Minimality Effects in Syntax · The MLC and Derivational Economy ...
Minimality Effects in Syntax · The MLC and Derivational Economy ...
Minimality Effects in Syntax · The MLC and Derivational Economy ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> <strong>Derivational</strong> <strong>Economy</strong><br />
Gisbert Fanselow<br />
Introduction<br />
<strong>The</strong>re is a certa<strong>in</strong> tension between the role which the M<strong>in</strong>imal L<strong>in</strong>k Condition<br />
(<strong>MLC</strong>, (1)) plays <strong>in</strong> at least the m<strong>in</strong>imalist theories of syntax, <strong>and</strong> the existence<br />
of numerous (apparent or real) counterexamples such as (2) that arise<br />
<strong>in</strong> multiple questions. For such questions, the <strong>MLC</strong> seems to imply strict<br />
superiority effects. In particular, wh-objects should not be able to cross whsubjects<br />
on their way to Spec,CP. More often than not, this prediction fails<br />
to be observed. Put differently, the question arises as to why the <strong>MLC</strong> is<br />
respected strictly by head movement, <strong>and</strong> more of less so by A-movement,<br />
while it is a fairly poor predictor for grammaticality when the proper way of<br />
carry<strong>in</strong>g out operator movement is at stake.<br />
(1) M<strong>in</strong>imal L<strong>in</strong>k Condition (<strong>MLC</strong>)<br />
α cannot move to γ if there is a β that can also move to γ <strong>and</strong> is closer<br />
to γ than α<br />
(2) Constructions violat<strong>in</strong>g the superiority condition<br />
a. which book did which person read?<br />
b. was hat wer gelesen (German)<br />
what has who read<br />
“what was read by whom?”<br />
If correct, this characterization of the problem already suggests a solution:<br />
the <strong>MLC</strong> must be <strong>in</strong>terpreted as a pr<strong>in</strong>ciple that is sensitive to <strong>in</strong>terpretation/<br />
expressivity (cf. also Kitahara (1993), (1994), Re<strong>in</strong>hart (1995), Sternefeld<br />
(1997)). Whenever it does not make a semantic difference whether the <strong>MLC</strong><br />
is respected or not, the <strong>MLC</strong> must be obeyed strictly. However, the <strong>MLC</strong> is<br />
never (by itself) able to block a movement operation that is <strong>in</strong>evitable for<br />
express<strong>in</strong>g a certa<strong>in</strong> mean<strong>in</strong>g. Consequently, to the extent that head movement<br />
does not have any semantic effects, the <strong>MLC</strong> governs head movement
74 Gisbert Fanselow<br />
<strong>in</strong> a strict <strong>and</strong> exceptionless way. To the extent that different ways of carry<strong>in</strong>g<br />
out operator movement are crucial <strong>in</strong> establish<strong>in</strong>g different semantic relations,<br />
the <strong>MLC</strong> effects we observe <strong>in</strong> this doma<strong>in</strong> are modulated by considerations<br />
of <strong>in</strong>terpretation.<br />
Orig<strong>in</strong>ally, the idea that the <strong>MLC</strong> decides between those structural alternatives<br />
only that have identical mean<strong>in</strong>gs was motivated by data <strong>in</strong>volv<strong>in</strong>g<br />
different scope assignments to wh-operators (see sect. 1). <strong>The</strong> present paper<br />
argues that the required mean<strong>in</strong>g identity must also <strong>in</strong>volve dist<strong>in</strong>ctions of<br />
<strong>in</strong>formation structure (sections 3 <strong>and</strong> 4), which expla<strong>in</strong>s why many (if not<br />
most) languages are like German <strong>in</strong> not show<strong>in</strong>g simple superiority effects<br />
at all. Languages like English <strong>and</strong> Bulgarian fit <strong>in</strong>to such a picture as well –<br />
there is no variation among languages <strong>in</strong> this respect. Furthermore, we concur<br />
with Sternefeld (1997) <strong>in</strong> the claim that the <strong>MLC</strong> must be applied <strong>in</strong> a<br />
cyclic rather than global fashion (section 2.4), <strong>and</strong> we argue that it <strong>in</strong>volves<br />
reference to LF-identity rather than mean<strong>in</strong>g identity <strong>in</strong> a broad sense.<br />
1. <strong>The</strong> <strong>MLC</strong> <strong>and</strong> wh-phrase scope<br />
<strong>The</strong> <strong>MLC</strong> is a core pr<strong>in</strong>ciple of current syntactic theoriz<strong>in</strong>g, <strong>and</strong> has been<br />
made responsible for a wide variety of syntactic generalizations, such as the<br />
Head Movement Constra<strong>in</strong>t of Travis (1984), the <strong>in</strong>tervention effects<br />
restrict<strong>in</strong>g A-movement to subject position (Chomsky 1993, 1995, Stepanov<br />
2001, this volume), <strong>and</strong> the superiority effect govern<strong>in</strong>g the formation of<br />
multiple questions. In spite of the important role it plays <strong>in</strong> determ<strong>in</strong><strong>in</strong>g<br />
whether syntactic computations are formally correct, some aspects of multiple<br />
questions require that the <strong>MLC</strong> is sensitive to the <strong>in</strong>terpretation of the<br />
structures or derivations that it compares.<br />
Before we discuss this fact, let us consider some simple superiority<br />
effects <strong>in</strong> English. Object wh-phrases cannot cross c-comm<strong>and</strong><strong>in</strong>g subject<br />
wh-phrases (3), as was observed by Kuno <strong>and</strong> Rob<strong>in</strong>son (1972). Haider<br />
(this volume) argues that the contrast <strong>in</strong> (3) <strong>in</strong>volves a grammatical constra<strong>in</strong>t<br />
that bans wh-phrases occupy<strong>in</strong>g the subject position of f<strong>in</strong>ite clauses<br />
(such as the Empty Category Pr<strong>in</strong>ciple of Chomsky 1981).<br />
(3) Simple subject-object asymmetry<br />
a. (It does not matter) who bought what<br />
b.(It does not matter) *what who bought _
Independent of whether such a factor contributes to mak<strong>in</strong>g (3b) worse than<br />
(3a), the special status of the subject position cannot be the only source for<br />
superiority effects: wh-objects must not cross wh-subjects even when the<br />
latter are lexically governed, as <strong>in</strong> (4). Likewise, a wh-object from a lower<br />
clause cannot cross a wh-object from a higher clause on its way up to<br />
Spec,CP (5). <strong>The</strong> <strong>in</strong>teraction of clausemate objects yields identical <strong>in</strong>tervention<br />
effects, as evidenced by the contrasts <strong>in</strong> (6).<br />
(4) Subject-object asymmetry not <strong>in</strong>volv<strong>in</strong>g proper government<br />
a. who do you expect _ to do what?<br />
b.*what do you expect who to do _<br />
(5) Biclausal object-object-asymmetry<br />
a. who do you persuade _ to do what<br />
b.*what do you persuade who to do _<br />
(6) Superiority effects among objects<br />
a.*what did you give who _<br />
b. who did you give _ what<br />
c. what/which check did you send _ to who<br />
d.*who(m) did you send what/which check to _<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 75<br />
As Hendrick <strong>and</strong> Rochemont (1982) correctly po<strong>in</strong>t out, data such as (4) – (6)<br />
are <strong>in</strong>compatible with the view that the superiority effect can be completely<br />
reduced to the ECP or a similar pr<strong>in</strong>ciple. What is called for is an account<br />
along the l<strong>in</strong>es orig<strong>in</strong>ally proposed by Kuno <strong>and</strong> Rob<strong>in</strong>son (1972): A wh-DP<br />
a cannot cross a structurally higher wh-DP b when mov<strong>in</strong>g to Spec,CP. This<br />
generalization derives from the <strong>MLC</strong> <strong>in</strong> (1) straightforwardly.<br />
One notorious difficulty of purely formal accounts of the superiority<br />
condition derives from the fact that pairs of wh-phrases that take different<br />
semantic scope need not obey the <strong>MLC</strong>, as (7) illustrates (see, e.g., Huang<br />
1982, Lasnik <strong>and</strong> Saito 1992). If the lower occurence of who <strong>in</strong> (7a) takes<br />
matrix scope, the sentence is f<strong>in</strong>e, although the movement of what across<br />
who fails to obey the <strong>MLC</strong>. If the lower who takes scope over the complement<br />
clause only, (7a) is as ungrammatical as (3b). <strong>The</strong> effect is not conf<strong>in</strong>ed<br />
to clausemate wh-phrases. Of ten English native speakers (all l<strong>in</strong>guists) that I<br />
consulted, seven accepted (7c), <strong>and</strong> five did not even f<strong>in</strong>d (7d) objectionable.
76 Gisbert Fanselow<br />
(7) Absence of superiority effects for wh-phrases with different scope<br />
a. who wonders what who bought?<br />
b. who wonders who bought what?<br />
c. who wonders what John persuaded who to buy __ ?<br />
d. who wonders what John told who that he should buy __ ?<br />
Accord<strong>in</strong>g to Golan (1993), Kitahara (1993), <strong>and</strong> Re<strong>in</strong>hart (1995, 1998), such<br />
facts suggest that the <strong>MLC</strong> must be <strong>in</strong>terpreted as an economy constra<strong>in</strong>t<br />
related to LF-outputs (mean<strong>in</strong>gs). Whenever there is no other way to express<br />
a certa<strong>in</strong> mean<strong>in</strong>g, the <strong>MLC</strong> need not be respected. Let us consider (7a) <strong>in</strong><br />
more detail. Overt movement of a wh-phrase to Spec,CP fixes its scope. A<br />
wh-phrase merged <strong>in</strong> a complement clause can thus take matrix scope under<br />
two conditions only: it moves to the Spec,CP position of the matrix clause,<br />
or it stays <strong>in</strong> situ, <strong>and</strong> gets scope-bound by an element <strong>in</strong> the matrix clause.<br />
It must not, however, be placed <strong>in</strong>to the Spec,CP position of the complement<br />
clause, <strong>and</strong> still take matrix scope. Thus, the subject of the complement clause<br />
who can take matrix scope <strong>in</strong> (7a) only if it stays <strong>in</strong> situ. In other words, it<br />
can take matrix scope only if crossed by the lower wh-phrase what target<strong>in</strong>g<br />
the complement clause Spec,CP position. <strong>The</strong> mean<strong>in</strong>g (8a) of (7a) simply<br />
cannot be expressed differently – (7b) means someth<strong>in</strong>g else (viz. (8b)).<br />
Whether the <strong>MLC</strong> is respected or not is irrelevant when the structural alternatives<br />
differ <strong>in</strong> <strong>in</strong>terpretation.<br />
(8) a.For which persons x,y: x wonders what y bought<br />
b.For which person x, <strong>and</strong> for which z: x wonders who bought z<br />
In contrast to what holds for (7), the two derivational alternatives <strong>in</strong> (3) do<br />
not yield different <strong>in</strong>terpretations: there is only one scope option available<br />
for the two wh-phrases. In such a situation (<strong>and</strong> only <strong>in</strong> such a situation),<br />
the <strong>MLC</strong> filters out derivations that are not <strong>in</strong> l<strong>in</strong>e with it. Further English<br />
constructions illustrat<strong>in</strong>g that the application of the <strong>MLC</strong> depends on the<br />
<strong>in</strong>terpretation arrived at will be presented <strong>in</strong> sections 2.1. <strong>and</strong> 2.3.<br />
Given that the wellformedness of (7a) is of some theoretical importance,<br />
it is surpris<strong>in</strong>g that little evidence from other languages has entered the discussion<br />
of the <strong>in</strong>terpretation-sensitivity of the <strong>MLC</strong>. Accord<strong>in</strong>g to one of<br />
my <strong>in</strong>formants (Koyka Stoyanova, p.c.), (9a,b) are as f<strong>in</strong>e <strong>in</strong> Bulgarian as<br />
they are <strong>in</strong> English if the second occurence of koj is stressed, but <strong>in</strong> her<br />
dialect, the order kakvo koj is grammatical <strong>in</strong> simple multiple questions, too.
Penka Stateva, my second Bulgarian <strong>in</strong>formant, does not accept the order<br />
kakvo koj <strong>in</strong> a simple clause, <strong>and</strong> rejects (9) as well. No contrast such as the<br />
one between (3) <strong>and</strong> (7) exists <strong>in</strong> Bulgarian. <strong>The</strong> absence of this contrast<br />
will be expla<strong>in</strong>ed <strong>in</strong> section 4.1: we argue there that the order<strong>in</strong>g restrictions<br />
among Bulgarian wh-phrases are not caused by the <strong>MLC</strong>. <strong>The</strong> ungrammaticality<br />
of (9) <strong>in</strong> some dialects therefore does not bear on the issue of the<br />
<strong>in</strong>terpretation sensitivity of the <strong>MLC</strong>.<br />
(9) Anti-superiority <strong>in</strong> Bulgarian<br />
a.#koj se chudi, kakvo koj kupi?<br />
who wonders what who bought<br />
“who wonders what who bought?”<br />
b.#na kogo kaza, kakvo koj kupi?<br />
who.dat you-tell what who bought<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 77<br />
For other languages, it is not much easier to construct relevant evidence,<br />
because the simple superiority effect exemplified <strong>in</strong> (3) is not a widespread<br />
phenomenon. <strong>The</strong> follow<strong>in</strong>g data from German, however, provide further<br />
evidence for the <strong>in</strong>terpretation sensitivity of the <strong>MLC</strong>. (2b) has already<br />
shown that the formation of multiple questions is not affected by the <strong>MLC</strong><br />
<strong>in</strong> German (at least superficially) when clausemates are <strong>in</strong>volved, but it has<br />
been claimed frequently that a wh-phrase from a lower clause cannot cross<br />
a matrix wh-word.<br />
(10) Superiority for non-clausemates <strong>in</strong> st<strong>and</strong>ard German<br />
a.*wen hat wer gehofft, dass Ir<strong>in</strong>a e<strong>in</strong>lädt<br />
wh.acc has who.nom hoped that Ir<strong>in</strong>a <strong>in</strong>vites<br />
b. wer hat gehofft, dass Ir<strong>in</strong>a wen e<strong>in</strong>lädt<br />
“who has hoped that Ir<strong>in</strong>a will <strong>in</strong>vite who?”<br />
<strong>The</strong>re are reasons to doubt, however, that the ungrammaticality of (10a) (<strong>in</strong><br />
the st<strong>and</strong>ard dialect) is caused by the <strong>MLC</strong>. Superiority effects disappear<br />
when the wh-phrases are discourse-l<strong>in</strong>ked <strong>in</strong> the sense of Pesetsky (1987).<br />
However, (10a) does not improve <strong>in</strong> the st<strong>and</strong>ard language when d-l<strong>in</strong>ked<br />
wh-phrases are used. Thus, what rules out (10a) must be different from the<br />
<strong>MLC</strong>.
78 Gisbert Fanselow<br />
(11) *welchen Studenten hat welcher Professor gehofft,<br />
dass Ir<strong>in</strong>a e<strong>in</strong>lädt<br />
which.acc student.acc has which.nom professor hoped<br />
that Ir<strong>in</strong>a <strong>in</strong>vites<br />
“which professor has hoped that Ir<strong>in</strong>a <strong>in</strong>vites which student?”<br />
In less restrictive dialects (such as the one spoken by the author), all sentences<br />
<strong>in</strong> (12) are acceptable up to a certa<strong>in</strong> degree, but (12a) <strong>and</strong> (12b)<br />
have different <strong>in</strong>terpretations. If (12a) is completely wellformed at all, the<br />
sentence allows a s<strong>in</strong>gle-pair <strong>in</strong>terpretation only. A pair-list-read<strong>in</strong>g is available<br />
for (12b) only, i.e., for the structure which violates the <strong>MLC</strong>. In addition<br />
the “scope-mark<strong>in</strong>g” construction (12c) allows the pair-list-read<strong>in</strong>g as<br />
well.<br />
(12) Nonst<strong>and</strong>ard German: Subord<strong>in</strong>ate clause wh-elements cross<strong>in</strong>g matrix<br />
wh-phrases<br />
a.(?)wer hat gehofft, dass Ir<strong>in</strong>a wen e<strong>in</strong>lädt<br />
who has hoped that Ir<strong>in</strong>a who <strong>in</strong>vites<br />
b. wen hat wer gehofft, dass Ir<strong>in</strong>a e<strong>in</strong>lädt<br />
who.acc has who.nom hoped that Ir<strong>in</strong>a <strong>in</strong>vites<br />
c. was hat wer gehofft, wen Ir<strong>in</strong>a e<strong>in</strong>lädt<br />
what has who hoped who.acc Ir<strong>in</strong>a <strong>in</strong>vites<br />
How can these data be understood? 1 In quite a number of languages, <strong>in</strong> situ<br />
wh-phrases cannot take scope out of the m<strong>in</strong>imal (f<strong>in</strong>ite) clause they are<br />
conta<strong>in</strong>ed <strong>in</strong>. H<strong>in</strong>di is a case <strong>in</strong> po<strong>in</strong>t (see Mahajan 1990). <strong>The</strong> scope of an<br />
<strong>in</strong>-situ wh-phrase must be determ<strong>in</strong>ed by l<strong>in</strong>k<strong>in</strong>g it to a higher wh-phrase, or<br />
to a scope marker. <strong>The</strong> l<strong>in</strong>k<strong>in</strong>g might be arrived at <strong>in</strong> various ways (b<strong>in</strong>d<strong>in</strong>g,<br />
covert movement), but the important observation concern<strong>in</strong>g H<strong>in</strong>di <strong>and</strong> other<br />
languages is that l<strong>in</strong>k<strong>in</strong>g is subject to strong locality requirements. In contrast<br />
to what holds for overt movement (=wh-scrambl<strong>in</strong>g <strong>in</strong> the case of H<strong>in</strong>di),<br />
f<strong>in</strong>ite CPs are barriers for the l<strong>in</strong>k<strong>in</strong>g relation. Consequently, (13) is ungrammatical<br />
because the lower occurence of kis-ko must be l<strong>in</strong>ked to a whphrase<br />
or a scope marker, but cannot be so because it is embedded <strong>in</strong> an<br />
isl<strong>and</strong> for l<strong>in</strong>k<strong>in</strong>g.
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 79<br />
(13) Clauseboundedness of the b<strong>in</strong>d<strong>in</strong>g of <strong>in</strong> situ wh-phrases <strong>in</strong> H<strong>in</strong>di<br />
*Raam-ne kis-ko kahaa ki Sitaa-ne kis-ko dekhaa<br />
Raam-erg who.dat told that Sita-erg who saw?<br />
“who did Ram tell that Sita saw who?”<br />
Let us now come back to (12). First, we want to expla<strong>in</strong> why (12a) is out with<br />
a pair-list <strong>in</strong>terpretation. This follows if (the relevant version of) German<br />
resembles H<strong>in</strong>di <strong>in</strong> that f<strong>in</strong>ite CPs are barriers for the scope l<strong>in</strong>k<strong>in</strong>g of <strong>in</strong><br />
situ wh-phrases. Consequently, wen cannot be scope-l<strong>in</strong>ked to wer <strong>in</strong> (12a),<br />
which renders the structure ungrammatical under the <strong>in</strong>tended <strong>in</strong>terpretation.<br />
F<strong>in</strong>ite clauses are not, however, barriers for overt movement. <strong>The</strong>refore,<br />
there is a way of construct<strong>in</strong>g a Logical Form for (12) <strong>in</strong> which both whphrases<br />
take matrix scope, viz. by mov<strong>in</strong>g the wh-element from the complement<br />
clause <strong>in</strong>to the matrix-Spec-CP position, <strong>and</strong> by scope-l<strong>in</strong>k<strong>in</strong>g the<br />
matrix subject to the matrix Spec,CP position. This is what has happened <strong>in</strong><br />
(12b). None of the relations established there is <strong>in</strong> conflict with locality<br />
requirements – but the <strong>MLC</strong> is violated. Apparently, this <strong>MLC</strong>-violation is<br />
licensed because the relevant Logical Form cannot be arrived at <strong>in</strong> a different<br />
way – the structure (12a) respect<strong>in</strong>g the <strong>MLC</strong> is <strong>in</strong>compatible with the locality<br />
of the licens<strong>in</strong>g of wh-phrases <strong>in</strong> situ. (12) illustrates the same phenomenon<br />
as (7), but <strong>in</strong> a rather different context.<br />
<strong>The</strong> other examples <strong>in</strong> (12) illustrate two further po<strong>in</strong>ts. (12c) shows that<br />
German is like H<strong>in</strong>di <strong>in</strong> hav<strong>in</strong>g a wh-scope-mark<strong>in</strong>g construction, <strong>in</strong> which<br />
a scope marker (was) rather than the real wh-phrase appears <strong>in</strong> Spec,CP.<br />
(12c) is well-formed <strong>in</strong> all dialects of German, <strong>and</strong> expresses a pair-list<br />
<strong>in</strong>terpretation. A m<strong>in</strong>or po<strong>in</strong>t illustrated by this example is that f<strong>in</strong>ite clauses<br />
are isl<strong>and</strong>s for scope tak<strong>in</strong>g <strong>in</strong> German only for wh-phrases that do not<br />
occupy a Spec,CP position (note that the lower wh-phrase is fronted <strong>in</strong> the<br />
complement clause). <strong>The</strong>re are various ways of analys<strong>in</strong>g the construction<br />
(see, e.g., the contributions <strong>in</strong> Lutz, Müller <strong>and</strong> von Stechow 2000), but<br />
details are irrelevant for the more important po<strong>in</strong>t: long wh-movement <strong>in</strong><br />
(12b) <strong>and</strong> wh-scope mark<strong>in</strong>g <strong>in</strong> (12c) yield the same <strong>in</strong>terpretation, but the<br />
wh-scope-mark<strong>in</strong>g construction (12c) avoids an <strong>MLC</strong> violation, <strong>in</strong> contrast<br />
to (12b). This shows that the sensitivity of the <strong>MLC</strong> to <strong>in</strong>terpretation cannot<br />
<strong>in</strong>volve a simple, “global” concept of mean<strong>in</strong>g identity. If it would, the<br />
wellformedness of (12c) should imply that the <strong>MLC</strong> is able to rule out<br />
(12b). Given (12c), no <strong>MLC</strong>-violation is necessary for express<strong>in</strong>g the<br />
“mean<strong>in</strong>g” of (12b). <strong>The</strong> <strong>MLC</strong> must therefore not be sensitive to “mean<strong>in</strong>g<br />
identity” <strong>in</strong> a global sense. Rather, the identity of <strong>in</strong>terpretation that is rele-
80 Gisbert Fanselow<br />
vant for the applicability of the <strong>MLC</strong> must be a matter of identical (or closeto-identical)<br />
Logical Forms. <strong>The</strong> LF of (12c) is different from the one of (12b)<br />
(see <strong>in</strong> particular Fanselow <strong>and</strong> Mahajan (2000) for arguments), <strong>and</strong> therefore,<br />
(12c) does not count when the grammaticality of (12b) is established.<br />
Haider (1997: 221) exemplifies the claim that complement clause whphrases<br />
may cross matrix wh-phrases <strong>in</strong> German with examples such as (14).<br />
To me, (14) <strong>in</strong>vites a s<strong>in</strong>gle-pair answer only, so that (14) is not fully comparable<br />
to the multiple questions discussed so far. Furthermore (14) <strong>in</strong>volves<br />
apparent movement from a V2-complement clause, <strong>and</strong> the theoretical<br />
status of such an operation is quite unclear, see Reis (1996, 1997) for arguments<br />
that the construction is parenthetical. I therefore refra<strong>in</strong> from discuss<strong>in</strong>g<br />
such examples <strong>in</strong> more detail.<br />
(14) Superiority violations <strong>in</strong> a construction with extraction out of a V2<br />
complement<br />
wemi Bild<br />
hat wer<br />
verkauft]?<br />
gesagt [ei habe sie ei e<strong>in</strong><br />
who.dat hat who.nom said has.subjunctive she a<br />
picture sold<br />
“who said she had sold a picture to whom?”<br />
Our argumentation presupposes that s<strong>in</strong>gle-pair <strong>in</strong>terpretations of multiple<br />
questions (for which (12a) seems marg<strong>in</strong>ally acceptable) have a derivation<br />
different from the one for multiple questions with a pair-list read<strong>in</strong>g. This<br />
claim is supported by the observation that further constructions are ungrammatical<br />
with a pair-list read<strong>in</strong>g, but acceptable under a s<strong>in</strong>gle-pair <strong>in</strong>terpretation.<br />
E.g., most native speakers of German (<strong>in</strong>clud<strong>in</strong>g the author) reject<br />
(15) as a question ask<strong>in</strong>g for pair-lists, but the s<strong>in</strong>gle pair <strong>in</strong>terpretation is<br />
f<strong>in</strong>e.<br />
(15) Multiple adjunct question with a s<strong>in</strong>gle pair <strong>in</strong>terpretation<br />
wie hat er es warum geschrieben<br />
how has he it why written<br />
“how did he write it, <strong>and</strong> why”<br />
Examples such as (7) show that the applicability of the <strong>MLC</strong> depends on<br />
the <strong>in</strong>terpretation of the structure that it would block. German data such as
(12) constitute further evidence for this. At the same time, the data <strong>in</strong> (12)<br />
shows that the <strong>MLC</strong> is not sensitive to “mean<strong>in</strong>g” <strong>in</strong> a global sense – rather,<br />
it is the nature of the LF that a movement operation creates that determ<strong>in</strong>es<br />
whether the <strong>MLC</strong> must be respected.<br />
2. <strong>The</strong> <strong>MLC</strong> <strong>and</strong> expressivity<br />
<strong>The</strong> strongest conclusion one can draw from from the discussion <strong>in</strong> the preced<strong>in</strong>g<br />
paragraph is that requirements of semantic expressivity always override<br />
the <strong>MLC</strong>. A structure violat<strong>in</strong>g the <strong>MLC</strong> is ungrammatical only if the<br />
Logical Form it would express can be arrived at with a structure respect<strong>in</strong>g<br />
the <strong>MLC</strong>. In this section, we defend this strong conclusion aga<strong>in</strong>st potential<br />
counterexamples, <strong>and</strong> discuss how the <strong>MLC</strong> can be applied <strong>in</strong> local fashion.<br />
First, we discuss the <strong>in</strong>teraction of the <strong>MLC</strong> with the that-trace filter.<br />
Section 2.2 focuses argument-adjunct asymmetries, while section 2.3 is<br />
dedicated to nestedness effects, which have been related to the <strong>MLC</strong>.<br />
F<strong>in</strong>ally, we will briefly discuss what a cyclic application of the <strong>MLC</strong> might<br />
look like.<br />
2.1. Interactions with the ECP<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 81<br />
As one of the anonymous reviewers has po<strong>in</strong>ted out, the absence of a contrast<br />
<strong>in</strong> (16) might pose a problem for the idea that the <strong>MLC</strong> applies only if<br />
that does not prevent a certa<strong>in</strong> <strong>in</strong>terpretation from be<strong>in</strong>g expressed:<br />
(16) Two wh-phrases merged <strong>in</strong> a f<strong>in</strong>ite complement clause<br />
a.*who do you th<strong>in</strong>k that _ bought what<br />
b.*what do you th<strong>in</strong>k that who bought _<br />
(16b) violates the <strong>MLC</strong>, so its ungrammaticality is expected. However, the<br />
constellation that respects the <strong>MLC</strong>, viz., (16a), is ungrammatical as well<br />
because of a that-trace-filter violation. In contrast to what we saw <strong>in</strong> section 1,<br />
the <strong>MLC</strong> violation of (16b) is not tolerated by the grammatical system of<br />
English, <strong>in</strong> spite of the fact that this renders the <strong>in</strong>terpretation of (16b) <strong>in</strong>expressible.<br />
<strong>The</strong> absence of a contrast <strong>in</strong> (16) does not show, however, that the <strong>MLC</strong><br />
is able to block structures even if the compet<strong>in</strong>g structure respect<strong>in</strong>g the
82 Gisbert Fanselow<br />
<strong>MLC</strong> violates a further condition on LF. Aoun et al. (1987) <strong>and</strong> others have<br />
argued that the pr<strong>in</strong>ciple Q responsible for the that-trace effect applies at<br />
PF, <strong>and</strong> not at LF. Consequently, Q cannot <strong>in</strong>teract with the <strong>MLC</strong>: the <strong>MLC</strong><br />
applies to LFs, <strong>and</strong> compares derivations that yield (close-to-) identical LFs.<br />
It is bl<strong>in</strong>d to what happens <strong>in</strong> other branches of the derivation. A structure<br />
that has an optimal LF <strong>and</strong> is accepted by the <strong>MLC</strong> need not be <strong>in</strong> l<strong>in</strong>e with<br />
further PF-requirements, render<strong>in</strong>g the LF unpronouncable. Given this relevance<br />
of PF-constra<strong>in</strong>ts, (16) does not exclude an <strong>in</strong>terpretation of the <strong>MLC</strong><br />
that compares different ways of arriv<strong>in</strong>g at essentially the same LFs – while<br />
it falls <strong>in</strong> l<strong>in</strong>e with an other conclusion arrived at <strong>in</strong> section 1: the <strong>MLC</strong> is not<br />
a pr<strong>in</strong>ciple that takes care of “expressivity” <strong>in</strong> a literal sense.<br />
<strong>The</strong> follow<strong>in</strong>g observation leads to a modification of our analysis of (16),<br />
which leaves the crucial po<strong>in</strong>t <strong>in</strong>tact, however: the <strong>MLC</strong> responds to the<br />
need of respect<strong>in</strong>g further LF-constra<strong>in</strong>ts, but it is bl<strong>in</strong>d to what happens <strong>in</strong><br />
the PF-branch of grammar. Haider (this volume) argues that there is an extra<br />
constra<strong>in</strong>t bann<strong>in</strong>g wh-phrase occupy<strong>in</strong>g the specifier position of a f<strong>in</strong>ite IP<br />
<strong>in</strong> English. <strong>The</strong> constra<strong>in</strong>t is <strong>in</strong>dependent of the <strong>MLC</strong>, s<strong>in</strong>ce it shows its<br />
force even <strong>in</strong> constructions that do not <strong>in</strong>volve a cross<strong>in</strong>g wh-dependency,<br />
as was already observed by Chomsky (1981). Interest<strong>in</strong>gly, as (17) illustrates,<br />
the relative degree of (un-)acceptability <strong>in</strong>volves dimensions such as<br />
discourse-l<strong>in</strong>k<strong>in</strong>g (see 17b), <strong>and</strong>, as Bresnan (1972) has observed, <strong>in</strong> situ<br />
wh-subjects are much better when they appear <strong>in</strong> subjunctive clauses.<br />
(17) Wh-subjects <strong>in</strong> situ<br />
a.*who believes that who loves Ir<strong>in</strong>a?<br />
b.?who believes that which man loves Ir<strong>in</strong>a?<br />
c.?who dem<strong>and</strong>s that who be arrested?<br />
<strong>The</strong> ungrammaticality of (16b) might therefore also be caused by the presence<br />
of an <strong>in</strong> situ wh-subject <strong>in</strong> a f<strong>in</strong>ite clause quite <strong>in</strong>dependent of the <strong>MLC</strong>.<br />
Given the contrasts <strong>in</strong> (17), one would expect that structures like (16b)<br />
improve if, e.g., the complement clause appears <strong>in</strong> the subjunctive mood. In<br />
such a construction, the overt movement of the subject of the complement<br />
clause still implies a that-trace filter violation, but the additional ban aga<strong>in</strong>st<br />
<strong>in</strong> situ wh-subjects is now much less strict. Accord<strong>in</strong>g to Anthony Green <strong>and</strong><br />
Sue Olsen (p.c.), (18a) is <strong>in</strong>deed much better that (16b).
(18) Miss<strong>in</strong>g superiority effect for extraction out of a subjunctive complement<br />
a.(?)what do they require that who buy?<br />
b.*who do they require that buy what<br />
c.*what do you expect who to buy<br />
d. who do you expect to buy what<br />
If the contrast between (16b) <strong>and</strong> (18a) generalizes, we have a further example<br />
from English that shows that the <strong>MLC</strong> does not block a construction<br />
(viz., (18a)) if the structure that conforms to the <strong>MLC</strong> (viz., (18b)) violates a<br />
different pr<strong>in</strong>ciple. <strong>The</strong> contrast between (18a) <strong>and</strong> (18b) would force upon<br />
us the assumption that the that-trace filter bann<strong>in</strong>g overt subject movement<br />
<strong>in</strong> fact applies at LF, <strong>and</strong> not at PF. Otherwise, its effects would not be visible<br />
to the <strong>MLC</strong>, as necessary for (18a). Consequently, the PF-located constra<strong>in</strong>t<br />
that is <strong>in</strong>visible to the <strong>MLC</strong> (as required for (16)) is rather the further ban<br />
aga<strong>in</strong>st <strong>in</strong> situ wh-subjects argued for by Haider (this volume) <strong>and</strong> not the<br />
that-trace filter.<br />
It should f<strong>in</strong>ally be noted that the contrast between (18a) <strong>and</strong> (18c) is due<br />
to the fact that the <strong>MLC</strong>-respect<strong>in</strong>g competitor is well-formed <strong>in</strong> the case of<br />
(18c), but not <strong>in</strong> the case of (18d).<br />
2.2. Adjuncts<br />
Multiple questions with adjunct wh-pronouns constitute a second doma<strong>in</strong><br />
that is relevant for the status of the <strong>MLC</strong> as an economy constra<strong>in</strong>t. None of<br />
the structures <strong>in</strong> (19) is grammatical – although there is no other (monoclausal)<br />
way of express<strong>in</strong>g the <strong>in</strong>tended <strong>in</strong>terpretations.<br />
(19) Adjunct effects <strong>in</strong> English<br />
a.*who came why<br />
b.*why did who come<br />
c.*who spoke how<br />
d.*how did who speak?<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 83<br />
<strong>The</strong> <strong>MLC</strong> clearly picks (19a,c) rather than (19b,d), <strong>and</strong> correctly so <strong>in</strong> the<br />
light of (20). (19a,c) are blocked by some requirement (see, e.g., Haider, this
84 Gisbert Fanselow<br />
volume, Re<strong>in</strong>hart 1995, Hornste<strong>in</strong> 1995, among many others) that excludes<br />
the adjuncts how <strong>and</strong> why <strong>in</strong> any position but Spec,CP.<br />
(20) a. who spoke when?<br />
b. who spoke <strong>in</strong> what way?<br />
Aga<strong>in</strong>, the question arises as to why the <strong>MLC</strong> cannot be overriden <strong>in</strong> this<br />
context – yield<strong>in</strong>g (19b,d), which do not violate the strong constra<strong>in</strong>t aga<strong>in</strong>st<br />
how <strong>and</strong> why appear<strong>in</strong>g <strong>in</strong> situ. Note that (21) is ungrammatical: only one out<br />
of some twenty l<strong>in</strong>guists with English as a native language who I consulted<br />
accepted the sentence with a downstairs <strong>in</strong>terpretation of how. Unlike what<br />
we saw <strong>in</strong> the preced<strong>in</strong>g section <strong>in</strong> the context of (18), the ungrammaticality<br />
of (19b,d) can not be expla<strong>in</strong>ed <strong>in</strong> terms of an additional constra<strong>in</strong>t filter<strong>in</strong>g<br />
out wh-phrases <strong>in</strong> the subject position of non-subjunctive clauses.<br />
(21) *how does the police dem<strong>and</strong> that who be treated _<br />
We will propose two accounts of (19) that allow us to ma<strong>in</strong>ta<strong>in</strong> that the<br />
<strong>MLC</strong> is ignored when a certa<strong>in</strong> LF cannot be constructed otherwise.<br />
As Haider (this volume) has po<strong>in</strong>ted out, adjunct effects of the sort exemplified<br />
<strong>in</strong> (19) are absent <strong>in</strong> OV languages, as (22) illustrates. This observation<br />
excludes the idea that (19a,c) are ungrammatical on simple semantic grounds.<br />
(22) Miss<strong>in</strong>g Adjunct <strong>Effects</strong> <strong>in</strong> OV-languages<br />
a. wie het hoe gedaan heeft (Dutch complement question)<br />
b. wer es wie gemacht hat (German complement question)<br />
who it how done has<br />
Haider suggests that higher-order wh-operators such as how <strong>and</strong> why must<br />
c-comm<strong>and</strong> the head of the phrase they are applied to. Higher order adverbs<br />
range over events, so how <strong>and</strong> why should c-comm<strong>and</strong> the element that situates<br />
the proposition <strong>in</strong> time, i.e., how <strong>and</strong> why must c-comm<strong>and</strong> the (f<strong>in</strong>ite)<br />
verb. This condition is fulfilled <strong>in</strong> (22a,b), but not <strong>in</strong> (19a,c). Movement of<br />
the f<strong>in</strong>ite verb to Comp does not render wh-adjuncts <strong>in</strong> situ ungrammatical<br />
<strong>in</strong> Dutch or German. This is <strong>in</strong> l<strong>in</strong>e with the general observation that verb<br />
second movement is <strong>in</strong>visible at the level of Logical Form, either because it<br />
is reconstructed, or because it applies <strong>in</strong> the phonological component of<br />
grammar.
(22) Miss<strong>in</strong>g Adjunct <strong>Effects</strong> <strong>in</strong> OV-languages<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 85<br />
c. wie heeft het hoe gedaan (Dutch matrix question)<br />
d. wer hat es wie gemacht (German matrix question)<br />
who has it how done<br />
<strong>The</strong> account suggested by Haider (this volume) cannot be fully correct,<br />
however, because Swedish is not <strong>in</strong> l<strong>in</strong>e with it. All of my five <strong>in</strong>formants<br />
accepted (23a), <strong>and</strong> three of them found (23b) grammatical, <strong>in</strong> spite of the<br />
VO-nature of Swedish.<br />
(23) Miss<strong>in</strong>g adjunct effects <strong>in</strong> Swedish<br />
a. vem skrattade varfoer<br />
who laughed why<br />
b. Det spelar <strong>in</strong>gen roll vem som skrattade varfoer<br />
it plays no role who that laughed why<br />
“it does not matter who laughed for what reason”<br />
Similarly, Richards (2001: 18–19) reports adjunct effects for the SOV language<br />
Tibetan. <strong>The</strong>refore, a different solution is called for. Rizzi (1990: 47)<br />
has proposed that certa<strong>in</strong> wh-adjuncts (correspond<strong>in</strong>g to sentence-level<br />
adverbs) are based-generated <strong>in</strong> Comp. One way of translat<strong>in</strong>g this proposal<br />
<strong>in</strong>to the current discussion consists of the assumption that certa<strong>in</strong> wh-elements<br />
are required to appear at the left periphery of clauses on a language<br />
particular <strong>and</strong> item-specific basis. Because of (22) – (23), this idiosyncrasy of<br />
how <strong>and</strong> why (<strong>and</strong> French pourquoi) cannot be reduced to semantic considerations<br />
alone. One way of spell<strong>in</strong>g this idea out lies <strong>in</strong> the assumption that the<br />
<strong>MLC</strong> applies cyclically (see below for details), while the constra<strong>in</strong>ts forc<strong>in</strong>g<br />
how <strong>and</strong> why <strong>in</strong>to Spec,CP are representational pr<strong>in</strong>ciples check<strong>in</strong>g the wellformedness<br />
of completed Logical Forms. <strong>The</strong> <strong>MLC</strong> would therefore apply<br />
prior to the constra<strong>in</strong>ts affect<strong>in</strong>g higher order wh-phrases, with the desired<br />
effect: the <strong>MLC</strong> picks (19a,c), <strong>and</strong> these sentence are blocked at too late a<br />
po<strong>in</strong>t <strong>in</strong> the derivation for undo<strong>in</strong>g the impact of the <strong>MLC</strong>.<br />
<strong>The</strong> account sketched so far predicts the data as judged <strong>in</strong> (19) <strong>and</strong> (24).<br />
<strong>The</strong> <strong>MLC</strong> forces the subject to move to Spec,CP <strong>in</strong> a multiple question <strong>in</strong>volv<strong>in</strong>g<br />
subjects <strong>and</strong> adjuncts, but the result<strong>in</strong>g structure is blocked because<br />
why <strong>and</strong> how cannot appear <strong>in</strong> any position but Spec,CP. On the other h<strong>and</strong>,<br />
when adjuncts <strong>in</strong>teract with objects, the <strong>MLC</strong> will make (24b) block (24a).<br />
(24b) is also <strong>in</strong> l<strong>in</strong>e with the requirement that English wh-adjuncts appear at<br />
the left periphery.
86 Gisbert Fanselow<br />
(24) Adjunct-object <strong>in</strong>teraction <strong>in</strong> multiple questions<br />
a.*what did Bill buy why<br />
b. why did Bill buy what<br />
Hornste<strong>in</strong> (1995: 147–149) reports further data such as (25) that may <strong>in</strong> fact<br />
lead to a simpler analysis. If his judgements are correct, why (unlike its<br />
Dutch, German <strong>and</strong> Swedish counterparts) cannot appear at all <strong>in</strong> multiple<br />
questions, quite <strong>in</strong>dependent of the position it appears <strong>in</strong>.<br />
(25) wh-adjuncts blocked <strong>in</strong> multiple questions<br />
a.*I wonder why Bill left when<br />
b.*I wonder why Bill lives where<br />
c.*I wonder why which person came<br />
d.*I wonder why you bought what<br />
e.*why does John expect who to w<strong>in</strong><br />
If Hornste<strong>in</strong> is correct, wh-adjuncts come <strong>in</strong> two varieties. German wie<br />
“how” <strong>and</strong> warum “why” are l<strong>in</strong>ked to a semantic representation that makes<br />
them eligible for multiple questions, whereas how <strong>and</strong> why cannot appear<br />
there. Under such an account, all sentences <strong>in</strong> (19) are simply gibberish, <strong>and</strong><br />
we need not care about what the <strong>MLC</strong> would predict for them. Whether this<br />
simplification is tenable or not depends on the status of (24b). If grammatical,<br />
this sentence is <strong>in</strong>compatible with the idea that why cannot appear <strong>in</strong> multiple<br />
questions. <strong>The</strong> simplification thus presupposes that (24b) <strong>in</strong>volves an “illusion<br />
of acceptablity” (Hornste<strong>in</strong> 1995: 148). We need not settle the issue here,<br />
because the idea that the <strong>MLC</strong> is an economy constra<strong>in</strong>t can be ma<strong>in</strong>ta<strong>in</strong>ed<br />
<strong>in</strong> the account discussed earlier as well.<br />
2.3. Nestedness<br />
A third doma<strong>in</strong> sheds light on the question of whether the <strong>MLC</strong> is sensitive<br />
to LF-identity or not: nestedness effects. It has been suggested that the nestedness<br />
effect can be derived from the <strong>MLC</strong>, see Richards (2001) for a<br />
detailed proposal. If this suggestion is correct, the application of the <strong>MLC</strong><br />
could not be conf<strong>in</strong>ed to structural c<strong>and</strong>idates yield<strong>in</strong>g the same LF.
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 87<br />
In English, the <strong>in</strong>teraction of two wh-phrases mov<strong>in</strong>g to two different<br />
Spec,CP positions is governed by a nestedness effect (see Fodor 1978,<br />
Pesetsky 1982): the dependencies formed by the two wh-cha<strong>in</strong>s must not<br />
cross – one path must be embedded <strong>in</strong> the other. <strong>The</strong> nestedness condition is<br />
respected, even when it blocks the expression of a certa<strong>in</strong> <strong>in</strong>terpretation, as<br />
it does <strong>in</strong> (26b) <strong>and</strong> (27b). 2<br />
(26) Nestedness <strong>Effects</strong><br />
a.?Which viol<strong>in</strong>j do you wonder which sonatai to play _i on _j<br />
b.*Which sonataj do you wonder which viol<strong>in</strong>i to play _i on _j<br />
(27) a.?Whatj did you decide [whoi [to persuade ti [to buy tj]]] (Oka 1993: 255, (2a))<br />
b.*Whoi did you decide [whatj [to persuade ti [to buy tj]]] (Oka 1993: 255, (2b))<br />
<strong>The</strong> constra<strong>in</strong>t responsible for nestedness is respected even though the mean<strong>in</strong>gs<br />
of (26b, 27b) are different from the one expressed by (26a, 27a). Such<br />
observations are relevant for the present discussion to the extent that claims<br />
made by Richards (2001) <strong>and</strong> others are correct that the nestedness condition<br />
reduces to the <strong>MLC</strong>. If it does, (26) <strong>and</strong> (27) would not be <strong>in</strong> l<strong>in</strong>e with the<br />
idea that the <strong>MLC</strong> is ignored when a certa<strong>in</strong> LF could not be formulated<br />
otherwise.<br />
Under what conditions does the <strong>MLC</strong> imply nestedness effects? Consider<br />
an abstract representation such as (28a), with two wh-phrases that both could<br />
be attracted by either of Comp A <strong>and</strong> Comp B. When the derivation reaches the<br />
po<strong>in</strong>t at which Comp B attracts a wh-phrase (at which the specifier of Comp B<br />
must be filled by a wh-phrase), a “bl<strong>in</strong>d” application of the <strong>MLC</strong> implies<br />
that wh 1 only can move, form<strong>in</strong>g (28b). At a later stage <strong>in</strong> the derivation,<br />
Comp A attracts (the specifier of Comp A must be filled by a wh-phrase). Let<br />
us conf<strong>in</strong>e our attention to a situation <strong>in</strong> which wh 1 has already reached its<br />
scope position <strong>in</strong> (28b). <strong>The</strong>refore, it cannot undergo further movement.<br />
What will happen <strong>in</strong> such a situation?<br />
(28) a.[CompA … [CompB [ .. wh1 .. [ wh2 … ]]]]<br />
b.[CompA … wh1 [CompB [ .. wh1 .. [ wh2 … ]]]]<br />
c.[wh2 CompA … wh1 [CompB [ .. wh1 .. [ wh2 … ]]]]
88 Gisbert Fanselow<br />
If the <strong>MLC</strong> applies bl<strong>in</strong>dly irrespective of whether the LF it generates is wellformed<br />
or not, then only wh 1 can move to Comp A (wh 2 cannot move because<br />
of <strong>in</strong>terven<strong>in</strong>g wh 1), which implies that the derivation breaks down, because<br />
a wh-phrase is required to move that must not do so. In this way, aspects of<br />
the wh-isl<strong>and</strong> condition might be derived, see Chomsky (1995). This would<br />
constitute a case <strong>in</strong> which the <strong>MLC</strong> rules out a mean<strong>in</strong>g that cannot be<br />
expressed otherwise. It is not advisable, however, to derive the wh-isl<strong>and</strong><br />
effect from the <strong>MLC</strong>. In spite of the fact that it respects the superiority condition,<br />
English is sometimes quite liberal with respect to wh-isl<strong>and</strong>s, as the<br />
status of e.g. what do you wonder how to fix suggests. German respects the<br />
wh-isl<strong>and</strong> condition, but fails to show superiority effects. <strong>The</strong> two phenomena<br />
simply are not correlated with each other.<br />
If wh 1 is frozen <strong>in</strong> its position <strong>in</strong> (28b), i.e., if it cannot move further, <strong>and</strong><br />
if that is taken <strong>in</strong>to account <strong>in</strong> the computation of <strong>MLC</strong> effects, then wh 1<br />
does not constitute a b is the sense of (1) repeated below that could go to<br />
Comp A<br />
(1) M<strong>in</strong>imal L<strong>in</strong>k Condition (<strong>MLC</strong>)<br />
α cannot move to γ if there is a β that can also move to γ <strong>and</strong> is closer<br />
to γ than α<br />
<strong>The</strong>refore, wh 2 can move to Comp A (as <strong>in</strong> (28c)). <strong>The</strong> derivation lead<strong>in</strong>g to<br />
(28c) is well-formed, yield<strong>in</strong>g a nested structure, because the lower of two<br />
Comps (which attracts first) only attracts the higher of two wh-phrases if<br />
movement respects the <strong>MLC</strong>. In this way, the nestedness condition is derivable<br />
from the <strong>MLC</strong>.<br />
Obviously, this reduction of the nestedness condition to the <strong>MLC</strong> presupposes<br />
that the applicability of the <strong>MLC</strong> does not depend on the existence<br />
of a different way of construct<strong>in</strong>g the <strong>in</strong>tended Logical Form. <strong>The</strong> LF (28d)<br />
is different from (28c), so that the fact that (28c) cannot be arrived at <strong>in</strong> a<br />
derivation respect<strong>in</strong>g the <strong>MLC</strong> is irrelevant for the wellformedness of (28d).<br />
(28) d.[wh 1 Comp A … wh 2 [Comp B [ .. wh 1 .. [ wh 2 … ]]]]<br />
If one wants to stick to the idea that the <strong>MLC</strong> triggers nestedness effects,<br />
one has to offer alternative accounts of the data presented sections 1 <strong>and</strong> 2.1<br />
that suggest an <strong>in</strong>terpretation-sensitivity of the <strong>MLC</strong>. <strong>The</strong>re are, however,<br />
good reasons for not endors<strong>in</strong>g such an <strong>MLC</strong>-based account of nestedness.<br />
Superiority <strong>and</strong> nestedness do not go h<strong>and</strong> <strong>in</strong> h<strong>and</strong>, as one would expect if<br />
the two phenomena were due to the same pr<strong>in</strong>ciple of UG.
For example, Swedish respects the nestedness condition (see Mal<strong>in</strong>g <strong>and</strong><br />
Zaenen 1982: 238f) although it fails to show superiority effects, see (29)<br />
<strong>and</strong> (ii) <strong>in</strong> endnote 3. Thus, at least <strong>in</strong> Swedish, nestedness cannot be<br />
reduced to the <strong>MLC</strong>.<br />
(29) Absence of superiority effects <strong>in</strong> Swedish<br />
Vad koepte vem<br />
what bought who<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 89<br />
At least certa<strong>in</strong> varieties of Spanish (see (30)) <strong>and</strong> Catalan exemplify what<br />
appears to be an anti-nestedness effect for extractions from wh-clauses: the<br />
wh-phrase that is merged <strong>in</strong> the higher position must also be the one moved<br />
to the higher of the two Spec,CP slots. Thus, wh-subjects <strong>and</strong> wh-<strong>in</strong>direct<br />
objects may cross wh-objects, but not vice versa.<br />
(30) Anti-nestedness <strong>in</strong> Spanish<br />
a.*qué libros no sabes quién ha leido<br />
which books not you know who has read<br />
b. quién no sabes qué libros ha leido<br />
who not you know which books has read<br />
c. a quién no sabes qué libros ha devuelto Celia<br />
to who not you know which books has returned Celia<br />
d.*qué libros no sabes a quíen ha devuelto Celia<br />
Likewise, Richards (2001: 27) claims that there is an anti-nestedness effect<br />
<strong>in</strong> Bulgarian. Aga<strong>in</strong>, the constra<strong>in</strong>t seems uncorrelated with superiority,<br />
s<strong>in</strong>ce simple superiority effects are observed <strong>in</strong> Bulgarian only, <strong>and</strong> not <strong>in</strong><br />
Spanish.<br />
(31) Anti-nestedness <strong>in</strong> Bulgarian<br />
a. Koj1 se opitvat da razberat kogo2 t1 e ubil t2 Who self try to f<strong>in</strong>d out whom is killed<br />
b.*Kogo 1 se opitvat da razberat koj 2 t 2 e ubil t 2<br />
In any event, it is hard to draw firm theoretical conclusions from such contrasts,<br />
s<strong>in</strong>ce there is considerable <strong>in</strong>dividual variation among speakers of
90 Gisbert Fanselow<br />
Bulgarian (see Richards 2001: 28) <strong>and</strong> of Spanish (at least among the<br />
speakers we consulted). This variation suggests that process<strong>in</strong>g factors contribute<br />
to generat<strong>in</strong>g (anti-)nestedness effects (see also Fodor 1978).<br />
Furthermore, nestedness effects have properties are different from those of<br />
superiority. Norwegian shows nestedness effects, but only if three (or more)<br />
dependencies are <strong>in</strong>volved (Mal<strong>in</strong>g & Zaenen 1982). This is unexpected from<br />
an <strong>MLC</strong> perspective: the addition of a third wh-phrase elim<strong>in</strong>ates superiority<br />
effects <strong>in</strong> English. Likewise, at least <strong>in</strong> English, there is no discourse-l<strong>in</strong>k<strong>in</strong>g<br />
<strong>in</strong>fluence on nestedness: (26b) is bad although both wh-phrases are d-l<strong>in</strong>ked<br />
<strong>in</strong> the sense of Pesetsky (1987). Superiority effects fail to show up, however,<br />
when the wh-phrases are d-l<strong>in</strong>ked. To sum up, there is a number of reasons<br />
for not deriv<strong>in</strong>g (anti-)nestedness from the <strong>MLC</strong>.<br />
2.4. Cyclic application of the <strong>MLC</strong><br />
Our discussion corroborated the view that the <strong>MLC</strong> is an economy constra<strong>in</strong>t:<br />
it does not apply when the relevant LF cannot be generated without violat<strong>in</strong>g<br />
it. <strong>The</strong> <strong>MLC</strong> is, however, <strong>in</strong>sensitive to the issue of whether other components<br />
of grammar (such as PF) might prevent the structure selected by it from surfac<strong>in</strong>g.<br />
<strong>The</strong> target LFs that the <strong>MLC</strong> compares must be very similar to each other.<br />
Otherwise, we could not underst<strong>and</strong> the data discussed <strong>in</strong> section 1: the availability<br />
of a wh-scope-mark<strong>in</strong>g constructions was shown to be irrelevant for<br />
the applicability of the <strong>MLC</strong> <strong>in</strong> a structure <strong>in</strong>volv<strong>in</strong>g long wh-movement.<br />
From a conceptual po<strong>in</strong>t of view, the <strong>MLC</strong> should be a derivational pr<strong>in</strong>ciple<br />
that applies when a phrase moves, or when a phase is completed. A<br />
cyclic application of the <strong>MLC</strong> may be called for on empirical grounds as<br />
well: if the pr<strong>in</strong>ciples that block <strong>in</strong> situ wh-subjects <strong>in</strong> English non-subjunctive<br />
clauses, <strong>and</strong> wh-adjuncts <strong>in</strong> non-left peripheral positions do not apply<br />
to PF, but rather at LF, then we must guarantee that the application of the<br />
<strong>MLC</strong> is not affected by them. This would hold if the <strong>MLC</strong> is applied cyclically,<br />
while the two constra<strong>in</strong>ts are representational restrictions on completed<br />
LFs.<br />
<strong>The</strong> simplest (but <strong>in</strong>sufficient) way of apply<strong>in</strong>g the <strong>MLC</strong> cyclically <strong>and</strong><br />
captur<strong>in</strong>g <strong>in</strong>terpretation effects at the same time works with the assumption<br />
that attract<strong>in</strong>g Comps come with some <strong>in</strong>dex that must be shared by the whphrase<br />
to be attracted. <strong>The</strong> <strong>in</strong>dex <strong>in</strong>dicates the target scope of the wh-phrases.<br />
Comp can attract a wh-phrase only if the <strong>in</strong>dices borne by the two elements<br />
are identical. <strong>The</strong>refore, under a strict <strong>in</strong>terpretation of (1), a wh-phrase can
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 91<br />
skip another wh-phrase if they have different <strong>in</strong>dices. See, e.g., Sternefeld<br />
(1997) for a discussion. What 3 can move across the wh-subject <strong>in</strong> (32a),<br />
s<strong>in</strong>ce who bears the scope <strong>in</strong>dex A of the matrix Comp.<br />
(32) a. whoA CompA wonders whatB CompB whoA bought _<br />
b. whoA CompA wonders whoB CompB bought whatA c. whoA CompA wonders whoB CompB bought what<br />
d.*who CompA wonders whatB CompB whoB bought<br />
In such a model, the <strong>MLC</strong> can be hard-wired <strong>in</strong>to the def<strong>in</strong>ition of movement<br />
(as proposed by Chomsky 1995): the attract<strong>in</strong>g Comp always triggers the<br />
movement of the closest wh-phrase with the same <strong>in</strong>dex. Exceptions to the<br />
<strong>MLC</strong> such as (32a) are more apparent than real: who A cannot be attracted by<br />
Comp B at all.<br />
While be<strong>in</strong>g attractive from a conceptual po<strong>in</strong>t of view, this model does<br />
not account for a number of data we have considered. In (33), the wh-phrases<br />
must bear the same <strong>in</strong>dex, because they take scope over the same proposition<br />
(viz., the whole sentence). <strong>The</strong>refore, if Comp attracts the closest wh-phrase<br />
with the same <strong>in</strong>dex, the sentences <strong>in</strong> (33) cannot be generated at all – contrary<br />
to what is necessary.<br />
(33) a.?what do they require that who buy?<br />
b. wen hat wer gehofft, dass Ir<strong>in</strong>a e<strong>in</strong>lädt<br />
who.acc has who.nom hoped that Ir<strong>in</strong>a <strong>in</strong>vites<br />
For (33a), it might suffice to assume that the that-trace effect is hard-wired<br />
<strong>in</strong>to the def<strong>in</strong>ition of movement as well, so that who is <strong>in</strong>visible to the<br />
attract<strong>in</strong>g matrix Comp <strong>in</strong> (33a). Such a solution cannot be applied for (33b),<br />
however, s<strong>in</strong>ce matrix subjects easily reach Spec,CP <strong>in</strong> German questions.<br />
(33b) <strong>and</strong> (to a lesser extent) (33a) thus show that a local version of compar<strong>in</strong>g<br />
different derivations cannot be avoided <strong>in</strong> a successful theory of the<br />
<strong>MLC</strong>. This can be made precise as follows.<br />
Let us assume that wh-phrases <strong>in</strong> situ receive their scope by be<strong>in</strong>g bound<br />
(as was first suggested by Baker (1970), see Dayal (2003) for an overview<br />
of non-movement theories of wh-phrases <strong>in</strong> situ), <strong>and</strong> that the b<strong>in</strong>d<strong>in</strong>g<br />
process is itself cyclic. When the syntactic object (34a) has been constructed,<br />
a wh-phrase must move to the specifier of Comp A if Comp A has a<br />
feature attract<strong>in</strong>g a wh-phrase. <strong>The</strong>re are four derivations to be considered,
92 Gisbert Fanselow<br />
then: either wh 1 or wh 2 moves to the specifier of Comp A , <strong>and</strong> the wh-phrase<br />
rema<strong>in</strong><strong>in</strong>g <strong>in</strong> situ may or may not be scope-bound by Comp or the other whphrase.<br />
(34) a.[Comp A - - - [wh 1 - - - [ wh 2 - - - ]]]]<br />
Suppose that the <strong>in</strong>-situ wh-phrase is not scope bound after movement<br />
with<strong>in</strong> (34a). <strong>The</strong>n we arrive at the representations (34b,c), which are different<br />
from each other. <strong>The</strong>refore, the <strong>MLC</strong>-respect<strong>in</strong>g structure (34b) cannot<br />
block (34c), if an application of the <strong>MLC</strong> presupposes that the relevant LF<br />
can be generated otherwise.<br />
(34) b.[wh1 [CompA - - - [wh1 - - - [ wh2 - - - ]]]]]<br />
c.[wh 2 [CompA - - - [wh1 - - - [ wh2 - - - ]]]]]<br />
Wh-phrases that are not scope-bound at all are illegal at LF. Consequently,<br />
the two partial derivations <strong>in</strong> (34b,c) will end up as grammatical only if the<br />
wh-phrase left unbound so far is later bound by a higher Comp, or by a<br />
higher wh-phrase. This is exactly what happens <strong>in</strong> (32a,b). <strong>The</strong>se examples<br />
show that neither of (34b,c) should be able to block the other. If the derivation<br />
proceeds beyond (34b,c), the cyclic nature of wh-b<strong>in</strong>d<strong>in</strong>g implies that the<br />
scope of the <strong>in</strong> situ wh-phrase must not be conf<strong>in</strong>ed to the doma<strong>in</strong> of Comp A .<br />
Suppose now that the <strong>in</strong> situ wh-phrase is scope bound after movement.<br />
This yields the representations (34d,e):<br />
(34) d.[wh1 [ Σ CompA - - - [wh A<br />
1 - - - [ wh2 A - - - ]]]]]<br />
e.[wh 2 [ Σ CompA - - - [wh A<br />
1 - - - [ wh2 A - - - ]]]]]<br />
<strong>The</strong> two syntactic objects <strong>in</strong> (34d,e) are certa<strong>in</strong>ly not identical, but they differ<br />
<strong>in</strong> a specific way only. <strong>The</strong> presence or absence of a phonetic matrix should<br />
be irrelevant for the constitution of a Logical Form. If we abstract away from<br />
the distribution of phonetic features <strong>in</strong> a syntactic object (<strong>and</strong> call the result a<br />
“partial Logical Form”), then the categories Σ are fully identical <strong>in</strong> (34d,e).<br />
Consequently, the <strong>MLC</strong> is applicable if it sensitive to the identity of the partial<br />
Logical Forms under construction, <strong>and</strong> if it selects the most economical<br />
one of the legal derivations. Normally, the <strong>MLC</strong> will pick (34d) <strong>and</strong> block<br />
(34e) because the closest phrase must be attracted. However, if there is a<br />
factor that applies cyclically <strong>and</strong> renders (34d) illegal, the <strong>MLC</strong> will let (34e)<br />
pass, s<strong>in</strong>ce there is no better compet<strong>in</strong>g structure left. <strong>The</strong> that-trace filter
(33a) <strong>and</strong> the locality requirements for b<strong>in</strong>d<strong>in</strong>g (33b) are examples of factors<br />
that imply a vacuous application of the <strong>MLC</strong>.<br />
3. Pragmatic effects<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 93<br />
In the majority of languages, there are no simple superiority effects for<br />
clausemate wh-phrases. <strong>The</strong> purpose of this section is to <strong>in</strong>tegrate the<br />
description of these languages <strong>in</strong>to our <strong>in</strong>terpretation of the <strong>MLC</strong>. Section<br />
3.1 presents the core facts, discusses potential process<strong>in</strong>g <strong>in</strong>fluences, <strong>and</strong><br />
conta<strong>in</strong>s further remarks on argument-adjunct asymmetries. Section 3.2.<br />
refutes the idea that the absence of simple superiority effects is due to a<br />
relaxed def<strong>in</strong>ition of closeness, while section 3.3 argues that we also cannot<br />
be content with the proposal that the superiority violations are absent<br />
because scrambl<strong>in</strong>g may precede wh-movement. <strong>The</strong> economy account<br />
envisaged here is discussed <strong>in</strong> section 3.4.<br />
3.1. <strong>The</strong> absence of simple superiority effects: some general remarks<br />
In a surpris<strong>in</strong>gly large number of languages, <strong>in</strong>tervention effects of the k<strong>in</strong>d<br />
exemplified <strong>in</strong> (3) do not show up <strong>in</strong> s<strong>in</strong>gle clauses. Consider, e.g., the<br />
examples given <strong>in</strong> (35), all illustrat<strong>in</strong>g (apparent) violations of (1). Other<br />
languages belong<strong>in</strong>g to this group are Mohawk, Kashmiri, Malayalam, <strong>and</strong><br />
the Slavic languages except Bulgarian.<br />
(35) Apparent violations of the <strong>MLC</strong> for clausemate arguments<br />
a. Vad koepte vem (Swedish)<br />
what bought who<br />
b. hva# keypti hver (Icel<strong>and</strong>ic)<br />
what bought who<br />
c. qué dijo quién (Spanish)<br />
what said who<br />
d. co kto robił (Polish)<br />
what who did<br />
e. nani-o dare-ga tabeta no (Japanese)<br />
what who ate<br />
f. was hat wer gesagt (German)<br />
what has what said
94 Gisbert Fanselow<br />
Two remarks are <strong>in</strong> order before we can discuss possible analyses for (35).<br />
First, it is often hard to determ<strong>in</strong>e whether a language tolerates superiority<br />
violations or not. When I asked 22 Dutch l<strong>in</strong>guists via the <strong>in</strong>ternet to rate<br />
(36), five accepted it <strong>and</strong> seven found it questionable, while ten speakers<br />
rejected the sentence. It not very plausible that this judgment pattern lends<br />
support to the claim that there is a categorial difference between, say, Dutch<br />
<strong>and</strong> German with respect to superiority. Likewise, it is not obvious what the<br />
marg<strong>in</strong>ality of (37) implies for the status of superiority <strong>in</strong> French.<br />
(36) Dutch superiority<br />
#ik weet niet wat wie gekocht heeft<br />
I know not what who bought has<br />
“I do not know who has bought what”<br />
(37) French superiority<br />
?Je me dem<strong>and</strong>e à qui a parlé qui<br />
I me wonder to whom has talked who<br />
“I wonder who has talked to whom”<br />
Instead of forc<strong>in</strong>g (36) <strong>and</strong> (37) <strong>in</strong>to one or the other category, the graded<br />
nature of such <strong>MLC</strong> violations should figure <strong>in</strong> the analysis of the construction.<br />
3 This is particularly true <strong>in</strong> the light of experimental f<strong>in</strong>d<strong>in</strong>gs concern<strong>in</strong>g<br />
judgements by l<strong>in</strong>guistically naive <strong>in</strong>formants. We compared structures such<br />
as (38a) <strong>and</strong> (38b) <strong>in</strong> a questionnaire study <strong>and</strong> found a highly significant<br />
difference between multiple questions that respect the <strong>MLC</strong> <strong>and</strong> those that<br />
do not. Structures violat<strong>in</strong>g the <strong>MLC</strong> were rated worse than those respect<strong>in</strong>g<br />
it (4.8 vs. 2.34) on a 1-6 scale (1: perfect, 6: completely ungrammatical) by<br />
l<strong>in</strong>guistically naive <strong>in</strong>formants.<br />
(38) a. Wer besucht wen <strong>in</strong> der Villa? 2.34<br />
who visited whom <strong>in</strong> the villa<br />
b. wen besucht wer <strong>in</strong> der Villa? 4.80<br />
Given that the syntax literature states more or less unanimously that German<br />
lacks simple superiority effects, such f<strong>in</strong>d<strong>in</strong>gs are a bit surpris<strong>in</strong>g at first<br />
glance, but they are <strong>in</strong> l<strong>in</strong>e with those obta<strong>in</strong>ed by Featherston (2002a,b),<br />
<strong>and</strong> they reappeared <strong>in</strong> a very similar shape <strong>in</strong> our questionnaire studies<br />
concern<strong>in</strong>g Polish <strong>and</strong> Russian.
<strong>The</strong> key to an underst<strong>and</strong><strong>in</strong>g of this difference between the syntacticians’<br />
wisdom <strong>and</strong> empirical f<strong>in</strong>d<strong>in</strong>gs lies <strong>in</strong> the observation that acceptability<br />
judgements are <strong>in</strong>fluenced by a variety of factors, among them be<strong>in</strong>g process<strong>in</strong>g<br />
difficulty. Object <strong>in</strong>itial structures are harder to process than their<br />
subject-<strong>in</strong>itial counterparts (as was already shown by Krems 1984 <strong>and</strong> Frazier<br />
<strong>and</strong> Flores d’Arcais 1989, see also Hemforth 1993, among many others),<br />
<strong>and</strong> it seems to be for exactly this reason that object-<strong>in</strong>itial structures are <strong>in</strong><br />
general rated worse than subject <strong>in</strong>itial ones <strong>in</strong> German, irrespective of<br />
whether a potential superiority violation is <strong>in</strong>volved or not (see Featherston<br />
2002b). <strong>The</strong> rat<strong>in</strong>g difference between (38a) <strong>and</strong> (38b) is thus not a proof<br />
that there is some underly<strong>in</strong>g <strong>MLC</strong>-based superiority effect <strong>in</strong> German, but<br />
if this l<strong>in</strong>e of reason<strong>in</strong>g is correct, it is hard to see on what basis one would<br />
have to assume a grammatical rather than a process<strong>in</strong>g account for the rat<strong>in</strong>g<br />
profile for Dutch (36).<br />
<strong>The</strong> second remark concerns the reappearance of argument-adjunct<br />
asymmetries <strong>in</strong> structures violat<strong>in</strong>g superiority. Wh-Objects may cross whsubjects<br />
<strong>in</strong> Swedish (35a), but wh-adjuncts do not have such a freedom: my<br />
five <strong>in</strong>formants unanimously rejected (39b), <strong>and</strong> accepted (39a) only.<br />
(39) Swedish adjunct superiority<br />
a. Vem skrattade varfoer<br />
who laughed why<br />
b.*Varfoer skrattade vem<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 95<br />
German, on the other h<strong>and</strong>, imposes no real restrictions on multiple questions<br />
<strong>in</strong>volv<strong>in</strong>g warum, ‘why’. We asked 17 non-l<strong>in</strong>guist native speakers of<br />
German to rate the grammaticality of (40). 15 of these accepted (40a), <strong>and</strong><br />
10 found (40b) grammatical as well.<br />
(40) Absence of superiority effects for German adjuncts<br />
a. wer lachte warum<br />
who laughed why<br />
b. warum lachte wer<br />
Presumably, this contrast is related to a further difference between Swedish<br />
<strong>and</strong> German. Multiple questions <strong>in</strong>volv<strong>in</strong>g two adjuncts were unanimously<br />
rejected by the Swedish <strong>in</strong>formants. German shows someth<strong>in</strong>g rem<strong>in</strong>iscent<br />
of a superiority effect <strong>in</strong> such multiple questions : (42a) was accepted by 9
96 Gisbert Fanselow<br />
of 17 <strong>in</strong>formants, while (42b) was judged as grammatical by three <strong>in</strong>formants<br />
only. To my ears, (42a) allows a pair-list read<strong>in</strong>g, while (42b) is restricted to<br />
a s<strong>in</strong>gle pair/echo <strong>in</strong>terpretation. See Haider (this volume) for an analysis of<br />
languages (not necessarily true for German) <strong>in</strong> which multiple questions<br />
must not <strong>in</strong>volve two adjuncts. Below, we will comment on the apparent<br />
superiority effect <strong>in</strong> (42).<br />
(41) Swedish multiple questions <strong>in</strong>volv<strong>in</strong>g two adjuncts<br />
a.*Varfoer bettedde sig barnen hur<br />
why behaved refl the children how<br />
b. *Hur betedde sig barnen varfoer<br />
(42) German multiple questions <strong>in</strong>volv<strong>in</strong>g two adjuncts<br />
a. Warum benahmen sich die K<strong>in</strong>der wie?<br />
why behaved refl the children how?<br />
b.*Wie benahmen sich die K<strong>in</strong>der warum?<br />
3.2. <strong>The</strong> absence of simple superiority effects: caused by low subject<br />
positions?<br />
At least two types of formal accounts for the absence of superiority effects <strong>in</strong><br />
(35) can be found <strong>in</strong> the literature, <strong>and</strong> we will discuss them <strong>in</strong> turn before<br />
we consider a pragmatic explanation. First, the def<strong>in</strong>ition of “closeness”<br />
central to the <strong>MLC</strong> might be modified, so that two phrases can be “equidistant”<br />
from a target position even if one of them asymmetrically c-comm<strong>and</strong>s<br />
the other. Second, additional movement operations might reverse the ccomm<strong>and</strong><br />
relations between wh-phrases before wh-movement.<br />
Whether a wh-phrase α may cross another wh-phrase β c-comm<strong>and</strong><strong>in</strong>g<br />
α depends on the def<strong>in</strong>ition of closeness <strong>in</strong> (1): If the <strong>MLC</strong> is def<strong>in</strong>ed as <strong>in</strong><br />
(43), cross<strong>in</strong>g is excluded <strong>in</strong> general, but if closeness is made precise <strong>in</strong> a<br />
more liberal way, as <strong>in</strong> (44), the <strong>MLC</strong> does not restrict the movement of<br />
phrases with<strong>in</strong> the same maximal projection.<br />
(43) <strong>MLC</strong>: Strict Version<br />
α cannot move to γ if there is a β that can also move to γ <strong>and</strong> that<br />
c-comm<strong>and</strong>s α
(44) <strong>MLC</strong>: Liberal Version<br />
α cannot move to γ if there is a β that can also move to γ <strong>and</strong> that<br />
asymmetrically m-comm<strong>and</strong>s α<br />
Suppose, then, that <strong>MLC</strong> effects are computed relative to (44). Whether a whobject<br />
may be moved across a wh-subject then depends on the hierarchical<br />
position of the subject. Subjects are base-generated <strong>in</strong> the VP. If the subject<br />
moves to Spec,IP as <strong>in</strong> (45a), it asymmetrically m-comm<strong>and</strong>s the object.<br />
<strong>The</strong>refore, an object cannot pass it on its way up to Spec, CP. If the subject<br />
stays <strong>in</strong> VP, as <strong>in</strong> (45b), the condition for the application of (44) is not met,<br />
so that the presence of a wh-subject does not <strong>in</strong>terfere with the prepos<strong>in</strong>g of<br />
a wh-object.<br />
(45) a.[ IP subject [ verb phrase [V object]]<br />
b.[ IP [ verb phrase subject [V object]]<br />
(44) thus l<strong>in</strong>ks the presence or absence of simple superiority effects to an<br />
<strong>in</strong>dependent parameter, viz., the location of the subject. Indeed, subjects<br />
need not move to Spec, IP <strong>in</strong> many of the languages (among them Spanish<br />
or German) that disrespect superiority. <strong>The</strong> “free <strong>in</strong>version” of subjects <strong>and</strong><br />
verbs <strong>in</strong> Spanish has always been taken as evidence that Spec,TP can be<br />
filled by an empty pleonastic pro, which allows the subject to stay <strong>in</strong> the<br />
verbal projection.<br />
(46) Free Inversion <strong>in</strong> Spanish<br />
le regalaron los estudiantes un libro<br />
her gave the students a book<br />
“the students gave her a book as a present”<br />
<strong>The</strong> view that thematic subjects need not leave the VP <strong>in</strong> German either is<br />
corroborated by constructions <strong>in</strong> which the VP precedes the second position<br />
auxiliary, as was noted by Haider (1986, 1990, 1993): <strong>The</strong> subject can be<br />
part of such VPs (47b,c), a fact suggest<strong>in</strong>g that it need not move to Spec, IP<br />
<strong>in</strong> overt syntax.<br />
(47) a.[Mädchen geküsst] hat er noch nie<br />
girls kissed has he not yet<br />
“he has not yet kissed any girls”<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 97
98 Gisbert Fanselow<br />
b.[Häuser gebrannt] haben hier noch nie<br />
houses burnt have here yet never<br />
“houses have never burnt here”<br />
c.[Mädchen geküsst] haben ihn noch nie<br />
girls kissed have him yet never<br />
“girls have not kissed him yet”<br />
Thus, there is <strong>in</strong>dependent evidence that (45b) is a legal constellation of<br />
German <strong>and</strong> Spanish. An <strong>MLC</strong> formulated as <strong>in</strong> (44) will not prevent the<br />
object from mov<strong>in</strong>g across the subject <strong>in</strong> (45b). In contrast, subjects must go<br />
to Spec,IP <strong>in</strong> English. Here, (45a) is the only constellation that can underlie<br />
multiple questions such as (3). Even <strong>in</strong> its liberal version (44), the <strong>MLC</strong><br />
prevents an object from cross<strong>in</strong>g a subject. <strong>The</strong> choice between (44a) <strong>and</strong><br />
(44b) is thus a good c<strong>and</strong>idate for an explanation of crossl<strong>in</strong>guistic variation<br />
concern<strong>in</strong>g simple superiority effects (see, e.g., Haider, this volume).<br />
While such an approach successfully captures basic superiority facts,<br />
more complex data are not readily expla<strong>in</strong>ed along these l<strong>in</strong>es. Consider<br />
Icel<strong>and</strong>ic first. One may want to relate the absence of superiority effects <strong>in</strong><br />
this language to (44), s<strong>in</strong>ce the existence of so-called transitive expletive<br />
constructions <strong>in</strong> Icel<strong>and</strong>ic suggests that thematic subjects may be placed<br />
<strong>in</strong>to lower positions than <strong>in</strong> English (see, e.g., Bobaljik <strong>and</strong> Jonas (1996)).<br />
Haider (2000, this volume) observes that movement to Spec,IP is an option<br />
for thematic subjects <strong>in</strong> Icel<strong>and</strong>ic, German, <strong>and</strong> Spanish, <strong>and</strong> notes that an<br />
explanation of the absence of superiority effects <strong>in</strong> terms of a low subject<br />
position predicts that one should observe English-type asymmetries whenever<br />
the position of adverbial material makes it clear that the thematic subject<br />
occupies a high position. Haider cites contrasts such as (48) (which he<br />
attributes to Ottósson (1989), <strong>and</strong> H. Sigurdsson, p.c.) as evidence for the<br />
claim that this prediction is borne out:<br />
(48) Superiority effects <strong>in</strong> Icel<strong>and</strong>ic <strong>and</strong> different subject positions<br />
a. Hva# hefur hver gefi# börnunum?<br />
what has who given the-children?<br />
b.*Hva# hefur hver oft gefi# börnunum?<br />
what has who often given the-children?<br />
“who has often given what to the children?”<br />
<strong>The</strong> availability of a low position for the subject hver <strong>in</strong> the verbal projection<br />
expla<strong>in</strong>s the grammaticality of (48a). In (48b), however, the subject precedes
oft ‘often’, i.e., it precedes an element adjo<strong>in</strong>ed to VP, <strong>and</strong> occupies a high<br />
position <strong>in</strong> the clause. <strong>The</strong> ungrammaticality of (48b) suggests, then, that<br />
the position of the thematic subject is the crucial factor govern<strong>in</strong>g superiority<br />
effects, as Haider argues. <strong>The</strong> argument presupposes, however, that<br />
(48b) becomes perfect when the order of the subject <strong>and</strong> the adverb is<br />
reversed. Accord<strong>in</strong>g to my <strong>in</strong>formant (-orste<strong>in</strong>n Hjaltason), this expectation<br />
is not fulfilled. Rather, we get the follow<strong>in</strong>g array of relative judgements:<br />
(49) Superiority effects <strong>in</strong> Icel<strong>and</strong>ic <strong>and</strong> different subject positions<br />
a. Hva# hefur hver gefi# börnunum?<br />
what has who given the children<br />
b.?Hva# hefur hver oft gefi# börnunum?<br />
what has who often given the children<br />
c.?Hva# hefur hver oft gefi# hverjum?<br />
what has who often given whom?<br />
d.*Hva# hefur oft hver gefi# börnunum?<br />
e. Hva# hefur hva#a fa#ir oft gefi# börnunum?<br />
what has which father often given the children<br />
f.*Hva# hefur oft hva#a fa#ir gefi# börnunum?<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 99<br />
(49a) is grammatical because Icel<strong>and</strong>ic shows no superiority effects. (49b)<br />
is less acceptable, but this effect is not elim<strong>in</strong>ated by the addition of a third<br />
wh-phrase (49c), as it should be if the phenomenon is related to the <strong>MLC</strong>.<br />
Most importantly, the structure becomes fully ungrammatical when the<br />
order of the subject <strong>and</strong> the adverb is reversed. <strong>The</strong> status of (49d) is quite<br />
unexpected, because the order of subject <strong>and</strong> adverb seems to imply a low<br />
position for the former. <strong>The</strong> ungrammaticality of (49d) is matched by the<br />
one of (49f), which <strong>in</strong>volves a d-l<strong>in</strong>ked wh-phrase. Whatever may be<br />
responsible for the contrasts <strong>in</strong> (48) <strong>and</strong> (49) – the <strong>MLC</strong> is not likely to<br />
come <strong>in</strong>to play.<br />
Similarly, the grammaticality of (50a) might be related to the low position<br />
occupied by the unaccusative subject <strong>in</strong> this example. 4 However, the structure<br />
does not degrade dramatically when the subject is placed <strong>in</strong>to the slot preced<strong>in</strong>g<br />
the object pronoun (that is, when it presumably moves to Spec,IP):<br />
five out of a total of eight Dutch l<strong>in</strong>guists I consulted found (50b) completely<br />
unobjectionable.
100 Gisbert Fanselow<br />
(50) Different subject positions <strong>and</strong> Dutch superiority<br />
a. wanneer is hem wat overkomen<br />
when is him what happened<br />
b.wanneer is wat hem overkomen<br />
<strong>The</strong> position of the subject is also not completely irrelevant for the wellformedness<br />
of multiple questions <strong>in</strong> German. Consider the contrasts <strong>in</strong><br />
(51) 5 , <strong>in</strong> which a non-subject has been placed <strong>in</strong> front of a wh-subject. Such<br />
constructions fail to be fully grammatical (to different degrees) when the<br />
subject precedes a clitic object pronoun (51b,d), an unstressed object pronoun<br />
(51f), or one of the particles like denn which have been claimed to<br />
mark the VP boundary <strong>in</strong> German (51h) (see Dies<strong>in</strong>g 1992, Me<strong>in</strong>unger<br />
1995, for a discussion of VP boundaries).<br />
(51) Different subject positions <strong>in</strong> German multiple questions<br />
a. wann hat’s wer gesehen<br />
when has it who seen<br />
b.?*wann hat wer’s gesehen<br />
“who saw it when?”<br />
c. wem hat`s wer gegeben<br />
who.dat has it who given<br />
d.?*wem hat wer’s gegeben<br />
e. wem hat es wer gegeben<br />
f.?*wem hat wer es gegeben<br />
“who gave it to whom”<br />
g. was hat denn wer gesagt<br />
what has ptc.who said<br />
h.?*was hat wer denn gesagt<br />
“who said what”<br />
Multiple questions are less grammatical when a wh-phrase crosses a whsubject<br />
that has moved to Spec,IP. It is tempt<strong>in</strong>g to expla<strong>in</strong> such contrasts <strong>in</strong><br />
terms of the assumption that the wh-subject asymmetrically c-comm<strong>and</strong>s<br />
the trace of the wh-object <strong>in</strong> the ungrammatical examples, so that the <strong>MLC</strong><br />
(44) would block the bad structures.
Such an analysis is not conv<strong>in</strong>c<strong>in</strong>g, however. It does not take <strong>in</strong>to account<br />
the fact that the same or similar contrasts show up <strong>in</strong> constructions for<br />
which the <strong>MLC</strong> cannot be relevant. German wh-words are ambiguous<br />
between an <strong>in</strong>terrogative <strong>and</strong> an <strong>in</strong>def<strong>in</strong>ite <strong>in</strong>terpretation. <strong>The</strong> restrictions on<br />
the placement of <strong>in</strong>terrogative wh-subjects exemplified <strong>in</strong> (51) are exactly<br />
mirrored by comparable restrictions on the placement of <strong>in</strong>def<strong>in</strong>ite wh-subject<br />
pronouns, as (52) shows. <strong>The</strong> <strong>in</strong>def<strong>in</strong>ite subjects <strong>in</strong> (52) share the distribution<br />
of wh-phrases, but they do not <strong>in</strong>teract with any other element <strong>in</strong> the<br />
clause <strong>in</strong> terms of the <strong>MLC</strong>. <strong>The</strong>refore, the <strong>MLC</strong> cannot expla<strong>in</strong> (52), <strong>and</strong> it<br />
would be strange if it accounted for the same distribution of data <strong>in</strong> (51).<br />
(51) <strong>and</strong> (52) show that German syntax imposes restrictions on the placement<br />
of subjects that are not def<strong>in</strong>ite. <strong>The</strong> <strong>MLC</strong> is not responsible for these.<br />
(52) <strong>Effects</strong> of the subject position for <strong>in</strong>def<strong>in</strong>ite pronouns<br />
a. dann hat’s wer gesehen<br />
then has it someone seen<br />
“then, someone saw it”<br />
b.??dann hat wer’s gesehen<br />
c. dem hat`s wer gegeben<br />
him.dat has it someone given<br />
“someone gave it to him”<br />
d.?*dem hat wer’s gegeben<br />
e. dem hat es wer gegeben<br />
f.?*dem hat wer es gegeben<br />
g. hat denn wer angerufen<br />
has ptc.someone called<br />
“did someone call?”<br />
h.?*hat wer denn angerufen?<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 101<br />
A central prediction of an account of (absent) superiority effects that exploits<br />
differences <strong>in</strong> the placement of subjects is not borne out: <strong>in</strong> a number<br />
of languages that fail to show superiority effects (German, Icel<strong>and</strong>ic, <strong>and</strong><br />
perhaps Dutch), the actual position of the subject does not <strong>in</strong>fluence the<br />
grammaticality of multiple question <strong>in</strong> the expected way.
102 Gisbert Fanselow<br />
3.3. <strong>The</strong> absence of simple superiority effects: caused by scrambl<strong>in</strong>g?<br />
A second attempt of captur<strong>in</strong>g (35) assumes that the object <strong>in</strong> fact c-comm<strong>and</strong>s<br />
the subject at the po<strong>in</strong>t of derivation when movement to Spec, CP is<br />
carried out. Under this circumstance, the <strong>MLC</strong> does not have to be relaxed<br />
<strong>in</strong> order to expla<strong>in</strong> (35): Given that the order object > subject is <strong>in</strong> pr<strong>in</strong>ciple<br />
always grammatical <strong>in</strong> a German (53a,b), the question arises whether (53c)<br />
really is not <strong>in</strong> l<strong>in</strong>e with even the strictest version of the <strong>MLC</strong>. After all,<br />
(53c) might be derived from (53d) rather than (53e). In the former case, the<br />
highest wh-phrase is moved to Spec,CP <strong>in</strong> (53c), as predicted by the <strong>MLC</strong>.<br />
(53) Object-subject order <strong>in</strong> German <strong>and</strong> the <strong>MLC</strong><br />
a. dass fast jeden jem<strong>and</strong> angerufen hatte<br />
that nearly everyone.acc someone.nom called had<br />
“that someone had called nearly everyone”<br />
b. dass fast jeden wer angerufen hatte<br />
that nearly everone.acc someone.nom called had<br />
“that someone had called nearly everyone”<br />
c. wen hat wer e<strong>in</strong>geladen<br />
who.acc has who.nom <strong>in</strong>vited<br />
“who has <strong>in</strong>vited whom?”<br />
d. hat [wen [wer e<strong>in</strong>geladen]]<br />
e. hat [wer [wen e<strong>in</strong>geladen]]<br />
In other words, (53c) might be grammatical because additional movement<br />
operations (scrambl<strong>in</strong>g) can change the c-comm<strong>and</strong> relations established by<br />
Merge. 6 If the object can <strong>in</strong> general be placed <strong>in</strong> front of the subject, structures<br />
such as (53d) can be derived <strong>in</strong> which the wh-object c-comm<strong>and</strong>s the<br />
wh-subject. Even <strong>in</strong> its strictest version, the <strong>MLC</strong> cannot block the subsequent<br />
movement of the wh-object to Spec, CP. See, e.g., Fanselow (1998,<br />
2001), Haider (1986), Wiltschko (1998), among others, for different versions<br />
of this account.<br />
Accord<strong>in</strong>g to Fanselow (1998), the contrasts <strong>in</strong> (54) corroborate the view<br />
that apparent violations of superiority are licensed by scrambl<strong>in</strong>g. Certa<strong>in</strong><br />
wh-phrases such as wen von den Studenten (54a) or was für Frauen (54d)<br />
can either move to Spec,CP as a whole, or be split up <strong>in</strong> simple <strong>and</strong> multiple<br />
questions (54b,e). In the latter case, only the wh-part of the phrase under-
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 103<br />
goes front<strong>in</strong>g, whereas the rema<strong>in</strong><strong>in</strong>g part is str<strong>and</strong>ed. <strong>The</strong> str<strong>and</strong>ed material<br />
<strong>in</strong>dicates the position from which the phrase has been attracted to Spec, CP.<br />
<strong>The</strong> ungrammaticality of (54c,f) suggests, then, that a wh-phrase cannot<br />
cross another one <strong>in</strong> German, either. Objects may undergo overt wh-movement<br />
<strong>in</strong> multiple questions, but only if movement starts <strong>in</strong> a position c-comm<strong>and</strong><strong>in</strong>g<br />
a wh-subject.<br />
(54) Superiority <strong>and</strong> Splitt<strong>in</strong>g<br />
a. wen von den Studenten hat heute wer e<strong>in</strong>geladen?<br />
who.acc of the students has today who.nom <strong>in</strong>vited<br />
b. wen hat [von den Studenten] heute wer heute e<strong>in</strong>geladen?<br />
c.*wen hat heute wer abends von den Studenten<br />
e<strong>in</strong>geladen<br />
who has today who <strong>in</strong> the even<strong>in</strong>g of the students<br />
<strong>in</strong>vited<br />
“who has <strong>in</strong>vited which of the students today (<strong>in</strong> the even<strong>in</strong>g)s”<br />
d. was für Frauen hat wer heute e<strong>in</strong>geladen<br />
what for women has who.nom today <strong>in</strong>vited<br />
e.<br />
“who has <strong>in</strong>vited which k<strong>in</strong>d of women today”<br />
was hat für Frauen wer heute e<strong>in</strong>geladen<br />
f.??was hat wer für Frauen heute e<strong>in</strong>geladen<br />
Pesetsky (2000) po<strong>in</strong>ts out that contrasts such as the ones <strong>in</strong> (54) f<strong>in</strong>d an<br />
explanation <strong>in</strong> terms of the <strong>in</strong>tervention effects analysed by Beck (1996), see<br />
also Mathieu (2002). (55) shows that the parts of a discont<strong>in</strong>uous wh-phrase<br />
must not be separated by any k<strong>in</strong>d of operator <strong>in</strong> German. An <strong>in</strong>tervention<br />
account can expla<strong>in</strong> (54) <strong>and</strong> (55) at the same time, while the <strong>MLC</strong>-based<br />
explanation for (54) cannot be easily extended to (55).<br />
(55) Intervention effects <strong>and</strong> Split noun phrases<br />
a. was hat er für Frauen nicht getroffen<br />
what has he for women not met<br />
“what k<strong>in</strong>d of woman did he not meet?”<br />
b.*was hat er nicht für Frauen getroffen<br />
Pesetsky’s observation certa<strong>in</strong>ly establishes that data such as (54) cannot be<br />
used to show that object wh-movement cannot orig<strong>in</strong>ate below a wh-subject
104 Gisbert Fanselow<br />
<strong>in</strong> a multiple question. Notice, however, that (54b,e) still show that whextraction<br />
of an object may start <strong>in</strong> a position c-comm<strong>and</strong><strong>in</strong>g a wh-subject.<br />
Reference to (53b), i.e., to the grammaticality of structures <strong>in</strong> which the<br />
object occupies a higher position than the subject, thus seems to be <strong>in</strong> general<br />
a sufficient 7 (though not a necessary) condition for the absence of simple<br />
superiority effects <strong>in</strong> a language.<br />
Unfortunately, the scrambl<strong>in</strong>g solution for (35) cannot be applied <strong>in</strong> all<br />
languages <strong>in</strong> which superiority effects are absent, because quite a number of<br />
them (Swedish, Icel<strong>and</strong>ic, French, Dutch) do not have free constituent order<br />
generated by scrambl<strong>in</strong>g! 8<br />
3.4. Pragmatics<br />
In spite of its shortcom<strong>in</strong>gs, the scrambl<strong>in</strong>g account has an attractive feature:<br />
it implies that the choice between the object- <strong>and</strong> the subject-<strong>in</strong>itial versions<br />
of a multiple question is never arbitrary <strong>in</strong> the languages that tolerate (35).<br />
Scrambl<strong>in</strong>g can place an object <strong>in</strong> front of a subject only if the latter is more<br />
focal than the former. <strong>The</strong>refore, the scrambl<strong>in</strong>g account of miss<strong>in</strong>g superiority<br />
predicts that apparent superiority violations are acceptable under certa<strong>in</strong><br />
pragmatic circumstances (those that would license scrambl<strong>in</strong>g) only. This<br />
prediction is borne out. <strong>The</strong> pragmatic conditions of use of (56a) <strong>and</strong> (56b)<br />
are different. <strong>The</strong>y require different “sort<strong>in</strong>g keys” (Comorovski 1996).<br />
Answers to (56b) are well-formed if the object of the clause represents a<br />
contrastive topic. <strong>The</strong>re are no comparable restrictions on the wellformedness<br />
of (56a).<br />
(56) Absence of superiority <strong>in</strong> German<br />
a. wer hat wen gesehen<br />
who.nom has who.acc seen<br />
b. wen hat wer gesehen<br />
“who has seen whom?”<br />
This pragmatic dependency becomes evident when one considers the m<strong>in</strong>itexts<br />
<strong>in</strong> (57). <strong>The</strong> a.- <strong>and</strong> b. examples <strong>in</strong>troduce the referents of the subject<br />
<strong>and</strong> the object, respectively, as known to the speaker. <strong>The</strong>se referents constitute<br />
the “sort<strong>in</strong>g keys” for the multiple questions a’ <strong>and</strong> b’, they are discoursel<strong>in</strong>ked<br />
(see Pesetsky 1987). (57a) can only be cont<strong>in</strong>ued by (57a’), <strong>and</strong><br />
(57b) only by (57b’).
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 105<br />
(57) Discourse <strong>in</strong>fluence on superiority violations <strong>in</strong> German<br />
wir haben bereits herausgefunden<br />
we have already found out<br />
a. wer jem<strong>and</strong>en gestern anrief, und wer nicht<br />
who.nom someone.acc yesterday called <strong>and</strong> who.nom not<br />
b. wen jem<strong>and</strong> gestern anrief, und wen nicht<br />
who .acc someone.nom yesterday called <strong>and</strong> who.acc not<br />
Aber wir s<strong>in</strong>d nicht eher zufrieden, bis wir auch wissen<br />
But we are not earlier content until we also know<br />
a’. wer WEN angerufen hat<br />
who.nom who.acc called has<br />
b’. wen WER angerufen hat<br />
In other words, a wh-object can precede a wh-subject <strong>in</strong> German if the former<br />
is more topical than the latter. Out of the blue wh-questions allow subject ><br />
object order, only. This is particularly clear when the predicate is symmetric<br />
(such as treffen, “meet”) as <strong>in</strong> (58), so that discourse-l<strong>in</strong>ked differentiations<br />
of subjects <strong>and</strong> topics are very hard to imag<strong>in</strong>e.<br />
(58) Superiority effects <strong>in</strong> out of the blue contexts<br />
Erzähl mir was über die Party. “Tell me someth<strong>in</strong>g about the party”<br />
a. Wer hat wen getroffen?<br />
who.nom has who.acc met?<br />
b.??Wen hat wer getroffen<br />
“who met who?”<br />
Ste<strong>in</strong>itz (1969) was the first to observe that modal or sentence level adverbs<br />
resist reorder<strong>in</strong>g <strong>in</strong> the <strong>in</strong>terest of <strong>in</strong>formation structure. <strong>The</strong> adverbial<br />
“superiority” effects discussed <strong>in</strong> (42) can be accounted for <strong>in</strong> these terms.<br />
<strong>The</strong> languages that lack simple superiority effects do not differ <strong>in</strong> this<br />
respect: constituent order reflects <strong>in</strong>formation structure. Different types of<br />
operations conspire to guarantee that focal <strong>in</strong>formation is preceded by topical<br />
one: scrambl<strong>in</strong>g (German, Japanese, Polish), topicalization to Spec, CP<br />
(Swedish, Icel<strong>and</strong>ic, German), or subject placement <strong>in</strong> Spec, IP or VP<br />
(Spanish, German). In the most parsimonious account, these operations are<br />
driven by a constra<strong>in</strong>t C-INF that requires that topical material c-comm<strong>and</strong>
106 Gisbert Fanselow<br />
focal elements (but more luxurious theories of <strong>in</strong>formation structure would<br />
have the same effect). If C-INF plays a role <strong>in</strong> determ<strong>in</strong><strong>in</strong>g the well-formedness<br />
of partial LFs <strong>in</strong> multiple questions as well, then (34d) (repeated here<br />
for convenience) is able to block (34e) only if this does not prevent a particular<br />
distribution of focality/topicality among the wh-phrases from be<strong>in</strong>g expressed<br />
with<strong>in</strong> the limits imposed by C-INF. If the higher degree of topicality<br />
of wh 2 must be expressed, (34e) can be chosen. Information structure overrides<br />
the <strong>MLC</strong>.<br />
(34) d.[wh1 [ Σ CompA A - - - [wh1 - - - [ wh2 A - - - ]]]]]<br />
e.[wh 2 [ Σ CompA - - - [wh A A<br />
1 - - - [ wh2 - - - ]]]]]<br />
4. <strong>The</strong> Nature of Exceptions<br />
While the absence of simple superiority effects <strong>in</strong> the <strong>in</strong>terest of <strong>in</strong>formation<br />
structure is a widespread phenomenon, it is far from be<strong>in</strong>g universal, as evidenced,<br />
e.g., by the relevant English data. <strong>The</strong> contrast between English <strong>and</strong><br />
German <strong>in</strong> the formation of multiple questions might be <strong>in</strong>dicative of the<br />
different importance the languages attribute to C-INF: <strong>in</strong> German, its effects<br />
are stronger than the <strong>MLC</strong>, while it is the other way round <strong>in</strong> English. <strong>The</strong><br />
<strong>in</strong>teraction of the <strong>MLC</strong> <strong>and</strong> the constra<strong>in</strong>s of <strong>in</strong>formation structure would<br />
thus be rem<strong>in</strong>iscent of an optimality theoretic framework (for OT accounts<br />
of <strong>MLC</strong>-effects, see, e.g., Müller 2001, <strong>and</strong> the constributions by Hale <strong>and</strong><br />
Legendre, Lee, <strong>and</strong> Vogel, this volume). We will argue, however, that such a<br />
conclusion is not warranted. <strong>The</strong> <strong>MLC</strong> is never stronger than C-INF.<br />
4.1. Bulgarian<br />
Roumanian <strong>and</strong> Bulgarian are languages cited frequently when one wants to<br />
substantiate the claim that superiority effects are not conf<strong>in</strong>ed to English.<br />
(59) Simple superiority effects <strong>in</strong> Roumanian <strong>and</strong> Bulgarian<br />
a. c<strong>in</strong>e ce cumpara<br />
who what buys<br />
b.*ce c<strong>in</strong>e cumpara<br />
c. koj kogo vizda<br />
who whom sees<br />
d.*kogo koj vizda
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 107<br />
<strong>The</strong> Slavic languages other than Bulgarian such as Czech, Polish or Russian<br />
allow superiority violations, however. <strong>The</strong> sem<strong>in</strong>al study of Rud<strong>in</strong> (1988)<br />
<strong>in</strong>itiated an impressive series of studies that try to account for this <strong>and</strong> other<br />
differences among the Slavic languages, cf., Błaszczak <strong>and</strong> Fischer (2002)<br />
for an overview. <strong>The</strong> proposal advanced by Bo‰koviç (2002) (see also<br />
Bo‰koviç 1997, Stepanov 1998) is the most <strong>in</strong>terest<strong>in</strong>g one <strong>in</strong> the context of<br />
the preced<strong>in</strong>g section. Accord<strong>in</strong>g to him, wh-phrases move to specifier positions<br />
def<strong>in</strong>ed <strong>in</strong> terms of <strong>in</strong>formation structure (focus) <strong>in</strong> Polish, Russian, or<br />
German, while movement targets a pure [+wh] specifier <strong>in</strong> Bulgarian or<br />
English. This might fit <strong>in</strong>to the preced<strong>in</strong>g discussion <strong>in</strong> the follow<strong>in</strong>g way:<br />
when phrases move to [+wh]-specifier, only the attract<strong>in</strong>g feature is grammatically<br />
visible, so that additional features of <strong>in</strong>formation structure will not<br />
<strong>in</strong>terfere with the application of the <strong>MLC</strong>. However, when XPs are attracted<br />
to heads def<strong>in</strong>ed <strong>in</strong> terms of <strong>in</strong>formation structure, it is the distribution of<br />
the pert<strong>in</strong>ent features that determ<strong>in</strong>es how attraction is carried out.<br />
It is doubtful, however, that a model draw<strong>in</strong>g a sharp l<strong>in</strong>e between<br />
Bulgarian <strong>and</strong> the other Slavic languages is adequate. <strong>The</strong> <strong>in</strong>tuition represented<br />
<strong>in</strong> (59c,d) is not shared by all native speakers of Bulgarian: two of the<br />
five native speakers that I have consulted accept a sentence such as kakvo<br />
koj pravi? “what who did” provided that koj “who” is stressed. It is not<br />
obvious, then, that the judgement pattern for Bulgarian multiple questions is<br />
qualitatively different from the one for German or Dutch.<br />
Even if we disregard the empirical issue of whether (59d) is really ungrammatical<br />
<strong>in</strong> Bulgarian (<strong>and</strong> not just rejected by some speakers, similar to<br />
what holds for German or Polish superiority violations), the contrast between<br />
(59c-d) is not identical with the one we f<strong>in</strong>d <strong>in</strong> English. A number of differences<br />
to English come out clearly. In Bulgarian, strict superiority effects can<br />
be found for animate subjects only. When the subject is <strong>in</strong>animate, <strong>and</strong> the<br />
object animate, both orders are f<strong>in</strong>e, as Bill<strong>in</strong>gs <strong>and</strong> Rud<strong>in</strong> (1996: 38) have<br />
observed.<br />
(60) Absence of superiority effects with <strong>in</strong>animate subjects of transitive<br />
verbs<br />
a. Kogo kakvo e udarilo?<br />
whom.acc what.nom CL hit<br />
b. Kakvo kogo e udarilo?<br />
“What hit whom?”
108 Gisbert Fanselow<br />
No superiority effects show up with psychological predicates, as (61) illustrates.<br />
Sometimes, subject-<strong>in</strong>itial sentences even seem worse than sentences<br />
beg<strong>in</strong>n<strong>in</strong>g with the dative wh-phrase:<br />
(61) Absence of superiority effects with psychological predicates<br />
a. Koj na kogo mu xaresva?<br />
who.nom whom.dat CL-dat 3.sg is-pleas<strong>in</strong>g<br />
(literally) “Who is likeable to whom?”<br />
b. Na kogo koj mu xaresva?<br />
c. ??Kakvo na kogo mu xaresva?<br />
what.nom to whom.das CL-dat.3.sg is-pleas<strong>in</strong>g<br />
(literally) “what is likeable to whom?’<br />
d. Na kogo kakvo mu xaresva?<br />
Superiority effects are therefore restricted to external arguments of transitive<br />
verbs, <strong>and</strong> even for them, the only defensible generalization is the one offered<br />
by Bill<strong>in</strong>gs <strong>and</strong> Rud<strong>in</strong> (1996: 46) “If the wh-external argument is human<br />
(i.e., koj), then it must appear first <strong>in</strong> the wh-cluster.”<br />
That such a constra<strong>in</strong>t on wh-clusters may be necessary quite <strong>in</strong>dependent<br />
of any considerations of superiority is suggested by the fact that Bulgarian<br />
differs from English with respect to ternary questions as well. Kayne<br />
(1983), Hornste<strong>in</strong> (1995), Pesetsky (2000) <strong>and</strong> others have observed that<br />
superiority need not be respected <strong>in</strong> ternary questions: even the lowest whphrase<br />
can be fronted.<br />
(62) Absence of superiority effects <strong>in</strong> English ternary questions<br />
a. what did who buy where?<br />
b. what did who persuade who to buy<br />
Cancellation effects due to the addition of a third wh-phrase exist <strong>in</strong><br />
Bulgarian, too (see (63)), but the examples used <strong>in</strong> the literature <strong>and</strong> the<br />
<strong>in</strong>tutions of my <strong>in</strong>formant Penka Stateva suggest that the liberaliz<strong>in</strong>g effect<br />
never affects subject koj.
(63) Restricted liberalization of superiority <strong>in</strong> Bulgarian<br />
a. Koj kogo kakvo e pital?<br />
who whom what is asked<br />
“Who asked whom what?’<br />
b. Koj kakvo kogo e pital?<br />
c. Koj kogo kak e tselunal?<br />
who whom how is kissed<br />
“Who kissed whom how?”<br />
d. Koj kak kogo e tselunal?<br />
e. Koj kogo kŭde e vidjal?<br />
who whom where is seen<br />
f. Koj kŭde kogo e vidjal?<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 109<br />
Recall also that the order<strong>in</strong>g restrictions of Bulgarian koj do not show the<br />
<strong>in</strong>terpretation-sensitivity of the English superiority effect. <strong>The</strong> judgements<br />
for (9) – repeated for convenience – seem to correlate with the jugdements<br />
for simple kakvo koj kupi. If the <strong>MLC</strong> would be responsible for the ungrammaticality<br />
of (59d), it would be unclear why the condition is not <strong>in</strong>terpretation-sensitive<br />
<strong>in</strong> Bulgarian, whereas it is <strong>in</strong> English <strong>and</strong> German.<br />
(9)Anti-superiority <strong>in</strong> Bulgarian<br />
a.#koj se chudi, kakvo koj kupi?<br />
who wonders what who bought<br />
“who wonders what who bought?”<br />
b.#na kogo kaza, kakvo koj kupi?<br />
who.dat you-tell what who bought<br />
<strong>The</strong> <strong>MLC</strong> is thus not a likely cause for the order<strong>in</strong>g restrictions <strong>in</strong><br />
Bulgarian. A simple account can, however, be formulated <strong>in</strong> terms of the fact<br />
that Bulgarian is a multiple front<strong>in</strong>g language. One of the crucial <strong>in</strong>sights of<br />
Rud<strong>in</strong> (1988) was that the peculiarities <strong>in</strong> the behavior of Bulgarian (as compared<br />
to other Slavic languages) can be related to the fact that Bulgarian is a<br />
“multiple filler” language: all wh-phrases must be preposed <strong>in</strong> a multiple<br />
question (unless they are discourse-l<strong>in</strong>ked). Suppose that sequences of whpronouns<br />
form a cluster, <strong>and</strong> that the morphophonological realization of this<br />
cluster is subject to the k<strong>in</strong>d of rules that also govern the l<strong>in</strong>ear arrangement
110 Gisbert Fanselow<br />
of sequences of clitics <strong>and</strong> <strong>in</strong>flectional affixes. As Bonet (1991), Halle (1992)<br />
<strong>and</strong> Noyer (1992) show, the order of elements <strong>in</strong> such clusters cannot be<br />
exclusively predicted from syntax. Rather, <strong>in</strong>dependent pr<strong>in</strong>ciples of morphology<br />
are needed, a view that is well-established nowadays <strong>in</strong> the theory<br />
of distributed morphology (Halle <strong>and</strong> Marantz 1993,1994).<br />
<strong>The</strong>re is <strong>in</strong>dependent evidence that the composition of wh-phrases <strong>in</strong><br />
clusters is governed by non-syntactic pr<strong>in</strong>ciples <strong>in</strong> Bulgarian. Bill<strong>in</strong>gs <strong>and</strong><br />
Rud<strong>in</strong> (1996: 43) suggest that (64b) is ungrammatical because *na kogo<br />
kogo ‘to whom whom’ violates a ban aga<strong>in</strong>st consecutive wh-homophones.<br />
In colloquial Bulgarian, na kogo can be replaced by na koj, <strong>in</strong> which case<br />
both orders of the objects are f<strong>in</strong>e:<br />
(64) Phonological restrictions <strong>in</strong> wh-clusters <strong>in</strong> Bulgarian<br />
a. Koj kogo na kogo e pokazal?<br />
who.nom whom.acc to whom.dat CL showed<br />
b. *Koj na kogo kogo e pokazal?<br />
c. Koj kogo na koj e pokazal?<br />
who.nom whom.acc to who.dat CL showed<br />
d. Koj na koj kogo e pokazal?<br />
“Who po<strong>in</strong>ted out whom to who?”<br />
It natural to assume that further templatic constra<strong>in</strong>ts determ<strong>in</strong>e the arrangement<br />
of wh-pronouns <strong>in</strong> the cluster, among them a requirement that koj must<br />
come first <strong>in</strong> a truly transitive construction. This requirement implies the<br />
contrast <strong>in</strong> (59). S<strong>in</strong>ce it is a PF constra<strong>in</strong>t, considerations of expressivity<br />
will not play a role, as required.<br />
Some observations from other languages lend support to the view that<br />
cluster formation is crucial <strong>in</strong> establish<strong>in</strong>g order<strong>in</strong>g restrictions that resemble<br />
(but fail to be) superiority effects. Languages <strong>in</strong> which cluster formation is<br />
optional are of particular relevance here. In Yiddish, multiple front<strong>in</strong>g of whphrases<br />
to a position preced<strong>in</strong>g the verb is possible, but not m<strong>and</strong>atory. In<br />
wh-clusters, word order is strict (65a,b) but it is free when only one wh-phrase<br />
is placed <strong>in</strong>to preverbal position (65c,d), see Hoge (2000) for discussion.<br />
Likewise, <strong>in</strong> Hebrew, superiority can be violated only if the verb is placed<br />
between the two wh-phrases, although <strong>in</strong>version <strong>in</strong> wh-questions is not necessary<br />
as such (66):
(65) Multiple questions <strong>in</strong> Yiddish<br />
a. ver vemen hot kritikirt?<br />
who whom has criticised<br />
b.*vemen ver hot kritikirt?<br />
c. ver hot vemen kritikirt?<br />
who has whom criticise<br />
d. vemen hot ver kritikirt?<br />
whom has who criticised<br />
“who criticised whom?”<br />
(66) Superiority <strong>in</strong> Hebrew<br />
a. ma kana mi<br />
what bought who<br />
b.*ma mi kana<br />
For obvious reasons, wh-pronouns cannot form a cont<strong>in</strong>uous cluster when<br />
they are separated by a verb. <strong>The</strong> data <strong>in</strong> (65) <strong>and</strong> (66) can be captured easily<br />
<strong>in</strong> a model that allows for templatic order<strong>in</strong>g restrictions of wh-phrases which<br />
apply when syntax is spelt out. Grewendorf (1999, 2001) <strong>and</strong> Hoge (2000)<br />
account for superiority <strong>in</strong> Bulgarian by cluster formation as well, but <strong>in</strong> a<br />
fairly different way.<br />
4.2. English<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 111<br />
English superiority effects are difficult to account for <strong>in</strong> the model we propose.<br />
This is not necessarily a negative aspect: superiority effects <strong>in</strong> English<br />
are distributed <strong>in</strong> a very complex way, for which it is not clear at all how it<br />
could be captured <strong>in</strong> a simple <strong>MLC</strong> account.<br />
Intervention effects disappear <strong>in</strong> English when the wh-phrases allow a<br />
context-related <strong>in</strong>terpretation. Pesetsky (1987) shows that (67a) is f<strong>in</strong>e because<br />
it has a “discourse-l<strong>in</strong>ked” <strong>in</strong>terpretation: a wh-phrase is discoursel<strong>in</strong>ked<br />
if its <strong>in</strong>terpretation relates to a contextually given set of objects <strong>and</strong><br />
persons, from which one tries to pick a relevant one with the wh-phrase.<br />
Thus, the d-l<strong>in</strong>ked wh-phrase <strong>in</strong> (67a) generates s contrastive topic for the<br />
answers, as it does <strong>in</strong> German. As Bol<strong>in</strong>ger (1978) observes, proper contexts<br />
even license the absence of <strong>in</strong>tervention effects for wh-pronouns, as <strong>in</strong> (67b).
112 Gisbert Fanselow<br />
(67) Absence of superiority effects <strong>in</strong> certa<strong>in</strong> contexts <strong>in</strong> English<br />
a. which book did which person read<br />
b. I know what everyone was supposed to do. But what did who<br />
actually do?<br />
However, reference to (67) allone does not expla<strong>in</strong> why (68a) sounds bad to<br />
the English ear, while its one-to-one translation <strong>in</strong>to German (68b) is grammatical.<br />
(68) a.*what will who see<br />
b. was wird wer sehen<br />
<strong>The</strong> key to an underst<strong>and</strong><strong>in</strong>g of this contrast lies <strong>in</strong> the observation that who<br />
is a topic <strong>in</strong> (68a), while wer can be focal <strong>in</strong> (68b), <strong>and</strong> bear focal stress. If<br />
wh-pronouns are <strong>in</strong>herently <strong>in</strong>def<strong>in</strong>ite, <strong>and</strong> constitute bad topics, the different<br />
status of (68a,b) can be understood. A number of facts support this view. First,<br />
the acceptability of a cross<strong>in</strong>g structure depends of the degree to which the<br />
subject wh-phrase can be <strong>in</strong>terpreted as a referential category, as discoursel<strong>in</strong>ked,<br />
as a potential topic.<br />
(69) Cross<strong>in</strong>g effects as a function of the potential topicality of the subject<br />
a. what did a friend of who say to Bill?<br />
b. what did whose friends say to Bill?<br />
c.*what did each friend of who say to Bill (Hornste<strong>in</strong> 1995: 147)<br />
d.*what did how many men buy?<br />
Second, Erteschik-Shir (1997: 190) observes that object <strong>in</strong>itial questions are<br />
<strong>in</strong> general relatively bad <strong>in</strong> English when the subject is an <strong>in</strong>def<strong>in</strong>ite, a weak<br />
quantifier. Obviously, the unacceptability of (70b) (with a non-generic nonspecific<br />
read<strong>in</strong>g of a boy) cannot be expla<strong>in</strong>ed <strong>in</strong> terms of the <strong>MLC</strong>. But if<br />
<strong>in</strong>def<strong>in</strong>ites are bad as such <strong>in</strong> the subject position of questions, one does not<br />
need to additionally <strong>in</strong>voke the <strong>MLC</strong>.
(70) Non-referential subjects <strong>in</strong> wh-questions <strong>in</strong> English<br />
a. what did two boys f<strong>in</strong>d?<br />
b.*what did a boy f<strong>in</strong>d?<br />
c. which book did two boys f<strong>in</strong>d?<br />
d.?which book did a boy f<strong>in</strong>d?<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 113<br />
Summ<strong>in</strong>g up, there is reason to believe that the difference between (68a) <strong>and</strong><br />
(68b) stems from the fact that a wh-subject must be topical <strong>in</strong> English when<br />
it is <strong>in</strong> situ, while this does not hold for German.<br />
Zubizarretta (1998) develops a prosodic theory for accent <strong>and</strong> focus placement<br />
<strong>in</strong> English which implies that the predicate will be <strong>in</strong> focus <strong>in</strong> double<br />
questions of English <strong>in</strong> which the subject is left <strong>in</strong> situ. Erteschik-Shir<br />
(1997) proposes a model of the syntax-<strong>in</strong>formation structure <strong>in</strong>terface which<br />
also implies topichood for the subject when certa<strong>in</strong> formal dependencies are<br />
built up <strong>in</strong> a clause. In the <strong>in</strong>terest of space, I will not try to assess the merits<br />
of these approaches, but conf<strong>in</strong>e myself to po<strong>in</strong>t<strong>in</strong>g out that the connection<br />
between topichood <strong>and</strong> <strong>in</strong> situ wh-subjects apparently need not be stipulated<br />
for English.<br />
Explanations borrowed from Zubizarretta <strong>and</strong> Erteschik-Shir may help<br />
expla<strong>in</strong><strong>in</strong>g the status of (68a) – but do they also fit the general model we try to<br />
defend here, viz. that the <strong>MLC</strong> is an <strong>in</strong>terface economy constra<strong>in</strong>t that blocks<br />
structures only if their (partial) LF can be arrived at <strong>in</strong> a more economical<br />
way? What is the proper way of express<strong>in</strong>g questions <strong>in</strong> which an object<br />
wh-pronoun is the sort<strong>in</strong>g key for answers? It is worthwhile to compare the<br />
constellations which lead to cross<strong>in</strong>g effects with wh-pronouns <strong>in</strong> English<br />
with those that do not:<br />
(71) Structural constellations lead<strong>in</strong>g to cross<strong>in</strong>g effects: passive<br />
a. who bought what?<br />
a’.*what did who buy?<br />
a”. what was bought by whom?<br />
b. who did you give _ what<br />
b’.*what did you give who _<br />
c. what did you give _ to whom<br />
c’: *who did you give what to _
114 Gisbert Fanselow<br />
<strong>The</strong> contrasts <strong>in</strong> (71) are related to the fact that English expresses <strong>in</strong>formation<br />
structure dist<strong>in</strong>ctions <strong>in</strong> a way different from scrambl<strong>in</strong>g <strong>and</strong> topicalization.<br />
(71) shows that English bans cross<strong>in</strong>g wh-pronouns primarily <strong>in</strong> those contexts<br />
<strong>in</strong> which it offers an alternative way of mak<strong>in</strong>g a lower (wh-) phrase<br />
more topical than the higher one. For subjects <strong>and</strong> objects, this alternative<br />
way is the passive construction. <strong>The</strong> conditions of <strong>in</strong>formation structure that<br />
license counterparts to (71a’) <strong>in</strong> German are therefore not <strong>in</strong>expressible <strong>in</strong><br />
English. Rather, they imply the use of a passive.<br />
(71b-c) illustrate that one can front both the direct <strong>and</strong> the <strong>in</strong>direct object<br />
<strong>in</strong> a multiple question, but the options (related to <strong>in</strong>formation structure) are<br />
l<strong>in</strong>ked to the dative alternation. (71b) is unobjectionable because it is <strong>in</strong> l<strong>in</strong>e<br />
with the <strong>MLC</strong>. <strong>The</strong> <strong>MLC</strong>-violation <strong>in</strong> (71b’) would have to be motivated on<br />
grounds of <strong>in</strong>formation structure (who be<strong>in</strong>g more focal than what), but <strong>in</strong> a<br />
dative shift construction, the <strong>in</strong>ner object (who) must be more topical than<br />
the outer object. <strong>The</strong>refore, (71b’) is ill-formed on pragmatic grounds. (71c)<br />
is grammatical s<strong>in</strong>ce it conforms to the <strong>MLC</strong>. <strong>The</strong> <strong>in</strong>formation structure<br />
requirements that would license the <strong>MLC</strong>-violation <strong>in</strong> (71c’) are those that<br />
trigger the dative shift alternation. (71c’) is illicit because the proper way to<br />
express its <strong>in</strong>formation structure is (71b).<br />
In an OT-framework, one may feel tempted to expla<strong>in</strong> the data <strong>in</strong> (71) by<br />
assum<strong>in</strong>g a grammatically visible competition between active <strong>and</strong> passive<br />
sentences, or between the constructions V NP PP <strong>and</strong> V NP NP at the po<strong>in</strong>t<br />
when the <strong>MLC</strong> is evaluated, but a more conservative solution is also at h<strong>and</strong>:<br />
we can assume that the <strong>in</strong>formation structure constellation needed to override<br />
the <strong>MLC</strong> <strong>in</strong> (71a’, b’, c’) cannot be l<strong>in</strong>ked to the construction <strong>in</strong> question <strong>in</strong><br />
English (because of the structural alternatives passive <strong>and</strong> dative shift).<br />
Other constellations do not yield a cross<strong>in</strong>g effect. English has no special<br />
way of express<strong>in</strong>g <strong>in</strong>formation structure <strong>in</strong>teractions of objects <strong>and</strong> adverbs<br />
<strong>and</strong> adverbial PPs. <strong>The</strong>re be<strong>in</strong>g no restrictions on the distribution of topicality,<br />
the <strong>in</strong>formation structure needed to override the <strong>MLC</strong> <strong>in</strong> either (72a) or (72a’)<br />
can l<strong>in</strong>ked easily with to the sentences, so that both ways of formulat<strong>in</strong>g the<br />
multiple question are wellformed.<br />
(72) Constellations without cross<strong>in</strong>g effects<br />
a. what did you see where?<br />
a’. where did you see what?<br />
b. to whom did you give what?<br />
b’. what did you give to whom?
<strong>The</strong> absence of a contrast between (72b) <strong>and</strong> (72b’) forces upon us the assumption<br />
that the construction V NP [to NP] comes <strong>in</strong> two varieties: to may<br />
be a dative marker, or the head of a PP. If <strong>in</strong>formation structure restrictions<br />
favor<strong>in</strong>g the dative alternation affect the former version only, the absence of<br />
a contrast is predicted. Alternatively, we may assume that wh-PPs may<br />
always cross wh-DPs. For English, the approach just sketched implies that<br />
the topical nature of <strong>in</strong> situ subjects <strong>in</strong> multiple questions must be the block<strong>in</strong>g<br />
factor for sentences with wh-pronouns <strong>in</strong> subject position.<br />
(73) a. who arrived when?<br />
b.*when did who arrive<br />
5. Conclud<strong>in</strong>g Remarks<br />
In the theory defended here, the <strong>MLC</strong> is a constra<strong>in</strong>t that applies cyclically <strong>in</strong><br />
a derivation: if more that one category can be attracted to a certa<strong>in</strong> position<br />
P, only the one closest to P can move. However, the <strong>MLC</strong> cannot prevent a<br />
movement operation from apply<strong>in</strong>g if that movement step is <strong>in</strong>evitable <strong>in</strong><br />
generat<strong>in</strong>g the (partial) LF-representation <strong>in</strong> question. Given that considerations<br />
of <strong>in</strong>formation structure play a role <strong>in</strong> this context, the fact that the <strong>MLC</strong><br />
decides between syntactic objects with the same partial LF only renders the<br />
pr<strong>in</strong>ciple quite weak <strong>in</strong> the doma<strong>in</strong> of operator movement.<br />
<strong>The</strong> predictions are quite different for head movement, if head movement<br />
does not have semantic effects. Consequently, the two syntactic objects <strong>in</strong><br />
(74) (with A <strong>and</strong> B be<strong>in</strong>g heads attracted to X) do not yield different partial<br />
LFs, because they differ <strong>in</strong> the location of the phonetic matrix of A <strong>and</strong> B<br />
only. In the model advocated here, this is equivalent to say<strong>in</strong>g that noth<strong>in</strong>g<br />
will prevent the <strong>MLC</strong> from block<strong>in</strong>g (74b).<br />
(74) a.[[ X A ] [ … A … [… B … ]]]<br />
b.[[ X B ] [ … A … [… B … ]]]<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 115<br />
Phrasal A-movement has semantic consequences <strong>in</strong> many theories, <strong>and</strong> the<br />
pragmatic implications of different options of fill<strong>in</strong>g the subject position are<br />
obvious. <strong>The</strong> current proposal therefore implies that <strong>MLC</strong>-effects should be<br />
<strong>in</strong>fluenced by considerations of <strong>in</strong>terpretation <strong>in</strong> the doma<strong>in</strong> of A-movement<br />
as well, i.e., one should be able to observe apparent <strong>MLC</strong>-violations. This<br />
prediction is borne out. E.g., Hestvik (1986) observes that both objects can
116 Gisbert Fanselow<br />
be attracted to the subject position <strong>in</strong> the passive version of double object<br />
constructions <strong>in</strong> Norwegian:<br />
(75) Passive formation <strong>in</strong> Norwegian double object constructions<br />
a. det ble gitt ham en gave<br />
there was given him a present<br />
b. han ble gitt en gave<br />
he was given a present<br />
c. en gave ble gitt ham<br />
<strong>The</strong> st<strong>and</strong>ard assumption concern<strong>in</strong>g English is that the direct object must<br />
not cross the <strong>in</strong>direct one <strong>in</strong> the passive of a double object construction, but<br />
this does not characterize all dialects of the language. After all, McCawley<br />
(1988: 79) observes that (76) sounds acceptable to speakers of British<br />
English.<br />
(76) a car was sold my brother __ for $200 by Honest Oscar<br />
Phrasal A-movement thus seems to have properties comparable to the one<br />
of operator movement with respecr to the <strong>MLC</strong>. One needs to identify the<br />
<strong>in</strong>terpretive conditions that license (75c) or (76), <strong>and</strong> offer an account as to<br />
why <strong>in</strong>formation structure does not seem to modulate <strong>MLC</strong>-effects <strong>in</strong> certa<strong>in</strong><br />
languages or dialects of languages (such as American English).<br />
German shows that additional formal aspects comes <strong>in</strong>to play that do not<br />
figure <strong>in</strong> operator movement: both objects may be promoted to subject status<br />
<strong>in</strong> a passive construction, but different auxiliaries are used for the promotion<br />
of direct <strong>and</strong> <strong>in</strong>direct objects:<br />
(77) Passive formation <strong>in</strong> German double object constructions<br />
a. jem<strong>and</strong> stiehlt dem K<strong>in</strong>d e<strong>in</strong>en Schlüssel<br />
someone.nom steals the.dat child a.acc key<br />
b. e<strong>in</strong> Schlüssel wird dem K<strong>in</strong>d gestohlen<br />
a.nom key is the.dat child stolen<br />
c. das K<strong>in</strong>d bekam e<strong>in</strong>en Schlüssel gestohlen<br />
the.nom child got a.acc key stolen<br />
“someone stole a key from the child”
Similarly, noun phrases with an oblique Case must not move to the subject<br />
position <strong>in</strong> many languages, <strong>and</strong> they may be skipped by A-movement to<br />
Spec,IP (see, e.g., Stepanov, this volume). <strong>The</strong>re is no comparable array of<br />
facts with A-bar movement. <strong>The</strong> data show that the application of the <strong>MLC</strong><br />
is not only sensitive to questions of identity of (partial) Logical Forms, but<br />
also constra<strong>in</strong>ted by purely formal factors. A discussion of these is beyond<br />
the scope of the present paper.<br />
I have argued that the <strong>MLC</strong> must be considered an economy constra<strong>in</strong>t<br />
that compares (partial) derivations <strong>and</strong> selects the one that fulfils check<strong>in</strong>g<br />
requirements with the shortest movements possible. However, the set of<br />
c<strong>and</strong>idate derivations which the <strong>MLC</strong> compares is constra<strong>in</strong>ed by formally<br />
encoded expressivity conditions: a derivational step B lead<strong>in</strong>g from structure<br />
S* to a partial LF S is blocked by the <strong>MLC</strong> only if S can also be<br />
reached from S* <strong>in</strong> a way that respects the <strong>MLC</strong>.<br />
Acknowledgments<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 117<br />
<strong>The</strong> research reported here was supported by grants of the Deutsche<br />
Forschungsgeme<strong>in</strong>schaft to the Forschergruppe Konfligierende Regeln<br />
(FOR 375), <strong>and</strong> to the Innovationskolleg Formale Modelle kognitiver<br />
Komplexität (INK 12).<br />
I want to thank Joanna Błaszczak, Eva Engels, Susann Fischer, Stefan<br />
Frisch, Hans-Mart<strong>in</strong> Gärtner, Andreas Haida, Hubert Haider, Gereon Müller,<br />
Doug Saddy, Matthias Schlesewsky, Penka Stateva, Arthur Stepanov, Koyka<br />
Stoyanova, Ralf Vogel, <strong>and</strong> the two anonymous referees for helpful comments.
118 Gisbert Fanselow<br />
Notes<br />
1.Superiority can be violated <strong>in</strong> similar contexts <strong>in</strong> Swedish, too. (ii) was<br />
accepted by two of my five <strong>in</strong>formants, two rejected it, one found it questionable.<br />
All five <strong>in</strong>formants considered (i) grammatical.<br />
(i) Vem tror att Johan gjorde vad<br />
Who believes that John did what<br />
(ii) Vad tror vem att Johan gjorde<br />
2.Contrast between structures rated as “?” with others rated as “*” may not be too<br />
impressive, but examples <strong>in</strong>volv<strong>in</strong>g different k<strong>in</strong>ds of A-bar-movement yield<br />
clearer contrasts:<br />
(i) which viol<strong>in</strong>-1 is this sonata-2 easy to play t-2 on t-1<br />
(ii) *which sonata is this viol<strong>in</strong> easy to play on<br />
3.And the model proposed below does so by l<strong>in</strong>k<strong>in</strong>g the acceptability of a cross<strong>in</strong>g<br />
constellation to the expression of a non-st<strong>and</strong>ard <strong>in</strong>formation structure.<br />
4. As suggested by Hubert Haider, p.c.<br />
5.<strong>The</strong> relevance of such examples has been brought to my attention by Gereon<br />
Müller.<br />
6.One may wonder, why scrambl<strong>in</strong>g is able to create structures <strong>in</strong>compatible with<br />
a simple <strong>MLC</strong>. Fanselow (2001) suggests that this problem is part of an argument<br />
<strong>in</strong> favor of the base-generation of scrambl<strong>in</strong>g structures.<br />
7.Tibetan shows at least some of the contrasts one is familiar with from English<br />
(Seele p.c, Chungda Haller, p.c), <strong>in</strong> spite of the fact that it is a free constituent<br />
order language. I have no explanation for this.<br />
(i) a. su ga re nyos pa red?<br />
who what bought<br />
b.*ga re su nyos pa red<br />
8.One might claim that these languages nevertheless allow scrambl<strong>in</strong>g, but only<br />
as an <strong>in</strong>termediate step followed by further movements. It is difficult to assess,<br />
however, which data could possibly refute such an account. Its empirical force<br />
is thus limited, <strong>and</strong> we refra<strong>in</strong> from consider<strong>in</strong>g it.
References<br />
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 119<br />
Aoun, Joseph, Norbert Hornste<strong>in</strong>, David Lightfoot <strong>and</strong> Amy We<strong>in</strong>berg<br />
1987 Two Types of Locality. L<strong>in</strong>guistic Inquiry 18: 537–577.<br />
Aoun, Joseph, <strong>and</strong> Audrey Li<br />
1989 Scope <strong>and</strong> constituency. L<strong>in</strong>guistic Inquiry 20: 141–172.<br />
1993 <strong>The</strong> <strong>Syntax</strong> of Scope. Cambridge, Mass: MIT Press,<br />
2002 Essays on the Representational <strong>and</strong> <strong>Derivational</strong> Nature of Grammar.<br />
Cambridge, Mass.: MIT-Press.<br />
Baker, C.<br />
1970 Notes on the description of English questions: the role of an abstract<br />
question morpheme. Foundations of Language 6: 197–219.<br />
Beck, Sigrid<br />
1996 Quantified structures as barriers for LF movement. Natural Language<br />
Semantics 4, 1–56.<br />
Bill<strong>in</strong>gs, Loren <strong>and</strong> Cather<strong>in</strong>e Rud<strong>in</strong><br />
1996 Optimality <strong>and</strong> Superiority: A new approach to overt multiple-wh<br />
order<strong>in</strong>g. In; Proceed<strong>in</strong>gs of Annual Workshop on Formal Approaches<br />
to Slavic L<strong>in</strong>guistics. <strong>The</strong> College Park Meet<strong>in</strong>g 1994, J<strong>in</strong>drich Toman<br />
(ed), 35–60. Michigan Slavic Publications. Ann Arbor.<br />
Błaszczak, Joanna <strong>and</strong> Susann Fischer<br />
2002 Multiple Wh-Konstruktionen im Slavischen. L<strong>in</strong>guistics <strong>in</strong> Potsdam<br />
14.<br />
Bobaljik, Jonathan <strong>and</strong> Diane Jonas<br />
1996 Subject Positions <strong>and</strong> the Role of TP’. L<strong>in</strong>guistic Inquiry 27:<br />
195–236.<br />
Bol<strong>in</strong>ger, Dwight<br />
1978 Ask<strong>in</strong>g more than one th<strong>in</strong>g at a time. In; Questions, Henry Hiz (ed.),<br />
107–150. Dordrecht: Reidel.<br />
Bonet, Eulalia<br />
1993 Morphology after syntax: Pronom<strong>in</strong>al clitics <strong>in</strong> Romance. Doctoral<br />
dissertation, MIT, Cambridge, Mass.<br />
Bo‰koviç, Îeljko<br />
1997 Superiority effects with multiple wh-front<strong>in</strong>g <strong>in</strong> Serbo-Croatian.<br />
L<strong>in</strong>gua 102: 1–20.<br />
2002 On multiple wh front<strong>in</strong>g. L<strong>in</strong>guistic Inquiry 33: 351–383.<br />
Bresnan, Joan<br />
1972 <strong>The</strong>ory of Complementation <strong>in</strong> English <strong>Syntax</strong>. Doctoral dissertation,<br />
MIT, Cambridge, Mass.<br />
Chomsky, Noam<br />
1981 Lectures on Government <strong>and</strong> B<strong>in</strong>d<strong>in</strong>g. Dordrecht: Foris.<br />
1993 A M<strong>in</strong>imalist Program for L<strong>in</strong>guistic <strong>The</strong>ory. In; <strong>The</strong> View from<br />
Build<strong>in</strong>g 20, Ken Hale <strong>and</strong> Samuel Keyser (eds), 1–58. Cambridge:<br />
MIT Press.<br />
1995 <strong>The</strong> m<strong>in</strong>imalist program. Cambridge, Mass.: MIT-Press.
120 Gisbert Fanselow<br />
Comorovski, I.<br />
1996 Interrogative Phrases <strong>and</strong> the <strong>Syntax</strong>-Semantics Interface, Dordrecht:<br />
Kluwer Academic Publishers.<br />
Dayal, Veneeta<br />
2003 Multiple wh-questions. Case 66, <strong>The</strong> <strong>Syntax</strong> Companion.<br />
Dies<strong>in</strong>g, Molly<br />
1992 Indef<strong>in</strong>ites. Cambridge, Mass.: MIT-Press.<br />
Erteschik-Shir, Nomi<br />
1997 <strong>The</strong> dynamics of focus structure. Cambridge, Cambridge UP.<br />
Fanselow, Gisbert<br />
1998 M<strong>in</strong>imal L<strong>in</strong>k <strong>Effects</strong> <strong>in</strong> German (<strong>and</strong> Other Languages). Paper, presented<br />
at the 1998 <strong>MLC</strong> conference, Potsdam.<br />
2001 Features, θ-roles, <strong>and</strong> free constituent order. L<strong>in</strong>guistic Inquiry 32, 3.<br />
Fanselow, Gisbert <strong>and</strong> Anoop Mahajan<br />
2000 Towards a m<strong>in</strong>imalist theory of wh-expletives, wh-copy<strong>in</strong>g, <strong>and</strong> successive<br />
cyclicity. In: Wh-scope mark<strong>in</strong>g, Uli Lutz, Gereon Müller <strong>and</strong><br />
Arnim von Stechow (eds). Amsterdam: Benjam<strong>in</strong>s.<br />
Featherston, Sam<br />
2002a Grammaticality <strong>and</strong> Universals. Wh-constra<strong>in</strong>ts <strong>in</strong> German. to appear<br />
<strong>in</strong> L<strong>in</strong>guistics.<br />
2002b Magnitude estimation <strong>and</strong> what it can do for your syntax.: some whconstra<strong>in</strong>ts<br />
<strong>in</strong> German. Ms., Tüb<strong>in</strong>gen.<br />
Fodor, Janet Dean<br />
1978 Pars<strong>in</strong>g strategies <strong>and</strong> constra<strong>in</strong>ts on transformations. L<strong>in</strong>guistic<br />
Inquiry 9: 427–473.<br />
Frazier, Lyn <strong>and</strong> Giovanni Flores d’Arcais<br />
1989 Filler-driven pars<strong>in</strong>g: A study of gap fill<strong>in</strong>g <strong>in</strong> Dutch. Journal of<br />
Memory <strong>and</strong> Language 28: 331–344.<br />
Golan, Yael<br />
1993 Node cross<strong>in</strong>g economy, superiority <strong>and</strong> D-l<strong>in</strong>k<strong>in</strong>g. Ms., Tel Aviv<br />
University.<br />
Grewendorf, Günther<br />
1999 <strong>The</strong> additional-wh effect <strong>and</strong> multiple wh-front<strong>in</strong>g. In: Specifiers, D.<br />
Adger, S. P<strong>in</strong>tzuk, B. Plunkett <strong>and</strong> G. Tsoulas (eds.), 146–162. Oxford:<br />
Oxford University Press,<br />
2001 Multiple wh-movement. L<strong>in</strong>guistic Inquiry 32: 87–122.<br />
Haider, Hubert<br />
1986 Deutsche <strong>Syntax</strong> – Generativ. Habilitation thesis. Vienna.<br />
1990 Topicalization <strong>and</strong> other puzzles of German syntax. In: Scrambl<strong>in</strong>g<br />
<strong>and</strong> Barriers. Günter Grewendorf <strong>and</strong> Wolfgang Sternefeld (eds.),<br />
93–112. Amsterdam: Benjam<strong>in</strong>s.<br />
1993 Deutsche <strong>Syntax</strong>-generativ. Tüb<strong>in</strong>gen: Narr.<br />
1997 <strong>Economy</strong> <strong>in</strong> syntax is projective economy. In; <strong>The</strong> Role of <strong>Economy</strong><br />
Pr<strong>in</strong>ciples <strong>in</strong> L<strong>in</strong>guistic <strong>The</strong>ory. Chris Wilder, Hans-Mart<strong>in</strong> Gärtner<br />
<strong>and</strong> Manfred Bierwisch (eds.), 205–226. Berl<strong>in</strong>: Akademie Verlag.
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 121<br />
2000 Superiority Revisited – Dutch, English, German, Icel<strong>and</strong>ic Contrasts.<br />
A representational account. Ms., University of Salzburg.<br />
This vol. <strong>The</strong> Superiority Conspiracy – Four Constra<strong>in</strong>ts <strong>and</strong> a Process<strong>in</strong>g Effect.<br />
Halle, Morris<br />
1992 Latvian declension. In Yearbook of Morphology Geert Booij <strong>and</strong> Jaap<br />
van der Marle (eds.), 33–47. Kluwer, Dordrecht.<br />
Halle, Morris <strong>and</strong> Alec Marantz<br />
1993 Distributed Morphology <strong>and</strong> the pieces of <strong>in</strong>flection. In <strong>The</strong> View from<br />
Build<strong>in</strong>g 20. Ken Hale <strong>and</strong> S. Jay Keyser (eds.), 111–176. MIT Press,<br />
Cambridge.<br />
1994 Some key features of Distributed Morphology. In MITWPL 21: Papers<br />
on Phonology <strong>and</strong> Morphology. Andrew Carnie & Heidi Harley (eds.),<br />
275–288. MIT Press, Cambridge.<br />
Hemforth, Barbara<br />
1993 Kognitives Pars<strong>in</strong>g: Repräsentation und Verarbeitung sprachlichen<br />
Wissens. Sankt August<strong>in</strong>: Infix.<br />
Hendrik, R. & Michael Rochemont<br />
1982 Complementation, multiple wh, <strong>and</strong> echo questions. Ms., University<br />
of North Carol<strong>in</strong>a, Chapel Hill, N.C. <strong>and</strong> University of California,<br />
Irv<strong>in</strong>e, California.<br />
Hestvik. Arild<br />
1986 Case <strong>The</strong>ory <strong>and</strong> Norwegian Impersonal Constructions: Subjekt-<br />
Object Alternations <strong>in</strong> Active <strong>and</strong> Passive Verbs. Nordic Journal of<br />
L<strong>in</strong>guistics 9: 181–197.<br />
Hoge, Kerst<strong>in</strong><br />
2000 Superiority. Doctoral dissertation. Oxford.<br />
Hornste<strong>in</strong>, Norbert<br />
1995 Logical Form. Cambridge, Mass.: Blackwell.<br />
Huang, James T.<br />
1982 Logical relations <strong>in</strong> Ch<strong>in</strong>ese <strong>and</strong> the theory of grammar. Doctoral dissertation,<br />
MIT, Cambridge, Mass.<br />
Kayne, Richard<br />
1983 Connectedness. L<strong>in</strong>guistic Inquiry 14: 223–249.<br />
Kitahara, Hisatsugu<br />
1993 Deduc<strong>in</strong>g ‘superiority’ effects from the Shortest Cha<strong>in</strong> Requirement.<br />
Harvard Work<strong>in</strong>g Papers <strong>in</strong> L<strong>in</strong>guistics 3: 109–119.<br />
Kitahara, Hisatsugu<br />
1994 Target α. Doctoral dissertation, Harvard University.<br />
Krems, Josef<br />
1984 Erwartungsgeleitete Sprachverarbeitung. Frankfurt/Ma<strong>in</strong>: Lang.<br />
Kuno, Susumo <strong>and</strong> J. Rob<strong>in</strong>son<br />
1972 Multiple wh-questions. L<strong>in</strong>guistic Inquiry 3: 463–487.<br />
Lasnik, Howard <strong>and</strong> Mamoru Saito<br />
1992 Move a: Conditions on its Application <strong>and</strong> Output, Cambridge, Mass.:<br />
MIT Press.
122 Gisbert Fanselow<br />
Lee, Hanjung<br />
This vol.<strong>M<strong>in</strong>imality</strong> <strong>in</strong> a Lexicalist OT.<br />
Lutz, Uli, Gereon Müller <strong>and</strong> Arnim von Stechow (eds.)<br />
2000 Wh-scope mark<strong>in</strong>g. Amsterdam: Benjam<strong>in</strong>s.<br />
Mahajan, Anoop<br />
1990 <strong>The</strong> A/A-bar dist<strong>in</strong>ct<strong>in</strong> <strong>and</strong> movement theory. Doctoral dissertation,<br />
MIT, Cambridge, Mass.<br />
Mal<strong>in</strong>g, Joan <strong>and</strong> Annie Zaenen<br />
1982 A phrase structure account of Sc<strong>and</strong><strong>in</strong>avian extraction phenomena. In:<br />
<strong>The</strong> Nature of Syntactic Representations, Paul<strong>in</strong>e Jacobson & Geoffrey<br />
Pullum. (eds), 229–282. Dordrecht: Reidel.<br />
Mathieu, Eric<br />
2002 <strong>The</strong> <strong>Syntax</strong> of Non-Canonical Quantification: A Comparative Study.<br />
Doctoral dissertation. London.<br />
McCawley, James<br />
1988 <strong>The</strong> Syntactic Phenomena of English. Chicago: <strong>The</strong> University of<br />
Chicago Press.<br />
Me<strong>in</strong>unger, André<br />
1995 Discourse dependent DP de-)placement. Doctoral dissertation,<br />
University of Potsdam.<br />
Müller, Gereon<br />
2001 Order preservation, parallel movement, <strong>and</strong> the emergence of the<br />
unmarked. In: Optimality <strong>The</strong>oretic <strong>Syntax</strong>, Gerald<strong>in</strong>e Legendre, Jane<br />
Grimshaw, <strong>and</strong> Sten Vikner (eds), 279–313. Cambridge, Mass.: MIT-<br />
Press.<br />
Noonan, Maire<br />
1988 Superiority <strong>Effects</strong> : How do antecedent government, lexical government<br />
<strong>and</strong> V2 <strong>in</strong>teract. McGill Work<strong>in</strong>g Papers <strong>in</strong> L<strong>in</strong>guistics 1988:<br />
192–214.<br />
Noyer, Robert<br />
1992 Features, Positions <strong>and</strong> Affixes <strong>in</strong> Autonomous Morphological<br />
Structure. Doctoral dissertation, MIT, Cambridge, Mass.<br />
Oka, T.<br />
1993 Shallowness. MIT Work<strong>in</strong>g Papers <strong>in</strong> L<strong>in</strong>guistics 19: 255–320.<br />
Pesetsky, David<br />
1982 Paths <strong>and</strong> categories. Doctoral dissertation, MIT, Cambridge, Mass.<br />
1987 Wh-<strong>in</strong>-situ: movement <strong>and</strong> unselective b<strong>in</strong>d<strong>in</strong>g. In: <strong>The</strong> Representation<br />
of (In)def<strong>in</strong>iteness, Eric Reul<strong>and</strong> <strong>and</strong> Alice. ter Meulen (eds),<br />
98–129. Cambridge, Mass.: MIT Press.<br />
2000 Phrasal Movement <strong>and</strong> Its K<strong>in</strong>. Cambridge, Mass.: MIT-Press.<br />
Re<strong>in</strong>hart, Tanya<br />
1995 Interface Strategies. OTS work<strong>in</strong>g papers <strong>in</strong> L<strong>in</strong>guistics.<br />
1998 Wh-<strong>in</strong>-situ <strong>in</strong> the Framework of the M<strong>in</strong>imalist Program. Natural<br />
Language Semantics 6: 29–56.
<strong>The</strong> <strong>MLC</strong> <strong>and</strong> derivational economy 123<br />
Reis, Marga<br />
1996 Extractions from Verb-Second Clauses <strong>in</strong> German?. In: On Extraction<br />
<strong>and</strong> Extraposition <strong>in</strong> German, Uli Lutz <strong>and</strong> Jürgen Pafel (eds.), 45–88.<br />
Amsterdam: Benjam<strong>in</strong>s.<br />
1997 Zum syntaktischen Status unselbständiger Verbzweit-Sätze. In:<br />
Sprache im Fokus, Christa Dürscheid, Karl He<strong>in</strong>z Ramers, <strong>and</strong><br />
Monika Schwarz (eds.), 121–144. Tüb<strong>in</strong>gen: Niemeyer.<br />
Richards, Norv<strong>in</strong><br />
2001 Movement <strong>in</strong> Language. Oxford & New York: Oxford University<br />
Press.<br />
Rizzi, Luigi<br />
1990 Relativized <strong>M<strong>in</strong>imality</strong>. Cambridge, Mass.: MIT Press.<br />
Rud<strong>in</strong>, Cather<strong>in</strong>e<br />
1988 On multiple questions <strong>and</strong> multiple wh front<strong>in</strong>g. Natural Language<br />
<strong>and</strong> L<strong>in</strong>guistic <strong>The</strong>ory 6: 445–501.<br />
Ste<strong>in</strong>itz, Renate<br />
1969 Adverbialsyntax. Berl<strong>in</strong>: Akademie-Verlag,<br />
Stepanov, Arthur<br />
1998 On Wh-Front<strong>in</strong>g <strong>in</strong> Russian – Proceed<strong>in</strong>gs of NELS 28. Pius N.<br />
Tamanji & Kiyomi Kusumoto (eds), 453–467.<br />
2001 Cyclic doma<strong>in</strong>s <strong>in</strong> syntactic theory. Doctoral dissertation, University<br />
of Connecticut, Storrs.<br />
This vol.Ergativity, Case, <strong>and</strong> the M<strong>in</strong>imal L<strong>in</strong>k Condition.<br />
Sternefeld, Wolfgang<br />
1997 Compar<strong>in</strong>g Reference Sets. In: <strong>The</strong> Role of <strong>Economy</strong> Pr<strong>in</strong>ciples <strong>in</strong><br />
L<strong>in</strong>guistic <strong>The</strong>ory, Chris Wilder, Hans.-Mart<strong>in</strong> Gärtner <strong>and</strong> Bierwisch<br />
(eds.), 81–114. Berl<strong>in</strong>: Akademie Verlag.<br />
Travis, Lisa<br />
1984 Parameters <strong>and</strong> <strong>Effects</strong> of Word Order Variation. Doctoral dissertation,<br />
MIT, Cambridge, Mass.<br />
Vogel, Ralf<br />
This vol.Correspondence <strong>in</strong> OT syntax <strong>and</strong> m<strong>in</strong>imal l<strong>in</strong>k effects.<br />
Wiltschko, Mart<strong>in</strong>a<br />
1998 Superiority <strong>in</strong> German. In: Proceed<strong>in</strong>gs of the Sixteenth West Coast<br />
Conference on Formal L<strong>in</strong>guistics, E. Curtis, J. Lyle & G. Webster<br />
(eds.). 431–445. Stanford, Cal.: CSLI Publications.<br />
Zubizaretta, Maria Luisa<br />
1998 Prosody, Focus <strong>and</strong> Word Order. Cambridge, Mass.: MIT Press.