Syntactic Productivity
Syntactic Productivity
Syntactic Productivity
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Syntactic</strong> <strong>Productivity</strong>: Extending Argument<br />
Structure Constructions to New verbs<br />
Jóhanna Barðdal
2<br />
This research was conducted in the period<br />
2003–2007 and is coming out in the Constructional<br />
Approaches to Language series of Benjamins<br />
• Barðdal, Jóhanna. 2008. <strong>Syntactic</strong> <strong>Productivity</strong>: Evidence<br />
from Case and Argument Structure in Icelandic<br />
[Constructional Approaches to Language]. Amsterdam: John<br />
Benjamins.<br />
Bergen Jóhanna Barðdal June 5, 2008
Overview<br />
3<br />
• <strong>Productivity</strong><br />
• New verbs in Icelandic: A general outline<br />
• Nonce verbs: A psycholinguistic experiment<br />
• New Verbs of communication: A questionnaire<br />
• Old and Modern Icelandic: A frequency comparison<br />
Bergen Jóhanna Barðdal June 5, 2008
Senses and Synonyms of “productive” and<br />
“productivity” (1)<br />
4<br />
1. ‘frequent’<br />
2. ‘regular’<br />
3. ‘rule-based’<br />
4. ‘operative’<br />
5. ‘easily combinable’<br />
6. ‘transparent or compositional’<br />
7. ‘having a wide coverage’<br />
8. ‘default’<br />
9. ‘schematically open or non-restricted, i.e. general’<br />
Bergen Jóhanna Barðdal June 5, 2008
Senses and Synonyms of “productive” and<br />
“productivity” (2)<br />
5<br />
10. ‘schematic vs. listable’<br />
11. ‘lexically and semantically modifiable’<br />
12. ‘syntactically manipulatable/flexible vs. frozen or idiomatic’<br />
13. ‘having a meaning/function vs. being historical relics’<br />
14. ‘developing new functions’<br />
15. ‘living vs. dead’<br />
16. ‘occurring or existing’<br />
17. ‘occurring with new/novel/nonce items’<br />
18. ‘spreading to already existing items’<br />
19. ‘deviating from adult language’<br />
Bergen Jóhanna Barðdal June 5, 2008
Examples of Different Usages (1)<br />
6<br />
• <strong>Productivity</strong> as syntactic ‘flexibility’ or ‘manipulatability’<br />
Many linguists have also noted that the noncompositional<br />
nature of idioms explains why idioms tend to be limited in<br />
their syntactic and lexical productivity (Gibbs 1995: 98).<br />
• <strong>Productivity</strong> as schematicity<br />
In spite of the fact that it is host to a large number of fixed<br />
expressions the form [the X-er the Y-er] has to be recognized<br />
as fully productive. Its member expressions are in principle<br />
not listable (Fillmore, Kay & O'Connor 1988: 507).<br />
Bergen Jóhanna Barðdal June 5, 2008
Examples of Different Usages (2)<br />
7<br />
• Productive semantics as semantic transparency<br />
Thus, it appears that very young children do not have access to<br />
the productive semantics of the noun class system, but treat it<br />
rather as a formal grammatical system. (Demuth 2000: 284<br />
[Emphasis added])<br />
• <strong>Productivity</strong> as meaningfulness/functionality<br />
... we also have to bear in mind that not all contrasts and<br />
distributions are meaningful or functional. Some patterns<br />
represent a lexically-arbitrary residue of formerly productive<br />
patterns. (Bybee & Thompson 1997: 384 [Emphasis added])<br />
Bergen Jóhanna Barðdal June 5, 2008
Bybee’s (1995) Usage<br />
8<br />
• <strong>Productivity</strong> as extensibility<br />
... the ability to be extended to new items ... (1995: 426)<br />
• <strong>Productivity</strong> as regularity<br />
The main issue to be discussed is the role of type frequency in<br />
the determination of regularity or productivity. (1995: 426<br />
[Emphasis added])<br />
• <strong>Productivity</strong> as default status<br />
I will argue that default status or productivity is not<br />
necessarily associated with source-oriented rules. (1995: 444<br />
[Emphasis added])<br />
Bergen Jóhanna Barðdal June 5, 2008
Tomasello’s (2000) Usage<br />
9<br />
• <strong>Productivity</strong> as children’s use of language that goes beyond what they have<br />
heard from adults (Tomasello 2000: 211)<br />
... both of which focus on children’s productivity, that is, their use of<br />
language in ways that go beyond what they have heard from adults.<br />
• <strong>Productivity</strong> as production of language (Tomasello 2000: 210)<br />
Thus, on the basis of hearing just The window broke ... they cannot go over<br />
to use He broke it ... This lack of productivity ...<br />
• <strong>Productivity</strong> as schematicity (Tomasello 2000: 236)<br />
This construction is highly abstract in the sense that it is not dependent on<br />
any particular word or phrase, and it is highly productive in the sense that<br />
any fluent speaker of English can generate innumerable further exemplars.<br />
Bergen Jóhanna Barðdal June 5, 2008
The Concepts of <strong>Productivity</strong><br />
10<br />
1. <strong>Productivity</strong> as generality<br />
2. <strong>Productivity</strong> as regularity<br />
3. <strong>Productivity</strong> as extensibility<br />
Bergen Jóhanna Barðdal June 5, 2008
Making Sense of Different Senses of “Productive”<br />
11<br />
Bergen Jóhanna Barðdal June 5, 2008
Definitions of productivity<br />
12<br />
1. The probability of a word formation pattern accepted as the<br />
model for new/potential words (Aronoff 1983: 163, fn. 1).<br />
2. The likelyhood that a pattern will apply to new forms<br />
(Bybee 1995: 430, Bybee and Thompson 1997: 384).<br />
3. The interaction between the potential of a morphological<br />
process to generate repetitive non-creative forms and the<br />
degree to which it is utilized in language use to yield new<br />
lexical items (Bauer 2001: 211).<br />
Bergen Jóhanna Barðdal June 5, 2008
Measurements of productivity (1)<br />
13<br />
1) The number of words a morphological process may apply to<br />
(Lieber 1981: 114-115).<br />
2) The number of new words coined by a morphological process<br />
(during a specific period of time) (Anhsen and Aronoff 1989).<br />
3) The proportion between actual items and potential items<br />
generated by a morphological process (Aronoff 1976: 36).<br />
4) The low token frequency of items as compared to the higher<br />
mean token frequency of other items in a corpus (Aronoff<br />
1983).<br />
Bergen Jóhanna Barðdal June 5, 2008
Measurements of productivity (2)<br />
14<br />
5) The proportion between the number of words, formed by a<br />
morphological process, occurring only once in a corpus and<br />
the total token frequency of all the words in the same corpus<br />
formed by that particular morphological process (Baayen<br />
1992, Baayen and Lieber 1991).<br />
6) The proportion between the number of items of a category<br />
which only occur once in a corpus and the total amount of all<br />
items occurring only once (Baayen 1993).<br />
7) The proportion between type and token frequency.<br />
Bergen Jóhanna Barðdal June 5, 2008
Predicting productivity<br />
15<br />
8) The type frequency of a schema and its coherence (Bybee<br />
1995: 430, Clausner and Croft 1997, Croft and Cruse 2004).<br />
Bergen Jóhanna Barðdal June 5, 2008
<strong>Syntactic</strong> productivity<br />
16<br />
1. The language user’s ability to generate and understand<br />
sentences never encountered before.<br />
2. The ability of a syntactic pattern or argument structure<br />
construction to be extended to new lexical verbs.<br />
1. = Regularity Concept<br />
2. = Extensibility Concept<br />
<strong>Productivity</strong> is a matter of how available a pattern is for the<br />
sanction of novel expressions (Langacker 2000: 26<br />
[emphasis added]).<br />
Bergen Jóhanna Barðdal June 5, 2008
“Two kinds of ‘creativity’”<br />
17<br />
We can distinguish two kinds of “creativity” in language.<br />
In one case there is the ability of speakers, using existing<br />
resources in the language, to produce and understand novel<br />
expressions. In the other case, the one for which we use the<br />
term coining, a speaker uses existing patterns in the<br />
language for creating new resources (Fillmore 2002).<br />
Bergen Jóhanna Barðdal June 5, 2008
Regular performance vs. extensibility in<br />
morphology and syntax<br />
18<br />
Bergen Jóhanna Barðdal June 5, 2008
Types, tokens and patterns in morphology and<br />
syntax<br />
19<br />
Bergen Jóhanna Barðdal June 5, 2008
The extensibility of regularity in morphology and<br />
syntax<br />
20<br />
Bergen Jóhanna Barðdal June 5, 2008
Definition of syntactic productivity used here<br />
21<br />
The extension of syntactic patterns or argument structure<br />
constructions to new verbs is a function of the correlation<br />
between the construction's type frequency and semantic<br />
coherence.<br />
Bergen Jóhanna Barðdal June 5, 2008
Predicting factor<br />
22<br />
Inverse correlation between type frequency and semantic<br />
coherence:<br />
– High type frequency – Low degree of semantic coherence<br />
– Low type frequency – High degree of semantic coherence<br />
Bergen Jóhanna Barðdal June 5, 2008
The inverse correlation between type frequency<br />
and semantic coherence<br />
23<br />
Bergen Jóhanna Barðdal June 5, 2008
Different definitions and degrees of productivity<br />
24<br />
Bergen Jóhanna Barðdal June 5, 2008
The productivity cline<br />
25<br />
Bergen Jóhanna Barðdal June 5, 2008
The productivity cline<br />
26<br />
Bergen Jóhanna Barðdal June 5, 2008
High type frequency vs. open schema<br />
27<br />
• Can these be teased apart?<br />
Bergen Jóhanna Barðdal June 5, 2008
The schematicity–lexicality hiararchy<br />
28<br />
Bergen Jóhanna Barðdal June 5, 2008
Degrees of productivity<br />
29<br />
• Assuming different levels of schematicity automatically entails<br />
gradient productivity<br />
• The productivity domain of each category is a function of its type<br />
frequency and coherence<br />
• The highest level of schematicity of each construction thus<br />
determines its productivity<br />
• BUT if the highest level of schematicity is derived from the type<br />
frequency of a construction, can’t productivity then just be reduced<br />
to type frequency<br />
• No, not really, because a construction’s level of schematicity cannot<br />
be derived from the type frequency of low-type frequency<br />
constructions.<br />
Bergen Jóhanna Barðdal June 5, 2008
Gradient productivity (Clausner & Croft 1997)<br />
30<br />
Clausner and Croft (1997) take productivity to be a derivative of the<br />
relation between a schema and its instantiations. That is, higher type<br />
frequency contributes to a fuller exploitation of a schemas range, resulting<br />
in a more entrenched schema than entrenched instantiations.<br />
Bergen Jóhanna Barðdal June 5, 2008
Modified version of C&C’s gradient productivity<br />
31<br />
Bergen Jóhanna Barðdal June 5, 2008
Type vs. token frequency (Bybee 1995)<br />
32<br />
• High type frequency contributes to productivity<br />
• High token frequency detracts from productivity<br />
– Strong verbs are high in type frequency<br />
– Irregular verbs are high in token frequency<br />
• Two classes of irregular verbs, the strung vs. the<br />
swept class, of same type frequency (13 vs.14)<br />
show differences in productivity correlating with<br />
token frequency. The swept class is of high token<br />
frequency and non-productive, while the strung<br />
class is of intermediate token frequency and semiproductive<br />
Bergen Jóhanna Barðdal June 5, 2008
strung vs. swept verbs<br />
33<br />
• It is not clear how the type frequency of 13 vs. 14 (based<br />
on Moder 1992) has been calculated<br />
• Bybee and Slobin (1982: 288) list 27 strung verbs.<br />
• Hornby, Cowie and Gimson’s monolingual English<br />
dictionary (1974: 1003–1006) lists 21 verbs of the strung<br />
class and 22 of the swept class<br />
• A short glance at that list reveals that the strung class is<br />
phonologically much more coherent than the swept class<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
34
Token frequency’s contribution to productivity?<br />
35<br />
Bergen Jóhanna Barðdal June 5, 2008
Token frequency’s contribution to productivity?<br />
36<br />
Bergen Jóhanna Barðdal June 5, 2008
The inverse correlation between type and token<br />
frequency for productivity<br />
37<br />
Bergen Jóhanna Barðdal June 5, 2008
Summary (1)<br />
38<br />
• The 19 different usages of the term productivity reveals a<br />
highly structured and systematic metaconcept of which several<br />
inferentially-related subconcepts can be discerned.<br />
• The general concept of productivity within linguistics can be<br />
broken down into three main subconcepts, i.e. those of (a)<br />
generality, (b) regularity and (c) extensibility.<br />
• <strong>Productivity</strong> is as a function of a construction’s type frequency<br />
and its coherence, and the inverse correlation between the two.<br />
Bergen Jóhanna Barðdal June 5, 2008
Summary (2)<br />
39<br />
• There two notions of productivity found within syntax, i.e. (a)<br />
speakers’ ability to generate sentences never heard before, and<br />
(b) speakers extensions of argument structure constructions to<br />
new verbs, reflect the regularity and the extensibility concepts,<br />
respectively, within syntax.<br />
• <strong>Productivity</strong> is gradient with high type frequency correlating<br />
with high degree of schematicity, while low type frequency<br />
presupposes a high degree of semantic coherence in order for a<br />
construction to be productive.<br />
• It is the highest level of schematicity at which a construction<br />
exists in speakers’ minds that determines its degree of<br />
productivity and not its most entrenched level.<br />
Bergen Jóhanna Barðdal June 5, 2008
Summary (3)<br />
40<br />
• Token frequency can contribute to the productivity of low type<br />
frequency constructions, since low-level lexically-filled verbspecific<br />
constructions, rather than abstract schematic<br />
constructions, are models for analogical extensions.<br />
• The current approach to syntactic productivity, focusing on the<br />
Extensibility concept, offers a unified account of productivity<br />
since it captures different degrees of productivity, ranging<br />
from highly productive patterns to various intermediate<br />
degrees of productivity to low-level analogical modeling. The<br />
Regularity/Generality concept can also be derived from this<br />
analysis.<br />
Bergen Jóhanna Barðdal June 5, 2008
Case study 1: New verbs in Icelandic: A general<br />
outline<br />
41<br />
• Overview of intransitive and transitive case and<br />
argument structure constructions in Icelandic<br />
• Type frequency, relative and absolute frequency<br />
• Semantics of the Nominative subject construction<br />
• Borrowed verbs from the area of Information technology<br />
• Problems for generative approaches<br />
• Analogy and token frequency<br />
Bergen Jóhanna Barðdal June 5, 2008
Case and argument structure constructions in<br />
Icelandic<br />
42<br />
Bergen Jóhanna Barðdal June 5, 2008
Type frequencies<br />
43<br />
• Nominative subject construction: Intermediate-sized,<br />
bilingual Icelandic–English dictionary (Hólmarsson,<br />
Sanders and Tucker 1989)<br />
• Oblique subject construction: List of predicates that take<br />
quirky subjects in Icelandic (Jónsson 1998)<br />
Bergen Jóhanna Barðdal June 5, 2008
Type frequency of the Nominative subject<br />
construction<br />
44<br />
• The differences between the three subconstructions across the two counts<br />
are highly significant, i.e. the chances are only one against 1,000 that they<br />
are due to coincidence (Pearson Chi-Square, p < 0.001)<br />
Bergen Jóhanna Barðdal June 5, 2008
Corpus count<br />
45<br />
• Modern Icelandic<br />
• Written material: 25,000 running words<br />
• Five different genres, 5,000 words each, ten 500-word texts in each<br />
genre:<br />
– Icelandic fiction<br />
– Translated fiction<br />
– Children’s literature<br />
– Biographies and memoirs<br />
– Non-fiction<br />
• Spoken material: 15,000 running words<br />
– Interactive ring-in radio program<br />
Bergen Jóhanna Barðdal June 5, 2008
The semantics of the Nominative subject<br />
construction<br />
46<br />
• Narrowly circumscribed verb classes:<br />
– 303 Nom-Acc verbs: 46 verb classes<br />
– 188 Nom-Dat verbs: 33 verb classes<br />
– 24 Nom-Gen verbs: five verb classes<br />
• On average 6.1 verbs in each class<br />
Bergen Jóhanna Barðdal June 5, 2008
Nom-Acc construction<br />
47<br />
• Basic event type categories:<br />
– Making: byggja ‘build’<br />
– Moving: afhenda ‘deliver’<br />
– Affectedness: lemja ‘hit’<br />
– Cognition: skilja ‘understand’<br />
– Emotion: elska ‘love’<br />
– Change (of state): byrja ‘begin’<br />
– Location: gista ‘be accommodated somewhere’<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
48
Nom-Dat construction<br />
49<br />
• Basic event type categories:<br />
– Change (of quantity): fjölga ‘increase’<br />
– Cognition: líkja saman ‘compare’<br />
– Emotion: kvíða ‘be anxious’<br />
– Superiority: stjórna ‘govern’<br />
– Motion: kasta ‘throw’<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
50
Animate vs. inanimate reference of the object in<br />
Nominative subject constructions in Icelandic<br />
51<br />
(Pearson Chi-Square, p < 0.000).<br />
Bergen Jóhanna Barðdal June 5, 2008
Nom-Gen construction<br />
52<br />
• No Basic event type categories:<br />
– Verb-subclass-specific constructions:<br />
• Asking/wishing: biðjast ‘ask’<br />
• Cognition: vera fullviss ‘be sure’<br />
• Emotion: njóta ‘enjoy’<br />
• Social influence: mega ‘have power’<br />
Bergen Jóhanna Barðdal June 5, 2008
Nom-Gen<br />
53<br />
Bergen Jóhanna Barðdal June 5, 2008
The Nominative subject construction – an open<br />
schema<br />
54<br />
Bergen Jóhanna Barðdal June 5, 2008
Recent borrowings in Icelandic (1)<br />
55<br />
• 107 borrowed predicates<br />
– a recent on-line slang dictionary (Magnússon 2004)<br />
still under construction<br />
– the Icelandic discussion forum for Mac-users<br />
(www.apple.is/umraedur)<br />
– 68 are Nom-Acc verbs<br />
– 39 are Nom-Dat verbs<br />
Bergen Jóhanna Barðdal June 5, 2008
Recent borrowings in Icelandic (2)<br />
56<br />
• Nom-Acc (68):<br />
– archive-a ‘archive’, battla ‘battle’, biddsslappa ‘bitchslap’, blasta ‘blast’, bojkotta ‘boycut’,<br />
builda ‘build’, bomba ‘bomb’, branda ‘brand’, browsa ‘browse’, bösta ‘bust’, compilera<br />
‘compile’, copy-a ‘copy’, digga ‘dig’, döbba ‘dub’, editera ‘edit’, erasa ‘erase’, fíla ‘like‘, fixa<br />
‘fix’, flexa ‘flex‘, formatta ‘format’, fótósjoppa ‘photoshop’, gúggla ‘google’, hakka ‘hack’,<br />
hössla ‘hussle’, kidda ‘kid‘, krakka ‘crack’, logga ‘log’, massa ‘finish with style’, meisa ‘spray<br />
with tear gas’, modda ‘modify’, mounta ‘mount’, muffa ‘bang’, mönnsa ‘munch’, offa ‘off’,<br />
óna ‘own’, paira ‘pair’, partiona ‘partition’, patcha ‘patch’, peista ‘paste’, pinga ‘ping’,<br />
plögga ‘plug’, pródúsera ‘produce’, prógrammera ‘program’, publisha ‘publish’, r[e]nta<br />
‘rent’, releasa ‘release’, render ‘render’, resetta ‘reset’, resolva ’resolve’, restora ‘restore’,<br />
rippa ‘rip’, rokka ‘rock’, skratsa ‘scratch’, skvassa ‘squash, break’, slamma ‘slam’, ssh-a ‘ssh’,<br />
stúdera ‘study’, supporta ‘support’, sörfa ‘surf’, synca ‘synchronize’, synkrónisera<br />
‘synchronize’, tagga ‘tag, write’, testa ‘test’, tóka ‘smoke hash’, trimma ‘trim’, updata<br />
‘update’, upgreida ‘upgrade’, verifya ‘verify’ …<br />
Bergen Jóhanna Barðdal June 5, 2008
Recent borrowings in Icelandic (3)<br />
57<br />
• Nom-Dat (39):<br />
– adda ‘add’, blasta ‘blast’, bomba ‘bomb’, bundla ‘bundle’, convertera<br />
‘convert’, downloada ‘download’, deleta ‘delete’, de-multiplexa<br />
‘demultiplex’, dumpa ‘dump’, droppa ’drop‘, ejecta ‘eject’, expandera<br />
‘expand’, exporta ‘export’, farta ‘fart‘, innstalla/innstallera ‘install’, krassa<br />
‘crash’, mnsa ‘msn‘, mökka ‘contaminate’, neimdroppa ‘namedrop’, offa<br />
‘off’, parkera ‘park’, peista ‘paste’, poppa ‘deliver when popping by’,<br />
pósta ‘post’, publisha ‘publish’, releasa ‘release’, resolva ’resolve’,<br />
restarta ‘restart’, rippa ‘steal’, sjera ‘share’, slamma ‘slam’, starta ‘start’,<br />
statta ‘stat, dublicate’, streyma ‘stream’, tilta ‘tilt’, umba ‘be agent for’,<br />
unzippa ‘unzip’, uploada ‘upload’, untara ‘unzip’ …<br />
Bergen Jóhanna Barðdal June 5, 2008
Problem for generative approaches to case<br />
58<br />
Nom-Acc is structural case<br />
Nom-Dat is lexical case (idiosyncratic)<br />
... we make a rough and tentative distinction between the core<br />
of a language and its periphery, where the core consists of<br />
what we tentatively assume to be pure instantiations of UG<br />
and the periphery consists of marked exceptions (irregular<br />
verbs, etc.). [emphasis original]<br />
(Chomsky and Lasnik 1995: 19–20)<br />
Bergen Jóhanna Barðdal June 5, 2008
<strong>Productivity</strong> is associated with rules and nonproductivity<br />
with memory<br />
59<br />
The theory that regular forms are generated by rule and<br />
irregular forms are retrieved by rote is pleasing ... because it<br />
explains the differences in productivity between the two<br />
patterns ... (Pinker 1999: 19)<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
60
Bergen Jóhanna Barðdal June 5, 2008<br />
61
Bergen Jóhanna Barðdal June 5, 2008<br />
62
The productivity of Nom-Acc and Nom-Dat (1)<br />
63<br />
• The borrowed verbs assigned Nom-Acc belong to the<br />
semantic classes of already existing Nom-Acc verbs<br />
• The borrowed verbs assigned Nom-Dat belong to the<br />
semantic classes of already existing Nom-Dat verbs<br />
• This suggests that the productivity of these two<br />
constructions is based on the extension of either verbspecific<br />
constructions (a.k.a. analogy) or verb-classspecific<br />
constructions<br />
Bergen Jóhanna Barðdal June 5, 2008
Case assignment based on verb-specific<br />
constructions<br />
64<br />
(1) Eftir að ég updataði tölvuna núna síðast þá ... Nom-Acc<br />
after I updated computer-the.acc now last then<br />
‘After I updated the computer last time, then ...’<br />
• uppfæra = update<br />
(2) … þá ákvað ég að offa því um 2–4 vikur. Nom-Dat<br />
then decided I to postpone it.dat about 2–4 weeks<br />
‘… then I decided to postpone it for 2–4 weeks.’<br />
• fresta = offa<br />
Bergen Jóhanna Barðdal June 5, 2008
Low-level verb-class generalizations<br />
65<br />
Nom-Dat<br />
(3) búin að importera öllum færslum þessa árs inn í MT …<br />
finished to import all.dat transactions.dat this year into in MT<br />
‘finished importing all transactions from this year into MT …’<br />
Nom-Dat<br />
(4) Farið að farta þessari plötu út svo ég ...<br />
start to fart this.dat record.dat out so I<br />
‘Can’t you start farting this record out so that I ...’<br />
Bergen Jóhanna Barðdal June 5, 2008
Semantic overlap<br />
66<br />
Nom-Acc<br />
(5) Þá voru Kanarnir að bomba Júgóslava.<br />
then were Americans to bomb Yugoslavians.acc<br />
‘Then the Americans were bombing the Yugoslavians.’<br />
Nom-Dat<br />
(6) Mér tókst að bomba tannkremi í augað á mér.<br />
I.dat managed to bomb tooth-paste.dat in eye on me<br />
‘I managed to bomb toothpaste into my eye.’<br />
Bergen Jóhanna Barðdal June 5, 2008
The productivity of Nom-Acc and Nom-Dat (2)<br />
• On the generative approach, this would at best be expected for the<br />
Nom-Dat construction, but it is not expected for the Nom-Acc<br />
construction, which is supposed to be default.<br />
• This means that borrowed verbs with the Nom-Acc construction are<br />
also assigned case and argument structure constructions “lexically,”<br />
exactly like borrowed verbs with the Nom-Dat construction<br />
• If the borrowed Nom-Acc verbs had not been assigned case and<br />
argument structure constructions on the basis of their synonymous<br />
verb-specific construction in Icelandic, but rather on some “default”<br />
basis, one would expect some more of the Nom-Dat verbs to have<br />
been assigned Nom-Acc<br />
67<br />
Bergen Jóhanna Barðdal June 5, 2008
Rule-based vs. analogical extensions<br />
68<br />
• It is generally assumed in the literature that there is a<br />
qualitative difference between rule-based extensions and<br />
analogical extensions:<br />
Rule-governed innovation is always based on an<br />
already existing pattern which is found across at<br />
least two items; rule-changing innovation is based<br />
on a pattern which is perceived for the first time at the<br />
point of innovation and which may exist in only one<br />
model. [Emphasis added]<br />
(Bauer 2001: 93)<br />
Bergen Jóhanna Barðdal June 5, 2008
Two sides of the same coin<br />
69<br />
• On the present usage-based constructional approach<br />
where constructions are assumed to exist at different<br />
levels of schematicity, there is no qualitative difference<br />
between an extension based on one model item and an<br />
extension based on two model items<br />
• It is only a difference in quantity or the strength of a<br />
schema<br />
• Hence, productivity and analogy are simply two sides of<br />
the same coin<br />
Bergen Jóhanna Barðdal June 5, 2008
Role of token frequency<br />
70<br />
• Everybody working on productivity, within usage-based cognitive<br />
and constructional approaches, argues that token frequency<br />
detracts from productivity<br />
• Everybody working on productivity, within usage-based cognitive<br />
and constructional approaches, agrees that productivity is gradient<br />
• Yet, all these scholars have only viewed the role of token frequency<br />
in relation to full productivity, not gradient or low productivity<br />
• It seems obvious that if one model item can be the basis for<br />
analogical extensions then token frequency can be a decisive factor<br />
when different model items are in competition<br />
Bergen Jóhanna Barðdal June 5, 2008
Choice of model items for analogical extension<br />
71<br />
Non-fictive conversation between two Icelandic<br />
speakers living abroad:<br />
– A: Ef þú ætlar að applisera þessari greiningu á efnið þá ...<br />
if you intend to apply this.dat analysis on material-the then<br />
– B: Þú meinar applisera þessa greiningu.<br />
you mean apply this.acc analysis<br />
– A: Nei, applisera einhverju eins og beita einhverju.<br />
no, apply something.dat like apply something.dat<br />
– B: Nei, applisera eitthvað eins og nota eitthvað.<br />
no, apply something.acc like use something.acc<br />
Bergen Jóhanna Barðdal June 5, 2008
Differences in token frequency<br />
72<br />
• Observe that beita ‘apply’ is the translational equivalent<br />
of applisera ‘apply’ in Icelandic, not nota ‘use’<br />
• In contrast, nota ‘use’ is more general<br />
• Neither verb collocates more strongly with the noun<br />
greining ‘analysis’ than the other<br />
• Token frequency:<br />
Bergen Jóhanna Barðdal June 5, 2008
Inverse correlation between type and token<br />
frequency<br />
73<br />
Bergen Jóhanna Barðdal June 5, 2008
Summary<br />
74<br />
• The nominative subject construction is highest in type frequency of<br />
all argument structure constructions containing two-place predicates<br />
• The Nom-Acc subconstruction is approximately 58–64% of<br />
nominative subject predicates<br />
• The Nom-Dat subconstruction is approximately 34–35% of<br />
nominative subject predicates<br />
• Both constructions exist at a high level of schematicity, although the<br />
semantic domain of the Nom-Acc construction is much wider than<br />
that of the Nom-Dat construction, and hence its productivity domain<br />
is also larger<br />
• Both constructions are productively assigned to borrowed verbs, by<br />
Icelandic speakers, within their semantic domains<br />
Bergen Jóhanna Barðdal June 5, 2008
Summary<br />
75<br />
• This case assignment clearly takes place on the basis of low-level<br />
constructions, either verb-specific constructions or verb-subclassspecific<br />
constructions<br />
• This fact is problematic for generative approaches, exactly as the<br />
fact that the proportion of Nom-Acc and Nom-Dat verbs among<br />
borrowed verbs mirrors their proportion in language use<br />
• Full productivity and analogy are two sides of the same coin<br />
• The role of token frequency for productivity is inverse to the role of<br />
type frequency for productivity<br />
Bergen Jóhanna Barðdal June 5, 2008
Case study 2: Nonce verbs: A psycholinguistic<br />
experiment<br />
76<br />
• 40 participants<br />
– 20 adults<br />
– 20 children (6–13 years)<br />
• Presentation:<br />
This is a Funny-game. I will be showing you some pictures, and then<br />
you will tell me what is happening in the pictures. The only thing is<br />
that you are not going to do it in Icelandic, you are going to tell me in<br />
Funny-language. Now, Funny-language is almost identical to Icelandic,<br />
there are only a few words that are different, and you don’t have to<br />
worry about that because I will teach you those words.<br />
Bergen Jóhanna Barðdal June 5, 2008
Five Screen Beans silhouettes<br />
77<br />
Bergen Jóhanna Barðdal June 5, 2008
78<br />
• Each picture was introduced with a formula<br />
corresponding to the following:<br />
In this picture we see a man. What he is doing is<br />
called grilla [‘barbeque’] in Icelandic, but in Funnylanguage<br />
it is called slobba. Could you now tell me<br />
what you see happening in the picture?<br />
Bergen Jóhanna Barðdal June 5, 2008
Lexical and semantic priming<br />
79<br />
Bergen Jóhanna Barðdal June 5, 2008
80<br />
• Priming verbs in the infinitive<br />
• One filler between each stimulus<br />
• Experimental comments on the morphological structure<br />
of the verb<br />
Bergen Jóhanna Barðdal June 5, 2008
Statistics<br />
81<br />
The differences between the nominative and dative on the subjects of<br />
flokast and tvíta are not significant, as the chances are 499 against 1,000<br />
that they are due to a coincidence (Pearson Chi-Square, p < 0.499).<br />
Bergen Jóhanna Barðdal June 5, 2008
Results (1)<br />
82<br />
• Three participants assigned case and argument<br />
structure to the nonce verbs on the basis of the priming<br />
verbs<br />
• 11 participants assigned Nom(-Acc) categorically to all<br />
nonce verbs<br />
• The remaining participants assigned a mixture of Nom-<br />
Acc and the priming verbs to the nonce verbs<br />
Bergen Jóhanna Barðdal June 5, 2008
Results (2)<br />
83<br />
• This suggests that speakers, when faced with unknown<br />
verbs, have two ways of assigning argument structure<br />
constructions:<br />
– On the basis of a synonymous verb<br />
– On the basis of the schematic Nom-Acc construction<br />
– Some speakers pendulate between the two methods<br />
Bergen Jóhanna Barðdal June 5, 2008
Response time<br />
84<br />
• Some of the speakers who assigned the case and argument<br />
structure of the priming verbs took an unusually long time to<br />
respond<br />
– It makes perfect sense that assignment of case and argument<br />
structure on the basis of a low-level verb-specific construction<br />
takes longer time, as it involves retrieval of a concrete formmeaning<br />
mapping and the assignment of a form on the basis of<br />
the meaning component<br />
• Some of the speakers who assigned the Nom-Acc construction<br />
simply rattled off their responses in no time<br />
– It makes perfect sense that assignment on the basis of a highlevel<br />
schematic construction will take shorter time as, only<br />
relational meaning is involved<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
85
86<br />
• Differences in subject case marking: Less than one chance against<br />
1,000 that the differences in subject case marking are due to a<br />
coincidence (Pearson Chi-Square, p < 0.000)<br />
• Higher mean age for the children assigning dative case than the<br />
ones assigning nominative case to the subject of flokast and tvíta, or<br />
9.47 vs. 8.03 years of age. The difference is highly significant, with<br />
only seven chances against 1,000 that it is due to a coincidence<br />
(Paired Sample T-test, p < 0.007)<br />
• Differences in object case marking: The chances are only three<br />
against 1,000 that the object case marking is due to a coincidence<br />
(Pearson Chi-Square, p < 0.003)<br />
Bergen Jóhanna Barðdal June 5, 2008
Generative vs. usage-based constructional approach 87<br />
• All the facts, presented here, about the assignment of case and<br />
argument structure to nonce verbs are compatible with both the<br />
generative approach and the present usage-based constructional<br />
approach, except one:<br />
– Assignment of Nom-Acc is structural case assignment<br />
– Assignment of Nom-Dat to one verb is “thematic” assignment of<br />
dative objects to motion verbs<br />
– Assignment of Dative subjects is thematic assignment of dative<br />
case to experience-based verbs<br />
– *If so, then one of the Nom-Acc verbs, the one meaning ‘love’<br />
should have been assigned a Dative subject<br />
Bergen Jóhanna Barðdal June 5, 2008
Summary<br />
88<br />
• The construction highest in type frequency in Icelandic, the Nom-Acc<br />
construction, is the only construction extended beyond its verb-specific<br />
construction, suggesting that it exists as a schematic construction in minds<br />
of speakers<br />
• This assumption is supported by the impression of lower response time for<br />
the participants who categorically assigned the Nom-Acc construction to all<br />
the nonce verbs in the experiment<br />
• Hence, Icelandic speakers assign case and argument structure<br />
constructions to unknown nonce verbs on the basis of (a) synonymous<br />
verb-specific constructions in Icelandic, or (b) high-level schematic<br />
constructions with little or only relational content, yielding default effects<br />
• The results of the experiment also contradict the predictions of generative<br />
grammar that the Dative subject construction should be productive, as it is<br />
not generally extended to nonce verbs construed as emotion verbs.<br />
Bergen Jóhanna Barðdal June 5, 2008
Case study 3: New Verbs of instrument of<br />
communication: A questionnaire<br />
89<br />
• Faxa ‘fax’<br />
• (E)meila ‘e-mail’<br />
• SMSa ‘text, send a text message’<br />
Bergen Jóhanna Barðdal June 5, 2008
New Verbs of Instrument of Communication<br />
90<br />
• These verbs have been subject to considerable interest<br />
in the literature (Goldberg 1995, Croft et al. 2001,<br />
Barðdal 2003).<br />
• The verb is usually derived from the noun for the<br />
instrument or the product of the technology.<br />
• These verbs have emerged during the last two decades<br />
or so concomitant with the relevant technology, and their<br />
development can thus easily be traced.<br />
Bergen Jóhanna Barðdal June 5, 2008
Why Icelandic?<br />
91<br />
• 86% of the Icelandic population has access to the<br />
Internet and 81% use it regularly. This makes Internet<br />
usage in Iceland highest in Europe.<br />
(National Statistical Institute of Iceland)<br />
• 96% of the Icelandic population use cell phones. Cell<br />
phone usage in Iceland is thus higher than it is in Finland<br />
which has been on the top of the cell phone usage<br />
statistics in the world for years.<br />
(Síminn: Icelandic National Telecommunication)<br />
Bergen Jóhanna Barðdal June 5, 2008
Constructional Choices<br />
92<br />
(1) Ég faxa þetta til þín. TRANSFER CONSTRUCTION<br />
I fax this.ACC to you<br />
SUBJ+VERB+OBJ ACC<br />
PP DIR<br />
‘I’ll fax this to you.’<br />
(2) Ég sendi þetta til þín í tölvupósti. TRANSFER CONSTRUCTION<br />
I send this.ACC to you in email SUBJ+VERB+OBJ ACC<br />
PP DIR<br />
‘I’ll send this to you with email.’<br />
(3) Ég sendi þér þetta í tölvupósti. DITRANSITIVE CONSTRUCTION<br />
I send you.DAT this.ACC in email SUBJ+VERB+OBJ DAT<br />
+OBJ ACC<br />
‘I’ll send you this with email.’<br />
Bergen Jóhanna Barðdal June 5, 2008
Constructional Choices<br />
93<br />
(4) … faxa þessu bréfi á … CAUSED-MOTION CONSTRUCTION<br />
fax this.DAT letter.DAT on SUBJ+VERB+OBJ DAT<br />
PP DIR<br />
‘… faxed this letter to …’<br />
(5) ... e-maila þér munstrið … DITRANSITIVE CONSTRUCTION<br />
email you.DAT pattern-the.ACC SUBJ+VERB+OBJ DAT<br />
+OBJ ACC<br />
‘... email you the pattern …’<br />
(6) … sms-a mér svefntöflu. DITRANSITIVE CONSTRUCTION<br />
text me.DAT sleeping-pill.ACC SUBJ+VERB+OBJ DAT<br />
+OBJ ACC<br />
‘... text me a sleeping pill.’<br />
Bergen Jóhanna Barðdal June 5, 2008
The Investigation<br />
94<br />
• Questionnaire survey<br />
• 32 participants<br />
• 18 females and 14 males<br />
• Age 14–85<br />
• April 2005 – Iceland<br />
Bergen Jóhanna Barðdal June 5, 2008
The Questionnaire<br />
95<br />
• Ég (e)meila þessu til þín 1 2 3 4 5 6 7<br />
• Ég faxa þetta til þín 1 2 3 4 5 6 7<br />
• Ég sms-a þér þetta 1 2 3 4 5 6 7<br />
• Ég (e)meila þér þetta 1 2 3 4 5 6 7<br />
• Ég sms-a þessu til þín 1 2 3 4 5 6 7<br />
• Ég (e)meila þetta til þín 1 2 3 4 5 6 7<br />
• Ég faxa þér þetta 1 2 3 4 5 6 7<br />
• Ég sms-a þetta til þín 1 2 3 4 5 6 7<br />
• Ég faxa þessu til þín 1 2 3 4 5 6 7<br />
Bergen Jóhanna Barðdal June 5, 2008
Judgments<br />
96<br />
1 This is impossible in Icelandic<br />
2 This is hardly possible in Icelandic<br />
3 I don’t know whether it is possible to say this but I would<br />
never do it<br />
4 I could say this but normally I wouldn’t<br />
5 I might perhaps say this<br />
6 I could very well use this formulation<br />
7 This is exactly how I would say it<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
97
Results (1)<br />
98<br />
1) All three construction are accepted with new verbs of instrument of<br />
communication in Icelandic to a varying degree<br />
2) All verbs are judged most acceptable in the Transfer construction<br />
(with an accusative object), which is the prototypical construction for<br />
verbs of sending<br />
3) The verb faxa, which was borrowed earliest of the three, is most<br />
accepted in the Transfer construction<br />
4) The Transfer construction is rated most acceptable with faxa<br />
5) The Ditransitive construction is rated most acceptable with smsa<br />
6) Caused-Motion construction is rated most acceptable with (e)meila,<br />
although these last two conclusions may be due to a coincidence<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
99
Results (2)<br />
100<br />
7) There is a clear correlation between judgments and the mean age of the participants.<br />
Younger speakers are generally more acceptant while the older ones are less<br />
acceptant.<br />
8) This correlation is strongest for the Caused-Motion construction for all three verbs.<br />
9) There is also a correlation between participant age, judgment and the age of the<br />
technology in the society. Faxing has existed longest and then email which has<br />
existed in Iceland for the last 15 years. Older speakers prefer the verbs faxa and<br />
(e)meila in the Transfer and the Ditransitive construction and disprefer them in the<br />
Caused-Motion construction. Older speakers also prefer sms-a, which represents the<br />
most recent technology, only in the Transfer construction. Whether this is because<br />
older speakers are more conservative, or because they have been less exposed to<br />
the technology and its accompanying language, is difficult to know.<br />
Bergen Jóhanna Barðdal June 5, 2008
Question<br />
101<br />
• Why is the Caused-Motion construction accepted at all with new verbs of<br />
instrument of communication, given that verbs of sending are<br />
conventionalized as occurring in the Transfer and the Ditransitive<br />
construction in Germanic?<br />
(8) Ég sendi þetta til þín. TRANSFER CONSTRUCTION<br />
I send this.ACC to you SUBJ+VERB+OBJ ACC PP DIR<br />
‘I’ll send this to you.’<br />
(9) Ég sendi þér þetta. DITRANSITIVE CONSTRUCTION<br />
I send you.DAT this.ACC SUBJ+VERB+OBJ DAT +OBJ ACC<br />
‘I’ll send you this.’<br />
(10) *Ég sendi þessu til þín. CAUSED-MOTION CONSTRUCTION<br />
I send this.DAT to you<br />
SUBJ+VERB+OBJ DAT PP DIR<br />
Bergen Jóhanna Barðdal June 5, 2008
Bergen Jóhanna Barðdal June 5, 2008<br />
102
Bergen Jóhanna Barðdal June 5, 2008<br />
103
Motion Verbs<br />
104<br />
• Nom-Acc (10 types)<br />
bera ‘carry’, draga ‘draw, pull’, flytja<br />
‘transport’, flæma burt ‘drive away’, færa<br />
‘move’, hreyfa ‘move’, senda ‘send’, taka út<br />
‘withdraw’, toga ‘pull’, (út)breiða ‘spread,<br />
unfold’ …<br />
• Nom-Dat (50 types)<br />
banda frá ‘wave away’, beina ‘direct’, benda ‘direct, point’,<br />
berja (hrömmunum) ‘thrash one’s paws’, blanda ‘mix’, bylta<br />
(sér) ‘turn around’, dreifa ‘spread’, feykja ‘blow’, forða ‘get<br />
sth away’, fresta ‘postpone’, fylgja ‘follow, accompany’,<br />
fyllast ‘become filled with’, henda ‘throw’, hella ‘pour’,<br />
hleypa ‘let, allow (sby into sth)’, hliðra ‘shirk’, kasta ‘throw’,<br />
klappa ‘pat’, koma (sér) ‘get off, go away’, mjaka ‘budge’,<br />
mæta ‘run into sby’, renna ‘glide’, ríða ‘ride’, róa ‘row’,<br />
safna ‘gather’, skella (e-u í sig) ‘slam (in oneself), eat’, skila<br />
‘return’, skjóta ‘shoot’, skjóta upp (kollinum) ‘pop up’, skutla<br />
‘drive’, slengja ‘throw’, sleppa ‘drop, release’, smeygja (sér)<br />
‘slide’, smjúga (fingrum) ‘slip’, snúa ‘turn’, sparka ‘kick’,<br />
spyrna ‘kick’, stilla upp ‘take a position’, stinga ‘insert’,<br />
sveipa ‘wrap’, svæla (í sig) ‘puff (in oneself), eat’, troða<br />
‘squeeze (one’s way)’, tylla (sér) ‘take a seat’, varpa<br />
‘throw’, velta ‘roll’, vefja ‘wrap’, víkja ‘make way’, ýta ‘push’,<br />
þrengja ‘push’, þrýsta ‘squeeze, thrust’…<br />
Bergen Jóhanna Barðdal June 5, 2008
Proportion of motion verbs among Nom-Acc and Nom-Dat<br />
verbs<br />
105<br />
Nom-Acc<br />
Nom-Dat<br />
N<br />
10/303<br />
50/188<br />
f<br />
0.33%<br />
26.6%<br />
Bergen Jóhanna Barðdal June 5, 2008
Borrowed Verbs<br />
106<br />
• 3 Verbs of translational motion (Nom-Acc):<br />
– hakka ‘hack’, logga ‘log’, ssh-a ‘ssh’ …<br />
• 23 Verbs of caused motion (Nom-Dat):<br />
– blasta ‘blast’, bomba ‘bomb’, bundla ‘bundle’, downloada<br />
‘download’, dumpa ‘dump’, droppa ‘drop’, ejecta ‘eject’,<br />
expandera ‘expand’, exporta ‘export’, farta ‘fart’, innstalla/<br />
innstallera ‘install’, neimdroppa ‘name drop’, parkera ‘park’,<br />
peista ‘paste’, poppa ‘deliver when popping by’, pósta<br />
‘post’, publisha ‘publish’, releasa ‘release’, slamma ‘slam’,<br />
streyma ‘stream’, uploada ‘upload’, untara ‘unzip’, unzippa<br />
‘unzip’ …<br />
Bergen Jóhanna Barðdal June 5, 2008
Conclusion: Analogy vs. High type frequency<br />
107<br />
• Icelandic speakers have a choice between the Caused-Motion<br />
construction and the Transfer construction because of:<br />
• the high degree of semantic overlap between senda and<br />
new verbs of instrument of communication<br />
• the high type frequency of the Caused-Motion construction in<br />
Icelandic in general<br />
• the boundaries between “analogy” and “productivity” cannot<br />
be maintained, as both give rise to extensibility, to varying<br />
degrees. As such they are simply two sides of the same coin.<br />
Bergen Jóhanna Barðdal June 5, 2008
Conclusion<br />
108<br />
Bergen Jóhanna Barðdal June 5, 2008
Summary<br />
109<br />
• All three new verbs of instrument of communication are accepted in<br />
all three constructions, although to a varying degree<br />
• The Transfer construction obtained the highest mean score for all<br />
the verbs, presumably on analogy with senda<br />
• Differences in the mean age of the participants who accepted and<br />
rejected the three verbs in the Caused-motion construction, with<br />
younger speakers being more acceptant<br />
• Suggests that the Caused-motion construction is gaining in<br />
productivity in Icelandic, presumably on the basis of the fact that it is<br />
a verb-subclass-specific construction, existing at a considerably<br />
higher level of schematicity than senda<br />
• The boundaries between “analogy” and “productivity” cannot be<br />
maintained, as both give rise to extensibility, to varying degrees<br />
Bergen Jóhanna Barðdal June 5, 2008
Thank You!<br />
110<br />
Bergen Jóhanna Barðdal June 5, 2008