above examples exhibit highly parallel surface structures, there
are a series of morphological, syntactic and semantic differences between these and
more complex examples of relative clauses in the two dialects. This chapter presents
a parametric study of the structural origins of these differences. I find that IQ more
closely resembles Japanese in that the relative clause is more sentence-like than in
CQ, and the head may be identified on a purely pragmatic level.
The organization of this chapter is as follows. In Section 4.2 I review some background
on relative clauses in the two dialects, pointing out several morphosyntactic
differences between the two languages. In Section 4.3 I propose a structure for IQ

107
subordinate clauses based on the morphsyntactic evidence. I go on in Section 4.4 to
show how an analysis in which head-raising is mandatory in CQ but not in IQ accounts
for four semantic differences between IQ and CQ internally headed relatives.
The next question to be addressed is whether head-raising ever occurs in IQ, and in
Section 4.5 I present data which I argue supports the possibility of head-raising in
certain IQ relative clauses. Section 4.6 is the conclusion.
4.2 Morphosyntactic comparison between the dialects
4.2.1 Review of similarities between the dialects
Some basic morphosyntactic properties of Quechua (nominalized) relative clauses
have been mentioned in the previous chapters and can be observed in the above
examples. To review briefly, in both dialects I am considering here a nominalizing
morpheme appears on the verb. 2
2 Interestingly, a new relative clause type appears to be emerging in Imbabura
Quechua and in some cases was strongly preferred by some consultants. In this
relative clause type, the verb appears fully inflected, as in main clauses, with no
nominalizing suffix at all. Examples are given in (i) and (ii).
(i) Juan randi-gri-ju-n wagra-ka yurak-mi.
Juan buy-inch-prog-3sg cow-top white-evid
‘The cow that Juan is buying is white.’
(ii)
Ñuka-ka [[kan randi-gri-ju-nki] wagra]-wan-ka yapu-rka-ni-mi.
I-top you buy-inch-prog-2sg cow-instr-top plow-past-1sg-evid
‘I plowed with the cow that you bought.’
Note that in both (i) and (ii) there is no nominalizer on the verb and in fact the
subordinate verb is fully conjugated and apparently finite, just as it would be in a
main clause.
These examples resemble complement clause examples mentioned in [Cole 1985]
for the sub-dialect of IQ spoken in the villages Ilumán and San Roque, northeast
of Otavalo. [Cole p.35] Peguche is in the same direction but even closer to Otavalo

108
The cognate nominalizers -sqa (CQ) and -shka (IQ) can be observed in (4.1) and
(4.2). Differences in the distribution of these morphemes are discussed in Section
4.2.2.4. Also in both dialects neither a wh- expression nor an overt complementizer
is present. Furthermore, the head can only appear internal (as in (4.2)) to the clause
or to the right of the clause (as in (4.1)). External heads can be associated with
both argument and adjunct positions within the relative clause (adjunct cases are
illustrated in (4.3)), but receive the Case marking associated with their role within
the matrix clause. 3
(4.3) [[Juan-pa
Juan-gen
qilqa-sqa-n]
write-nm-3sg
lapicero]-qa
pen-top
‘The pen Juan wrote with is blue.’
azul-mi.
blue-evid
(CQ)
(4.4) [[Juan
Juan
killka-shka]
write-nm
lapiz]-ka
pen-top
‘The pen Juan wrote with is blue.’
azul-mi.
blue-evid
(IQ)
Previous work on relative clauses has been discussed in the preceding chapters.
Regarding the idea of head-raising in relative clauses, in [Lefebvre & Muysken 1988]
on Cuzco Quechua and [Cole 1987a] on Imbabura Quechua it is assumed that external
heads are generated externally while internal heads raise at LF. In [Kayne 1994]
the Antisymmetry framework points to an analysis which is a modification of Cole’s
so very likely participates in this same subdialect. Cole provides the example in
(iii) of an embedded verb with subject agreement, and examples like this were also
accepted and proposed by my consultants.
(iii) Pedro ya-n [ñuka Agatu-pi kawsa-ni-ta]
Pedro think-3 I Agato-in live-1-acc
‘Pedro thinks that I live in Agato.’
I am unfortunately unable to treat this alternative relative clause type here but
leave its analysis to future work.
3 This Case-marking is subject to a matching condition for consultants of [Lefebvre
& Muysken 1988] but not for my consultants. See Section 3.4.4.

109
original analysis of Quechua that is compatible with the LCA. In Kayne’s analysis
both internal and external heads are raised at some point in the derivation.
4.2.2 Four morphosyntactic differences between the dialects
In this section I point to differences in agreement paradigms, Case marking, restrictions
on the head and nominalizing morphology between CQ and IQ. In the next
section I propose a structure of IQ nominalized clauses which reflect these differences.
I will suggest that the differences seen here add up to essentially a more
nominal relative clause in Cuzco Quechua, and in particular a nominal C dominating
the relative clause which probes for the head within the clause. By contrast IQ
relative clauses are more sentence-like and their C head is not (necessarily) a probe.
4.2.2.1 Agreement
In CQ, -sqa and -na nominalized verbs agree with their subjects. In IQ, they do not.
This conforms to a broader distinction between the two languages which has to do
with genitive constructions: CQ possessed nouns agree with their possessors, while
IQ possessed nouns do not. Simple possessives are illustrated in (4.5) and (4.6) for
Cuzco Quechua and Imbabura Quechua respectively.
(4.5) Juan-pa
Juan-gen
‘Juan’s house’
wasi-n
house-3sg
(CQ)
(4.6) Juan-pak
Juan-gen
‘Juan’s house’
wasi
house
(IQ)
Relative clause constructions (internally headed) are illustrated in (4.7) and
(4.8). Note that in (4.7) (from CQ) but not (4.8) (IQ) the verb agrees with the
subject ‘boy’.

110
(4.7) [wayna-q
boy-gen
waka
cow
‘the cow which the boy bought’
ranti-sqa-n]
buy-nm(past/nonsubj)-3sg
(CQ)
(4.8) [[wambra
boy
wagra-(ta)
cow-acc
randi-shka]]
buy-nm(past)
‘the cow which the boy bought’ (C162)
(IQ)
4.2.2.2 Case-marking
The last examples illustrate two other differences between CQ and IQ morphology.
One first difference that can be observed in (4.7) and (4.8) is that in CQ
the subject of a verb that is nominalized with -sqa or -na (nm(past/nonsubj)
and nm(fut/nonsubj) respectively) is marked genitive 4 while in IQ a nominalized
clause’s subject has no overt Case marker, apparently corresponding to the
nominative Case.
A second difference between the two dialects is that in CQ, the direct object
of a nominalized clause (here, waka ‘cow’) has no overt Case-marker. In IQ, it is
reported in [Cole 1985] that the accusative Case-marker -ta is obligatory unless the
direct object immediately precedes a nominalized verb, in which case it is optional
(as in (4.8)). (In both CQ and IQ, main clause direct objects are obligatorily
marked -ta.) I have found that my consultants’ intuitions are basically in line with
this generalization. However, the dropping of -ta on a direct object immediately
preceding the verb seems to be more acceptable in the case of a relative clause
head than in a complement clause. Furthermore, in certain apparently rare cases
of a locative argument relative clause head directly preceding the verb, a locative
marker may also be dropped, as in (4.9) where in fact the presence of the locative
4 As discussed in the previous chapter, [Lefebvre & Muysken 1988] found that the
genitive marking is optional in some cases. My Cuzco Quechua consultants felt the
genitive marking was necessary, however, and here I continue to just consider the
genitive-subject Cases.

111
marker is dispreferred.
(4.9) [[Ñuka
I
wasi-(??pi)
house-loc
‘The house I live in’
kausa-shka]]
live-nm
(IQ)
4.2.2.3 What can be a head
Whereas in IQ there seems to be no direct connection between the Case-marking on
a subordinate clause DP and its possible interpretation as an internal head, in CQ
there is a strict relationship between these things: we have seen in Chapter 3 that
only null-marked arguments can be CQ internal heads. Furthermore, CQ subjects
may be internal heads as long as the subject-head nominalizer -q appears on the
verb, since in these cases subjects have no overt Case marking.
That IQ allows a wide variety of internal heads is illustrated in (4.10) and (4.11)
below, showing a locative and a time adjunct head respectively.
(4.10) [[Ñuka
I
bus-pi
bus-loc
‘The bus I came in’
shamu-shka]]
come-nm
(IQ)
(4.11) [[ñuka
I
chay
that
punlla-pi
day-loc
‘the day that I arrived’ (C187a)
chaya-shka]]-ka
arrive-nm(past)-top
(IQ)
The equivalent IHRs in Cuzco Quechua are ungrammatical, as shown in (4.12)
and (4.13).
(4.12) *nuqa-q
I-gen
bus-pi
bus-loc
‘the bus I came in’
hamu-sqa-y
come-nm-1sg
(CQ)
(4.13) *nuqa-q
I-gen
chay
dem
‘the day I arrived’
p’unchay-pi
day-loc
chaya-mu-sqa-y
arrive-cis-nm-1sg
(CQ)

112
Furthermore, like CQ, IQ allows adjunct external heads in all of the above examples,
in which cases the head does not receive its subordinate clause case-marking.
Examples are given in (4.14) and (4.15).
(4.14) [[ñuka
I
shamu-shka]
arrive-nm
‘The bus I arrived in’
bus]
bus
(IQ)
(4.15) [[Chay
that
ñuka
I
chaya-shka]
arrive-nm
‘That day that I arrived’
punlla]
day
(IQ)
4.2.2.4 Nominalizing morphology
We have seen that CQ distinguishes by means of nominalization morphology on the
verb of a relative clause whether it is the subject or a non-subject of the clause which
is the head. In general, IQ does not make this distinction. Rather, the nominalizers
-shka, -k and -na relate to the tense of the subordinate clause as compared to
the matrix clause. Roughly, they correspond to past, present and future tense of
the subordinate clause compared with the main clause. 5 However, it appears that
5 The discussion in [Cole 1985 p.185] of this point implies that for his consultants,
in the present tense of an IHR the choice of suffix can in fact determine the identity
of the head (that is, -shka indicates a non-subject head while -k indicates a subject
head. This results in a situation in which IQ IHRs and EHRs can employ different
nominalizers even if they are identical in every other way except for the position of
the head. From the description in Cole [1985], this occurs in the case of presenttense,
non-subject relativization. Cole’s descriptions imply that (ii) should also have
the present tense reading.
(i) [ñuka ∅ kawsa-j] wayku
I live-nm(pres) mountain gap
‘the mountain gap where I live’ (C184b)
(ii) ñuka wayku-pi kawa-shka
I mountain gap-loc live-nm(past,pres/non-subj)
‘the mountain gap where I lived’ (C188b)
‘the mountain gap where I live’ (predicted by Cole’s description)
Note that in (i) the nominalizer is -j while in (ii) it is -shka.
(IQ)
(IQ)

113
the use of -na as a nominalizer indicating future tense is being lost in Peguche
Quechua since even consultants who accept or recognize this form generally prefer a
different version of the future which involves the suffix -gri (traditionally analyzed
as ‘inchoative’) along with the “present tense” nominalizer, -k. An example is given
in (4.16) (which does not necessarily imply that the climbing event is imminent).
(4.16) Chai
That
Alberto
Alberto
witsi-gri-ju-k
climb-inch-prog-nm
kiru-kuna-ka
tree-pl-top
‘The trees that Alberto is going to climb are small.’
uchilla-mi.
small-evid
(IQ)
4.2.3 Summary of CQ/IQ differences
The Case, Agreement and Morphological facts about IQ and CQ reviewed thus far
are summarized in Table 4.1. The basic intuition I would like to pursue regarding
these differences is that CQ relative clauses are somehow more nominal in nature
than IQ relative clauses. In the following sections I will ascribe this difference to
differences in the feature structure of CQ vs. IQ functional heads. I will trace the
effect of this basic difference both syntactically, through a proposal in which headraising
(either overt or covert) is mandatory in CQ but not IQ, and semantically as
I examine some systematic differences in IHR meanings between the two dialects.
To anticipate the analysis, however, I will briefly preview how the facts tabulated
in Table 4.1 will be explained. The lack of overt subject-agreement in IQ I attribute
to the lack of availability of agreement morphology in the NP as seen in the lack
Furthermore, Cole finds that the future (relative) tense is for some reason incompatible
with internal heads in IQ. My consultants’ judgments are somewhat different
on these points, however. Although my consultants preferred an external head in
all cases, they do not detect the potential tense shift associated with head position
as in (i) and (ii). Furthermore, their judgments do not correlate suffix choice in the
present tense internal head cases with head position choice. Instead, suffix choice
(in particular -shka vs. -k) is related to relative tense and apparently also to aspect
in some ways that are not clear to me at this point. However, both -shka and -k
seemed to be able to encode present (relative) tense on some occasions.

114
of possessor-noun agreement. The lack of subject-verb agreement in CQ subjectheaded
relatives was explained in Chapter 3 as a result of the defective nature of
the inflectional head that selects the nominalizer -q (n) in that dialect. However, in
IQ I assume that the nominative Case-marking on subjects of nominalized clauses
is associated with non-overt agreement between the subject and the verb. Genitivemarked
subjects I associate to the more nominal nature of CQ relative clauses.
In particular, to a requirement that elements merging to [Spec,n] and triggering
nominal agreement must be genitive-marked. The last three properties of IQ (overtly
Case-marked heads, adjunct internal heads, and disassociation of nominalizer from
head identity) I associate to the more sentence-like nature of relative clauses in that
language and in particular to the possibility of identifying the head pragmatically
and not structurally.
Table 4.1: Summary of CQ/IQ morphosyntactic differences
Property CQ IQ
RC subject-verb agreement? Yes (obj. head) No
No (subj. head)
DP possessor/noun agreement? Yes No
Genitive-marked subjects? Yes No
Can internal heads be overtly Case-marked? No Yes
Adjunct can be an internal head? No Yes
Nominalizer reflects head identity? Yes No

115
4.3 Syntax of IQ subordinate clauses
4.3.1 Explaining the morphological differences
In the previous section I have discussed several differences between CQ and IQ
morphological paradigms. Taken together, these differences suggest that IQ relative
clause functional heads are more similar to those in main clauses than is the case in
CQ. Since subordinate clause Case assignment is apparently potentially identical to
main clause Case assignment in IQ, 6 it is reasonable to assume that the IQ relative
clause determiner can select a complementizer which is identical to the main clause
C. This C head selects a nominative-assigning T head, which in turn dominates
a -n which can select a TrP, etc. Similarly, IQ Complement clauses also contain
functional heads Tr and T which are the same as those in main clauses. However,
the complementizer head of a complement clause must also have a Case feature to
reflect its Case marking within the main clause. The only other difference between
main clauses and complement clauses is n vs. v. It is not entirely clear to me
that this distinction is strictly necessary but here I adopt the convention that the
presence of n corresponds to the eventual Case-marking of the extended nP clause
(DP or CP) within the matrix clause.
4.3.2 Structure of IQ subordinate clauses
This view of IQ gives the structure (4.18) for the IQ complement clause in (4.17).
(4.17) ya-ni
think-1sg
[Marya
Maria
mishu
mestizo
shimi-ta
language-acc
parla-j]-ta
speak-nm-acc
‘I think that Maria speaks Spanish.’ (Cole 109b) (IQ)
6 I return to the possibility of null marking on subordinate clause objects and
relative clause heads in Section 4.5

116
(4.18) CP
✟ ✟✟✟ ❍ ❍ ❍
C TP
✟ ✟✟✟
❍
❍
❍ ❍
Maria i T’
✟ ✟✟✟ ✟
T
❍❍
❍
❍
❍
nP
❍ ❍
✟ ✟✟✟✟✟✟
❍
❍
❍ ❍
t i n’
❍ ✟ ✟✟✟✟
TrP
✟ ✟ ✟ ❍
❍
❍
mishu shimi-ta Tr’
Spanish-acc ✟❍
V Tr
❍
❍ ❍
n
parla-j
speak j -nm-
t j
t j
I have suggested in the preceding section that there is a connection between the
presence of genitive Case on a subject in CQ nominalized clauses and the nominal
agreement pattern found also in regular possessor constructions. Let us examine
this question here by explicitly comparing CQ and IQ in this regard. I first propose
the structure in (4.19) for a DP with a possessor. The presence of n provides the
position in which the possessor, which I take to function like a “subject” of the DP
is initially Merged.
DP
✟ ✟✟✟
❍ ❍ ❍
(4.19) DP D’
✟
Juan i -gen
✟✟ ❍
❍
❍
D nP
❍
✟ ✟✟✟ ❍
❍
❍
t i n’
✟ ✟✟ ❍ ❍
NP n
house(IQ)
house-3sg(CQ)

117
With this background let us turn to the question of what determines Casemarking
on a subject in CQ and in IQ. In Table 4.2 I summarize the relevant facts.
Table 4.2: Case-marking on subjects
Cuzco Quechua
Imbabura Quechua
Case Assigner n/v Agr? Case Assigner n/v Agr?
DP w/poss. Gen D n N-Agr Gen D n —
subj-head RC none T(defect.) n — Nom T n null
obj-head RC Gen T n N-Agr Nom T n null
Cmp. Clause Gen T n N-Agr Nom T n null
Main Clause Nom T v V-Agr Nom T v V-Agr
For each of Cuzco Quechua and Imbabura Quechua, and each construction type
(relative clauses, main clause, complement clause and DP with possessor) I have
indicated the Case-marking on the subject/possessor, the Case assigner (D or T), the
presence of n or v (recall that this is correlated with the presence or absence of Casemarking
on the constituent in question) and the presence or absence of agreement
with the verb. As mentioned earlier, and reflected in Table 4.2, I am assuming that
IQ subordinate clauses have null agreement with the subject, presumably associated
with the loss of nominal agreement morphology in this language. The justification
for this assumption is the idea that nominative Case assignment is associated with
(overt or null) subject-verb agreement.
Examination of the table reveals that in Cuzco Quechua, genitive Case marking
is correlated with nominal agreement on the verb. In Imbabura Quechua, however,
genitive Case-marking correlates with Case-assignment by D and not by T. Thus,
the simplest model attributes the particular reflex of structural Case of a subject

118
(merged with n or v) to n or v itself in CQ and to the particular Case assigner (T
or D) in IQ. To implement this idea I propose that in CQ, only a genitive-marked
nominal can merge with the complex head n-N (after head-movement of n to N). This
explains why in main clauses, which lack n, subjects cannot be genitive-marked. It
further explains why in subject-headed relative clauses, where there is no agreement
between noun and verb, genitive marking is also not possible. In IQ, I propose that
genitive marking on a noun phrase is licensed only when the Case of that nominal
is checked by a D head, not a T head. Looked at this way, the genitive marking
in CQ can be viewed as a kind of quirky Case-marking since it does not correlate
directly with the identity of the Case-checker itself. 7
I now turn to the structure of IQ relative clauses. In fact, this structure will be
largely elucidated in the following sections. However, as in CQ I will adopt the basic
relative clause structure of (3.30), repeated here as (4.20) for IQ. The question of
whether and when the head (DP i ) raises to [Spec,CP] will be addressed in Sections
4.4 and 4.5.
(4.20) DP
✟ ✟ ❍❍ ❍
D CP
✏ ✏✏
...DP i ...
Since the Case-marking properties of the CP in (4.20) in IQ are (at least potentially)
the same as in a complement clause, I will assume for the moment that the
functional heads of the relative clause are essentially the same as those of a complement
clause. Again, I will examine this issue more closely in the next sections.
7 In [Chomsky 2000a p.127], for instance, quirky Case is described as “inherent
Case with an additional structural Case feature”.

119
4.4 Semantics of IQ vs. CQ internal heads
The results of the previous section suggest that the complementizer in CQ is more
nominal in nature than that in IQ, as reflected in morphological differences between
the two dialects’ relative clauses. In this section I would like to take this idea one
step further by proposing that in fact in a relative clause the complementizer in IQ
may not even probe for the head at all. That is, that IQ relative clauses do not
(necessarily) involve head-raising. This is a significant departure from my proposal
for CQ in the previous chapter. The basic idea is summarized in (4.21).
(4.21) CQ: Head-raising mandatory at some level, for Case-based reasons.
IQ: Relative clause is more sentence-like. No head-raising is necessary.
The idea that IQ internally headed relative clauses are more sentence-like and
can be interpreted without head-raising, though I have initially motivated it in
connection with morphosyntactic facts, makes specific predictions regarding the
meaning of IHRs in IQ, which I will consider carefully in the next section.
In order to implement the basic intuition expressed in (4.21), I will proceed as
follows. In the next section I point out four differences between internal heads and
external heads in CQ and IQ. I will argue that all four differences can be explained by
the dichotomy suggested in (4.21). In the following section I will suggest a structural
implementation of (4.21) which will effectively show that IQ is a Japanese-like dialect
of Quechua with respect to relative clause interpretation.
4.4.1 Four differences between CQ and IQ internal heads
In this section I will show that CQ and IQ IHRs differ semantically with respect to
ambiguity of the head, adjunct internal heads, universally quantified internal heads

120
and in the pattern of Case-marking on the head. Each of these differences can be
explained by the more sentence-like nature of IQ relative clauses suggested in (4.21).
4.4.1.1 Ambiguity of the head
A first difference between the two dialects is that in CQ the identity of an internal
head is unambiguously determined by a combination of the choice of nominalizer
(subject or non-subject) and Case marking. By contrast in IQ the identity of the
head is often highly ambiguous.
To see this, compare example (4.22) from CQ with (4.23) from IQ. Notice that
in CQ the internal head is unambiguously the un-Casemarked direct object rumi
‘rock’, while in the same sentence in IQ the head could be understood as the subject
wawa ‘child’, the direct object rumi ‘rock’ or the indirect object alku ‘dog’.
(4.22) [[Irqi-q
Child-gen
ka-rqa-n.
be-past-3sg
alqu-man
dog-dat
rumi
rock
ch’anqa-sqa-n]]-qa
throw-nm(nonsbj)-3sg-top
‘The rock that the child threw at the dog at was black.’
yana
black
(CQ)
(4.23) [[Wawa
child
alku-man
dog-dat
rumi-ta
rock-acc
shita-shka]]-ka
throw-nm-top
‘The rock that the child threw at the dog is black.’
‘The child who threw the rock at the dog is black.’
‘The dog that the child threw the rock at is black.’
yana-mi.
black-evid
(IQ)
4.4.1.2 Possibility of adjunct internal heads
Next we come to the issue discussed extensively in the last chapter, which has to do
with the possibility of adjunct heads. Adjunct heads may be internal in Imbabura
Quechua, as illustrated again in (4.25). However, in Cuzco Quechua adjunct heads

121
cannot be internal, as shown in (4.24). 8
(4.24) *[[Lapicero-(wan)
pen-instr
qilqa-sqa-y]]
write-nm(nonsbj)-1sg
musuq.
new
Intended meaning: ‘The pen that I wrote with is new.’
(CQ)
(4.25) [[Juan
Juan
lapiz-wan
pen-instr
killka-shka]]-ka
write-nm-top
azul
blue
‘The pen Juan wrote with is a blue pen.’
lapiz-mi.
pen-evid
(IQ)
4.4.1.3 Interpretation of universal quantifiers on internal heads
A third difference between the dialects relates to an issue analyzed in detail for
Cuzco Quechua in Chapters 2 and 3. This was the fact that in Cuzco Quechua
a strong quantifier on an internal head always receives an external interpretation.
That is, the strong quantifier will have scope over the entire relative clause. This
fact is illustrated in (4.26).
By contrast, a strong quantifier on an internal head in Imbabura Quechua is
generally interpreted internally to the clause. The external interpretation is accepted
by some speakers but not by others. An example is shown in (4.27).
(4.26) Ana-qa
Ana-top
ukya-ra-ni.
drink-past-1sg
[[Jose-q
Jose-gen
tukuy
all
unu
water
‘Ana drank all the water that Jose brought.’
*’Jose brought all the water and Ana drank it.’
apa-mu-sqa-n]]-ta
bring-cis-nm-3sg-acc
(CQ)
(4.27) Ana-ka
Ana-top
[[Jose
Jose
tucuilla
all
yaku-ta
water-acc
‘Jose brought all the water and Ana drank it.’
? ‘Ana drank all the water that Jose brought.’
apa-mu-shka]]-ta
take-cis-nm-acc
upya-rka.
drink-past
(IQ)
8 The externally headed equivalents of (4.25) and (4.24) are acceptable in both
dialects.

122
4.4.1.4 Overt Case marking on an internal head
Another point discussed individually for CQ and IQ in 4.2.2.3 has to do with the
possibility of an overt Case marker on an internal head. The basic generalization
was that overt Case-marking on a head is possible in IQ (and obligatory if the head
is not adjacent to the verb) but not in CQ. Examples are shown in (4.28) for CQ
and (4.29) for IQ.
(4.28) Juan-pa
Juan-gen
waka-(*ta)
cow-acc
ranti-sqa-n-qa
buy-nm-3sg-top
‘The cow that Juan bought is black.’
yana-n.
black-evid
(CQ)
(4.29) [[Juan
Juan
wagra-(ta)
cow-(acc)
ranti-shka]]-ka
buy-nm-top
‘The cow that Juan bought is black.’
yana-mi.
black-evid
(IQ)
4.4.2 Structural implementation of the intuition
I will now turn to an implementation of the idea expressed in (4.21) in structural
terms. Observe that if we adopt the Hoshi/Shimoyama analysis of Japanese IHRs
(as presented in Chapter 2), we derive immediately that the relative clause can be
interpreted as a sentence. I claim that such an analysis not only reflects the lack of
a +wh/head-feature on C which I have associated with head-raising, but actually
predicts the four differences between internal heads outlined in the previous section.
Recall that under the Japanese-type analysis, the head is identified (purely)
pragmatically via E-type anaphora. This view was spelled out for Japanese by
[Hoshi 1995] and [Shimoyama 1999, 2001].
Analyses of particular examples from CQ and IQ are given in the following trees.
First, (4.30) shows an IQ internally headed relative clause with no head raising. This
tree corresponds to example (4.27).

123
(4.30) CP
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟
❍
❍
❍
❍
CP i
TP
✏
✏ ✏✏✏✏✏✏✏✏ ✟ ✟✟✟
❍ ❍
❍ ❍
DP VP
Juzi tukuilla yaku-ta apamu-shka-ta
✟ ✟✟ ❍ ❍
Jose brought all the water Ana-ka DP V
✟❍
pro t i upya-rka
drank
In (4.31), I show an example of a CQ IHR with head raising and determiner
incorporation as discussed in Chapter 3. This type of head movement is similar to
that proposed in [Bianchi 2000], and uses example (4.26) as a model.
(4.31) TP
❍ ❍
❍
❍
❍
❍
✟ ✟✟✟✟✟✟✟✟✟✟
❍
❍
❍ ❍
DP
VP
❍
❍
❍
Nuqa
❍
❍
I
✟ ✟✟✟✟✟✟✟✟
❍
❍
❍ ❍
DP
V
✟ ❍❍
✟ ✟✟✟✟✟✟
D
✟❍
all j D
❍
❍
❍
❍
❍
❍
CP
❍
✟ ✟✟✟✟
CP
✏ ✏✏✏✏✏✏
taytay-pa e i ruwa-sqa-n-ta
my father built
❍
❍
❍
❍
DP i
✏ ✏
t j wasi
houses
muna-ni
like

124
4.4.3 How the analysis explains the CQ/IQ semantic differences
Let us now reconsider the four distinctions between CQ and IQ from 4.4.1 in light
of the above structures.
First, given that the necessity for head-raising distinguishes IQ from CQ, the
fact that the internal head is highly ambiguous in Imbabura Quechua, while unambiguous
in Cuzco Quechua follows immediately. Essentially, in IQ, head identity is
established purely pragmatically through an anaphoric relationship with a nominal
in the embedded clause. In CQ, on the other hand, the head is identified structurally;
it is the un-Casemarked nominal, which must raise, as described in Chapter
3.
Next we turn to the observation that adjuncts can be internal heads in Imbabura
Quechua but not Cuzco Quechua. The reason for this under the current analysis
must be that adjuncts can be selected pragmatically but not structurally. Here I
assume that nominals which are not θ-marked by the subordinate verb must receive
non-structural Case within their clause.
The third difference between the two dialects is that an internal interpretation
of universally quantified heads is possible in IQ but not in CQ. I propose that
the reason for this is again that an IQ subordinate clause can be interpreted as a
sentence. Therefore, all RC-internal quantifiers remain relative clause-internal at
LF and have internal scope. In CQ, head-raising leads to scope of the Determiner
over the subordinate clause via determiner incorporation as discussed in Chapter 3.
Finally, recall that overt Case-marking on a head is possible in IQ but not in
CQ. This can be explained if in IQ, no element need raise, and hence all subordinate
DPs can be Case-marked within the subordinate clause.

125
4.5 Do IQ heads ever raise?
In the preceding sections I have argued that IQ heads, unlike CQ heads, do not have
to raise to clause-external position at any point in the derivation. I have shown that
the analysis of internally headed relative clauses posited for Japanese by [Shimoyama
1999, 2001] explains a variety of semantic effects in IQ and I have suggested that
there is a correlation between the less nominal nature of IQ morphology and the
lack of a D-feature on the complementizer C which is associated with raising of the
head in CQ at some point in the derivation. (In Chapter 3 I have proposed that the
[+D] complementizer has a [uwh] probe which Agrees with the probe).
However, the simple fact that externally headed relatives do exist in IQ leads us
to ask whether these heads must all now be supposed to originate clause-externally
or whether in fact any of them have raised from a clause-internal position. If the
latter explanation is correct, then we may wonder whether it is ever possible for
even an internal head to raise. Since IQ internal heads have been posited to raise
in past work such as [Cole 1985, 1987a], it is important to consider the arguments
for raising in light of the previous discussion.
In this section I address this variety of facts. I present evidence that some IQ
external heads are derived via movement, while other IQ external heads are basegenerated
externally. I will argue that in fact IQ is more similar to CQ than it
appears at first glance in that argument heads are the ones which may raise, while
adjunct heads may not.
4.5.1 Evidence from Islands
Whether a head can be extracted from an island is often used to argue for or against
the mechanism of raising in deriving a relative clause head. In this section I show
that in general IQ does show island effects in relative clauses.

126
Examples (4.32) to (4.39) show that the relative clause head cannot be associated
with a position within a complex DP. Example (4.32) and (4.33) show that neither
an internal nor an external head can be a possessor.
(4.32) Maria
[Maria
wagra-pak
cow-gen
chico-ta
baby-acc
ranti-shka
buy-nm]
chinka-shka
lose-past.rep
*‘The cow whose calf Maria bought is said to be lost.’
nin.
say.
Consultant suggests the following possible interpretation ‘The calf that
Maria bought for the cow got lost.’ (Under beneficiary reading of -pak)
(IQ)
(4.33) *[[Maria
Maria
chico-ta
baby-acc
ranti-shka]
buy-nm
wagra]-ka
cow-top
‘The cow whose calf Maria bought got lost.’
chinka-shka.
lose-past.rep
(IQ)
Recall that in Cuzco Quechua too, relativization was also not possible from
within a DP except for the cases of external possessor heads as in (3.62) and (3.63),
repeated here.
(4.34) [[Waka-q
cow-gen
uña-n
baby-3sg
suwa-pu-sqa-nku]]
steal-definitively-nm-3pl
‘the cow’s calf which they stole’
*‘the cow whose calf they stole’
(CQ)
(4.35) [[e
e
Uña-n
baby-3sg
suwa-pu-sqa-nku]
steal-definitively-nm-3pl
‘the cow whose calf they stole’
waka]
cow
(CQ)
Given the asymmetry in (4.34) and (4.35) I suggested in Chapter 3 that (4.35)
was not an example of an extracted head but rather of an external head which
corefers with an internal pro. For IQ I propose that the lack of possessor-agreement
means that a pro is not licensed in possessor position, making the relationship
between the external head waka ‘cow’ and the internal chico ‘baby’ insufficiently
salient to license a modificational structure of that sort. Evidence for the lack
of possessor pro comes from the contrast between (4.36) and (4.37) from CQ and

127
IQ respectively. Only in CQ can a bare noun (with nominal person marking) be
understood as having a pronominal possessor.
(4.36) wasi-n
house-3sg
‘his/her house’
(CQ)
(4.37) wasi
house
‘house’
*’his/her house’
(IQ)
Examples (4.38) and (4.39) show that neither an internal nor an external head
can be derived from an element of a subordinate DP.
(4.38) *[[Maria [[alku-ta
Maria dog-acc
doberman-mi.
doberman-evid
miku-chi-k]
eat-cause-nm
warmi]-ta
woman-acc
riksi-k]]-ka
know-nm-top
‘The dog that Maria knows the woman who feeds (it) is a doberman.’
Consultant’s comment: Maria’s friend is a doberman?
(4.39) *[[Maria [ DP [[e] miku-chi-k]
Maria eat-cause-nm
doberman-mi.
doberman-evid
warmi]-ta
woman-acc
riksi-k]
know-nm
alku]-ka
dog-top
‘The dog that Maria knows the woman who feeds (it) is a doberman.’
(IQ)
(IQ)
Similar facts to those above are noted and discussed in [Cole 1985], [Cole &
Hermon 1981, 1994] and these examples are taken as evidence that the head does
raise in IQ.
Cole [1985] also finds that the subject of an embedded clause cannot be a head.
Relevant examples are shown in (4.40) and (4.41). My consultants share the judgments
reported there.

128
(4.40) *[[Marya
Maria
[warmi
woman
Juan-ta
Juan-acc
riku-shka]]-ta
see-nm-acc
ni-shka]
say-nm
llugshi-rka.
leave-past
‘The woman who Maria said saw Juan left.’ (Cole 1985 p.54 195)
(IQ)
(4.41) *[Marya Juzi-man
María José-to
llugshi-rka.
leave-past
ni-shka
say-nm
[∅ Juan-ta
Juan-acc
riku-shka]-ta
see-nm-acc
warmi]
woman
‘The woman who Maria told José that saw Juan left.’ (Cole 1985 p.56 190)
(IQ)
However, Cole finds that an embedded clause object can head a relative clause.
In this case my consultants do not share these judgments. For my consultants, it
is impossible or extremely difficult (and equally difficult) to get either the subject
or an object of an embedded clause to be the head of a relative clause. Examples
of attempted subject relativization are shown in (4.40) and (4.41), and object relativization
in (4.42) and (4.43). I take these examples from [Cole 1985] and give
both the judgments of his consultants and of mine. I should note that in each case
(since the intended meaning was never readily apparent to my consultants) I set up
a context by affirming (in Quechua) the content of the relative clause to promote
the intended meaning of the sentence. The context is given in each case.
(4.42) Context: Maria believes that Jose saw a child.
[Chay [Marya
dem Maria
ri-rka.
leave-past
[Juzi
Jose
[e] riku-shka]-ta
see-nm-acc
kri-k]
believe-nm
wawa]
child
ña-mi
already-aff
‘The child that Maria believes that Jose saw already left.’ (Cole 1985 p.54
193) (IQ)
My consultants’ judgment *‘The child that Maria believes that Jose saw
already left.’
Consultant’s comment “It must be the child who believes that Maria saw
José.”

129
(4.43) Context: Maria believes Jose gave the book to a child.
[Chay [Marya
that Maria
ña-mi
already-aff
[Juzi
Jose
ri-rka.
leave-past
libru-ta
book-acc
kara-shka]-ta
give-nm-acc
kri-k]
believe-nm
wawa]
child
‘The child that Maria believes Jose gave the book to.’ (Cole 1985 p.54 194)
(IQ)
My consultants’ judgment: *
(4.44) Context: Maria said that Juan saw a child.
[[Marya
Marya
[Juan
Juan
wawa-ta
child-acc
riku-shka]-ta
see-nm-acc
ni-shka]]
say-nm
llokshi-rka.
leave-past
‘The child who Maria said that Juan saw left.’ (Cole 1985 p.54 196)
My consultants’ judgment: *
(IQ)
(4.45) Context: Maria said that Juan gave the book to a woman.
[[Marya [Juan
Maria Juan
llokshi-rka.
warmi-man
woman-dat
libru-ta
book-acc
kara-shka]-ta
give-nm-acc
ni-shka]]
leave-past
‘The woman who Maria said that Juan gave the book to left.’ (Cole 1985
p.54 197) (IQ)
My consultants’ judgment: *
To summarize, for my consultants at least there is a strong condition by which
not only must a relative clause head not be drawn from within a DP, but in fact
it cannot be drawn from a complement clause either. Similar facts for CQ were
discussed in Section 3.4.3. Cole’s report differs in that his consultants do allow the
relativization of an embedded clause non-subject.
The general fact that IQ obeys a DP-island constraint needs to be explained.
Certainly blocked movement would be a good way to derive the ungrammaticality

130
of (4.38) and (4.39). On the other hand, if the E-type anaphora analysis is available,
why can’t it apply in examples such as (4.39).
I propose that the externally headed examples are bad in IQ (but not Japanese)
because the mere presence of pro in an IQ relative clause is not sufficient to establish
the relationship between the head and the relative clause. In fact, an external head
must have a pragmatically reconstructable relationship to the clause which is its
direct sister, and not to a clause which is further embedded. On the other hand,
it appears that only the C which is selected by D can have the +wh/head feature
which is associated with head-raising. In this respect IQ and CQ are apparently
identical.
4.5.2 Evidence from quantifiers
The quantifier evidence that needs to be considered in IQ is truly vast and I am unable
to do it justice here. I would like to comment on just a few aspects of quantifier
behavior that are particularly relevant to the current question of movement.
First, as has already been noted, strong quantifiers can take internal scope in
relative clauses, as in (4.27), repeated here as (4.46).
(4.46) Ana-ka
Ana-top
[[Jose
Jose
tukuilla
all
yaku-ta
water-acc
‘Jose brought all the water and Ana drank it.’
? ‘Jose brought all the water that Ana drank.’
apa-mu-shka]]-ta
take-cis-nm-acc
upya-rka.
drink-past
(IQ)
I should note here that I proposed (4.46) to my consultants initially with the
expectation that IQ was likely to follow the pattern I was familiar with from earlier
work on CQ: the quantifier takes wide scope over the clause. Instead, the consultant
invented a context to go with this expression in which there was some quantity of
water in the kitchen, and Jose had brought all of this quantity to Ana, who had then
drunk it. Generally, consultants did not recognize any second reading of expressions

131
of this sort, but occasionally the second gloss was admitted or even suggested. To
the extent that the wide scope reading of (4.46) is possible, as suggested by the
second gloss, we have a very strong argument for head-raising being an option in
IQ.
However, on most occasions my consultants’ intuitions about quantifier scope
matched the surface position of the quantifier. This was true in cases of both weak
and strong quantifiers. Typical examples are given in (4.47) and (4.48), which
show that an internal quantified head can interact with another embedded-clause
quantifier, while an external head of the same sort cannot.
(4.47) [[Wakin
some
miku-rka.
eat-past
wawa-kuna
child-pl
(IQ)
ishkai
two
tanta-ta
bread-acc
ranti-shka]]-ta
buy-nm-acc
Ana-ka
Ana-top
‘Some children bought two rolls [together or each] and Ana ate them.’
(4.48) [[Wakin
some
miku-rka.
eat-past
wawa-kuna
child-pl
(IQ)
ranti-shka]
buy-nm
ishkai
two
tanta]-ta
bread-acc
Ana-ka
Ana-top
‘Ana ate two rolls that some children bought.’ (two rolls total)
The equivalent examples in CQ are judged to be paraphrases of one another, although
the distributive reading with wakin ‘some’ in the subject and no distributive
marker -nka on the direct object is apparently hard to get (so both internal and
external heads are typically associated with the non-distributive reading).
In IQ, externally headed relatives with an overt quantifier on the head were often
met with disfavor. For example, consider the following set of four relative clauses
in which the strong quantifier tukuilla is associated with the head waka ‘cow’ and
appears to the left of the relative clause, on an internal head, internal with an
external head and on an external head. Of these, all were judged as potentially

132
allowing an external interpretation of tukuilla ‘all’ but the last was deemed degraded
(or, to quote the consultant directly “understandable, but not perfect”). The first
was the preferred construction for this meaning.
(4.49) [Tukuilla [Paul feria-pi ranti-shka] waka]-ka yana-mi.
all Paul market-loc buy-nm cow-top black-evid
‘All the cows that Paul bought in the market were black.’
(4.50) [[Paul
Paul
feria-pi
market-loc
tukilla
all
waka-ta
cow-acc
ranti-shka]]-ka
buy-nm-top
yana-mi.
black-evid
(4.51) [[Paul
Paul
feria-pi
market-loc
tukuilla
all
ranti-shka]
buy-nm
waka]-ka
cow-top
yana-mi.
black-evid
(4.52) ?[[Paul
Paul
feria-pi
market-loc
ranti-shka]
buy-nm
tukuilla
all
waka]-ka
cow-top
yana-mi.
black-evid
Judgments like these make it appear that in IQ, like CQ, the head can raise to
an external position. 9 This movement is not linked to lack of overt Case marking,
as seen in (4.50).
The quantifier evidence therefore suggests that head-raising is possible in IQ,
but it is not directly related to Case-marking and in most cases the most prominent
reading matches an in-situ analysis of the head.
4.5.3 Evidence from arguments vs. adjuncts in headless
relatives
The previous section suggested that perhaps head-raising is possible in IQ after all.
In this section I show further evidence for head-raising in certain cases of relative
9 I note in passing that quantified external heads like (4.52) in CQ did not give
consultants any trouble, although IQ consultants often hesitated over them, as mentioned
above. I do not know the reason for this difference, although it seems likely
to be related to the difficulty with determiner incorporation in IQ.

133
clauses. The evidence is based on a contrast in behavior between argument and
adjunct relativization.
I begin with simple examples of internally and externally headed relatives. In
(4.53) and (4.54) we have an EHR/IHR pair illustrating argument relativization,
while (4.55) to (4.58) illustrate adjunct relativization.
(4.53) [[Juan
Juan
[e i ]
e
shuwa-shka]
steal-nm
‘the cow that Juan stole’
wagra i ]
cow
(IQ)
(4.54) [[Juan
Juan
wagra
cow
shuwa-shka]]
steal-nm
‘the cow that Juan stole’
(IQ)
(4.55) [[Juan
Juan
[e i ]
e
killka-shka]
write-nm
lapiz i ]
pen
‘the pen that Juan wrote with’
(IQ)
(4.56) [[Juan
Juan
lapiz-wan
pen-instr
killka-shka]]
write-nm
‘the pen that Juan wrote with’
(IQ)
In (4.55) and (4.56), the head ‘pen’ is an instrumental adjunct of the subordinate
verb ‘write’. In (4.57) and (4.58), the head ‘pot’ is a locative adjunct of the
subordinate verb ‘cook’.
(4.57) [[Maria
Maria
[e]
e
yanu-ju-c]]
cook-prog-nm
‘the pot that Maria is cooking in’
manga]
pot
(IQ)
(4.58) [[Maria
Maria
manga-pi
pot-loc
yanu-ju-c]]
cook-prog-nm
‘the pot that Maria is cooking in’
(IQ)

134
4.5.3.1 Distribution of pro
Since the evidence for movement of arguments will depend on cases of null relative
clause heads, I first provide in this section some background on the distribution of
pro in IQ.
Quechua allows both null subjects as in (4.59) and null objects as in (4.60) from
IQ, and in the question-answer pair (4.61) from CQ. Subjects are always marked
overtly on the verb, while in Ecuadorian Quechua objects generally are not. Other
dialects of Quechua regularly mark 1st and 2nd person objects on the verb. In
Imbabura, overt marking of the 1st person object on the verb, as in (4.59), is possible
but optional.
(4.59) ∅
pro
Nuka-ta
I-acc
riku-(wa)-rka.
see-(1sg)-past
‘He/she/it saw me.’
(4.60) Juzi
Jose
∅ riku-rka.
saw
‘Jose saw him/her/it.’ (Cole 1987b (1))
(IQ)
(IQ)
(4.61) Taruka-ta
Deer-acc
riku-ra-nki-chu?
see-past-2sg-q
... Mana
not
riku-ra-ni-chu.
see-past-1sg-neg
‘Did you see a deer?’ ... ‘No, (I) didn’t see (one).’
Cole [1987b] argues that the null object in Quechua is a pronominal (pro) and
not a null variable. Quechua contrasts with languages like Chinese, which allow
only null variable objects (in addition to overt objects). Sánchez [1999b] extends
this view and shows that the null pronominal in object position in Quechua can be
marked as [+/- specific] and [+/- definite]. She illustrates this through examples
such as (4.62):
(4.62) Na
not
∅ riku-rka-ni-chu.
see-past-1sg-neg
‘I didn’t see her/any.’

135
Following these analyses, I will assume that Quechua allows a null pronominal
pro in both subject and object position. I will assume further that pro is limited to
argument positions. One piece of evidence for the absence of null adjuncts is seen
in (4.63), which contrasts with (4.62):
(4.63) Na
not
∅
killka-rka-ni-chu.
write-past-1sg-neg
‘I didn’t write (it).’ (IQ)
*‘I didn’t write with it.’
Similarly in the CQ question-answer pair in (4.64), the adjunct ‘this pen’ must
be overtly present in order to be specifically negated in the answer.
(4.64) Kay
this
lapicero-wan-chu
pen-instr-q
qilqa-ra-nki?
write-past-2sg
... Mana
not
qilqa-ra-ni. (Huq-wan qilqa-ra-ni.)
write-past-1sg. (other-instr write-past-1sg
kay
this
lapicero-wan-chu
pen-instr-neg
‘Did you write with this pen?’ ... ‘No I didn’t write with this pen. (I wrote
with another one.)’
Omitting the adjunct and placing negation on the verb yields sentential negation,
as seen in (4.65).
(4.65) ... Mana
not
qilqa-ra-ni-chu.
write-past-1sg-neg
(#Huq-wan
other-instr
‘I didn’t write. (#I wrote with another one.)’
qilqa-ra-ni.)
write-past-1sg
In the following section I will discuss the absence of pro in adjunct position
further.
4.5.3.1.1 Arguments (not adjuncts) may be the understood head of a
headless RC In (4.66) to (4.68), the subordinate clause is interpreted as a (restrictive)
relative. However, there is no overt reflex of the head. The head is understood
to be the subject, the direct object and the locative argument respectively.

136
(4.66) [[[e] Mishki-ta
candy-acc
ranti-shka]]-kuna
buy-nm-pl
‘Those who bought the candy’
(4.67) [[wawa-kuna
child-pl
[e] ranti-shka]]-kuna
buy-nm-pl
‘those (candies) that the children bought’
(4.68) [[Juan
Juan
kayna
last
wata
year
[e] kausa-ju-shka]]
live-prog-nm
‘the one (house) Juan was living in last year’
(IQ)
(IQ)
(IQ)
By contrast adjuncts may not be relativized through headless relativization.
This is illustrated in (4.69) for an instrumental adjunct and in (4.70) for locative
adjuncts. In each case, the headless relative can be interpreted as a case of object
relativization, but not adjunct relativization. For example in (4.70), consultants
report that even if we are talking about pens, this expression cannot mean ‘the one
that Juan wrote with’.
(4.69) [[Juan
Juan
killka-shka]]
write-nm
‘The thing that Juan wrote’
*‘The one that Juan wrote with’ (Talking about pens)
(4.70) [[Maria
Maria
yanu-ju-k]]
cook-prog-nm
‘The thing that Maria is cooking’
*‘The one that Maria is cooking in’ (Talking about pots)
(IQ)
(IQ)
An explanation for the asymmetry presented here can be found in the fact that
Quechua is a null object language as discussed in section 4.5.3.1. The understood
heads in (4.66) to (4.68) are in fact syntactically present as null pronominals. I
claim that the unacceptability of (4.69) and (4.70) is due to the fact that null pro is
not licensed in adjunct position. In fact, it appears that the null argument is obligatory
in (4.66) to (4.68). Otherwise we would have no way to distinguish between

137
argument and adjunct relativization. This pro will be in the position marked by the
empty category [e] in the examples.
Now, simply positing the presence of pro does not immediately decide the issue
of whether raising takes place or not. Therefore the next question to be considered
is whether this pro gets raised out of the relative clause or interpreted within the
relative clause. I claim that the null pro head must raise out of the relative clause
by the level of interpretation.
A first piece of evidence for this is the position of the plural marker -kuna on
the verb in examples (4.66) and (4.67) which we just looked at. Since nominalized
verbs in Imbabura Quechua do not show subject agreement, it is the plurality of the
head which is being encoded here. Notice that the plural marker appears externally,
attached to the verb. This provides a first indication that the null pronominal has
an external presence as well as an internal presence in its base argument position.
However, we must consider the possibility that the external -kuna is in fact reflecting
the plurality of the external DP. That is, could the plurality marker be associated
with plurality of the outer DP (DP 1 in (4.71)), not the head itself?
(4.71) DP 1
❍
✟ ✟✟✟✟
❍
❍
❍ ❍
D
✏ CP
✏ ✏✏✏✏✏✏ wawa-kuna pro ranti-shka-kuna
child-pl pro buy-nm-pl
If the plural marker is associated to DP 1 , we might expect the same structure
to be possible with an overt internal head. However, this is not possible, whether
the head is singular or plural, as seen in (4.72):
(4.72) *wawa-kuna
child-pl
mishki(-kuna)
candy-pl
ranti-shka-kuna
buy-nm-pl
‘The candies that the children bought’
(IQ)

138
Positive evidence that the external -kuna may be associated with an external
pro is given by examples which contain a null proform and a plural marker, such as
(4.74).
(4.73) jatun
big
‘big children’
wawa-kuna
child-pl
(IQ)
(4.74) jatun-kuna
big-pl
‘big ones’
(IQ)
Another piece of evidence that the null pro raises out of the relative clause is that
the pragmatic construal analysis (that is, the Japanese-inspired E-type anaphora
analysis) does not predict the right meaning for a sentence like (4.78): ‘Those that
the children bought are good.’ The tree in (4.76) shows the LF structure of the nonhead-raising
analysis. The core of the interpretation scheme is that the external
proform is interpreted via E-type anaphora as coindexed with a noun phrase of
the embedded clause. For example, in the headed relative clause in (4.75), the
embedded CP is raised out of the IP and interpreted as a sentence. This sentence is
then conjoined to the matrix IP, which contains the proform, which we may think
of as a function from an index to a predicate of type < e, t >. The denotation of the
proform is established by a contextually-determined assignment g C as illustrated
below.
(4.75) [[wawa-kuna
child-pl
mishki-kuna
candy-pl
ranti-shka]]-ka
buy-nm-top
‘The candies that the children bought are good.’
‘The children bought candies and they are good.’
sumak-mi.
good-evid
(IQ)

139
(4.76) IP
❍ ❍
✟ ✟✟✟✟✟✟
❍
❍
❍ ❍
CP j
IP
✏
✏ ✏✏✏✏✏✏ ✟ ✟✟ ❍ ❍ ❍
DP VP
The children bought candies ✏ ✏
✏ ✏
pro 5 t j are good
g C : 5 → the candy that children bought
Now we turn to the case of the headless relative clause. Here again as in (4.76),
in order to capture the fact that the relative clause is read as a separate sentence it
is actually scoped out to the spec of the matrix IP. So the predicted interpretation
is: ‘The children bought them and they are good’, with ‘they’ representing the
external pronominal head which is supposed to pick up the first pro from the context.
However, this reading requires this first little pro to be independently licensed by the
context. That is, the reading would require some contextually prominent referent
in order for the identity of the little pro in the relative clause to be recovered by the
addressee. On the other hand, the head-raising analysis does not require a licensing
condition of this sort, but predicts the correct meaning as shown in the second tree
of (4.77).
(4.77) (a) Wrong meaning of (4.78): (b) Right meaning of (4.78):
‘The children bought them and
they are good.’
IP
❍ ✟ ✟✟✟✟✟
❍
❍
❍ ❍
CP j
IP
✏ ✏✏✏✏✏
✟ ✟✟ ❍ ❍
DP VP
The children bought pro
✏✏
✏ ✏
pro t j are good
‘The ones the children bought
are good.’
IP
❍
✟ ✟✟✟✟
❍
❍
❍
❍
DP
VP
✟ ✟ ✟ ❍
❍ ✏ ✏
❍ are good
D CP
✟ ✟✟✟
❍ ❍
❍ ❍
pro i CP
✏
✏ ✏✏✏✏
the children t i bought

140
For example, in a context which contains equally prominently candy bought by
both children and adults, sentence (4.78) is felicitous, but the use of the relative
clause in (4.78) as an independent sentence is not.
(4.78) [[wawa-kuna
child-pl
[e] ranti-shka]]-kuna-ka
buy-nm-pl-top
sumak-mi.
good-evid
‘Those (candies) that the children bought are good.’
(IQ)
4.5.4 A proposal for movement of arguments only
In the preceding sections I pointed to evidence for head-raising from islands, quantifiers
and argument vs. adjunct behavior. With regard to the island effects, I showed
that in fact raising was not necessary to explain the impossibility of “extracting”
a head from within an embedded DP given the impossibility of extracting a head
from within an embedded clause coupled with the proposal that external heads can
be licensed by a pragmatic relationship with the clause which is their direct sister.
This leaves the quantifier evidence and the argument/adjunct asymmetry. I find
that the evidence points to the possibility of head-raising of arguments but not of
adjuncts. Therefore, I propose that internal argument heads can be raised at LF to
the Spec of the relative clause CP, whereas adjunct heads cannot. Similarly, external
heads corresponding to internal arguments may have raised from the internal position,
whereas external heads corresponding to internal adjuncts are base-generated
externally. In the next section I discuss the difficult question of whether in IQ
head-raising should be seen as Case-based, and tentatively conclude that it should
not. In the remainder of this section I highlight a consequence of this proposal for
apparently similar pairs of internally and externally headed argument and adjunct
relative clauses and some supporting evidence from Cuzco Quechua.

141
4.5.4.1 The relationship between external and internal heads
In this section I discuss an implication of the analysis I have outlined in the previous
section for the externally/internally headed pairs of relative clauses presented in
Section 4.5.3. Here I have repeated the initial relative clause data which we looked
at in examples (4.53) to (4.56).
• Argument heads (repeated from (4.53) and (4.54):
(4.79) [[Juan
Juan
t i
shuwa-shka]
steal-nm
‘the cow that Juan stole’
wagra]
cow
(IQ)
(4.80) [[Juan
Juan
wagra
cow
shuwa-shka]]
steal-nm
‘the cow that Juan stole’
(IQ)
• Adjunct heads (repeated from (4.55) and (4.56)):
(4.81) [[Juan
Juan
killka-shka]
write-nm
lapiz]
pen
‘the pen that Juan wrote with’
(4.82) [[Juan
Juan
lapiz-*(wan)
pen-instr
killka-shka]
write-nm
pro]
‘the pen that Juan wrote with’
(also: ‘the thing Juan wrote with the pen’)
(also, perhaps: ‘Juan who wrote with the pen’)
(IQ)
(IQ)
Although the pairs of argument and adjunct relative clauses look very similar,
under my analysis only the argument relativization pair is related by a raising analysis.
The adjunct relativizations in (4.81) and (4.82) may represent a gapless relative
and a relative in which the head is an external null pronominal. To elaborate briefly
on these last two ideas, I suggest that the structure of (4.81) is like that proposed
for Japanese gapless relatives in [Murasugi 2000], as in (4.83). This same idea was
discussed in the context of Cuzco Quechua in Section 3.4.3.

142
(4.83) DP
❍
✟ ✟✟✟ ❍
❍
❍
IP
✏ ✏✏ ✏ DP
✟ ✟ ❍ ❍
Juan killka-shka D NP
Juan wrote
lapiz/*pro
pen/*pro
Furthermore, note that if indeed the external adjunct head in (4.81) has not
raised to this position but in fact was base-generated externally, then a pro must
not be licensed in this same position, otherwise null adjunct heads would be possible.
This seems to be due to the unrecoverability of the relationship between a null pro
and a nominalized phrase like ‘Juan wrote’, as opposed to the relationship between
the overt ‘pen’ and the phrase ‘Juan wrote’, but I leave it to future work to clarify
this intuition.
Turning now to (4.82), my proposed structure here is (4.84). The interpretation
of this stucture would proceed as proposed in Shimoyama’s [2001] analysis of
Japanese IHRs. That is, pro is an E-type pronoun whose identity is determined by
the context established by the relative clause.

143
(4.84) DP
✟ ✟✟ ✟ ❍❍
❍
❍
D CP
✟ ❍❍
✟ ✟✟✟
pro
❍
❍
❍
CP
✏ ✏✏✏✏✏✏
Juan lapiz-wan killka-shka
Juan pen-instr wrote
At LF, the relative clause scopes out of DP:
DP
✟ ❍
✟ ✟✟✟
CP j
✏ ✏✏✏✏✏✏
Juan lapiz-wan killka-shka
Juan pen-instr wrote
❍
❍
❍
❍
DP
✟ ✟❍
❍
D
CP
✟❍
pro t j
4.5.4.2 Support from Cuzco Quechua
Strong support for this analysis can be found in Cuzco Quechua. Recall that in CQ,
only arguments can be internal heads. Adjuncts cannot be internal heads. This is
illustrated in the next four examples.
(4.85) [[Juan-pa
Juan-gen
ranti-sqa-n]
buy-nm-3sg
‘The cow that Juan bought’
(4.86) [[Juan-pa
Juan-gen
waka
cow
waka]
cow
ranti-sqa-n]]
buy-nm-3sg
‘The cow that Juan bought’
(CQ)
(CQ)
(4.87) [[qilqa-sqa-y]
write-nm-1sg
lapicero]
pen
‘the pen I wrote with’
(4.88) *[[lapicero-(wan)
pen-instr
‘the pen I wrote with’
qilqa-sqa-y]]
write-nm-1sg
(CQ)
(CQ)

144
(4.85) and (4.86) are examples of argument relativization in which the head
‘cow’ appears both externally and internally to the clause. The meaning in both
cases is ‘the cow that Juan bought’. However when we try the same thing with an
adjunct, the internal head turns out to be impossible. Therefore in (4.87), pen is
an acceptable external head in the expression ‘the pen I wrote with’, but cannot
appear internal to the clause either with or without its Case marker.
Under the analysis I have been developing here, these facts can be explained if
in Cuzco Quechua an internal head must always raise. This view was espoused in
Chapter 3. Assuming as in Section 3.3.3 that in both dialects the adjunct head must
be Case-marked within its clause, it follows that no internal adjunct heads will be
possible in Cuzco Quechua.
4.5.5 Does the head raise for Case reasons?
Given my proposal that argument heads may raise in IQ, it is natural to ask whether
this raising is Case-based as I have proposed it is in CQ. However, in IQ the evidence
on this question is not so clear as in CQ and in this section I tentatively conclude
that in fact raising is not Case-based in IQ.
As discussed in previous sections, argument heads in IQ can in some cases have
their Case-marker omitted but unlike in CQ this omission is optional and is limited
to heads in the immediately pre-verbal position. Examples are shown in (4.89) and
(4.90).
(4.89) Direct object head:
[[Wawa
child
rumi-(?ta)
stone-(acc)
shita-shka]]
throw-nm
‘the stone that the child threw’
(IQ)

145
(4.90) Locative argument head:
[[kan
you
llakta-(??pi)
town-loc
‘the town you live in’
kausa-shka]]
live-nm
(IQ)
In [Cole 1985] the optionality of Case-marking such as (4.89) is explained by
suggesting that the object can incorporate into the verb. One piece of evidence
for this is that the optionality is seen just in case the direct object is found in the
immediate pre-verbal position. Furthermore, the Case-marker drop can occur in
complement clauses as well as relative clauses. This is illustrated in (4.91) to (4.92).
(4.91) [wagra-(ta)
cow-acc
randi-shka]
buy-nm
warmi
woman
‘the woman who bought the cow’ (Cole 1985 (159) p.49)
(IQ)
(4.92) Juzi-ka
Jose-top
[ñuka
I
kaya
yesterday
llama-(ta)
sheep-acc
randi-na]-ta
buy-nm-acc
‘Jose believes I will buy a sheep tomorrow.’ (Cole 1985 (118))
kri-n.
believe-3sg
(IQ)
Juzi-ka
Jose-top
[ñuka
I
llama-*(ta)
sheep-acc
kaya
yesterday
randi-na]-ta
buy-nm-acc
kri-n.
believe-3sg
‘Jose believes I will buy a sheep tomorrow.’ (Cole 1985 (119))
(IQ)
Cole notes, and I have found in my research also that consultants more often
accept or suggest null-marked objects in relative clauses like (4.91) than in complement
clauses like (4.92). Interestingly, this asymmetry is repeated in a much sharper
way in the case of a locative object. Here, the locative marker may not be omitted
in a complement clause (regardless of the position of the object) or on a non-head of
a relative clause, but must almost obligatorily be omitted in the case of the head of
a relative clause. These facts are illustrated in (4.93) and (4.94). However, examples
of this phenomenon are hard to find in IQ.

146
(4.93) [ñuka
I
wasi-*(pi)
house-loc
yanu-shka]
cook-nm
‘the food I cooked in the house’
mikuna
food
(IQ)
(4.94) [ñuka
I
wasi-(??pi)
house-loc
‘the house I lived in’
tiya-shka]
live-nm
(IQ)
Thus, although in IQ the nominalization of the verb may be responsible for the
optionality of an accusative marker in this case, such an explanation is not possible
for the absence of the locative marker on the head ‘town’ in (4.90) and ‘house’ in
(4.94). The absence of this object marker is only possible on the head of a relative
clause.
By contrast, in the case of adjunct relativization as in (4.95) and (4.96), internal
heads are acceptable but their Case markers are mandatory. In (4.95), the meaning
is ‘the pen that Juan wrote with’, and the instrumental marker on ‘pen’ is obligatory.
Similarly in (4.96), ‘the car Ana arrives in’, the locative marker on ‘car’ cannot be
omitted.
(4.95) [[Juan
Juan
lapiz-*(wan)
pen-instr
killka-shka]]-ka
write-nm-top
‘The pen that Juan wrote with is blue.’
azul-mi.
blue-evid
(IQ)
(4.96) [[Ana
Ana
karru-*(pi)
car-loc
‘the car Ana arrives in’
chayamu-k]]
arrive-nm
(IQ)
Although it is tempting, given this set of facts, to relate head-raising directly
to lack of Case-marking in IQ, the optionality of the Case-marker even in those
cases where, for example, a strong quantifier appears to be interpreted externally
to the clause, lead me to question this correlation. Also, the fact that the Casemarker
can only be dropped when the head is in the pre-verbal position again is not

147
immediately reconcilable with an analysis in which heads raise for Case purposes.
Therefore, I propose that an internal argument of the verb may Merge directly with
the (nominalized) verb. Under these circumstances it may be Case-checked directly
by the verb instead of by Tr. Subsequent overt scrambling is clearly not possible
from this position, while covert head-raising is possible. I leave the question of why
locative arguments must undergo head-raising from this position open for future
research.
4.6 Conclusion
This chapter provides a comparison between CQ and IQ relative clauses and an attempt
to explain a set of differences between the two dialects based on a structural
distinction that arises from a special Caseless determiner which selects CP only in
CQ relative clauses. Many questions remain open, including that of the historical
process by which such a distinction could have arisen between closely related languages.
Another question which relates to a discussion of CQ in Chapter 3 has
to do with the exact interpretive mechanism associated with the E-type anaphora
approach to IQ as opposed to CQ. It was noted in Chapter 3 that although E-type
anaphora was an important element in the interpretation of CQ relative clauses,
there was also a structural component whereby the identity of the head was established
via syntactic raising. The results of this chapter show that IQ lends itself
to a simpler E-type anaphora analysis, along the lines of the original proposal for
Japanese in [Hoshi 1995] and [Shimoyama 1999, 2001].
To summarize this chapter, I present below schematic versions of the logical
possibilities in CQ and IQ for the position of the head (internal or external) and the
nature of the gap within the RC (in the case of external heads: the gap could be a
trace or a coindexed element).

148
(4.97) Empty external head position at LF Head raised at LF External head
CQ: * DP
✟ ✟ ❍❍ ❍
CP e i
✏ ✏✏ ✏
...head i ...
IQ: DP
✟ ✟ ❍❍ ❍
CP e i
✏ ✏✏ ✏
...head i ...
DP
✟ ✟ ❍❍ ❍
CP head i
✏ ✏
...t i ...
DP
✟ ✟ ❍❍ ❍
CP head i
✏ ✏
...t i ...
DP
✟ ✟ ❍❍ ❍
CP head i
✏ ✏
...e i ...
DP
✟ ✟ ❍❍ ❍
CP head i
✏ ✏
...e i ...
I have attempted to explain the impossibility of the upper left-hand tree in the
preceding chapter by positing mandatory syntactic raising in CQ. In the current
chapter I have explained the three options found in IQ by proposing that raising is
possible (though not mandatory) for arguments but not for adjuncts.
Since many of the differences between the dialects, and between different constructions
within the dialects, have been attributed to features of functional heads,
I close here with a tabular summary of the feature sets which have been proposed
in each case. I include in this summary the cases of main clauses, subordinate complement
clauses, and relative clauses in each dialect. The arrows represent selection
relationships. The initials RC, CC and MC stand for relative clause, complement
clause and main clause respecively.
The main difference between CQ and IQ can be seen in the nature of the Determiner
which selects the CP of the relative clause: in CQ this Determiner has no
Case feature of its own, so requires determiner incorporation and hence head-raising.
In IQ, this Determiner has its own Case feature and no determiner incorporation or
head-raising is required. A second difference between IQ and CQ lies in the expression
of genitive Case. In CQ the genitive Case marking is associated with merge
to n+N or n+V (and nominal agreement). In IQ, genitive Case is associated with
Case-checking by D and not T.

149
(4.98) Cuzco Quechua:
D C T n/v
RC [-Case] → [-Case] → [-Case] → n → Lk → V
(subj) [+wh/head] none
RC [-Case] → [-Case] → [+Case] → n → Lk → V
(obj) [+wh/head] gen.
CC [+Case] → [+Case] → n → Lk → V
gen.
MC [-Case] → [+Case] → v → Tr → V
nom.
DP [+Case] → n → N
gen.
(4.99) Imbabura Quechua:
D C T n/v
RC [+Case] → [-Case] → [+Case] → n → Tr → V
w/HR [+wh/head] nom.
RC [+Case] → [-Case] → [+Case] → n → Tr → V
wo/HR
nom.
CC [+Case] → [+Case] → n → Tr → V
nom.
MC [-Case] → [+Case] → v → Tr → V
nom.
DP [+Case] → n → N
[gen.]

Chapter 5
Existential and Possessive
Sentences in Quechua
5.1 Introduction
This chapter investigates existential and possessive constructions in Cuzco Quechua. 1
It uses some of the results developed so far with respect to relative clause structure
and Case-marking of nominals to propose analyses of these construction types, both
of which employ the verb kay ‘to be’. I explain some of their unexpected syntactic
and morphological properties using relative clauses as one diagnostic for subjecthood
and argumenthood in general. I also begin an analysis of extraction from DP in CQ
which is continued in a wider context in Chapter 6. The two types of sentences
under study are illustrated in (5.1) and (5.2). I will refer to these constructions as
existential and possessive sentences respectively. Part of my claim will be that pos-
1 Both this chapter and the next deal once again only with Cuzco Quechua. The
reasons for this are rather different in each case. Chapter 6 deals with discontinuous
noun phrases, which are broadly attested across the Quechua languages and
appear to behave similarly in Cuzco and Imbabura Quechua. On the other hand
the existential and possessive constructions discussed here are very different in IQ.
See Footnote 5 for examples.
150

151
sessive sentences such as (5.2) can be viewed as a special case of existential sentences
such as (5.1).
(5.1) Llama-kuna urqu-pi ka-n.
llama-pl mountain-loc be-3sg
‘There are llamas in the mountain.’
(5.2) Maria-q llama-n-kuna
Maria-gen llama-3sg-pl
‘Maria has llamas.’
ka-n.
be-3sg
A question that arises in (5.1) (and indeed in the existential constructions of
many languages) is what element, if any, occupies the subject position of this sentence.
I propose that a null expletive appears in this position.
A potential problem with an analysis in which the two sentences above have
parallel structures is the fact that in many languages the presence of a pre-nominal
possessor is an indication of specificity of the noun phrase. For instance in English,
‘Maria’s llamas’ is specific whereas ‘llamas of Maria’s’ has a non-specific reading.
If this is the case in Quechua too then (5.2) presents the problematic structure of
a specific DP in a position normally subject to the definiteness effect of existential
constructions. In this chapter I argue for a resolution of this problem involving
possessor extraction as a covert movement operation.
The organization of this chapter is as follows. In the next section I present the
basic data of CQ existential constructions. Section 5.3 proposes a structural analysis
of CQ existential sentences. Section 5.4 does the same for CQ possessive sentences.
Section 5.5 is the conclusion.

5.2 The basic facts of Cuzco Quechua existentials
152
According to the grammar Gramática Quechua by Antonio Cusihuamán [1976, 2001],
a native speaker of Cuzco Quechua 2 , existential constructions in CQ are formed using
the verb kay ‘to be’. This verb also appears in copular constructions. 3
The basic CQ existential sentence from (5.1) is repeated in (5.3).
(5.3) Llama-kuna urqu-pi ka-n.
llama-pl mountain-loc be-3sg
‘There are llamas in the mountains.’
Although kay has both existential and copular uses, these uses can often be kept
apart through the observation that existential kay always appears in the third person
present singular form as in (5.3). This third person singular form is reported to
be obligatorily omitted (when uninflected other than in person marking) in copular
constructions while optionally present in existential constructions (Cusihuamán 2000
p90). These facts are illustrated in (5.4) and (5.5).
(5.4) Nuqa hatun
I big
‘I am tall.’
*(ka-ni).
be-1sg
(5.5) Llama-qa hatun-mi
llama-top big-evid
‘The llama is big.’
(*kan).
be-3sg
2 Cusihuamán was in fact from the village of Chinchero outside of Cuzco, where
most of my consultants are also from.
3 A second verb with existential uses mentioned by Cusihuamán is tiyay ‘to sit’,
which he reports to be almost out of use and limited to situations involving the
existence of a market (Cusihuamán [1976] 2000, p.15). In Ecuadorian Quechua, in
fact, tiyay is the standard existential verb, and this is also the case in dialects much
more closely related to CQ, such as Bolivian Quechua, where it is employed in a
manner which seems to parallel closely the uses of kay in CQ. In the current chapter
I limit my discussion of CQ existentials to those involving kay.

153
The above examples contain individual-level predicates. Since stage-level predicates
are usually accompanied by the progressive form of the verb kay, which bears
the progressive suffix -sha, the copular here is overt even with a present tense third
person singular subject as shown in (5.6). Furthermore, in tenses other than present,
the copular use of kay is overt even in third person singular as seen in (5.7).
(5.6) Maria urqu-pi ka-sha-n.
Maria mountain-loc be-prog-3sg
‘Maria is in the mountains.’
(5.7) Maria
Maria
yachachiq
teacher
ka-rqa-n.
be-past-3sg
‘Maria was a teacher.’
Note that the above generalizations predict that (5.6) will be bad with overt
kan since Maria is a definite description incompatible with existential contexts, and
indeed this is the case as seen in (5.8). 4
(5.8) *?Maria
Maria
urqu-pi
mountain-loc
ka-n.
be-3sg
However, the distribution of kan is not entirely governed by the above generalizations
and the definiteness effect in existential constructions. The details will be
discussed further in Section 5.4, but one very frequent use of existential ka-n is in
‘have’ constructions such as that in (5.9). This use of the existential is not found in
all Quechua dialects, and in particular is not allowed in Imbabura Quechua. 5
4 Contra Cusihuamán’s generalization, some Cuzco Quechua consultants will accept
an overt copula kan as an apparent emphatic. I take this to be a nonstandard
spellout of the usually null kan, and in my analysis nonetheless distinguish between
the copular null kan and the overt existential kan. This corresponds to most consultants’
judgments of sentences like (5.8).
5 Ecuadorian Quechua existential sentences contain the verb tiyana ‘to sit’ instead
of kay ‘to be’. An example from a folktale is given in (i).
(i) shuc manzana tarpushca tiya-shca
one apple orchard sit-past

154
(5.9) Maria-q llama-n-kuna
Maria-gen llama-3sg-pl
‘Maria has llamas.’
ka-n.
be-3sg
Example (5.9) is a very natural way to express Maria’s ownership of llamas. For
completeness, I include in (5.10) an also frequent paraphrase.
(5.10) Maria-qa
Maria-top
llama-yuq.
llama-possessor
‘Maria has llamas.’ (’Maria is a llama-owner.’)
In (5.9) as in the other existential examples above, plural marking on the verb is
not possible without changing the meaning of the sentence to the declarative ‘They
are Maria’s llamas’, even though Mariaq llamankuna ‘Maria’s llamas’ is a plural
noun phrase. This is illustrated in (5.11).
(5.11) Maria-q llama-n-kuna
Maria-gen llama-3sg-pl
‘They are Maria’s llamas.’
*‘Maria has llamas.’
ka-n-ku.
be-3-pl
‘There was an apple orchard.’ (Jara & Moya p189)
As in CQ, the existential verb does not seem to show number agreement in its
existential use. Unlike CQ kay, however, it is not phonologically null in its (third
person present tense singular) non-existential use.
Possessive constructions are quite different in CQ and IQ. IQ has a lexical verb
charina meaning ‘to have’. The possessed nominal appears in accusative Case and
the possessor appears in nominal Case, making this verb also the apparent syntactic
equivalent of English ‘to have’. Two examples are given below.
(i) Ñuka-ka ishkay wawa-ta-mi chari-ni.
I-top two child-acc-evid have-1sg
‘I have two children’
(ii) ñucanchic jatun taitacunami chai rimai-ta-ca chari-shca:
we big fathers-evid dem speech-acc-top have-past
‘Our ancestors had this speech: ...’ (Jara & Moya p129)
Constructions like “Juan-gen book sit-3sg” as a way to express “Juan has a
book,” which would parallel (5.9) from Cuzco Quechua are not a regular part of
Imbabura Quechua grammar. Consultants judge them to be understandable but
strange.

155
Finally, although the form kan is invariant with respect to plural marking, tense
inflection is possible, and occasionally the progressive suffix may appear with kay in
its existential use, as illustrated in (5.12) and (5.13) respectively.
(5.12) Urqu-pi wik’uña ka-rqa-n.
mountain-loc vicuña be-pst-3sg
‘There were vicuñas in the mountains.’
(5.13) kancha-y-kuna
corral-1sg-pl
ka-sha-n-mi
be-prog-3sg-evid
‘I have corrals.’ (Itier 1999 p.192)
(5.14) Qhatu-pi
market-loc
ashka
many
llama
llama
uwija
sheep
ka-sha-n.
be-prog-3sg
‘There are many llamas and sheep (currently) in the market.’ 6
5.3 Structure of CQ existential sentences
Given the basic existential paradigm described in the previous section, one question
we may ask is what structural position the nominal which is being declared to exist
holds in CQ existential sentences. (Although I do not presuppose the presence of an
expletive, for convenience I will refer to this nominal as the associate). For instance,
is kay more like to exist in English, with the relevant nominal in a regular subject
6 Examples like (5.13) are perhaps rather surprising given the impossibility of a
progressive existential in English such as “There are being llamas in the market.”
However, it is also impossible or strange in English to say “Llamas are being in the
market.” whereas in Quechua it is perfectly fine to use the progressive in a stative
predicate of this sort as in (i).
(i) Llama-kuna qhatu-pi ka-sha-n-ku.
llama-pl market-loc be-prog-3-pl
‘The llamas are (currently) in the market.’
From this I conclude that the progressive suffix -sha in Quechua is not directly
parallel to the progressive participle form in English, and this is quite a separate
issue from that of the existential construction itself.

156
position, or is kay more like English ‘(there) is’, with perhaps a null expletive in
subject position. In this section I will argue in favor of the null expletive analysis.
I begin by looking at number agreement in CQ existentials, mentioned briefly
in the previous section. Recall that the verb kay appears in third singular form
in existentials such as (5.3), although the associate (such as llama-kuna ‘llamas’ in
that example) is a plural. In fact, plural marking in Quechua is often optional and
apparently governed by rather complex pragmatic considerations . However, if the
verb in (5.3) were in unmarked agreement with the verb kay, we would expect the
verb to be at least optionally plural marked. In fact, the addition of the plural
marker on kay changes the meaning of the sentence. 7
7 I have seen one example which challenges this generalization. As mentioned
above, I will analyze possessive sentences as a special case of existentials in CQ. Anticipating
this analysis, consider now the following example from a Cuzco Quechua
textbook [Morató-Peña & Morató-Lara 1995], the following sentence appears twice,
once with and once without plural marking on the verb.
(i) ¿Wawqe-pana-yki-kuna ka-sqa-∅-(ku)-chu?
brother-sister-2pl-pl be-past.rep-3(sg)-(pl)-Q
‘Did you have siblings?’
I have checked this sentence with speakers of Cuzco Quechua who have verified
that it is grammatical (although perhaps better without plural agreement). However,
this fact appears to be due to a special property of the “reportative” (Cusihuamán
[1976/2000]) suffix -sqa, since the same speakers do not accept any plural
marking on the present tense existential. This can be seen in the following further
examples, which show as we expect that plural agreement on the verb is not possible
in an existential sentence in which the verb does not appear with the reportative
suffix -sqa.
(ii) ¿Wawqi-yki-kuna ka-n-ku?
brother-2pl-pl be-3-pl
‘Are they your brothers?’
*’Do you have brothers?’
(iii) ¿Wawqi-yki-kuna ka-n-chu?
brother-2sg-pl be-3sg-q
‘Do you have brothers?’
Note that in (ii), which shows plural agreement on the verb, the possessive sentence
reading is not available, as it was above in (i) and also in (iii). I currently do
not know why -sqa affects the plural marking on the verb in this way, but leave this

157
Consultants report that (5.15) indicates it is a property of llamas that they are
in the mountains, or that certain llamas are living in the mountains.
(5.15) Llama-kuna urqu-pi ka-nku.
llama-pl mountain-loc be-3sg.
‘The llamas are in the mountains.’
Omitting the overt locative allows us to perceive more sharply the difference
between (5.3) and (5.15). The relevant examples are in (5.16) and (5.17).
(5.16) Llama-kuna
llama-pl
(5.17) ∅
pro
ka-n.
be-3sg
‘There are llamas.’ (Spanish: Hay llamas.)
Llama-kuna
llama-pl
ka-nku
be-3pl
‘They are llamas.’ (Spanish: Son llamas.)
Sentence (5.16) shows an existential construction in which some sort of locative is
implicit. Sentence (5.17) can only be understood with llamakuna as a predicate, and
with the subject null. Since Quechua is a pro-drop language the result is perfectly
acceptable.
This lack of number agreement between the associate and the verb suggests that
the associate is not a true subject in these existential constructions, and the verb
may in fact be agreeing with some other element in the sentence. One possibility
(advocated cross-linguistically by [Freeze 1992]) is that the locative is in subject
position and some sort of default agreement applies. Another possibility is that a
null expletive in subject position triggers third person singular agreement on the
verb. In the next sections I will argue that the latter view is correct. I first use two
tests (involving adverbial and relative clause constructions) for subject to show that
neither the associate nor the locative behaves like a syntactic subject.
matter to future research.

158
5.3.1 Quechua adverbial evidence that associate is not a
Subject
One method by which Cuzco Quechua forms subordinate clauses is via the adverbializing
suffixes -qti and -spa. These suffixes appear on the verb as illustrated in
(5.18) and (5.19). Agreement with the subordinate clause is mandatory with -qti
and optional with -spa.
(5.18) [Juan chayamu-qti-n/*spa],
Juan arrive-bi.adv-3sg/*uni.adv
‘If Juan arrives I will be happy.’
nuqa
I
kusisqa
happy
ka-saq.
be-1sgfut
(5.19) [Juan hamu-*qtin/spa-(n)],
Juan come-*bi.adv-3sg/uni.adv-3sg
‘If Juan comes, he will be happy.’
(pay)
(he)
kusisqa
happy
ka-nqa.
be-3sgfut
The adverbializing suffixes differ in that -qti is used when matrix and subordinate
clauses have distinct subjects, as in (5.18), whereas -spa indicates that both clauses
have the same subject, as in (5.19). This generalization holds for all transitive and
intransitive verbs (including those typically classified as unaccusative). Therefore,
the choice of adverbializer provides one diagnostic for the identification of subjects.
Now, if the associate were in subject position we would expect sentences like
(5.20), in which the associate in an existential adverbial clause and the subject of the
main clause are coindexed, necessarily to trigger the -spa unipersonal adverbializer
on the subordinate verb. In fact we see that the bipersonal -qti is also possible.
(5.20) [Warmi-kuna ayllu-pi ka-qti-n/ka-spa]-qa,
woman-pl family-loc be-bi.adv-3sg/be-uni.adv-top
pay-kuna-puni
s/he-pl-always
wayk’u-nku.
cook-3pl
‘[When there are women in a family]/[When women are in a family], they
always cook.’

159
The two meanings are distinct but are nonetheless quite close. How can we
be sure that the first clause does indeed have an existential interpretation? Note
that in the case of a copular adverbial clause (‘when women are in a family’) we
would never expect -qti to be a valid option since the main clause subject paykuna
’they’ also refers to these women. Thus, the fact that -qti is also acceptable in the
adverbial clause must be attributed to a reading of that clause in which ’women’
is not the subject. This is the existential reading of that clause (‘when there are
women in a family’). In fact, this reading has a paraphrase which is shown in (5.21)
and suggested as an alternative by some consultants. Here, a different adverbializing
strategy is invoked: chay ‘that’ is employed to subordinate a finite version of the
existential clause.
(5.21) [[Warmi-kuna ayllu-pi ka-n] chay]-qa, pay-kuna-puni
woman-pl family-loc be-3sg that-top s/he-pl-always
‘When there are women in a family, they always cook.’
wayk’u-nku
cook-3pl
A further example of the same phenomenon is shown in (5.22) to (5.23). Here
too, both -qti and -spa are accepted adverbializers, and the alternative chay strategy
is also possible.
(5.22) [Wasi-kuna-pi
house-pl-loc
ashka
many
alqu-kuna
dog-pl
ka-spa]-qa
be-uni.adv-top
ruwa-nku.
make-3pl
‘When many dogs are in a house they make a lot of noise.’
nishu-ta
much-acc
bullu-ta
noise-acc
(5.23) [Wasi-kuna-pi
house-pl-loc
ashka
many
alqu-kuna
dog-pl
ka-qti-n/ka-n
be-uni.adv/be-3sg
chay]-qa,
that-top
nishu-ta bullu-ta ruwa-nku.
much-acc noise-acc make-3pl
‘When there are many dogs in a house, they make a lot of noise.’
Now if the associate is not in subject position, let us consider what might be
there. One possibility is that the locative is found in subject position and the verb

160
is agreeing in some default way with that element. Such an analysis for existentials
cross-linguistically is proposed by Freeze [1992]. However, examples like (5.24) and
(5.25) do not seem entirely consistent with this proposal. These sentences have locatives
which are identical in both clauses, but nonetheless the bipersonal adverbializer
is employed.
(5.24) [Qheswa allpa-pi kapuli
Temperate land-loc kapuli.berries
ka-qti-n/ka-n chay/*ka-spa]-qa,
be-bi.adv-3sg/be-3sg that/*be-uni.adv-top
ka-lla-n-taq.
be-CONJ-3sg-CONJ
chiwchiku-pas
chiwchiku.bird-also
‘When there are kapuli berries in the valley, there are chiwchiku (birds) also.’
(5.25) Urqu-pi
Mountain-loc
ichhu
grass
ka-qti-n-qa/*ka-spa-qa
be-bi.adv-3sg-top/be-uni.adv-top
ka-n-mi.
be-3sg-evid
‘When there is grass in the mountains, there are also animals.’
uywa-pas
animal-also
Another possibility is that there is a null expletive in the subject position. Although
I will conclude that this is in fact the case, some initial evidence that might
be used to argue against this idea can be found by testing whether two existential
clauses trigger unipersonal or bipersonal adverbializer. If a null expletive is the
subject of both clauses we might expect the unipersonal suffix under these circumstances.
That this is not the case is shown in (5.26) and (5.27).
(5.26) [Ashka turista-kuna ka-qti-n/*ka-spa-n]-qa, ashka-llataq
Many tourist-pl be-bi.adv-3sg/be-uni.adv-3sg-top many-also
vindiq-kuna
vendor-pl
ka-n.
be-3sg
‘When there are many tourists, there are also many vendors.’

161
(5.27) Mana
not
turista
tourist
ka-qti-n/*spa
be-bi.adv-3sg/uni.adv
mana
not
ka-n-chu.
be-3sg-neg
‘When there are no tourists there are no buyers.’
ranti-ku-q
buy-refl-agent
However, (5.26) and (5.27) do not immediately disprove the expletive theory.
In fact, this failure of the null expletive to be a “unipersonal” subject in the relevant
sense extends to other expletive constructions too. (5.28) shows that weather
expressions in both clauses also trigger the bipersonal adverbializer -qti, which is
consistent with our conclusions regarding (5.26) and (5.27).
(5.28) Para-qti-n/*?Para-spa
rain-bi.adv-3sg/*rain-uni.adv
‘When it rains it is very cold.’
nishu-ta
much-acc
chiri-n.
cold-3sg
Furthermore, a weather expression cannot be coordinated with an existential by
means of the unipersonal adverbializer -spa, as seen in (5.29).
(5.29) [Para-qti-n/*Para-spa]-qa, ashka
rain-bi.adv-3sg/*rain-uni.adv-top, many
‘When it rains, there are many flowers.’
t’ika-kuna
flower-pl
ka-n.
be-3sg
In fact, I believe the -spa suffix is ruled out in these cases because the “same
subject” criterion is one on coreference and not on some sort of formal similarity.
This claim is supported by (5.30). In this example, the subject of the two clauses
is papa ‘potato’, but clearly the subjects are nonetheless not coreferential and the
bipersonal adverbializers is employed as expected.
(5.30) papa
potato
chakra-kuna-pi ka-qti-n-qa, (papa)
field-pl-loc be-bi.adv-3sg-top potato
ka-lla-n-taq.
be-CONJ-3sg-CONJ
‘When there is potato in the fields there is potato in the stores.’
tienda-kuna-pi-pas
store-pl-loc-also
Thus the adverbial suffix evidence suggests that the associate is not in subject
position, and is compatible with the presence of a null expletive.

162
5.3.2 Relative clause morphology indicates associate is not
a Subject
I now present evidence from relative clauses which again shows that the associate
fails to exhibit another property of a prototypical subject.
Recall that in Cuzco Quechua, relative clauses are formed by adding a nominalizing
suffix to the verb of the subordinate clause. When the subject of the subordinate
clause is the head of the relative, the suffix -q appears on the verb (following any
aspectual/directional/causative/etc. suffixes) and the verb otherwise does not agree
with the subject and is neutral for tense. On the other hand when a non-subject argument
of the subordinate clause is the head of the relative clause, the nominalizing
suffixes -sqa or -na appears on the verb, followed by subordinate subject agreement
morphology (in the nominal agreement paradigm). These two suffixes reflect relative
tense information, indicating that subordinate tense precedes or follows matrix
tense, respectively. These basic facts are illustrated in (5.31)–(5.34). Note that the
subject of the typically unaccusative verb ‘arrive’ in (5.32) behaves like other subjects.
Although we have seen that Quechua allows both internally and externally
headed relative clauses, for the purposes of this section I will just discuss the latter,
since the verbal morphology in question is not affected by the position of the head.
(5.31) Nuqa
I
[[e
e
phawa-q]
run-nm(sbj)
wik’uña]-ta
vicuña-acc
muna-ni.
like-1sg
‘I like the vicuña that is running/ran/will run.’
(5.32) Nuqa [[e chaya-mu-q] warmi]-ta
I e arrive-cis-nm(sbj) woman-acc
‘I know the woman who is arriving.’
riqsi-ni.
know-1sg
(5.33) Nuqa [[Juan-pa e riku-sqa-n]
I Juan-gen e see-nm(non-subj/pst)-3sg
‘I like the vicuña that Juan saw.’
wik’uña]-ta
vicuña-acc
muna-ni.
like-1sg

163
(5.34) Nuqa [[Juan-pa e chaya-sqa-n]
I Juan-gen e arrive-nm(non-subj)-3sg
‘I saw the bus Juan arrived in.’
bus]-ta
bus-acc
riku-ra-ni.
see-past-1sg
Recall further that in examples like (5.33) and (5.34) the suffix -sqa indicates
that the subordinate action took place in the past since the matrix verb itself is in
present tense. The suffix does not change despite the differing theta roles that the
head (wik’uña and runa respectively) plays in the two subordinate clauses. Similarly
the use of -q in (5.31) and (5.32) is due to the fact that the head wik’uña is the
(surface) subject of the subordinate clause.
With these facts in hand we may now turn to existential sentences in which the
associate is also the head of a relative clause. One difficulty with studying such
constructions is their semantic similarity to constructions containing a subordinate
copular verb. English examples of a semantically similar pair of this sort are given
in (5.35) and (5.36).
(5.35) The [llamas [that there are e in my country]] live only on farms.
(5.36) The [llamas [that e are in my country]] live only on farms.
In CQ, given the resemblance between copular and existential constructions, they
are harder to distinguish. However, if the associate behaved like a true non-subject,
then there would be a clear difference between the two constructions: sentences such
as (5.35) would involve a non-subject head and hence trigger the -sqa/-na nominalizers
on the verb, while sentences such as (5.36) would involve the relativization of
a subject and would trigger the relativizer -q.
In fact, we find that only the second possibility is attested in CQ, as illustrated
in (5.37).

164
(5.37) Llaqta-y-pi ka-q/*ka-sqa-n
llama-kuna-qa
town-1sg-loc be-nm(sbj)/*be-nm(non-subj)-3sg llama-pl-top
hacienda-lla-pi tiya-nku.
farm-delim-loc live-3pl
‘The llamas that are in my country live only on farms.’
The starred possibility in example (5.37) shows that the morphology which would
reflect the relativization of a non-subject is impossible when the head of a relative
clause is the associate in the subordinate existential. There remains, however, the
possibility that (5.37) with ‘kaq’ can mean ‘The llamas that there are in my country
live only on farms’. That is, is (5.37) ambiguous between the readings in (5.35)
and (5.36). I discussed with my consultants a context in which a hypothetical
interlocutor had asked me if there are vicuñas in my country. My answer was to
be: ‘yes, but the vicuñas that there are (in my country) live only on farms’. I was
unable to find a construction that consultants felt to have this meaning. 8 Rather,
the copula construction in the subordinate clause was suggested. A few options are
shown below.
(5.38) Llaqta-y-pi ka-q (wik’uña)-kuna
town-1sg-loc be-nm vicuña-pl
‘The vicuñas that are in my country’
(5.39) ??Ka-q
be-nm
wik’uña-kuna
vicuña-pl
‘The same vicuña’ (under a different reading of kaq)
8 One consultant did feel that (i) may have the desired meaning (‘those that there
are’, as opposed to ‘those that are (there)’), but stated that she could not find a
way to put wik’uña explicitly into the expression and maintain this meaning.
(i) ka-q-kuna
be-nm-pl
‘those that there are’ (Consultant comment: ‘vicuña’ can’t appear here.)

165
(5.40) wik’uña
vicuña
ka-q-kuna
be-nm-pl
‘The ones that are vicuñas’ (Consultant comment: as opposed to the ones
that are not vicuñas.)
Examples (5.39) and (5.40) show that even the copular subordinate clause is not
compatible with an internal head. I do not have an explanation for this fact and it
requires further investigation, but recall from Chapter 3 that subject internal heads
in general were sometimes judged as marginal by consultants.
One fact that clearly emerges from these data is simply that vicuña, the associate
to the target existential, is not behaving like a typical subject at all in this
relativization test. Only the subject of the copula reading of vicuña behaves in an
expected subject-like manner as observed in (5.38).
It is important to note further, however, that the associate also does not behave
as a non-subject would be expected to behave with regard to relativization. Note
that relativization of the associate is completely impossible with the -sqa (nonsubject)
nominalizer as seen in (5.37). In light of the discussion in Chapter 3, this
observation therefore leads us to the conclusion that the associate must be neither
a subject nor an internal argument of the embedded verb (‘be’). I take this to mean
that the associate cannot merge directly with the embedded verb and then raise to
be assigned Case within the matrix clause as discussed extensively for embedded
objects in Cuzco Quechua in Chapter 3. This conclusion will be elaborated below in
the context of an explicit proposal regarding the syntactic structure of existentials.
To summarize our results thus far, there is evidence that in CQ, the associate
is not a subject since it fails to behave like a typical subject in the formation of
adverbial and relative clauses. These results point to the presence of a null expletive
in subject position in existential constructions.

166
5.3.3 The syntax of CQ existential sentences
In this section I propose a structure of Cuzco Quechua existentials with a null
expletive in subject position.
5.3.3.1 Theoretical assumptions
In this subsection I mention some basic assumptions regarding clause structure and
principles of interpretation which I am adopting (to supplement the brief discussion
of this topic from Chapter 1). On the syntactic side I follow recent work of Chomsky
[1998, 2000, 2001a] and others in assuming that a Tense head merges with an
intermediary vP which itself dominates the VP. I assume that this basic structure,
illustrated in (5.41), is present in both transitive and unaccusative clauses. Although
in the trees I adopt the convention of representing the earlier positions of a lexical
item by a trace, these traces may simply be exact copies of the moved constituent
as in the Copy theory of movement.
(5.41) Structure of a transitive sentence Structure of an unaccusative sentence
TP
✟ ✟✟✟
❍ ❍ ❍
Subject T’
✟ ✟✟✟
❍ ❍
❍ ❍
T vP
❍
✟ ✟✟✟ ❍
❍
❍
t Subject v’
✟ ✟❍
❍
VP v
✟ ✟❍
❍
Object V
TP
✟ ✟✟ ✟ ❍❍
❍
❍
Subject T’
✟ ✟✟ ✟ ❍❍
❍
❍
T vP
✟ ✟❍
❍
VP v
✟ ❍
t Subject V
On the semantic side I adopt Diesing’s Mapping Hypothesis [Diesing 1992] which
states (in part) that material within the verbal projection (I take this to include vP)
is subject to existential closure at LF. Therefore definites and generic indefinites

167
must escape the vP before the level of interpretation. A statement of the Mapping
Hypothesis is as follows:
(5.42) Mapping Hypothesis [Diesing, 1992 p.10]:
Material from VP is mapped into the nuclear scope.
Material from IP is mapped into a restrictive clause.
5.3.3.2 Structure of existentials
Let us consider what these assumptions mean for CQ existential sentences like (5.3),
repeated here:
(5.43) Llama-kuna urqu-pi ka-n.
llama-pl mountain-loc be-3sg
‘There are llamas in the mountains.’
Llamakuna is an indefinite which is subject to existential closure and so must be
interpreted within the vP. The question of why llamakuna also does not behave like
an internal argument of the verb (shown in the previous section) will be addressed
below. Since I am not aware of any evidence of inversion of the locative over the DP,
I take it that llamakuna is merged to the verbal projection only after the locative
urqupi ‘in the mountain’ has been merged. (See [Fernández-Soriano 1999] for an
analysis in which the locative is actually the highest argument in Spanish impersonal
constructions.) I take urqupi to be a DP and not a PP (see Lefebvre & Muysken
[1988] for arguments that -pi is an affix and not a clitic). Therefore, -pi is a Case
marker on the nominal and not a postposition. The role of -pi may, however, be to
assign locative case to urqu ‘mountain’. I will therefore assume that kay does not
check the Case of the locative DP.
These observations suggest the following as a starting point for determining the
structure of (5.43).

168
(5.44) Initial view of the structure of (5.43):
llamakuna
llamas
TP
✟ ❍
✟ ✟✟✟
vP
❍ ❍
❍
❍
❍ ❍
✟ ✟✟✟✟✟
VP
❍
✟ ✟✟✟✟
❍
❍
❍
❍
V’
✟ ✟✟ ❍ ❍ ❍
urqu-pi t V
mountain-loc
v
✟ ❍
v
❍
❍
❍
❍
ka-n
be-3sg
There are several issues to be resolved. First, it will be observed that in (5.44)
there is no element in Spec of TP, although T is generally assumed cross-linguistically
to have an OCC (EPP) feature as described in [Chomsky 2001]. This issue I propose
to resolve through the addition of a null expletive pronoun which merges directly
with T (as in [Chomsky 2001]). Second, llamakuna in (5.44) may also undergo
internal Merge to [Spec,vP] . Finally, recall that number agreement is not possible
in a CQ existential sentence such as this one and if this indicates that the relation
Agree[T,llamas] does not hold, we may wonder how in fact the DP llamakuna gets
its Case checked.
With regard to Case assignment to llamakuna, one possibility (in the spirit of
[Chomsky 2001b]) is that in addition to Agree[Expl,T], T probes down and enters
the relation Agree[T,llamakuna], checking the case of llamakuna. In this case, the
following two problems arise, however. First, why doesn’t the verb agree with
llamakuna just as the verb agrees with the associate in English. Second, why can’t
the T be of the special non-Case assigning type proposed for subject relatives in
Chapter 3, and thus llamakuna would be expected to be a potential relative clause
head. I therefore turn to a second possibility, which is to follow the idea of Belletti
[1988], who first suggested that the associate of the there expletive in English bears
T

169
Partitive Case, based on data from Finnish. In her view this Case is an inherent
Case assigned by the copular verb. This idea was later taken up by Lasnik [1995,
1996]. Under this analysis, the impossibility of the associate being a relative clause
head must then be equated with the impossibility of certain other non-argument
DPs taking on this role. As we saw in Chapter 3, licensing of an external head can
take place either by raising (if the head is an argument of the subordinate verb)
or by a pragmatically reconstructable relationship between the subordinate clause
and the matrix clause. The latter case was illustrated by examples such as (3.66),
repeated here as (5.45).
(5.45) [[naranja
orange
ch’arwa-sqa-y]
squeeze-nm-1sg
jugo]
juice
‘the juice that I made by squeezing oranges’
In this case, the relationship between juice and orange-squeezing is apparent.
However, in examples like (5.46), the relationship between working hard and children
is not clear enough to license this construction, despite the existence of a
corresponding matrix sentence as in (5.47).
(5.46) *[[pro ancha-ta llank’a-sqa-n] wawa-kuna
() a.lot-ta(adv) work-nm-3sg child-pl
‘The children he worked hard on account of’
(5.47) Ancha-ta llank’a-rqa-n wawa-n–kuna-rayku.
a.lot-acc work-past-3sg child-3sg–pl-on account of
‘He worked hard on account of his children.’
Thus I conclude that the relative clause in (5.31), like (5.46), is not grammatical
because the pragmatic connection between the head and the clause is not sufficiently
reconstructable to license this sort of head, just as is the case with the proposed
head wawakuna ‘children’ in (5.46).

170
Adoption of the partitive Case analysis is compatible with the view in (5.44) in
which the associate is first Merged within VP. The important point given our results
of Chapter 3, however, is that the associate is not an argument of the verb. The
non-argument status of the associate is also consistent with the observation that
the associate cannot be the head of a non-subject -sqa relative clause as mentioned
above. Therefore, either the T to which the expletive merges directly also assigns
Case to the associate, or the associate receives inherent Case. Under the second
view the tree structure we end up with is shown again in (5.48).
(5.48) TP
✟ ✟✟ ❍
❍ ❍
Expl T’
✟ ✟❍
❍
vP T
❍
❍
✟ ✟✟✟✟✟
❍
❍
❍ ❍
VP
v
❍
✟ ✟✟✟✟
❍
✟
❍
❍
❍ v ka-n
❍
llamakuna V’ be-3sg
llamas ✟ ✟✟ ❍ ❍ ❍
urqu-pi t V
mountain-loc
5.4 Structure of CQ Possessive sentences
In this section I turn to the structure of possessive sentences in Cuzco Quechua.
5.4.1 Some problems: ‘Maria has llamas’, and ‘There are
those mountains’
The analysis presented in the previous section has, as one of its predictions, a
Quechua version of the definiteness effect. That is, the presence of an existential
construction seems to imply that the associate will always be an indefinite. Indeed,
in the vast majority of cases that I have examined this holds. For instance,

171
consultants’ intuitions indicate that (cross-linguistically typical) weak quantifiers
in existentials are perfectly acceptable while strong quantifiers are degraded. For
example we have the following contrast:
(5.49) Ashka llama-kuna urqu-pi ka-n.
many llamas-pl mountain-loc be-3sg
‘There are many llamas in the mountains.’
‘Many llamas are in the mountains.’
(5.50) ?*Llipin/Tukuy
every/all
llama-kuna
llama-pl
urqu-pi
mountain-loc
‘There are all the llamas in the mountains.’
ka-n.
be-3sg
Furthermore, just as in English it is generally considered to be ungrammatical
to put other types of definite noun phrases into the existential context such as the
demonstrative expression in (5.51).
(5.51) Chay
dem
llama
llama
kurral-pi
corral-loc
*’There is that llama in the corral.’
‘That llama is in the corral.’
*ka-n/ka-sha-n.
be-3sg/be-prog-3sg
However, there is a very common class of existential-like sentences which may
seem to constitute a set of counterexamples to this generalization. In the possessive
sentence shown in (5.52) the noun phrase which plays the role of the associate
appears to be the possessive expression Mariaq wawankuna ‘Maria’s children’. Although
such possessive noun phrases in English share properties of both definites
and indefinites (cf. [Barker 1995]), they at least cannot appear in existential contexts
in English (*There are Maria’s children).
(5.52) Maria-q wawa-n-kuna
Maria-gen child-3-pl
‘Maria has children.’
ka-n.
be-3sg

172
In the following sections I will propose an analysis of (5.52) in which the possessor
is raised out of its DP at LF.
There is another class of apparent definiteness effect violations which is considerably
less frequent and which I will not analyze here, but leave to future work. In
these constructions, a demonstratively-marked noun phrase appears in the existential
kan context. These examples seem to be rare but can be found in narratives
and my consultants confirm the grammaticality of these sentences. An example is
given in (5.53) from Gregorio Condori Mamani’s (GCM) narrative.
(5.53) chay-manta
that-abl
pacha-n
time-3sg
chay
that(those)
urqo-kuna
mountain-pl
‘from that time, those mountains exist.’ (GCM p.19)
ka-n.
be-3sg
The context of this sentence is quite clear. In (5.53), the narrator is telling a
folktale/myth about the construction of Cuzco. The wind at that time was said
to be so fierce that the great Inca had to construct mountains around the city to
protect it. So (5.53) refers to the specific mountains around Cuzco. Note that kan
is being used to express the existence of a definite set of mountains.
Note that it is not sufficient to say that this kan is a main verb meaning ‘exist’,
since plural marking is not possible. In fact, if the plural marker is added in either
case the meaning of the sentence will change, as shown in (5.54).
(5.54) chay-manta
that-abl
pacha-n
time-3sg
chay
that(those)
urqu-kuna
mountain-pl
ka-nku
be-3pl
‘from that time, those are mountains (and not something else)’
In the next section I will focus on the particularly productive examples like
(5.52) of possessive sentences with existential form. I leave the case of (5.54) to
future research. 9
9 Although (5.54) seems to be rare it is not considered particularly anomalous.
In a constructed context I described some particularly large, beautiful and unusual

173
5.4.2 Some past approaches to unifying existentials with
possessives
While the particular CQ possessive construction illustrated in (5.52) is itself somewhat
mysterious, some sort of connection between possessive and existential constructions
has been posited for a variety of languages.. In this section I will briefly
summarize certain past approaches to a unified analysis of these constructions.
5.4.2.1 Freeze 1992
As part of a general theory in which existentials are viewed as locatives, Freeze
(1992) proposes that in English ‘have’ constructions, the subject is essentially a
locative and therefore (5.55) is parallel in structure to (5.56).
(5.55) [[ P ′ on [ DP the table] ] is [ PP [ DP a cat] t P ′]
(5.56) [[ P ′ [ I ]] have [ PP [ DP a cat] t P ′]
houses to my consultants. I asked them how they would ask me if those houses
(the one I had described) really exists. My consultants agree that under these
circumstances it would be quite normal to ask:
(i) chay wasi-kuna ka-n-(*ku)-chu?
that house-pl be-3sg-pl-q
‘Do those houses (really) exist?’
Furthermore, the example in (ii) indicates that there may in fact be an ambiguity
between an existential and non-existential version of the kan construction. This is
also part of the narrative of Gregorio Condori Mamani [1977], and is uttered by
individuals whose cows, unbeknownst to them, have been dyed a different color.
They seek their cows everywhere, saying “What has become of my cows?” followed
by (ii). They go on to comment that there are some other cows around but not of
the same color as their cows. The problem with this sentence is that the speaker
seems to be referring to some specific cows, which have (apparently) disappeared.
(ii) Mana waka-y-kuna ka-n-chu.
Not cow-1sg-pl be-3sg-NEG
‘My cows are gone.’ (GCM p58)
‘I don’t have any cows.’ (most natural reading out of context)

174
According to Freeze, there is no overt preposition in (5.56) because the preposition
has been incorporated into Infl, yielding the Spell-out “have” (=is+P).
In fact, Russian seems to be an example of a language which overtly reveals the
parallel structures suggested in (5.55) and (5.56), as shown in (5.57).
(5.57) U
near
menya
me
yest’
is
‘I have a book.’
kniga
book
5.4.2.2 Kayne 1994
As part of a general analysis of possessor constructions such as English “three books
of John’s” and French “la voiture de Jean”, Kayne adopts an analysis in which such
pre-possessor nominals are raised from a position below the possessor, rather than
supposing that the of phrase is a complement of the head nominal. Specifically, he
suggests the following structure for the French expression. (The D/P head is Kayne’s
notation for a “prepositional determiner” which he describes as “comparable to a
prepositional complementizer” (p.102)).
(5.58) la [ D/PP [ NP voiture j ] [de [ IP Jean [I o [e] j ...]]]] (Kayne p103,84)
For Kayne, the English example is essentially the same but the ’s is in I o . Now,
Kayne suggests that ‘have’ constructions are derived from a structure which is similar
to (5.58) but with empty D/P o head and instead of the definite determiner la,
it is the abstract copula BE which takes the D/PP as its complement. Specifically,
he gives the following initial structure for ‘Jean a une voiture.’
(5.59) ...BE[ D/PP [D/P o [ IP Jean [I o [voiture] j ...]]]] (Kayne p102,82)
Here, Jean moves to [Spec,D/PP] and then to [Spec,BE], while D/P o is incorporated
to BE. The overt form of ‘have’ is seen as the Spell-out of D/P o +BE. In this

175
respect the analysis is similar to that of Freeze. However, Kayne differs from Freeze
in that for Kayne voiture ‘car’ is originally predicated of ‘Jean’ while for Freeze,
‘car’ would be in Spec of a PP of which ‘Jean’ is the complement.
One question that arises at this point is how the semantic interpretation of
the resulting sentence is achieved. Kayne states simply that “the relation between
voiture and Jean would presumably be established within IP.” (p.102)
5.4.2.3 Szabolcsi 1994
Szabolcsi [1994] presents a theory of the Hungarian noun phrase which highlights
the parallels between CP and DP structure. Her view differs from Abney [1987] in
that, among other things, the D head is equated with the C head, as opposed to
Abney’s suggestion that D is parallel to I.
Hungarian DPs, like CQ DPs, show overt possessor agreement morphology on
the noun head. The actual position of the possessor varies: it can follow the definite
article in which case it is null (nominative) Case-marked, or it may precede the definite
article in which case it is dative marked. These two possibilities are illustrated
in (5.60) and (5.61).
(5.60) (a)
(the)
Mari
Mari(-nom)
‘Mari’s hats’ (Szabolcsi 2)
kalap-ja-i
hat-poss-pl(-3sg)
(5.61) Mari-nak
Mari-dat
a-
the
‘Mari’s hats’ (Szabolcsi 3)
kalap-ja-i
hat-poss-pl(-3sg)
Szabolcsi proposes that the Hungarian ‘have’ construction, illustrated in (5.62),
is derived from an existential construction (as in (5.63)), together with a possessor
construction. Thus, the dative subject is an instance of possessor extraction.

176
(5.62) Mari-nak
mari-dat
van-nak
be-3pl
‘Mari has hats.’ (Szabolcsi 109)
(5.63) Van-nak kalap-ok
be-3pl hat-pl(-nom)
‘There are hats.’ (Szabolcsi 109)
kalap-ja-i.
hat-poss.3sg-pl(-nom)
Possessor extraction in this case is required due to the definiteness effect, according
to Szabolcsi’s analysis. In Hungarian, possessor extraction is associated
with non-specificity, similar to the difference between English ‘Chomsky’s poem’ (a
specific poem by Chomsky) and ‘a poem of Chomsky’s’ (specific or non-specific).
5.4.2.4 Sánchez 1999a
In Sánchez (1999a) it is proposed that Southern Quechua possessor constructions
such as Maria-q llama-n (Maria’s llama-3sg) are reduced forms of complement
clauses. Briefly, the proposal is that this expression would be derived from the
nominalized clause meaning ‘Maria is a llama-owner,’ with verb and possessive suffix
deleted as follows:
(5.64) Maria-q llama-yuq
Maria-gen llama-possessor
> Maria-q llama-n
Maria-gen llama-3sg
ka-sqa-n
be-nm(nonsbj)-3sg
This proposal raises the following question: Why is it that the reduced form,
Maria’s llama does not have the same distribution as a complement clause but
rather has the distribution of other nominals. In other words, if Maria-q llama-n
‘Maria’s llama’ is a kind of contraction of the expression ‘Maria is a llama-owner’,
one might expect to be able to say (5.65), since (5.66) is good.
(5.65) *[Maria-q llama-n]-ta
Maria-gen llama-3sg-acc
*’I know Maria’s llama.’
yacha-ni.
know-1sg

177
(5.66) [Maria-q llama-yuq ka-sqa-n]-ta
Maria-gen llama-poss be-nm-3sg-acc
‘I know Maria is a llama-owner.’
yacha-ni.
know-1sg
5.4.3 A proposal for CQ possessive sentences
I summarize the above proposals as follows:
(5.67) • Freeze (cross-linguistic)
[BE [ PP llamas (near) Maria]] > Maria has llamas
• Kayne (French, English)
BE [ D/PP D/P o [ IP Maria I o llamas]] > Maria has llamas
[ D/PP of [ IP Maria ‘s [two llamas]]] > two llamas of Maria’s
• Szabolcsi (Hungarian)
BE [ DP To-Maria llamas] > To-Maria are llamas. (=Maria has llamas)
• Sánchez (Quechua)
(that) Maria llama-owner is / (that) Maria has llamas > Maria’s llamas
In short, although a number of analyses unify possessive sentences with possessive
nominals, the nature and directionality of the proposed relationship is rather varied.
Since the CQ existential construction at Spellout closely resembles the proposed
underlying structure of Szabolcsi’s and Kayne’s analyses, I would like to take these
analyses as a point of departure in analyzing CQ. That is, I propose that the CQ
existential structure is in some sense more basic than those analyzed in the above
papers in that it is close to the underlying structure proposed above (essentially, ‘Is
Maria’s llamas.’). This is contra Sánchez’s analysis in which Quechua possessives
are being derived from a complement possessor clause. It is also contra Freeze’s
proposal that in fact it is the locative (P’ constituent) which ends up in subject
position.

178
I will argue that the overt presence in Quechua of constructions which reflect
the posited underlying structure in the Kayne/Szabolcsi brand of analysis reinforces
the correctness of this line of reasoning.
Given this basic approach, we must address the question of the specific version of
the Kayne/Szabolcsi analysis which the CQ data seem to indicate, and the question
of the correct LF structure of the CQ possessive sentences. That is, how does
CQ avoid a violation of the definiteness effect despite the surface appearance of a
prenominal possessor of the associate in an existential? To answer this question we
must look more closely at the structure of a possessive nominal.
Although CQ does not show different Case-marking paradigms for different
possessor-positions as Hungarian does, there is some evidence that the possessor
can appear in two different DP-internal positions, since the genitive-marked nominal
may occur either before or after a (strong) quantifier 10 and before or after a
demonstrative, as follows:
(5.68) Maria-q tukuy
Maria-gen all
‘All Maria’s llamas’
(5.69) Tukuy Maria-q
all Maria-gen
‘All Maria’s llamas’
llama-n
llama-3sg
llama-n
llama-3sg
(5.70) Maria-q chay llama-n
Maria-gen dem llama-3sg
‘Those llamas of Maria’s’
(5.71) chay Maria-q llama-n
dem Maria-gen llama-3sg
‘Those llamas of Maria’s’
10 The pre-possessor position is not available to weak quantifiers.

179
These facts, and the basic insights of Abney [1987] and Szabolcsi [1994] which
suggest parallels between nominal and clausal structure, suggest the following structure
for CQ possessives. 11
(5.72) DP
❍ ✟ ✟✟✟✟ ❍
❍ ❍
Maria-gen D’
❍ ✟ ✟✟✟✟ ❍
❍ ❍
D nP
❍ ✟ ✟✟✟✟ ❍
❍ ❍
t Maria−GEN n’
✟ ✟✟ ❍
❍ ❍
NP n
✟ ✟❍
❍
n llama-3sg
t llama−3sg
11 I am simplifying here. In fact we cannot immediately jump to the conclusion
that the possessor appears before or after a single fixed Determiner position, since
quantifiers and demonstratives can also co-occur as in (i) and (ii). These examples
contain CQ’s strong ‘some (of)’, wakin.
(i) Chay wakin llama-kuna
dem some llama-pl
‘That subgroup of llamas.’
(ii) Wakin chay llama-kuna
some dem llama-pl
‘Some of those llamas.’
In these expressions, some consultants sense a difference in interpretation which
seems to reflect a fixed scope relationship between the elements involved. Thus,
(ii) seems to be good in contexts where there is a group of llamas which have
been previously mentioned or are physically present, and we are now referring to a
subgroup of these. On the other hand (i) is used where it is the subgroup which
has already been mentioned or is physically present. More research is necessary to
sharpen this intuition, but I will take this to suggest that the S-structure relative
order of strong quantifiers and demonstratives is also the LF order. Such is not the
case with possessors. I have not been able to uncover semantic differences between
(5.68) and (5.69) or (5.70) and (5.71). I take this to mean that demonstratives and
quantifiers are heads, perhaps Determiner heads of stacked DPs, but for present
purposes I ignore this complication and conflate these categories to the position of
the D head.

180
Although differing in some details 12 , the possessor positions in this structure are
those of the nominative and dative possessor positions suggested by Szabolcsi for
Hungarian. The difference between (5.68) and (5.69) would then be a matter of the
position in which the possessor is spelled out.
Now we turn to the question: for the non-specific reading which is forced in the
existential construction, is the possessor extracted (as in Hungarian) at LF? Or, is
the possessed noun llama raised to Spec of a functional head as in Kayne’s analysis of
the English construction “two llamas of Maria’s”? Or is a non-specific interpretation
already available with the possessor and the possessed noun remaining in their base
positions within the DP?
I propose that possessor extraction occurs in CQ at LF, following Szabolcsi’s
analysis of overt possessor movement in Hungarian. I aim to justify this analysis on
both empirical and theoretical grounds by showing that overt possessor extraction
is often associated with non-specific or indefinite readings in Quechua, and that
(following an idea of Kayne’s [1994, p.85]) the possessor cannot remain DP-internal
at LF for Case reasons.
Note that even in a simple possessive sentence like (5.73), it is not immediately
evident from the surface configuration whether Ana-q ‘Ana’s’ and alqunkuna ‘dogs’
forms a surface constituent or not. Based on the mandatory agreement between
these two elements, however, I assume that they do form a DP at some level of
the derivation. Of course if extraction has already occurred then I have nothing
further to show. However, there is some evidence that the surface form in (5.73)
does contain the constituent Ana-q alqunkuna. The evidence comes from consultant
responses to examples in which the possessor appears separated from the noun by
12 In particular (as in the previous chapters), in the suggestion of an n head,
parallel to the clause’s v, in whose Spec position the Subject (possessor) of the
nominal is generated (as advocated in [Chomsky 1998] and [Marantz 2001]).

181
the verb. This is shown in (5.74), which is considered to be degraded. In fact, these
judgments are similar to those in (5.75) and (5.76) where Ana-q alqunkuna is a true
subject, shown continuously and discontinuously.
(5.73) Ana-q alqu-n-kuna
Ana-gen dog-3sg-pl
‘Ana has dogs.’
ka-n-mi.
be-3sg-evid
(5.74) ??Alqu-n-kuna
dog-3sg-pl
‘Ana has dogs.’
ka-n
be-3sg
Ana-q.
Ana-gen
(5.75) Ana-q alqu-n-kuna kani-sha-n.
Ana-gen dog-3sg-pl bark-prog-3sg
‘Ana’s dogs are barking.’
(5.76) ??Alqu-n-kuna kani-sha-n
dog-3sg-pl bark-prog-3sg
‘Ana’s dogs are barking.’
Ana-q.
Ana-gen
However, it is not the case that Ana-q and llaman must obligatorily appear in
adjacent positions in a possessive sentence. In fact, very frequently consultants
suggest variants on the basic order in (5.73) to express this meaning. The following
examples have all been offered by consultants presented with a drawing in which
an individual is shown as the owner of a number of animals, and I ask how to
express ‘Maria has three dogs’, etc. In the examples (5.77), (5.78) and (5.79),
an evidential marker appears on the possessor, and in the second two of these the
possessor is actually not contiguous with the rest of the DP. In (5.80) a topic marker
appears on the possessor. The syntax of evidentially marked constituents is not well
understood, 13 but these constructions can be seen as evidence that the subject has
been extracted. A concrete reason for this is that evidentially-marked elements
13 [Faller 2002] is an in-depth study of the semantics and pragmatics of evidentiality
in Cuzco Quechua.

182
are typically either full DPs or co-Case-marked elements extracted from a DP. DPinternal
elements do not support evidential suffixes. In each case the basic word
order ‘possessor-gen possessed-AGR kan’ is also available.
(5.77) Mariya-q-(mi) kinsa
Maria-gen-(evid) three
‘Maria has three dogs.’
alqu-n-kuna
dog-3sg-pl
(ka-n)
(be-3sg)
(5.78) Pay-pa-n ka-lla-n-taq
S/he-gen-evid be-CONJ-3sg-CONJ
‘She also has many sheep.’
ashka
many
obeja-n-kuna.
sheep-3sg-pl
(5.79) Ana-q-mi
Ana-gen-evid
ka-n
be-3sg
‘Ana has (many) goats.’
ch’ipu-n-kuna.
goat-3sg-pl
(5.80) Ana-q-qa pisqa-n
Ana-gen-top five-evid
‘Anna has five llamas.’
llama-n
llama-3sg
(kan).
(be-3sg)
I would like to suggest that the difference between (5.74) and, say, (5.79) is that
the lack of evidential or topic marking in (5.74) makes it difficult to understand that
the relevant construction is one with an extracted possessor. The examples in (5.77)
to (5.80) have moved the possessor to a focus position in each case. In CQ extraction
of a DP-internal element is indicated by mandatory co-Case-marking of the extracted
element and the DP remnant ([Lefebvre & Muysken 1988]). 14 However, in the above
examples we notice that since the associate receives null Case marking in existential
constructions, an extracted possessor cannot be immediately recognized as such by
this test. 15
14 In Chapter 6 I look at this phenomenon rather more generally with regard
to DPs in direct object position. For current purposes I assume that the genitivemarked
noun phrase has been extracted when it is discontinuous from a noun phrase
containing a noun which agrees with it.
15 This is as expected. However, the following surprising constructions which seem
to violate this expected pattern, have been suggested to me by a speaker of Bolivian

183
In fact, we find other situations in which the extraction of a possessor is associated
with non-specific or indefinite readings also. Examples (5.81) and (5.82) show
an extracted and unextracted possessor. The extracted possessor construction lends
itself more readily to an indefinite interpretation. The unextracted possessor in
(5.82) can also be interpreted as ‘I drank only a part of Gabriela’s cornbeer (which
happened to be a lot),’ but less readily. Since word order in CQ is quite free, the
relative positions of the extracted element and the remnant are also quite free.
(5.81) Nuqa Gabriela-q(pa)-ta [e ashka aqha-n-ta]
I Gabriela-gen-acc e much cornbeer-3sg-acc
‘I drank a lot of cornbeer of Gabriela’s.’
ukya-ra-ni.
drink-past-1sg
Quechua. This speaker uses the construction occasionally in conversational Bolivian
Quechua, and having lived extensively also in Cuzco felt that it also exists in that
dialect. However, I was not able to confirm this construction for Cuzco Quechua
with my consultants. Rather mysteriously, the examples involve the addition of
the suffix -ta to the possessor in an existential construction. This suffix would be
expected if we were extracting from an Accusative-marked DP.
(i) Kay warmi-q-(ta) ka-sqa kinsa wawa-n-kuna.
this woman-gen-acc be-pst,rep. three child-3sg-pl
‘This woman had (reportedly) three children.’
(ii) Maria-q-*(ta) qulqi-n ka-n chay-qa, qhatu-man ri-n.
Maria-gen-acc money-3sg be-3sg that-top market-dat go-3sg
‘When Maria has money she goes to the market.’
In (i) an optional -ta (accusative) marker appears on the possessor. In (ii), in
which the possessive clause is a subordinate clause adverbialized via the subordinator
chay ‘that’, accusative marking on the possessor is apparently obligatory. If
these examples are indeed acceptable for some Cuzco speakers, they provide strong
support for Kayne’s idea that extraction of the possessor is related to Case marking
of the possessor. That is, that when a possessor is in a DP with an indefinite
determiner, another Case licensor which agrees with the possessor must be present.
These examples seem to illustrate that the second Case licensor can overtly trigger
-ta marking on the possessive. Contrary to Hungarian, in which the secondary Case
licensor is associated with the [Spec,DP] position, CQ possessors cannot bear alternate
Case-marking when inside the DP. More work is necessary in order to establish
the exact conditions under which this secondary Case marking is triggered, however.

184
(5.82) Nuqa
I
[Gabriela-q
Gabriela-gen
ashka
much
‘I drank Gabriela’s (lot of) cornbeer.’
‘? I drank a lot of Gabriela’s cornbeer.’
aqha-n-ta]
cornbeer-3sg-acc
ukya-ra-ni.
drink-past-1sg
Furthermore, a Bolivian consultant provides the following strong evidence from
Bolivian Quechua (very closely related to Cuzco Quechua) that extraction is associated
with non-specificity. Comparing sentences (5.83) and (5.84) we find it is only
with an extracted possessor that the nonspecific reading of uj wawa-n ‘a/one child’
is available.
(5.83) Sapa
each
uj
one
riqsi-nchis.
know-2plincl
(nuqanchiq-manta)
(2plincl-abl)
Maria-q
Maria-gen
uj
one
wawa-n-ta
child-3sg-acc
‘Each one of us knows Maria’s child.’ (same child for each person)
(5.84) Sapa
each
uj
one
riqsi-nchis.
know-2plincl
(nuqanchiq-manta)
(2plincl-abl)
[e
e
uj
one
wawa-n-ta]
child-3sg-acc
Maria-q-ta
Maria-gen-acc
‘Each one of us knows a child of Maria’s.’ (could be different child each time)
The equivalent sentences in Cuzco Quechua are a bit different since in this dialect
the distributive suffix -nka is available, 16 while the expression sapa uj ‘each one’ is
not. Nonetheless, a similar generalization can be made (possessor extraction is
associated with the non-specificity of the noun phrase) based on the data shown in
(5.85) and (5.86). Similar examples with the distributive suffix -nka are shown in
(5.87) and (5.88).
16 The distributive or group-forming version of this suffix is discussed by Faller
[2001]. This suffix was also discussed in Chapter 2, Section 2.5. In Chapter 6 I look
at the version of -nka which appears on sapa ‘each’, and is here glossed simply as
-nka.

185
(5.85) Sapa-nka-yku
each-nka-2pl
mana
not
riqsi-yku
know-2pl
riqsi-yku-chu
know-2pl-neg
[Mariya-q
Maria-gen
Mariya-q
Maria-gen
huk
one
wakin
some(other)
irqi-n]-ta.
child-3sg-acc.
irqi-n-ta.
child-3sg-acc
Nuqa-yku
I-2pl
‘Each of us knows a (specific) child of Maria’s. We don’t know Maria’s other
child(ren).
(5.86) Sapa-nka-yku nuqa-yku
each-nka-2pl I-2pl
Maria-q-ta.
Maria-gen-acc
riqsi-yku
know-2pl
[e huk
one
irqi-n-kuna]-ta
child-3sg-pl-acc
‘Each of us know some (maybe different) children of Maria’s.’
(5.87) *[Maria-q
Maria-gen
huk-ni-nka
one-euph-distr
wawa-n-(kuna)]-ta
child-3sg-pl-acc
sapa-nka-yku
each-nka-2pl
pusa-yku.
bring-2pl
(5.88) [e Huk-ni-nka
one-euph-distr
Maria-q-ta.
Maria-gen-acc
wawa-n]-ta
child-3sg-pl-acc
sapa-nka-yku
each-nka-2pl
‘Each of us brought one of Maria’s children (to the party).’
pusa-yku
bring-2pl
Further motivation for a DP-extraction analysis can be found in some of the
past work on possessive sentences summarized earlier. Szabolcsi [1994] does not
offer an explanation for why possessor extraction should be associated with nonspecificity
in Hungarian (while a post-nominal PP like ‘of John’s’ should serve this
function in English) but points to this as an interesting question open for further
investigation. She comments that in English and Hungarian the relevant constraint
seems to be that no [Spec,DP] material may be present for a non-specific reading
to be available. Kayne [1994], however, sees both the Hungarian and English facts
as based in Case theory (p.85). He suggests that a definite but not indefinite D o
licenses Case of the possessor. In Hungarian this problem is solved for indefinites by
moving the possessor to the dative-Case position in [Spec,DP]. From this position
extraction can occur, licensed (say both Kayne and Freeze [1992]) by incorporation

186
of the preposition into the existential BE. In English the problem is solved by the
insertion of the Case-licensing preposition of into the D/P o position accompanied
by the raising of the possessed nominal.
In Quechua, although I am not aware of direct evidence that the [Spec,DP]
position is a Case position as opposed to the [Spec,nP] position, I will accept Kayne’s
notion that an indefinite determiner is not sufficient to assign Case to the possessor.
I suggest that Quechua, like Hungarian, has a higher position within a DP which is
a Case position. This position is probably the same escape hatch in which Co-case
marking of any extracted element is licensed. This analysis is reminiscent of the
proposal in Lefebvre & Muysken [1988] that there is a “COMP-like Case position”
within a DP, where the moved element picks up Case. However, relative clause
data provided by Lefebvre and Muysken intended to show that an element may
appear overtly in this position is ungrammatical for my consultants. These data
were discussed in Chapter 3, Section 4.4. Therefore for my consultants it appears
that this “COMP-like Case position” within the DP is only a position from which
movement out of the DP may occur. I discuss this position further in Chapter 6. 17
An explicit implementation of the notion of a DP-external Case position is as
follows. Suppose that there is a functional head K that selects DP and which is
a probe with nominal φ features. K is optionally present in instances of optional
extraction from DP, but in the case of DPs with indefinite D heads and a possessor,
as in possessive sentences, K must be present in order to check the Case of the
17 There is a further issue that this analysis brings up which I am currently unable
to explain. That is, how does Spellout occur in CQ before the possessor has had
its Case fully checked, since current assumptions suggest that all uninterpretable
features must be deleted before the construction is sent to PF? For the moment I
will simply suggest that this apparent violation may be related to the double Case
marking on the possessor (the combination of genitive Case marking plus the Case
marking associated with clause extraction). Similarly in English, the expression
“books of John’s” also contains two apparent Case-licensors: ’s and of.

187
possessor. The possessor merges with n as usual, the genitive Case marker being
consistent with this position of Merge and the agreement between the possessor and
the possessed noun (presumably due to the head-to-head match of n and N). All
this is just as we saw in Chapter 4. Next D merges with nP. I assume that the
indefinite D is a kind of defective probe: it has no uninterpretable Case feature
but it does have a +D feature which matches the possessor and EPP feature which
results in the Internal Merge of the possessor to [Spec,DP]. Next K selects the DP.
The resulting configuration is shown in (5.89).
(5.89) KP
✟ ✟✟✟ ❍ ❍ ❍
K DP
✟ ✟✟ ✟ ❍❍
❍
❍
DP D’
✟
Maria-q ✟✟ ❍
❍ ❍
i D nP
✟ ✟✟✟
❍ ❍
❍ ❍
t i n’
✟ ✟ ❍❍
NP
❍
n
llama-n-kuna
Now, K has an uninterpretable Case feature and a +D feature. Note that either
the DP Mariaq or the DP Mariaq llamankuna are potential goals for this probe, and
neither D head is an intervener. K agrees with Mariaq, checks the Case of Mariaq,
and Mariaq moves to [Spec,KP]. I assume that the particular reflex of the Case
marking by K will be determined by the matrix-clause Case-assigner, and that the
[Spec,KP] is a position from which further extraction to a matrix-clause A’ position
typically occurs. Again in these respects, [Spec,KP] would be the “COMP-like Case
position” posited in [Lefebvre & Muysken 1988].
Note that under this analysis Quechua looks very similar to Hungarian except
that Hungarian possessive sentences are associated specifically with dative Case

188
marking and with overt possessor extraction.
In the next two trees I compare the proposed structural analyses of the CQ
versions of ‘There are llamas’ and ‘Maria has llamas’ respectively.
(5.90) TP
✟ ✟✟✟
❍
❍
❍ ❍
Expl T’
❍
✟ ✟✟✟✟
❍
❍
❍ ❍
T
vP
✟ ✟ ❍❍
VP
✟ ✟✟ ❍ ❍
DP V
❍
v
llama-kuna i
kan
(5.91) TP
✟ ✟✟ ❍ ❍ ❍
Expl T’
✟ ✟✟ ❍ ❍ ❍
T vP
✟ ✟✟✟ ❍ ❍ ❍
VP v
❍ ✟ ✟✟✟✟ ❍
❍ ❍
KP i
V
✟ ✟✟ ❍
❍
❍
Maria-q j K’
kan
✟ ✟ ❍❍ ❍
K DP
✟ ✟✟ ❍ ❍
t j D’
✟ ✟✟ ❍
❍ ❍
D nP
✟ ✟✟✟
❍ ❍
❍ ❍
t j n’
✟ ✟ ❍❍
NP
❍
n
llama-n-kuna
The above trees have been simplified in a number of ways. For example, I
ignore head movement of V and of N. Furthermore, I put aside the question of what

189
exactly the final position of Maria-q will be. In both trees, however, the associate
of the expletive is assumed, as discussed in Section 5.3, to receive inherent partitive
Case within the matrix clause. In both trees, the associate is indefinite. The KP
dominating the DP in (5.91) checks the Case of Maria-q and I assume licenses the
copying of (null) partitive Case marking onto the genitive-marked possessor Maria-q.
5.4.4 Possessive adverbial and relative clauses
Recall that in Section 5.3 we saw examples of adverbial clauses and relative clauses
which were revealing of the (non)-subjecthood of the associate in an existential
construction. In this section I will look at similar data containing possessive clauses
in light of the syntactic structures I have just presented. A first example of a
possessive adverbial clause from [Valderrama & Escobar (Condori Mamani) 1977] is
shown in (5.92).
(5.92) [familiar-ni-n-kuna
relative-euph-3-pl
ka-qti-n],
be-adv/bi-3sg
p’anpa-chi-pu-nku.
bury-cause-permanently-3pl
pay-kuna
s/he-pl
reclama-spa
reclaim-uni.adv
‘If he has relatives, they, relaiming (his body), have him buried.’ (in the
generic case of the death of a porter) (Condori Mamani p.88)
In (5.92), the possessive clause “he has relatives” appears as a subordinate adverbial
clause, and note that the bipersonal adverbializing suffix -qti appears on
the subordinate verb. This indicates that the subjects of the two sentences are not
co-referential. This is expected given the tree in (5.91) since the expletive (and not,
say, the possessor), a non-referring expression, appears as the structural subject of
the possessive clause. Note that in this case the possessive DP (’his relatives’) and
the subject (’they’) of the matrix sentence are, in fact, coreferent.
A more complete set of constructed examples are shown in (5.93) to (5.95).

190
(5.93) Pilar-pa mihuna wasi-n ka-qti-n-qa pay-pa
Pilar-gen food house-3sg be-bi.adv-3sg-top she-gen
qulqi-n ka-ra-n.
money-3sg be-past-3sg
‘When Pilar had a restaurant, she had a lot of money.’
nishu
much
(5.94) Pilar-pa mihuna wasi-n ka-qti-n-qa,
Pilar-gen food house-3sg be-bi.adv-3sg-top
mihuna-ta mihu-ra-n.
food-acc eat-past-3sg
‘When Pilar had a restaurant, she ate good food.’
allin-ta
good-acc
pay
she
(5.95) Pilar-pa
Pilar-gen
mihuna
food
wasi-n
house-3sg
ka-qti-n-qa,
be-bi.adv-3sg-top
ka-q.
be-past(habitual)
‘When Pilar had a restaurant, it always used to be full.’
hunt’a-puni
full-always
Note that in each of the above examples, the bipersonal adverbial is selected,
indicating that the subject of the subordinate possessive clause (‘When Pilar had a
restaurant’) is neither Pilar (5.94) nor the restaurant (5.95), nor is it the case that
the expletive in two possessive sentences (5.93) co-refer in the relevant sense.
Now turning to relative clause data of a similar sort, recall further that the choice
of relative clause nominalizer (correlated with subjects vs. non-subjects) indicated
that the associate was not a subject nor was it an argument of the verb for the
purposes of relativization as shown in Section 5.32.
Thus, in light of the present study of possessor sentences it is interesting to
consider the CQ translations of relativized forms of English ‘have’ constructions. In
(5.96) and (5.97) I give a standard CQ possessive sentence followed by a construction
which in English involves object relativization from a ‘have’ sentence. Note that in
CQ the relative clause morphology indicates that a subject has been extracted,
though no overt external head is present. It is possible that wawa-y ‘child-1sg’ is
an internal head, but consultants do not allow a reversal of the order of the verb

191
and this nominal (*kaq wawa-y-kuna) and hence it seems most likely that waway
‘child-1sg’ is predicative here (cf. the discussion of wik’uña in (5.38) to (5.40)). That
is, the construction seems to be more along the lines of the English “Those (ones)
who are my children give me a lot of work.”
(5.96) Wawa-y-kuna ka-n.
child-1sg-pl be-3sg
‘I have children.’
(5.97) [[Wawa-y ka-q]-kuna] ancha-ta llank’a-chi-wa-nku.
child-1sg be-nm.subj-pl much-acc work-cause-1sg-3pl
‘The children that I have make me work a lot.’ (...so I don’t want any more)
It also turns out that the equivalent of English subject relativization from a
‘have’ clause is impossible in Quechua and we are forced to resort to the alternative
possessive construction, which uses the suffix -yuq ‘possessor of’. This contrast is
illustrated in (5.98) and (5.99).
(5.98) Wawa-yuq warmi
child-possessor woman
‘The woman who has children.’ (’the child-possessor woman’)
(5.99) *[[Wawan
child
ka-q/ka-sqa-n]
be-nm.subj/be-nm.nonsubj-3sg
warmi]
woman
Intended meaning: the woman who has children
Thus, although I have claimed that the possessor in a CQ possessive sentence
can be extracted from its DP at LF, this possessor cannot go on to become the head
of a relative clause. This again provides strong evidence that the raised possessor is
neither a subject nor an internal argument of the verb (and so is necessarily Casemarked
within its clause rather than via relativization). It is interesting to compare
examples such as (5.99) with cases in which a possessor can appear as an external
(though not internal) head of a relative clause. In Chapter 3 we saw examples of
this in (3.62) and (3.63), repeated here as (5.100) and (5.101).

192
(5.100) [[[Waka-q
cow-gen
uña-n]
baby-3sg
suwa-pu-sqa-nku]]
steal-definitively-nm-3pl
‘the cow’s calf which was stolen’
*‘the cow whose calf was stolen’
(5.101) [[e
e
Uña-n
baby-3sg
suwa-pu-sqa-nku]
steal-definitively-nm-3pl
‘the cow whose calf they stole’
waka]
cow
Example (5.100) was impossible under the theory proposed in Chapter 3 because
only arguments of the subordinate verb could be internal heads in Cuzco Quechua.
To account for (5.101) it was necessary to propose that an internal pro was licensing
the third person singular on ‘baby’, and that the external head was base-generated
in that position. This sort of head was proposed to be licensed via the pragmatic
connection between the relative clause and this head. Clearly to be correctly formulated
this condition must rule out a similar licensing in the -sqa version of (5.99). I
would like to suggest that the relevant difference between (5.99) and (5.101) has to
do with the licensing of pro. Notice that in (5.101), pro i would have to move out of
the DP and have its Case checked by K, whereas in (5.99) the Case of pro can be
checked by the definite D head. It seems, then, that only an in situ pro is sufficient
to license the external coindexed head. I leave to future work the exact nature of
this restriction on the distribution of pro.
5.5 Conclusion
In this chapter I have proposed structures for Cuzco Quechua existential and possessive
constructions in which each of these sentence types contains a null expletive and
an indefinite associate to the expletive. The null expletive proposal is motivated by
the necessarily existential interpretation of CQ sentences containing the overt verb
form kan (‘be-3sg’), together with data suggesting that the associate does not be-

193
have like a subject in other constructions. This syntactic movitation permitted
adherence to the Mapping Hypothesis, which says that an indefinite which is subject
to existential closure must be interpreted within the verb phrase (which I take
to be vP). The indefiniteness of the associate is in line with the definiteness effect
seen in existential constructions cross-linguistically. I have analyzed CQ possessive
sentences, which seem to have a specific/definite DP (with an explicit prenominal
possessor) in an existential construction, by suggesting that the possessor raises out
of its DP at LF, in parallel to overt movement which has been postulated in English
and French by Kayne [1994] and in Hungarian by Szabolcsi [1994].

Chapter 6
The Semantics of Discontinuous
Noun Phrases in Quechua ∗
6.1 Introduction
In this chapter I take a closer look at the issue of extraction from DP in Cuzco
Quechua, with the aim of developing a more detailed understanding of DP semantics
and syntax. In Chapter 5 I looked at the particular case of possessor extraction,
but in fact various other elements (especially quantifiers and adjectives) can also
appear separated from the noun phrase that they seem to be associated with. Thus
the study of discontinuous noun phrases in general allows us to continue to study
the behavior of quantifiers outside of the specific context of relative clauses. In
particular, in this chapter I compare the interpretation of the continuous and discontinuous
version of noun phrases, and propose a structural explanation for the
pattern of semantic differences observed.
Quechua noun phrases, like English noun phrases, typically occur as single constituents
with a fairly fixed internal word order. Quechua has overt Case markers
∗ This chapter is based on [Hastings 2003].
194

195
which appear at the end of the string of noun phrase-internal elements. An example
from Cuzco Quechua is shown in (6.1).
(6.1) [Hatun
big
wasi]-ta
house-acc
‘I saw a/the big house.’
riku-rqa-ni.
see-past-1sg
In (6.1), the noun phrase ‘big house’ is expressed as the constituent hatun wasi
which shows the standard Quechua word order (adjective + noun) and is marked
with the accusative Case marker -ta. However, unlike English, Quechua also allows
discontinuous noun phrases, in which different parts of an apparent single noun
phrase each receive their own Case marker. These parts may be separated by the
verb (or some other clausal constituent) as in (6.2) and (6.4) or adjacent to one
another as in (6.3). I will describe all these cases as involving discontinuous noun
phrases.
(6.2) Wasi-ta riku-rqa-ni
house-acc see-past-1sg
‘I saw a big house.’
(6.3) Pisi-ta mikhuna-ta
a.little-acc food-acc
‘I ate a little food.’
hatun-ta.
big-acc
mikhu-rqa-ni.
eat-past-1sg
(6.4) Qulqi-y-ta
money-1sg-acc
tari-rqa-ni
find-past-1sg
llipi-n-ta
all-3sg-acc
‘I found all my money.’ (Muysken 1989 15a)
This phenomenon is described as “floating” of the modifier by Lefebvre and
Muysken [1988], who further note that the directionality of the float is not fixed
[p.163]. This variability of word order can also be observed in (6.2) to (6.4). In
addition to examples like those above, which involve quantifiers and adjectives,
there are discontinuous noun phrases in Quechua in which possessors or wh-words

196
appear separated from the noun. Here I will limit my discussion to quantifier and
adjective discontinuities.
In this chapter I look at the semantics and syntax of the discontinuous noun
phrase construction in Cuzco Quechua. I consider the question of whether a continuous
DP is constructed (or perhaps re-constructed) as a single unit at the level
of interpretation (LF). I also ask more generally what semantic relationship is established
or indicated via the “co-Case marking” of the different parts of the noun
phrase. In addressing these questions I look at some semantic differences between
the continuous and discontinuous versions of noun phrases and point to implications
for the LF structure of discontinuous NPs. In particular, I argue that the indefiniteness
of certain discontinuous noun phrase constructions points to an interpretive
configuration in which the modifying element appears outside of the definiteness
head of the DP which contains the noun itself.
The organization of the chapter is as follows. In the next section I provide more
discontinuous noun phrase data and background on Quechua quantifiers. In Section
6.3 I discuss previous syntactic analyses of discontinuous noun phrases in Quechua
and present data illustrating semantic effects of the discontinuity. In Section 6.4
I propose an analysis which essentially states that co-Case marking of a modifier
indicates scope outside the DP. In Section 6.5 I discuss an apparently misbehaved
strong quantifier, sapa ‘each’. Section 6.6 is the conclusion.
6.2 More data
6.2.1 Co-Case marking
I begin with further examples of the phenomenon of “co-Case marking” (to borrow a
term used in [Lefebvre & Muysken 1988]). In (6.5) to (6.7) I show further examples

197
involving the three categories of discontinuous noun phrases which I address in this
chapter: weak quantifier, adjective and strong quantifier discontinuities. In (6.8) and
(6.9) I show examples of possessor and wh-word discontinuities. Possessor extraction
was discussed in the previous chapter, and wh-extraction closely parallels adjective
or quantifier extraction.
(6.5) Weak quantifier:
...mikhuna-ta-qa
food-acc-top
ashka-ta-n
a.lot-acc-evid
‘...they gave me a lot of food.’ (GCM p.25)
qu-wa-q-ku.
give-1sg-past(habitual)-3pl
(6.6) Adjective:
Runa-ta
man-acc
riqsi-ni
know-1sg
kallpa-yuq-ta
strength-with-acc
‘I know a man with strength.’ [‘I know a strong man.’] (Lefebvre & Muysken
1988 p.142)
(6.7) Strong quantifier:
...llipin-ta
all-acc
manka-kuna-ta
pot-pl-acc
chhalara-pu-ni.
change–1sg
‘I changed all the pots.’ (GCM p.27)
(6.8) Possessor:
Nuqa Gabriela-q-ta ashka aqha-n-ta
I Gabriela-gen-acc a.lot cornbeer-3sg-acc
‘I drank a lot of Gabriela’s cornbeer.’
ukya-ra-ni.
drink-past-1sg
(6.9) Wh-word:
Hayk’a-ta riku-rqa-nki Maria-q hatun
How.many-acc see-past-2sg Maria-gen big
‘How many of Maria’s big houses did you see?’
wasi-n-kuna-ta?
house-3sg-pl-acc

198
The phenomenon I am considering here is largely limited to the direct object position.
Certainly it is most common and productive when involving the -ta marker.
I will return a possible explanation for this restriction in Section 6.4. The basic incompatibility
of co-Case marking with subjects is shown in (6.10) and with locatives
is shown in (6.11).
(6.10) *?Ashka rima-sha-nku
many talk-prog-3pl
‘Many people are talking.’
(6.11) *Hatun-pi tiya-ni
big-loc live-1sg
‘I live in a big house.’
runa.
person
wasi-pi.
house-loc
Subjects of typically unaccusative verbs behave like other subjects in this regard,
as seen in (6.12).
(6.12) *?Ashka chaya-mu-ra-nku
many arrive-cis-past-3pl
‘Many people arrived.’
runa.
person
However, the status of discontinuous subjects is somewhat unclear. Muysken
[1989], who basically assumes that co-Casemarking is limited to -ta positions, notes
this same issue and provides the following example [Muysken p.636], which again is
not entirely unacceptable.
(6.13) *?[e i runa-kuna]
man-pl
hamu-n
come-3
llipi-n i
all-3
‘The men all come.’ (Muysken 1989 (22))
My consultants are similarly uncertain regarding these constructions. Furthermore,
recall that in Chapter 5 I looked at extraction of possessors from DPs in
existential constructions. Certainly this was not a -ta-marked position. In this

199
chapter I focus on -ta co-Casemarking since this phenomenon provides the most
frequent and robust data. However, my analysis will not depend crucially on the Tr
head (as discussed in Chapter 3) being the sole licensor of the double Case-marking,
and in fact this is unlikely to be the case. I will return to the question of the
environment in which the apparent extraction from DP is licensed in Section 6.4.3.
Besides the cases of quantification and (restrictive) modification considered here,
there are other situations in which the same Case marker may be used twice in
Quechua, most of which I am unable to address here in detail. These include ambiguous
Case markers, conjunctions, appositives, and secondary predicates. Ambiguous
Case-marking is discussed in Section 6.3. Conjunctions I leave aside as
clearly distinct from the phenomenon under study here. Appositives and secondary
predicates, however, can sometimes appear syntactically rather similar to the constructions
under study here. In fact, besides the semantic distinctions one piece of
evidence showing that they do constitute different constructions altogether is that
they allow co-Casemarking of oblique as well as structural Case. In some cases co-
Casemarking constructions may in fact be structurally ambiguous, and an example
and brief discussion are provided below in footnote 2.
6.2.2 Quantifiers
In this section I give further background on the Quechua quantifiers in question. I
claim that the basic use of the quantifiers is as D-quantifiers, or “determiner-like”
quantifiers (in the vocabulary of [Bach et al. 1995]). An example is shown in (6.14).
(6.14) wakin ‘some (of)’
ayni-ta-qa
work.exchange-acc-top
ru-ra-yku
do-past-1pl(excl.)
[waki-lla-n
some-delim-3
‘...only some peasants did the work exchange.’ (GCM p.36)
paisano]...
peasant

200
Other quantifiers that appear DP-internal in constructions such as these are pisi
‘a few/a little’, ashka ‘a lot/many’, tukuy ‘all/every’, llipin ‘all/every’, sapa ‘each’,
and numbers like huk ‘one’. All these quantifiers behave like classic D-quantifiers
in a variety of ways. For example, within the noun phrase they must appear prenominally,
as illustrated in (6.15). Furthermore, they can be scrambled along with
the DP as in (6.16). Finally, these modifiers do not have their own Case markers in
these common usages, and are not ambiguous with respect to their associates.
(6.15) *[Aqha
cornbeer
pisi]-ta
a.little-acc
ukya-rqa-ni.
drink-past-1sg
(6.16) Ukya-rqa-ni
drink-past-1sg
[pisi
a.little
aqha]-ta.
cornbeer-acc
‘I drank a little cornbeer.’
Although the most common use of these quantifiers is as D-quantifiers, in DPinternal
position, one issue which complicates this picture is that they sometimes
also have adverbial (A-quantificational) uses. These uses have been noted by [Cusihuamán
1976/2000] among others. Clear adverbial uses of the weak quantifier pisi
‘a little’ are given in (6.18) and (6.19). Note that here there is no direct object
(implicit or explicit) which could be construed as co-Case-marked with the adverb
pisi-ta. I have not provided a gloss for the adverbial suffix -ta, but I consider it to
be simply homophonous with the accusative -ta. One reason for this is that many
adverbials do not have a -ta marker at all. An example is shown in (6.17).
(6.17) Nuqa usqaylla
I quickly
‘I came quickly.’
hamu-ra-ni.
come-past-1sg
Thus, it appears that -ta is a derivational affix (and not a Case-marker) which
forms adverbs from adjectives (and some quantifiers), along the lines of English ‘ly’.

201
(6.18) Pisi-ta llank’a-rqa-ni.
a.little-ta work-past-1sg
‘I worked a little.’
(6.19) Wawa-yki
child-2sg
pisi-lla-ta-chu
a.litte-delim-ta-q
hamu-n?
come-3sg
‘Does your son come only a little (infrequently)?’
Non-quantificational adjectives can also participate in this adverbial construction.
An example is illustrated in (6.20). Here, however, there is an ambiguity
present. The ambiguity is between a verb-modifier use of sumaq-ta (‘well’) and a
noun-modifier use of sumaq-ta (‘good’). Because of the coindexation between wasi
‘house’ and sumaq ‘good’ in the latter case, and the adjectival interpretation of
‘good’, this reading falls into my category of discontinuous noun phrase.
(6.20) Nuqa
I
wasi-ta sumaq-ta
house-acc good-ta
‘I am watching over the house well.’
? ‘I am watching over a nice house.’
qhawa-sha-ni.
watch-prog-1sg
In (6.20), consultants vary on whether one or both readings are prominent. However,
the consultant who suggested the above example with the first reading also
proposed the following two examples in which sumaq ‘good’ is construed as adjectival
(modifying a noun) and felt that (6.21) and (6.22) have the same meaning.
(6.21) Machu
Machu
Picchu-pi
Picchu-loc
sumaq
good
mikhuna-ta
food-acc
‘At Machu Picchu, we ate good food.’
mikhu-ra-yku.
eat-past-2pl(excl.)
(6.22) Machu Picchu-pi sumaq-ta mikhuna-ta
Machu Picchu-loc good-acc food-acc
‘At Machu Picchu, we ate good food.’
mikhu-ra-yku.
eat-past-2pl(excl.)
One hypothesis which we might entertain at this point is that the adjective
or quantifier is some sort of unselective binder. That is, that this element may

202
be associated to any element within the VP. This would be a way to view both
meanings of sumaq-ta ‘good-ta’ in (6.20) as essentially adverbial, with the difference
correlating with the identity of the bindee. That this is not the case is illustrated in
examples in which an oblique appears within the VP but cannot be associated with
the -ta-marked modifier in the same way as the discontinuous noun phrase reading
of examples like (6.20) associates the adjective sumaq ‘good’ with the noun wasi
‘house’. This is illustrated in (6.23) and (6.24), where the -ta-marked modifier fails
to generate a reading in which it is understood as modifying a locatively-marked
oblique (wasi-pi, ‘in a house’).
(6.23) *Hatun-ta
big-acc
tiya-ni
live-1sg
wasi-pi.
house-loc
(intended: ‘I live in a big house.’)
(6.24) ?Pisi-ta
a.little-acc
tiya-rqa-ni
live-past-1sg
*‘I lived in a few houses.’
wasi-kuna-pi
house-pl-loc
(One consultant did suggest the adverbial reading of pisi-ta: ‘I lived in the
houses a little (for a short time).’)
What (6.23) and (6.24) show is that the -ta-marked adjective must be associated
with the -ta-marked object, and not with just any noun phrase in the predicate.
Recall further that we saw in (6.11) that an adjective such as that in (6.23) also
cannot be locatively co-Case-marked with wasi-pi (‘house-loc’). Thus I conclude
that the -ta marker on the adjective or quantifier is not marking an unselective
binder/adverbial but rather is indeed co-Case-marked with the direct object, and
that this co-Casemarking option does not extend to locative Case.

203
6.3 Interpretation and structure
6.3.1 Previous work
A small body of previous work has considered the syntax of discontinuous noun
phrases in Quechua. In this section I discuss a range of these past approaches.
The approach to what I have been calling discontinuous noun phrases adopted by
many Quechua grammarians is that these are examples of the adverbial construction
mentioned in the previous section. Antonio Cusihuamán provides the following
examples in the course of illustrating adverbials.
(6.25) Hatun-ta-n
big-ta-foc
chakra-ta-qa
field-acc-top
muna-yku.
want-2pl(excl.)
‘We want bigger plots.’ (Cusihuamán 1976/2000 p.128)
Another similar reading suggested by a consultant: ‘We want a big field.’
(6.26) Sumaq-ta-n
good-ta-foc
papa-qa
potato-top
wiña-mu-sha-n.
grow-cis-prog-3sg
‘The potato is growing well.’ (Cusihuamán 1976/2000 p.128)
Here, (6.25) contains what I consider to be a discontinuous noun phrase since the adjective
‘big’ modifies the noun ‘field’, whereas (6.26) I consider to be a true adverbial
construction.
Other approaches to this construction are found in [Lefebvre & Muysken 1988],
[Muysken 1989] and [Sánchez 1996]. Since each of these works presents a different
analysis of discontinuous noun phrases I will briefly discuss each of these in turn.
Each will be shown to answer the question “Is the discontinuous structure achieved
through syntactic extraction?” in a different way.
Lefebvre and Muysken [1988] look at a wide range of phenomena involving co-
Case marking. They posit that instances of discontinuous quantifiers and adjectives
are a result of extraction of these modifying elements from the noun phrase. A

204
sample analysis for an instance of “adjective float” is given in (6.27). According
to this theory there is a Case position in the periphery of the noun phrase which
functions as an escape hatch from the NP, and the floated element picks up its Case
marker in that position.
(6.27) [t i Runa]-ta riqsi-ni hatun i -ta
[t i man]-acc know-1sg tall i -acc
‘I know a tall man.’ (from Lefebvre & Muysken (6) p.143)
In [Sánchez 1996], an extraction analysis of a different sort is posited (in particular
for the case of disjoint adjectives and nouns). Here, the idea is that the modifier
is left behind and the noun phrase raises. Sánchez adopts the idea of Lefebvre and
Muysken that the Spec of the noun phrase is a Case position, where the extracted
element picks up its Case marking.
(6.28) [ FocP Runa i -ta [ Foc riqsi-ni [ AgrP [ DP [t’ i hatun j -ta] [ D ′ [ PredP [t i t j ]]]]]]]
[ FocP Man i -acc [ Foc know-1sg [ AgrP [ DP [t’ i tall j -acc] [ D ′ [ PredP [t i t j ]]]]]]]
‘I know a tall man.’ (from Sánchez 1996 pp128-131)
(In the above example note that the adjective ultimately moves to the [Spec,DP]
position also. However, unlike the noun it never actually leaves DP.)
Thirdly, the analysis in [Muysken 1989] suggests that in a variety of constructions
1 involving a double -ta Case-marker there is no literal extraction of one element
out of another, but rather there is a co-indexation between the NP and
another phrase (“XP”, which could represent a variety of categories) and that this
1 This study involves a rather different range of constructions from the ones I
am considering here, including for example apparent small clauses and apparent
extracted subjects from subordinate clauses. However, it does encompass co-Casemarked
quantifiers, like the ones I discuss in this chapter.

205
co-indexation is what establishes the semantic relationship between the two constituents.
In certain cases, including that of “quantifier float”, Muysken (p.634)
proposes that an empty operator moves from the base position of a quantifier in
the NP to the Comp position. This operator is coindexed with the (external) quantifier
itself, which allows for the quantifier to be interpreted at LF as the element
filling the gap in the NP resulting from the operator movement. This analysis is
presented as part of a theory of “predication chains”, in which co-Case marking is
one method of establishing a predication relationship between the NP and the XP
(rather than, say, a purely structural relationship like C-command). An illustration
of this analysis is shown in (6.29).
(6.29) [ V P ...XP i ...NP i ...]
[ V P [e i qulqi-y]-ta tari-rqa-ni llipin-ta i ]
[ V P [e i money-1sg]-acc find-past-1sg all-acc i ]
‘I found all my money.’ (Muysken 1989 (15a))
To summarize, all three basic syntactic options are represented in the literature:
discontinuous noun phrases have been claimed to be the result of modifier extraction,
noun extraction, and no extraction at all. Before returning to a comparison of
these approaches, I will look more closely at the meaning of discontinuous versus
continuous noun phrases. I will then consider the implications of the semantics for
the three types of analysis presented here. My eventual proposal will be that even
if the modifier does originate within the noun phrase, it is nonetheless interpreted
externally, and not in its base position.

206
6.3.2 Semantic effects of co-Case marking
In this section I will provide data showing that continuous and discontinuous noun
phrases are not identical in meaning. In particular, weak quantifiers and adjectives
in discontinuous constructions yield indefinite meanings of the “complete” noun
phrase. I consider each of the three cases of adjectives, weak quantifiers and strong
quantifiers in turn.
6.3.2.1 Adjectives
When an adjective and a noun appear independently Case-marked, this discontinuous
noun phrase receives an indefinite interpretation. By contrast the continuous
expression can be either indefinite or definite. One context in which this contrast
becomes evident is illustrated in the examples in (6.30). Here, the speaker and I
(the addressee) have previously discussed a particular big house in the speaker’s
village. I subsequently visit her village, and when I return the speaker questions me
about my visit. Under these circumstances my consultants find only (6.30a) to be
appropriate (‘Did you see the big house?), not (6.30b) (‘Did you see a big house?’).
(6.30) (a) [Hatun
big
wasi]-ta
house-acc
‘Did you see a/the big house?’
riku-rqa-nki-chu?
see-past-2sg-Q
(b) #Hatun-ta wasi-ta riku-rqa-nki-chu?
big-acc house-acc see-past-2sg-Q
‘Did you see a big house?’
It is interesting to note that this same paradigm is reflected also in the glosses
of examples such as (6.31) from (Sánchez 1996).
(6.31) (a) [Hatun
big
runa]-ta...
man-acc
‘A/the big man’ (Sánchez 1996 (263) p.129)

207
(b) Runa-ta
man-acc
riqsi-ni
know-1sg
hatun-ta
big-acc
‘I know a big man.’ (Sánchez 1996 (257) p.126)
In (6.31) we again see that the discontinuous noun phrase is interpreted as if it
were a continuous but indefinite noun phrase. 2
6.3.2.2 Weak quantifiers
A co-Case-marked weak quantifier also can provide a strategy for forcing an indefinite
interpretation. Examples comparing a continuous noun phrase with a discontinuous
noun phrase are shown in (6.32(a)) and (6.32(b)). Note that the best
English translation of the discontinuous noun phrase in (6.32(b)) involves the partitive
expression ‘a few of Ana’s llamas’. Again the discontinuous version is necessarily
indefinite. In this example the indefinite reading is not available for the continuous
version ((6.32)(a)), and so the discontinuity is forced if in fact Ana has llamas that
I didn’t see.
2 I should note that some consultants do accept certain examples consisting of a
definite noun phrase and an adjective, each with their own Case marker. However, in
such cases the adjective seems to be interpreted as a depictive secondary predicate,
which I take to be a construction distinct from the cases of restrictive modification
found in my examples of discontinuous noun phrases. Consider (i), which is similar
to (6.31(b)) except a demonstrative is associated with the noun. Also, the verb is
in the past tense, which one consultant proposed to make the only possible reading
more plausible. Note that the adjective is now interpreted as a depictive secondary
predicate.
(i) ?[Chay runa]-ta hatun-ta riqsi-ra-ni.
that man-acc big-acc know-past-1sg
‘I knew that man as a big person.’
Consultant’s comment: Perhaps he is sick now, and no longer big?
We now see that since a secondary predicate reading of ‘big’ is not salient in
the context given above for (6.31(b)), that example is not saved by this alternative
structure. For examples and discussion of ambiguity between discontinuous noun
phrases and depictive secondary predicates in Australian languages see [Schultze-
Berndt & Himmelman 2004].

208
(6.32) (a) [Ana-q pisi llama-n]-ta
Ana-gen a.few llama-3sg-acc
‘I saw Ana’s few llamas.’
*‘I saw a few of Ana’s llamas.’
riku-rqa-ni
see-past-1sg
(b) [Ana-q llama-n]-ta pisi-ta
Ana-gen llama-3sg-acc a.few-acc
‘I saw a few of Ana’s llamas.’
riku-rqa-ni.
see-past-1sg
It is interesting to compare (6.32(b)) with the paraphrase in (6.33). Consultants
feel that these two examples have the same meaning, each indicating that I saw a
small number of llamas among the total (larger) number of llamas owned by Ana.
However, the expression in (6.33) contains an ablative noun phrase as an adjunct
and I assume a null noun (llama) in the object position noun phrase containing pisi
‘a little’.
(6.33) [Ana-q llama-n]-manta pisi-ta
Ana-gen llama-3sg-abl a.few-acc
‘Of Ana’s llamas, I saw a few.’
riku-rqa-ni.
see-past-1sg
Another example of a discontinuous noun phrase with a weak quantifier and
an overt demonstrative is shown in (6.34). 3 Note that the interpretation is again
indefinite.
(6.34) Kinsa-ta [chay manka-kuna]-ta
three-acc that pot-pl-acc
‘I bought three of those pots.’
*‘I bought those three pots.’
ranti-rqa-ni.
buy-past-1sg
3 In some cases of co-Case marking between a weak quantifier and a noun phrase
consisting of demonstrative+noun as in (6.34), some consultants have a strong preference
for an ablative marker on the noun phrase instead of -ta. I do not at this
point know what circumstances lead to such a preference in some examples but not
others.

209
6.3.2.3 Strong quantifiers
In the previous two sections we have seen that in the cases of weak quantifiers
and adjectives, discontinuous noun phrases force indefinite readings. However, it
is not simply always the case that co-Case marking can be associated with the
indefiniteness of the (complete) DP in question. Consider the following examples
in which a strong quantifier tukuy ‘all’ appears in noun phrase-internal position (in
(6.35(a))) and then in a disjoint position (in (6.35(b))). I am not aware of any
semantic differences between these two sentences.
(6.35) (a) Nuqa [Gabriela-q tukuy aqha-n]-ta
I Gabriela-gen all cornbeer-3sg-acc
‘I drank all of Gabriela’s cornbeer.’
ukya-ra-ni.
drink-past-1sg
(b) Nuqa [Gabriela-q aqha-n]-ta
I Gabriela-gen cornbeer-3sg-acc
‘I drink all of Gabriela’s cornbeer.’
tukuy-ta
all-acc
ukya-ra-ni.
drink-past-1sg
A similar pair is shown in (6.36(a)) and (6.36(b)).
(6.36) (a) [Tukuy manzana]-ta
all apple-acc
‘I ate all the apples.’
miku-ra-ni.
eat-past-1sg
(b) Tukuy-ta manzana-ta
all-acc apple-acc
‘I ate all the apples.’
miku-ra-ni.
eat-past-1sg
6.3.3 Addressing the indefiniteness of discontinuous noun
phrases
Before giving a meaning-oriented analysis of the LF structure of discontinuous noun
phrases, in this section I present briefly and reject two candidate explanations for
the semantic effects just discussed.

210
Sánchez [1996] suggests that when adjectives are “stranded” the noun phrase
moves out of the DP to a focus position above TP. Hence she suggests that in
sentence (6.37), there is focus on runa ‘man’.
(6.37) Runa-ta
man-acc
riqsi-ni
know-1sg
hatun-ta
big-acc
‘I know a big MAN.’ (Sánchez 1996 p.129 (265))
However, in general I am unable to find evidence that the noun is necessarily
focused, and indeed a different element in the sentence may just as easily appear
with a focus marker, as seen in (6.38).
(6.38) Nuqa-n
I-foc/evid
pisi-lla-ta
a.few-delim-acc
michi-mu-na-y-paq
herd–nm-1sg-purpose
‘I want a few llamas so I can herd them.’
llama-ta
llama-acc
muna-ni
want-1sg
I conclude that while discontinuity may facilitate focus, since it permits what are
normally subconstituents of the DP to be independently focus marked within the
sentence, there is no direct dependence between focus and discontinuous noun
phrases.
Another possibility which is suggested by the kind of data seen in 3.2.1 and 3.2.2
when viewed as a restriction on extraction is that what we have here is an instance
of the cross-linguistic Specificity Effect. This phenomenon has been studied in a
variety of languages, for instance in Germanic (e.g. [Diesing 1992]) and Turkish
(e.g. [Kornfilt 2002]). Various explanations have been offered including the idea
that it is a kind of island effect: that specific/definite noun phrases have an extra
layer of structure (possibly DP) which non-specific indefinites do not have (these
may be bare NPs) (see e.g. [Bowers 1990]). The basic generalization is that no
extraction is allowed from specific DPs.

211
Of course, we must have some way to recognize specific DPs in order to see if this
is the relevant constraint in Quechua. In [Enç 1990] it is proposed that specificity
should be understood in terms of membership in a contextually prominent group
(as opposed to definiteness, which requires that the particular referent of a definite
noun phrase be contextually prominent). In [Diesing 1992] this specificity criterion
is analyzed in terms of presuppositionality. An example provided by Enç of a specific
noun phrase in Turkish is ‘two girls’ in (6.39).
(6.39) Odam-a birkaç çocuk girdi. Iki kiz-i taniyordum.
my-room-dat several child entered two girl-acc I-knew
‘Several children entered my room. I knew two girls.’ (Enç 1990 (16,17))
In (6.39), the presence of the accusative marker makes ‘two girls’ unambiguously
specific, and hence the two girls are included in the contextually prominent set of
children who entered the room. Without the accusative marker this noun phrase is
non-specific and hence the two girls may not be members of the original set.
If we look at a similar example in Quechua we find that at least under this
definition of specificity a discontinuous noun phrase may be specific, as illustrated
in (6.40). Here consultants report that the discontinuous ‘two girls’ may be members
of the group who arrived or not.
(6.40) Ashka
A.lot
irqi-kuna chaya-mu-rqa-nku.
child-pl arrive-cis-past-3pl.
riqsi-ra-ni.
know-past-1sg
‘A lot of children arrived. I knew two girls.’
Iskay-ta
two-acc
warmi warmacha-ta
girl-acc
Now if the semantic differences between continuous and discontinuous noun
phrases in Quechua could be explained via the Specificity Effect we would expect the
two girls in (6.40) to be outside of the original group. Thus we cannot immediately
attribute the indefiniteness of the co-Case-marked noun phrases to a restriction on
extraction from specific noun phrases.

212
Also striking in this regard are the examples of apparent partitives as seen above
in (6.32) and the case of the co-Case marking of strong quantifiers as in (6.35).
These examples cannot be assimilated to the view of a Specificity Effect outlined
above, since in both (a) examples the noun phrase with the quantifier internal to it
seems to be definite.
In the following section I will therefore pursue a different line of reasoning in order
to explain the indefiniteness of certain discontinuous noun phrases. In contrast
to an analysis in terms of the Specificity Effect, where essentially this indefiniteness
is attributed to a condition on extraction, my analysis below will be that the
indefiniteness arises as a result of the external interpretation of the modifier.
6.4 Co-Case marking indicates scope outside the
DP
6.4.1 The basic interpretive structure
Recall that one of the basic questions surrounding the interpretation of discontinuous
noun phrases was whether these noun phrases are interpreted as a single unit at LF.
The semantic differences between the discontinuous and continuous versions of the
same noun phrase indicate that even if some sort of constituent reconstruction takes
place, it cannot be the case that the LF structures of the two versions are identical.
However, the lack of ambiguity of the association between the adjective or quantifier
and the noun, and the similarity in meaning because the two surface versions of the
noun phrase do suggest that the two parts are interpreted as a unit. To solve
this problem I posit an interpretive structure in which the quantifier or adjective
(labeled Mod) is located outside the scope of the definiteness head D of the noun
phrase at LF. This solution is illustrated schematically in (6.41). I return below to

213
the question of whether the Modifier is in a DP-internal position ([Spec,DP]) or a
position immediately dominating the DP.
(6.41)
VP
✟ ✟✟ ❍
❍
❍
V
✟ ✟✟ ❍ ❍ ❍
Mod-ta DP
✟ ✟❍
❍
D NP
✟ ✟❍ ❍
Noun-ta
This structure is based on the notion that an external modifier is interpreted
externally to the noun phrase. I claim that it captures not only the empirical fact
that the modifier appears outside of the noun phrase at surface structure, but also
the indefiniteness effect described in Section 3. In the next sections I show how this
structure correctly predicts the interpretations noted for discontinuous noun phrases
containing adjectives, weak quantifiers and strong quantifiers. I then return to some
syntactic issues in this analysis.
6.4.2 LF structures
6.4.2.1 Adjectives
I assume that adjectives are modifiers of type . When an adjective appears
outside of the head responsible for definiteness, the resulting interpretation of course
depends on the type of the sister to the adjective. Consider first the case of an
indefinite DP. I represent the indefiniteness by a [-definite] feature on the head D
(though a reasonable alternative would be to assume that no D is present at all in
this case, and we have simply a bare NP). I assume that an indefinite DP is also
of type (and presumably can be eventually subject to existential closure).
Thus I represent the indefinite D head as semantically vacuous. The interpretive

214
configuration of a discontinuous noun phrase of this sort is shown in (6.42(a)). In
(6.42(b)) I show the disallowed structure in which the adjective modifies a definite
DP. Intuitively, the problem here is that the adjective cannot further modify a
constituent of type e. This is expressed formally by the fact that the result of such
a combination could only be of type t, and hence this constituent could not function
as the (nominal) direct object of the verb (‘see’).
(6.42) (a) VP
❍
✟ ✟✟✟ ❍
❍
❍
V
✟ ❍❍
✟ ✟✟✟ ❍
❍
❍
hatun-ta
tall-acc
✟ ✟✟
D(indef)
DP
❍
(b) * VP
✟ ❍
✟ ✟✟ ❍
❍
❍
V
❍
❍
NP
wasi-ta
house-acc
‘(I saw) a big house.’ (from (6.30))
❍ ✟ ✟✟✟✟ ❍
❍ ❍
DP
e
✟ ✟✟ ✟ ❍❍
D(def)
hatun-ta
tall-acc
❍
❍
NP
wasi-ta
house-acc
* ‘(I saw) the big house.’
6.4.2.2 Weak quantifiers
We now turn to the case of weak quantifiers, which behave similarly to the adjectives
but in fact allow more structural options due to what I take to be their more flexible
types. Again I start with the case in which the determiner is indefinite, and consider
the structure in (6.43(a)) in which the weak quantifier is interpreted outside of the
definiteness head.

215
(6.43)(a) VP
✟ ❍❍
✟ ✟✟✟ ❍
❍
❍
V
❍ ✟ ✟✟✟✟ ❍
❍ ❍
pisi-ta DP
a.few-acc
✟ ✟✟ ❍
D(indef)
‘(I saw) a few llamas.’
(b) VP
❍
✟ ✟✟✟ ❍
❍
❍
V
❍
✟ ✟✟✟
pisi-ta
a.few-acc
❍
❍
NP
llama-ta
llama-acc
❍
❍
❍
DP
e
✟ ✟✟ ✟ ❍❍
❍
❍
NP
✏
✏ ✏✏✏✏ Ana-q llama-n-ta
Ana‘s llama-3sg-acc
D(def)
‘(I saw) a few of Anas llamas’. (from 6.32(b))
As in the case of adjectives, the interpretation here is the one expected in the
case of a DP-internal weak quantifier too, again because of the semantic vacuousness
of the indefinite D. However, I propose to capture the differing behavior of
weak quantifiers and adjectives by assuming that the apparently partitive readings
allowed in the case of weak quantifier discontinuous noun phrases are a result of the
composition of the weak quantifier with the definite noun phrase as illustrated in
(6.43(b)).
Therefore I propose that a weak quantifier can combine directly with a type e
DP and the result is interpreted as an indefinite, partitive expression. In fact, an
analysis in which quantifiers regularly and perhaps universally take complements of
type e can be found in [Matthewson 2001]. Here I adopt a slightly more flexible
approach in that I presume that this is only one option for weak quantifiers in
Quechua, and still assume that quantifiers can also combine directly with a type
nominal as we saw above in (6.43(a)).
6.4.2.3 Strong quantifiers
Finally we turn to the case of strong quantifiers. Recall that strong quantifiers
did not obey the generalization that seemed to prohibit discontinuous noun phrases

216
from encoding definite DPs. In fact, strong quantifiers are perfectly able to appear
separated from the noun they are associated to. To explain this fact under the
current analysis, we may simply posit the configuration in (6.44) for the LF structure
of a discontinuous noun phrase with the strong quantifier llipin ‘every’.
(6.44)
VP
✟ ❍❍
✟ ✟✟✟ ❍
❍
❍
V
❍ ✟ ✟✟✟✟ ❍
❍ ❍
llipin-ta DP
every-acc
✟ ✟ ✟ ❍
D(indef)
❍
❍
NP
manzana-ta
apple-acc
‘(I ate) every apple.’
Note that in (6.44) I represent llipin ‘every’ as combining directly with an indefinite
DP and the overall unit is interpreted as ‘every apple’. It is quite possible,
however, that just as in the case of weak quantifiers, a strong quantifier can also take
a definite DP complement. The syntax of (6.44) will be discussed in more detail
below.
6.4.3 Revisiting the syntactic options
In the preceding section I suggested that the semantic differences between continuous
and discontinuous noun phrases can be explained by positing an LF structure in
which the modifier appears outside of the definiteness head of the DP. In this section
I consider the syntactic issues raised by this configuration, particularly in light of the
three syntactic analyses previously discussed. These three analyses are summarized
again in (6.45).

217
(6.45) (a) Modifier stranding (b) Modifier floating (c) Predication chain
[Sánchez 1996] [Lefebvre & Muysken 1988] [Muysken 1989]
FocP
✟ ✟ ❍❍ ❍
man i .
DP
✟ ❍
t’ i D’
✏✏
big t i
S
✟ ✟ ❍❍ ❍
big j .
DP
✟ ✟❍
❍
t’ j D’
✏ ✏
t j man
S
✟ ✟✟ ❍ ❍
VP
✏ ✏✏✏
money k ...all k
Among these three, my analysis is most immediately incompatible with (6.45(a)),
modifier stranding. Omitting details, the basic idea is as shown: the NP moves out
of the DP, leaving the modifier stranded. The problem for me here is that the
only way for the adjective to have scope over the noun at LF, which is part of
what I am arguing, is via reconstruction of ‘man i ’ to base position. This may very
well be what Sánchez has in mind, but of course this solution predicts that (aside
from the proposed focus effect, which I discussed in Section 3), the continuous
and discontinuous noun phrases should have the same meaning. The indefiniteness
restriction would not be captured.
There is another piece of evidence against (6.45(a)), which is that although the
basic word order in a noun phrase is numeral+adjective+noun, as in (6.46) it is not
possible for the numeral+adjective combination to appear alone in the discontinuous
version proposed in (6.47). 4 This seems surprising if the noun phrase containing just
4 The question of whether the number of pieces in the discontinuous noun phrase
is necessarily limited to two still needs to be explored. [Muysken 1989] took it that
predication chains (signalled by two -ta markers) were limited to two elements, and
my analysis here assumes too that this limitation is correct. However, sometimes
my consultants would accept examples like (i). At this point the status of such
examples is not clear to me.
(i) ?Kinsa-ta hatun-ta wasi-ta riku-rani.
three-acc big-acc house-acc see-past-1sg
’I saw three big houses.’

218
wasi ‘house’ can be extracted. On the other hand, the combination adjective+noun
can appear alone as in (6.48), with the expected indefinite meaning.
(6.46) [Kinsa
three
hatun
big
wasi]-ta
house-acc
‘I saw three big houses.’
riku-rqa-ni.
see-past-1sg
(6.47) *Wasi-ta [kinsa hatun]-ta riku-rqa-ni.
house-acc three big-acc see-past-1sg
Intended meaning: ‘I saw three big houses.’
(6.48) Kinsa-ta riku-rqa-ni
three-acc see-past-1sg
‘I saw three big houses.’
[hatun
big
wasi]-ta.
house-acc
A similar paradigm is illustrated in (6.49) to (6.51) in the case in which the
complete noun phrase contains a possessor and a numeral.
(6.49) Juan-pa iskay libru-n-ta
Juan-gen two book-3sg-acc
‘I read Juan’s two books.’
lii-rqa-ni.
read-past-1sg
(6.50) *Juan-pa iskay-ta libru-n-ta
Juan-gen two-acc book-3sg-acc
‘I read Juan’s two books.’
lii-rqa-ni.
read-past-1sg
(6.51) Juan-pa-ta iskay libru-n-ta
Juan-gen-acc two book-3sg-acc
‘I read two books of Juan’s.’
lii-rqa-ni.
read-past-1sg
It would be possible to view the ungrammaticality of (6.50) and (6.51) as due
to a restriction on the size of the constituent which is raising (under the modifierstranding
analysis). So, perhaps these examples are ruled out because the constituent
containing the single noun ‘house’ or ‘book’ is not large enough to be extracted
on its own. This analysis would raise the question of why a single noun can
be extracted in cases with fewer modifiers present, such as (6.52).

219
(6.52) Iskay-ta libru-ta
two-acc book-acc
‘I read two books.’
lii-rqa-ni.
read-past-1sg
For all these reasons I do not find the modifier stranding analysis to be compatible
with the Quechua data. Turning now to the structures sketched in (6.45(b)) and
(6.45(c)), I believe that my analysis is potentially compatible with either of these
syntactic options. I will comment briefly on each.
The LF structure I have proposed is compatible with a structure in which the
modifier and noun are generated as independent constituents and become associated
through some sort of coindexation as in (6.45(c)). This analysis raises a number of
theoretical questions, however. Assuming that accusative Case marking is associated
with a particular structural configuration in association with a Case assigner (Tr), it
seems that the two -ta-marked elements would still have to be in a local configuration
at some earlier syntactic level, as in (6.53). Thus the LF configuration would be
presumably a post-scrambling reconstruction to this basic configuration. Again the
question of how the independent generation of a modifier phrase and a noun phrase
in the scope of the Case assigner comes about remains open.
(6.53)
S
✟ ✟ ❍❍ ❍
TrP
✟ ✟✟ ❍ ❍
Tr VP
✟ ✟✟ ❍
❍ ❍
V
✟ ✟ ❍❍ ❍
Mod-ta DP-ta
The option in (6.45(b)), in which the modifier leaves the noun phrase (possibly
picking up Case marking in an escape hatch as proposed in [Lefebvre & Muysken
1988] and supported by [Sánchez 1996]), could also be adopted in conjunction with

220
the analysis I have outlined here. However, the modifier could not be construed as
reconstructing to its base position. Clearly, if the role of the base position of the
modifier is to provide the position at which the moved element is to be interpreted
then there is no expected interpretation difference between the continuous and discontinuous
versions of the noun phrase. (The possibility that there is an island effect
associated with specific noun phrases here has already been discussed and rejected.)
However, it could be that reconstruction takes place to the [Spec,KP] escape hatch
as proposed in Chapter 5, which would give us the exact LF configuration I have
posited. The question remains, however, of why reconstruction would only be to
this intermediary position.
One possible resolution of this issue is that (weak) quantifiers and adjectives can
be generated in a position above DP, from which they are extracted and to which
they are reconstructed at LF. In fact, from the discussion in Chapter 5, we have a
position of just this sort: the [Spec,KP] position was proposed to be the position from
which possessors could be extracted and in which the Case marking from the matrix
clause environment could be copied onto a DP-internal element. If modifiers can in
fact merge initially with KP instead of appearing in their DP-internal position, we
have an explanation both for the Case-marking on the modifier and the obligatory
semantic association of the modifier with the noun phrase it dominates. If adjectives
and quantifiers do not have their own Case feature, then their Case-markers would
reflect the Case-copying property of the [Spec,KP] position and not a feature match
with K. In this way these modifiers would differ from possessors, which actually
have their Case checked by K (cf. Section 5.4.3). Thus we have to assume that K
does not always come with this Case feature.
A perplexing question that still needs to be dealt with is how to reconcile the
view of a strong quantifier as an element that may be extracted at will from its noun

221
phrase, with my analysis which began in Chapter 2 that in fact strong quantifiers
are determiner heads of DPs. The problem here is that the “escape hatch” would
have to be simultaneously open to full phrases and to heads. One way to resolve
this problem is to assume that in the case of a strong quantifier, we have a stacking
of DPs, just as we saw overtly in (6.54).
(6.54) Llipin
every
chay
that
manzana
apple
‘Every one of those apples.’
Given this possibility, I would like to suggest that the entire inner DP can be
raised to [Spec,KP] and hence be co-Casemarked with the remaining element of
the DP, which in this case consists only of the remaining determiner. (The phrase
that moves will have to have a feature matching a feature of K in order to avoid a
MLC violation.) Under these circumstances the extracted DP interpreted in its base
position would give us exactly the interpretive configuration we need: the strong
quantifier would be outside of the determiner head of the reconstructed DP. This
possibility is illustrated in (6.55).
(6.55)
KP
❍
✟ ✟✟✟✟
❍
❍
❍
❍
DP i
K’
✟ ✟ ❍❍ ❍
✟
D NP
✟ ✟ ❍
❍
❍
K DP
✟
manzana-ta
✟ ❍ ❍
D
apple-acc
llipin-ta
every-acc
t i
The environments in which KP can be generated (and in which the Case-copying
operation can occur) need to be clarified. At the beginning of this chapter I mentioned
that discontinuous noun phrases of the type under study here are not allowed

222
in adjunct positions at all, and are often judged as degraded or ungrammatical in
subject position. If indeed the correct generalization is that only the -ta position
allows the DP-extraction operation it seems that this would be due to the probe Tr
(but not other Case-assigning heads) having a [+K] feature.
To summarize, the question of whether movement has occured–that is, whether
there is a trace within the DP–is addressed in my analysis with the idea that there is
no semantic gap within the DP (no position to which reconstruction takes place), but
there still may be a syntactic gap. Further, this apparent paradox can be resolved if
we assume that weak quantifiers and adjectives can be generated in [Spec,KP], the
position of co-Casemarking and ultimately of LF reconstruction.
6.5 Why sapa ‘each’ won’t behave
In this section I discuss a strong quantifier which does not seem to fit into the
pattern established in the previous sections, and offer an explanation.
6.5.1 The problem with sapa ‘each’
The basic problem is that sapa ‘each’ cannot be co-Case-marked like other universal
quantifiers. This is illustrated in the contrast between (6.56(a)) in which sapa
appears in a continuous noun phrase and (6.56(b)) in which it is disallowed as part
of a discontinuous noun phrase.
(6.56) (a) [Sapa/llapan [Inka-kuna-manta
each/every Inka-pl-abl
qu-wa-ra-nku.
give-1sg-past-3pl
rima-q
talk-nm
libru]]-ta
book-acc
‘They gave me each/every book (that talked) about the Inkas.’

223
(b) [Inka-kuna-manta
Inka-pl-abl
qu-wa-ra-nku.
give-1sg-past-3pl
rima-q
talk-nm
libru]-ta
book-acc
*sapa-ta/llapan-ta
each-acc/every-acc
‘They gave me *each/every book (that talked) about the Inkas.’
6.5.2 Two different lexical items?
To address the basic fact that sapa cannot appear in a discontinuous configuration,
we need to look more closely into the meaning of sapa. It appears in fact that there
are at least two uses of sapa relevant to the current discussion. 5 The first use (sapa 1 )
is as a D-quantifier meaning ‘each’, as shown in (6.56) above and also in (6.57) and
(6.58).
(6.57) Navidad-pi
Christmas-loc
[sapa 1
each
irqi]-man
child-dat
t’anta
bread
qu-y-qa,
give-inf-top
costumbre-y-mi
custom-1sg-evid
‘To give bread to each child at Christmas is my custom.’
nuqa-q
I-gen
(6.58) [Sapa 1
each
irqi]
child
urqu-man
mountain-dat
wicha-nqa.
climb-3sg/fut
‘Each child will climb the mountain.’
The second use, sapa 2 is as a quantifier over adverbial clauses. This version of
sapa can appear immediately before a clause whose verb is marked with one of the
adverbializing suffixes -spa (main and subordinate clause subjects are the same) or
-qti (main and subordinate clause subjects are different). Examples are shown in
(6.59), (6.60) and (6.61). Note that other universal quantifiers are not acceptable
in this pre-adverbial position, as shown in (6.59).
5 In fact there are also a variety of fascinating adverbial uses of sapa bearing
agreement morphology and discussed in [Muysken 1994] and [Sánchez 1996] which
I leave aside here.

224
(6.59) Sapa 2 /*llapan/*tukuy [pro rima-ri-qti-y],
each/every/all speak-inch-bi.adv-1sg
rima-n.
speak-3sg
‘Each time I talk, the parrot talks too.’
loro-pas
parrot-also
(6.60) ...chay lambe mayordomokunawan maqanakunay kaq
I had to fight with those wardens
sapa 2 [pro llant’a-ta
each firewood-acc
muna-wa-qti-nku].
want-1sg-bi.adv-3pl
otaq
or
q’ara-ta
cowdung-acc
qechu-y-ta
take-inf-acc
(each time) when they wanted to take my firewood or cowdung.’ (GCM p.27)
(6.61) Sapa 2
each
[pro galeta-ta
cracker-acc
mikhu-qti-y],
eat-bi.adv-1sg
loro
parrot
phawa-mu-n.
run-cis-3sg
‘Each time that I eat a cracker, the parrot comes running.’
6.5.3 Role of the suffix -nka
It is important to note at this point that the two versions of sapa do not have the
same status with all consultants. While sapa 2 is acceptable to all my consultants and
can be found fairly frequently in the narrative of Gregorio Condori Mamani, sapa 1
is unacceptable or marginal to some consultants and I have not found examples of
it in the narrative. However, sapa 1 can be “saved” by adding the suffix -nka, 6 No
-nka is allowed on instances of sapa 2 , however, as illustrated in (6.63).
6 I do not currently know how this use of -nka is related to the distributive use
of this suffix, discussed in 2.5.3. The two uses can be distinguished by noticing
that distributive -nka only appears with weak quantifiers, not strong quantifiers,
and as such it marks the distributive share (the quantity or product that is being
distributed). On the other hand the -nka under discusion here appears on sapa as
part of the distributive key (the recipients of the distribution), but it cannot other
strong quantifiers and furthermore seems not to affect the truth conditions of the
sentence.

225
(6.62) sapa 1 -?(nka)/llapa-(*nka)/tukuy-(*ni-nka)
each-nka/every-nka/all-euph-nka
‘Each/every/all llama(s)’
llama
llama
(6.63) Sapa 2 -(*nka)
each-nka
rimari-qti-y,
talk-vi.adv-1sg
loro-pas
parrot-also
‘Each time I talk, the parrot talks too.’
rima-n.
talk-3sg
Furthermore, a very common (and generally accepted) use of sapa is as a quantifier
over time and (rarely) space adverbials. In these cases, too, sapa-nka is disallowed,
and in this sense these uses can be assimilated with sapa 2 . Examples are
shown in (6.64), (6.65) and (6.66).
(6.64) [sapa-(*nka)
each-nka
‘every day’
p’unchay]
day
(6.65) [sapa
each
tuta]
night
huñuna-ku-q
gather-refl-past(habitual)
ka-yku
be-2pl(excl.)
‘Each night we would gather together...’ (GCM p.53)
(6.66) Kay mulaypa sutinmi Renunciable karan, payllawanmi maytapas purinay
kaq...
‘That mule of mine’s name was Renounceable, and with her I walked
everywhere...
[sapa
each
legua]-pi
league-loc
sama-spa,....
rest-uni.adv
resting every league...’ (GCM p.41)
6.5.4 Proposal for sapa
The collection of facts outlined thus far suggests that the basic use of sapa is as a
quantifier over adverbials. Not only is this a form that is acceptable to all consul-

226
tants, but the intervening suffix -nka is prohibited in this construction. Thus the
denotation of sapa would be as in (6.67).
(6.67) [[sapa]]= [λP. [λQ. ∀e P(e) → Q(e)]]
Here P(e) means that the proposition P is true of event e. Notice that only one
event is named in this denotation. Leaving aside details, I am adopting here what is
essentially Rothstein’s [1995] analysis of ‘every’, as in: ‘Every time I eat a cracker,
the parrot comes running’. Briefly, under this analysis Q in this example would
represent not the proposition that the parrot came running but the proposition that
there is an event e’ which is related to e by a certain matching function (M) and e’
is an event of the parrot running. For details the reader is referred to Rothstein’s
original exposition. Under this analysis, the denotation in (6.67) would yield the
following interpretation for sentence (6.61).
(6.68) [λP. [λQ. ∀e P(e)→ Q(e)] ]([[I eat a cracker]])([[The parrot comes running]])
=true iff ∀e, [I-eat-cracker(e)] → ∃ e’, [Parrot-running(e’)] and M(e’)=e
The role of the matching function can be seen clearly in this example: without
it, (6.61) would be true if I ate cracker after cracker and the parrot only happened
to come running once, perhaps because the cat had chased it.
In the case of pre-time-adverbial sapa we have a similar analysis, with an event
reading of the adverbial p’unchay ‘day’ in (6.69), for example. The interpretation
is given in (6.70).
(6.69) Sapa
each
p’unchay
day
llank’a-ni.
work-1sg
‘Each day I work.’
(6.70) [λP. [λQ. ∀e P(e)→ Q(e)] ]([[day]])([[I work]])=true iff ∀e, [day(e)→ ∃ e’,
I-work(e’)] ∧ M(e’)=e

227
We are now ready to consider the role of -nka in sapa-nka. I suggest that
these sapa-affixed versions of -nka are syntactic heads which mediate between the
adverb-binding sapa and the noun which follows. The relevant structure is shown
in (6.71(a)).
(6.71) (a) DP
✟ ✟✟ ❍
❍
sapa
✟
each
✟❍
❍
nka NP
llama
(b)
DP
✟ ✟ ❍ ❍
sapa
each
✟✟ ❍❍
∅ NP
llama
When consultants accept a D-quantifier use of sapa I suggest a null version of nka
as shown in (6.71(b)). These suffixes thus function as type-shifters which effectively
convert a noun phrase into an adverbial.
6.5.5 So why can’t sapa “float”?
We are now ready to answer the original question posed in this section: why does
sapa not behave like other strong quantifiers in appearing in discontinuous noun
phrases? I propose that the adverbial cannot be accusative-marked independently.
Since adverbials cannot be Case-marked, 7 a Case-marked noun with disjoint sapa
must be understood as a pure DP as in (6.72(a)). However, in this case the resulting
LF configuration would be as in (6.72(b)).
(6.72) (a) S
(b)*LF: S
✟ ✟ ✟ ❍❍
❍
❍
✟ ✟✟ ✟❍
❍
❍
Subject VP
Subject VP
✟ ✟✟ ❍
❍ ❍
✟ ✟✟ ❍
❍
❍
V
V
✏ ✏✏✏ ✏
✟ ✟✟ ❍ ❍
sapa-ta DP
sapa-ta...DP-ta
llama-ta
7 Recall from Section 6.2.2 that although -ta can be used to convert adjectives into
adverbs, many adverbs are not Case-marked and I thus consider this adverbializer
to be a derivational affix.

228
Without the intermediary level, at LF sapa ends up with scope immediately over
the DP, which gives us a type incompatibility (sapa combines directly with llama).
6.6 Conclusion
This chapter set out to address the problem of how discontinuous noun phrases
are interpreted in Quechua. The main data arguing for a structure different from a
simple reconstruction of external modifiers to “base” position within the DP involved
cases in which discontinuous and continuous noun phrases had different meanings.
The generalization was that when adjectives or weak quantifiers appeared outside
of their noun phrase, in a co-Case-marked position, the overall noun phrase received
an indefinite interpretation. Strong quantifiers, however, seemed to have the same
meaning in both positions. One apparently exceptional strong quantifier sapa ‘each’
was shown to fall outside of this basic pattern because its principal use is as a
quantifier over adverbials and not directly over nouns.
I argued that the meanings of continuous and discontinuous noun phrases could
be explained by positing an interpretive configuration in which the adjective or
quantifier appears outside of the determiner head, which I take to be the head
responsible for definiteness. In this sense, I have claimed that the surface position
of the modifying element determines its interpretive scope.

Bibliography
Abney, S. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation,
MIT.
Abusch, D. 1994. The scope of indefinites, Natural Language Semantics. 2:83-135.
Adelaar, W. 1977. Tarma Quechua, Leiden: Brill.
Bach, E. & R. Cooper. 1978. A note on the interpretation of adjoined relative
clauses. Linguistics and Philosophy, 5:473-482.
Bach, E. et al. 1995. Introduction to Bach et al. (eds.), Quantification in Natural
Languages. Dordrecht: Kluwer.
Barker, C. 1995. Possessive Descriptions. Stanford: CSLI Publications.
Barwise, J. and R. Cooper. 1981. Generalized quantifiers and natural language.
Linguistics and Philosophy, 4.159-219.
Basilico, D. 1996. Head position and internally headed relative clause. Language,
72.3 498-532.
Belletti, A. 1988. The Case of unaccusatives. Linguistics Inquiry, 19,1.
Bianchi, V. 1995. Consequences of Antisymmetry for the Syntax of Headed Relative
Clauses. Doctoral Dissertation, Pisa: Scuola Normale Superior.
Bianchi, V. 1999. Consequences of Antisymmetry: Headed Relative Clauses. Berlin:
Mouton de Gruyter.
Bianchi, V. 2000. The raising analysis of relative clauses: a reply to Borsley. Linguistic
Inquiry, 31.1.123-140.
Bianchi, V. 2002. Some issues in the syntax of relative determiners, in A. Alexiadou
et al. (Eds.) The Syntax of Relative Clauses. Amsterdam: John Benjamin.
229

230
Bittner, M. and K. Hale. 1995. Remarks on definiteness in Warlpiri. In Bach et al.
(eds.), Quantification in Natural Languages. Dordrecht: Kluwer.
Borsley, R. 1997. Relative clauses and the theory of phrase structure. Linguistic
Inquiry, 28.4.629-647.
Bowers, J. 1993. The syntax of predication, Linguistic Inquiry, 24:4.591-656.
Bowers, J. 1990. The syntax and semantics of nominals. Semantics and Linguistic
Theory, 1.1-31.
Bowers, J. 2002. Transitivity, Linguistic Inquiry, 33:2.183-224.
Cerrón-Palomino, R. 1987. Lingüística Quechua. Cuzco: Bartolomé de las Casas.
Choe, J. 1987. Anti-Quantifiers and a Theory of Distributivity. PhD dissertation,
University of Massachusetts, Amherst.
Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, N. 2000. Minimalist inquiries. In R. Martin et al. (eds.), Step By Step.
Cambridge: MIT Press.
Chomsky, N. 2001a. Derivation by phase. Ken Hale: a Life in Language, ed. Michael
Kenstowicz, 1-52. Cambridge, MA: MIT Press.
Chomsky, N. 2001b. Beyond explanatory adequacy. MIT Occasional Papers in
Linguistics 20..
Cole, P. 1985. Imbabura Quechua. Kent: Croom Helm.
Cole, P. 1987a. The structure of internally headed relative clauses. Natural Language
and Linguistic Theory, 5.277-302.
Cole, P. 1987b. Null objects in universal grammar. Linguistic Inquiry, 18.4: 597-
612.
Cole, P. & G. Hermon. 1981. Subjecthood and islandhood: evidence from Quechua,
Linguistic Inquiry, 12:1.1-30.
Cole, P. & G. Hermon. 1994. Is there LF wh-movement? Linguistic Inquiry,
25:2.239-262.
Cole, P., G. Hermon & W. Harbert. 1982. Headless relative clauses in Quechua,
IJAL 48, 113-125.

231
Collins, C. 2001. The Internal structure of verbs in Ju|’hoan and |=Hoan. In A. Bell
& P. Washburn, eds. Cornell Working Papers in Linguistics, V. 18: Khoisan.
Collins, C. 2003. The internal structure of vP in Ju|’hoansi and |=Hoan, Studia
Linguistica, 57, 1:1-25.
Culy, C. 1990. The Syntax and Semantics of Internally Headed Relative Clauses.
PhD dissertation, Stanford.
Cusihuamán, A. 1976a (2000). Diccionario Quechua Cuzco-Collao. IEP: Lima,
Peru.
Cusihuamán, A. 1976 (2000). Gramática del quechua de Cuzco. IEP: Lima, Peru.
Diesing, M. 1992. Indefinites. Cambridge: MIT Press.
Elbourne, P. 2003. Situations and Individuals. PhD dissertation, MIT.
Enç, M. 1990. The semantics of specificity, Linguistic Inquiry, 22:1.1-25.
Faller, M. 2001. The Quechua distributive suffix -nka, In A. Werle & S-Y Kim
(Eds.) Proceedings of SULA, UMOP 25. Amherst: GLSA Publications.
Faller, M. 2002. Semantics and Pragmatics of Evidentials in Cuzco Quechua. PhD
dissertation, Stanford University.
Fernández-Soriano, O. 1999. Two types of impersonal sentences in Spanish: locative
and dative subjects, Syntax, 2.2, 101-140.
Freeze, R. 1992. Existentials and other locatives. Language 68:553-595.
Gorbet, L.P. 1976. A Grammar of Diegueño Nominals. New York: Garland.
Grosu, A. 2000. Type-resolution in relative constructions. In A. Alexiadou et al.
(Eds.) The Syntax of Relative Clauses. Amsterdam: John Benjamin.
Grosu, A. 2002. Strange relatives at the interface of two millennia. GLOT International,
6:6.145-167.
Grosu, A. & F. Landman. 1998. Strange relatives of the third kind. Natural
Language Semantics, 6:125-170.
Hale, K. 1983. Warlpiri and the grammar of non-configurational languages, Natural
Languages and Linguistc Theory 1:5-48.

232
Hastings, R. 2001. The interpretation of Cuzco Quechua relative clauses. In Kim
& Werle (eds). Proceedings of SULA. U. Massachusetts Working Papers in
Linguistics. GLSA Publications, Amherst, MA.
Hastings, R. 2003. The semantics of discontinuous noun phrases in Quechua. In J.
Anderssen, P. Menéndez-Benito, and A. Werle (Eds.), Proceedings of SULA
2, UMWPL, GLSA Publications, Amherst, MA, pp 35-56.
Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD dissertation,
University of Massachusetts, Amherst.
Heim, I. & A. Kratzer. Semantics in Generative Grammar, Oxford: Blackwell.
Hiraiwa, K. 2001. Multiple Agree and the defective Intervention Constraint in
Japanese. MIT Working paper in Linguistics 40, 67-80.
de Hoop, H. 1995. On the characterization of the weak-strong distinction. In Bach
et al. (eds.) Quantification in Natural Languages. Dordrecht: Kluwer.
Hoji, H. 1985. Logical form constraints and configurational structures in Japanese.
Doctoral dissertation, University of Washington, Seattle.
Hoshi, K. 1995. Structural and Interpretive Aspects of Head-Internal and Head-
External Relative Clauses. Ph.D. dissertation, University of Rochester.
Itier, C. 1999. Karu
Ñankunapi. Cuzco: Bartolomé de las Casas.
Jake, J. 1985. Grammatical relations in Imbabura Quechua. New York: Garland.
Jara, F. & R. Moya. 1987. Taruca: Ecuador quichuacunapac rimashca rimaicuna.
Cuenca, Ecuador: Universidad de Cuenca.
Kaplan, T.I. & J. B. Whitman. 1995. The category of relative clauses in Japanese,
with reference to Korean, Journal of East Asian Linguistics, 4:1.29-58.
Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press.
Koizumi, M. 1995. Phrase Structure in Minimalist Syntax. PhD Dissertation, MIT.
Kornfilt, J. 2002. Relative clauses in Turkish, in A. Alexiadou et al. (Eds.) The
Syntax of Relative Clauses. Amsterdam: John Benjamin.
Kuno, S. 1973. The structure of the Japanese language, Cambridge: MIT.

233
Kuroda, S.-Y. 1974. Pivot-independent relativization in Japanese I. Papers in
Japanese Linguistics 3:59-93. Reprinted in Kuroda (1992).
Kuroda, S.-Y. 1992. Japanese Syntax and Semantics: Collected Papers. Kluwer
Academic Publishers, Dordrecht.
Lasnik, H. 1995. Case and expletives revisited: On greed and other human failings,
Linguistics Inquiry, 26.4.615-633.
Lasnik, H. 1996. Case and expletives: Notes toward a parametric account. In R.
Freidin (ed.) Current issues in comparative grammar, Kluwer. pp162-189.
Lefebvre, C. & P. Muysken. 1979. COMP in (Cuzco) Quechua, NELS 1978, cunyforum
Papers in Linguistics. 5-6:66-76.
Lefebvre, C. & P. Muysken. 1982. Raising as Move Case, The Linguistic Review 2,
161-210.
Lefebvre, C. & P. Muysken. 1988. Mixed Categories: Nominalization in Quechua.
Dordrecht: Kluwer.
Mannheim, B. 1991. The Language of the Inka since the European Invasion. Austin:
University of Texas Press.
Marantz, A. 2001. Words, LOT Summer School/WCCFL XX handout.
Matthewson, L. 2001. Quantification and the nature of cross-linguistic variation,
Natural Language Semantics, 9:145-189.
Milsark, G. 1977. Peculiarities of the existential construction in English. Linguistic
Analysis, 3:1.1-31.
Morató Peña, L. and L. Morató Lara. 1995. Quechua Qosqo-Qollaw, Nivel Básico
& Nivel Intermedio. Ithaca, NY: LASP, Cornell University.
Murasugi, K. S. 2000. Japanese complex NPs and antisymmetry, in Martin, R. et
al. Step by Step, Cambridge: MIT.
Muysken, P. 1977. Syntactic developments in the verb phrase of Ecuadorian Quechua.
Dordrecht: Foris.
Muysken, P. 1986. A note on passive-like statives in Quechua, Revue Québecoise
Linguistique, 15, 109-120.

234
Muysken, P. 1989. Predication chains: Case and argument status in Quechua and
Turkish. Linguistic Inquiry, 20.4:627-645.
Muysken, P. 1994. Inflection and agreement properties of quantifiers in Quechua.
In P. Cole et al. (Eds.) Language in the Andes. University of Delaware.
Parker, G. J. 1969-1971. Comparative Quechua phonology and grammar I-V, University
of Hawaii Working Papers in Linguistics.
Partee, B. 1976. Some transformational extensions of Montague grammar. In B.
Partee (ed.), Montague Grammar. New York: Academic Press.
Partee, B. 1987. Noun phrase interpretation and type-shifting principles. In J.
Groenendijk et al. (eds), Studies in Discourse Representation Theory and
the Theory of Generalized Quantifiers, Foris, GRASS 8. Dordrecht.
Payne, J. 1999. Cuentos cusqueños. Cuzco: CBC.
Pollard, C. & I. Sag. 1994. Head-Driven Phrase Structure Grammar, Chicago:
University of Chicago Press, and Stanford: CSLI.
Ross, J.R. 1973. You take the high node and I’ll take the low node. Papers from the
Comparative Syntax Festival. Chicago: U. of Chicago Linguistics Society.
Rothstein, S. 1995. Adverbial quantification over events, Natural Language Semantics,
3:1-31.
Saito, M. 1985. Some Asymmetries in Japanese and Their Theoretical Implications.
Doctoral dissertation, MIT: Cambridge.
Sánchez, L. 1996. Syntactic Structures in Nominals: A Comparative Study of Spanish
and Southern Quechua. PhD dissertation, U. Southern California.
Sánchez, L. 1999a. Why does Southern Quechua agree in person nominally? MIT
Occasional Papers in Linguistics, 17:131-147.
Sánchez, L. 1999b. Null objects in contact varieties of Spanish, in J-M Authier et
al., Eds., Formal Perspectives on Romance Linguistics, Amsterdam: John
Benjamins.
Schultze-Berndt, E. & N. P. Himmelmann. 2004. Depictive secondary predicates in
cross-linguistic perspective, Linguistic Typology, 8:1.
Sells, P. 1986. Coreference and bound anaphora: A restatement of the facts,’ NELS

235
16, Proceedings of the 16th Conference of the Northeast Linguistic Society,
McGill University, 1985, pp 434-446.
Sharvit, Y. 1996. Syntax and Semantics of Functional Relative Clauses. PhD dissertation,
Rutgers.
Shimoyama, J. 1999. Internally headed relative clauses in Japanese and E-type
anaphora. Journal of East Asian Linguistics, 8.147-182.
Shimoyama, J. 2001. Wh-constructions in Japanese. PhD dissertation, University
of Massachusetts.
Sigurdsson, H.A. 1996. Icelandic finite verb agreement. Working Papers in Scandinavian
Syntax 57, 1-46.
Solá, D. 1967. Gramática del quechua de Huánuco. Lima: PFL, Universidad Nacional
Mayor de San Marcos.
Srivastav, V. 1991. Wh Dependencies in Hindi and the Theory of Grammar. Phd
dissertation, Cornell.
Stark, L. 1973. El Quichua de Imbabura. Otavalo, Ecuador: Instituto Interandino
de Derarrollo.
Szabolcsi, A. 1994. The Noun Phrase. In Syntax and Semantics Vol 27: The
Syntactic Structure of Hungarian.
Torero, A. 1964. Los dialectos quechuas.
Agraria, 2, 446-478.
Anales científicos de la Universidad
Valderrama, R. and C. Escalante. 1977. Gregorio Condori Mamani, Autobiografia.
Cusco, Perú: Centro de Estudios Rurales Andinos.
Vergnaud, J.R. 1974. French relative clauses. Doctoral dissertation, MIT, Cambridge.
Weber, D. 1978. Relativization in Huallaga (Huánuco) Quechua. MA. thesis,
UCLA.
Weber, D. 1994. Relativizacion y Clausulas Nominalizadas en el Quechua Huallaguino.
Peru: Summer Institute of Linguistics.
Williamson, J. 1987. An indefniteness restriction on relative clauses in Lakhota. In
E. Reuland & A.G.B. ter Meulen (eds.), The Representation of Indefinites,
Cambridge, MA: MIT Press.