Psycholinguistic perspectives on grammatical representations

1
  Psycholinguistic perspectives on grammatical
representations
Harald Clahsen
2
Introduction

• Uriagereka (2005)The future of linguistics may
  lie in psycholinguistics.
• Chomsky (19819)Evidence from language
  acquisition along with evidence derived from
  psycholinguistic experimentation, the study of
  language use (e.g. processing), language deficit,
  and other sources should be relevant, in
  principle, to determining the properties of UG
  and of particular grammars. But such evidence is,
  for the moment, insufficient to provide much
  insight concerning these problems

3
Structure of the talk

• Part I How to bridge the gap between
  psycholinguistics and theories of grammar
• Part II Evidence from language acquisition
• Part III Evidence from language impairments

4
Mental representations of language

• Linguistics A grammar of a particular language
  is a mental structure consisting of grammatical
  representations which describe what it means to
  know a language.
• Language Processing
• Operations which transform a mental
  representation of a linguistic stimulus into a
  mental representation of a different form.
• Language Acquisition
• A sequence of transitional changes to the mental
  representations of language over time
• Language Impairments
• A normal set of mental representations of
  language minus impaired properties

5
Evaluating psycholinguistic results

• Are there any confounding factors or alternative
  explanations for a given psycholinguistic result?
• Is there converging evidence for a given finding
  from other sources?
• Does a given finding confirm/disconfirm a
  specific linguistic account?

6
Language acquisition The continuity hypothesis
(Weak) Continuity The childs grammar learning
device does not change over time and all
developmental changes are due to increases in the
childs lexicon, semantic and pragmatic
knowledge, and increases in cognitive resources
in general. Continuity makes sure that
developmental evidence will bear on the object of
inquiry that the linguist cares about, the study
of systems constrained by the human language
faculty (Rizzi 2000 269).
7
Three ways of representing regular and irregular
inflection

• Rules all the way down (e.g. Halle Mohanan
  1985)
• Associations all the way up (e.g. Bybee 1995)
• Rules and entries (e.g. Jackendoff 1997,
  Wunderlich 1996)

8
German Participles

• Regulars wischen gewischt 'to mop - mopped'
• holen geholt 'to fetch fetched'
• are affixed with t
• never exhibit any stem changes

• Irregulars fressen gefressen 'to eat - ate'
• trinken getrunken 'to drink - drunk'
• are affixed with (e)n
• sometimes exhibit (phonologically unpredictable)
  stem changes

• ge- prefixation prosodically determined, not
  morphologically
• occurs when stem is stressed on the first syllable

9
Stem formation in German
10
Participle formation in German child language

• gekommt 'come (correct gekommen)
• Age Total -t -n
• range children errors errors errors
• Existing verbs
• 14-39 9 116 108 (93) 8 (7)
• 36-611 51 88 77 (87,5) 11 (12,5
• 72-811 19 64 59 (92,2) 5 (7,8)
• Nonce words
• 310-810 41 454 422 (93) 32 (7)

11
Verb Frequency and Suffixation Errors
12
Stem formation errors in German child language

• 73 samples of spontaneous speech from 7 children
  covering the age period of 111 to 38
• Number of cases
• I. Overapplications of unmarked stem 84 (88.4)
• (a) er lauft he runs (correct
  läuft) (b) sie lest she reads (correct
  liest)
• II. Paradigmatic errors 11 (11,5)
• (c) alle fäll da runter (correct fall-en)
  everybody fall down there(d) ich gib dir
  das (correct geb(e)) I give you
  that(e) ich sieh (I see) (correct seh(e))
• III. Irregularization error non-existent
• sie tänzt (correct tanzt) she dances

13
Stem formation errors in German child language

• Auditory elicited production task with 26
  children (age 62 to 105)
• Martin will unbedingt den neuen Pokemon-Film
  sehen.M. definitely wants to see the new
  pokemon movie.
• Also gibt ihm seine Mutter Geld und Martin beep
  den Film. ? siehtHence his mother gives him
  some money, and Martin ___ the movie

14
Results of the elicited production task

• There were 168 errors out of 555 elicited forms
  all errors were overapplications of the unmarked
  stem.
• Low-frequency stems elicit significantly more
  stem errors than high-frequency ones
  Errors Correct Error stems
• -i- / high freq. 19 136 12.2
• -i- / low freq. 89 59 37.3
• -ä- / high freq. 9 118 7.0
• -ä- / low freq. 51 74 40.8
• Totals 168 387 30.2

Stem overregularizations in relation to age
15
Preliminary summary

• Regular/irregular contrasts in childrens
  inflectional errorsChildren overapply the
  regular t participle suffix and the unmarked
  stem to irregular verbs.
• Frequency effects in childrens inflectional
  errorsChildren produce more overregularizations
  for irregular verbs with low frequencies than for
  those with high frequencies.
• Age effects in childrens inflectional
  errorsOverregularization errors decrease with
  age.

16
Confounding factors?

• Regular rules of morphology usually have high
  type frequency and apply to a large number of
  different forms.
• The type frequency of the German t participle is
  much higher than that of irregulars.
• (Bybee 1999, Stemberger 1999)

Input frequencies (in types) (4 corpora, children
from 15 to 21, app. 40,000 words) -t
participle forms 45 -n participle
forms 55
17
Converging evidence?

• Within the same languagePlural formation in
  German child language (Clahsen et al. 1992,
  Bartke 1998)
• Across languages- Development of the English
  past-tense (Marcus et al. 1992)-
  Development of verb inflection in L1
  Spanish (Clahsen et al. 2002)

18
Specific theory?

• Rules all the way down
• Associations all the way up
• Rules and entries

Acquisition results provide evidence against (A)
and (B).
19
Specific theory?

• The basic distinction between combinatorial and
  frozen forms can be implemented in different
  ways
• Rules and entries (e.g. Wunderlich 1996)
• Rules that contain variables and those that have
  a constant output (e.g. Blevins 2001)
• a. ltV, 3sg, pres, ind, Xsgt
• b. ltV, 3sg, pres, ind, be, isgt
• - Probabilistic rules vs. default rules (Yang
  2000)

Acquisition results do not help to decide between
these accounts.
20
Language impairments
The breakdown-compatibility criterion Patterns
of impairment and sparing of linguistic ability
should be compatible with linguistic
theory. (Grodzinsky 1990 111) Double
dissociations Given two linguistic phenomena A
and B, if A is impaired in one population (where
B is spared) and B is impaired in another
population (where A is spared), then A and B are
likely to be supported by different mental
representations.
21
Passives in syntactic theory

• Transformational accounts (e.g. Chomsky 1981)
• The fishi is eaten ti by the man
• A?chain

• Lexicalist accounts (e.g. Bresnan 1982)
• The man eats the fish / The fish is eaten
  by the man
• SUBJ OBJ OBJ SUBJ
• Functional changes SUBJ ? OBJ (BY OBJ) OBJ ?
  SUBJ
• Morphological change V ? Vpart

22
Binding in syntactic theory

• Standard Binding Theory (Chomsky 1981)

• Binding of non-reflexive pronouns is based on
  principles of semantic interpretation, reflexive
  binding is defined of syntactic constraints on
  potential antecedent domains (e.g. Pollard Sag
  1992, Kiparky 2002).
• Deriving binding phenomena from independent
  syntactic principles (Hornstein 2001, Reuland
  2001). John believes that Maryi likes
  herselfi

A-chain
23
Downs Syndrome

• DS is a congenital disorder caused by an extra
  copy of a segment of Chromosome 21 that is
  associated with specific physical features and
  cognitive delay.
• Language abilities are relatively more impaired
  than other areas of cognition.
• Morphosyntax is more impaired than other
  linguistic domains.
• Patterns of morphosyntactic skill that are
  qualitatively different from those observed in
  normally developing children.

24
Participants
25
Method Passives

• Picture-pointing task (Van der Lely, 1996)
• 1. Active transitive The man eats the fish
• 2. Full verbal passive
• The fish is eaten by the man
• 3. Short progressive passive
• The fish is being eaten
• 4. Ambiguous passive
• The fish is eaten

26
Passives visual materials
The fish is eaten by the man
1
2

• 1. Correct
• 2. Adjectival
• 3. Reversal
• 4. Distracter

3
4
27
Method Binding

• 1. Name-pronoun
• Is Mowgli tickling him?
• 2. Name-reflexive
• Is Mowgli tickling himself?

• 3. Quantifier-pronoun
• Is every monkey tickling him?
• 4. Quantifier-reflexive
• Is every monkey tickling himself?

28
Results Binding
Percentages Correct DS vs. Controls
DS significantly lower scores on reflexive
conditions
29
Results Passives
Percentages Correct and Percentages of Reversal
Responses
DS chance performance on passives and sig. more
reversal responses than controls.
30
Preliminary Summary

• DS Binding
• good performance on pronouns
• poor performance on reflexives
• DS Passives
• low accuracy on passive structures
• high number of reversal errors

31
Confounding factors?

• Can the DS childrens difficulties with passives
  and reflexive binding be derived from their low
  IQ levels?
• Comparison with Williams Syndrome
• Can the DS childrens difficulties with passives
  and reflexive binding be explained in terms of
  delayed language development?
• Comparison with younger normal children

32
Binding in Williams Syndrome
Percentages Correct WS vs. Controls
WS high correctness scores and no significant
differences to controls
33
Passives in Williams Syndrome
Percentages Correct WS vs. Controls
WS high correctness scores and no significant
differences to controls
34
Binding in younger normal children

• Reflexives30 children (age range 26 53)
  achieved accuracy scores of gt90 (McKee 1992).
• Pronouns19 children (age range 40 51)
  achieved accuracy scores of lt50 (Thornton
  Wexler 1999).

35
Converging evidence?

• BindingPerovic (2004) tested four young adults
  with DS (CA 172 to 207 years) achieving near
  perfect accuracy scores of gt90 on non-reflexive
  pronouns and poor scores of lt60 on reflexives.
• PassivesBridges Smith (1984) tested 24 DS and
  24 non-retarded children matched to the DS
  children and found accuracy scores of over 80 on
  actives and of around 50 on passives.

36
Breakdown compatability?

• A double dissociation

37
Breakdown compatability?

• DS Binding and Passives
• - impaired reflexive binding
• - low accuracy on passive structures
• ? Binding of reflexives and passivization involve
  the same syntactic mechanism (A-chains).

38
Conclusion

• Has any theoretical linguist ever changed
  his/her theory in the face of psycholinguistic
  evidence?
• Common ground the search for the most
  appropriate mental representations for language.
• Three criteria for evaluating psycholinguistic
  results
• ?Psycholinguistic evidence may help to adjudicate
  between competing linguistic accounts.