The Quantificational Apparatus of Language: Integrating Theory, Development, and Pathology.
Julien MusolinoRutgers University
Quantification
Theoretical Linguistics
Developmental Psychology
Speech-language pathology
How does quantification work?
How does quantificational competence develop?
What happens in the case of atypical development?
Part 1: Developmental PsychologyPart 1: Theoretical Linguistics
Provide the ‘technical glue’ that is going to bind all three parts together.
Provide some general background on linguistic quantification.
Tell you why quantification is worth studying – why we should care.
Directly apply the theoretical notions discussed in part 1 to a classic developmental puzzle.
Part 2: Developmental Psychology
Present results that will lead to a reinterpretation of most previous research in that domain.
This will set the stage for part 3.
The approach developed in part 1 and 2 naturally extends to the study of atypical development.
Part 1: Developmental PsychologyPart 2: Linguistic TheoryPart 3: Speech-language pathology
Focus on the case of Williams Syndrome (WS).
Show that the integrative approach provides a unique way to address the central question of whether grammar is spared in WS.
To close the loop, I will show that new results on WS have implications for all three fields.
Part 1: Developmental PsychologyPart 1: Theoretical Linguistics
Provide the ‘technical glue’ that is going to bind all three parts together.
Provide some general background on linguistic quantification.
Tell you why quantification is worth studying – why we should care.
Quantifiers_______________________________________________________
Some, all, two, many, every, no …
Give us the power to express generalizations about quantities of individuals.
Quantifiers_______________________________________________________
“You can fool some of the people all of the time, and all of the people some of the time, but you cannot fool all of the people all of the time.”
“You can fool some of the people all the time, and those are the ones you want to concentrate on.”
Abraham Lincoln (attributed)
George W. Bush
Why Study Quantification?_______________________________________________________
Core property of natural language
Extremely complex phenomenon
Causes problems until late in development
Quantifiers_______________________________________________________
Scope
C-command
Covert displacement
Logical Form
Buying candy …
not everybody
The phenomenon_______________________________________________________
(1) Every N neg VP
a. ‘None’
b. ‘Not all’
Scope_______________________________________________________
(2 X 3) + 5
2 X (3 + 5)
Scope & C-command_____________________________
(2 x 3) + 52 x (3 + 5)
Scope = C-command domain
Scope_______________________________________________________
(1) Every horse didn’t jump over the fence
Every horse (not jump) ‘none’
• Isomorphic interpretation
• Every horse is interpreted outside the scope of negation
Scope and covert displacement_______________________________________________________
(1) Every horse didn’t jump over the fence
(Not every horse) jumped ‘not all’
• Every horse is interpreted within the scope of negation
• Non-isomorphic interpretation
Logical Form (LF)_______________________________________________________
(1) Every N neg VP
ISOMORPHIC LF = every > not NON-ISOMORPHIC LF = not > every
IP
Every N I’
NEG VP
V’
IP
I’
NEG VP
Every N V’
Covert displacement
Why should we care?_______________________________________________________
Because the mapping between form and meaning is complex.
Because the grammatical operations involved are undetectable in the surface form.
How can we learn about what we can’t detect?
Directly apply the theoretical notions discussed in part 1 to a classic developmental puzzle.
Part 2: Developmental Psychology
Present results that will lead to a reinterpretation of most previous research in that domain.
This will set the stage for part 3.
Early use_______________________________________________________
‘Because there no pictures’ (Eve, 2;1)
‘I drink all grape juice’ (Eve, 1;10)
‘Then Eve have some milk’ (Eve, 1;11)
‘Two knife out the box’ (Eve, 1;11)
_________________________________________________________ Cause problems until late
Adults: YES 5-year-olds: NO
Is every dog on a mat?
Not this one
Previous Accounts_________________________________________________________
Lack of conceptual knowledge (Inhelder & Piaget, 1964)
Lack of syntactic knowledge (Bucci, 1978; Roeper and deVilliers, 1991)
Lack of semantic knowledge (Philip 1995, Drozd & van Loosbroek, 1999)
Incomplete knowledge (Musolino, Crain and Thornton, 2000)
In sum_________________________________________________________
On most accounts, children’s ‘errors’ are taken to reflect a lack of knowledge (conceptual or linguistic)
Remarkable lack of consensus regarding the nature of the problem
Vast majority of studies have focused on one construction
Specific goals_________________________________________________________
Investigate a broader range of quantificational phenomena and show that:
Accounts based on ‘lack of knowledge’ are likely to be incorrect
The ‘errors’ that children make can be used to uncover their grammatical knowledge
Results to be presented_______________________________________________________
I won’t discuss all the details (number of
subjects, age range, types of analyses) but …
Most of the data have been published
Most of the data have been replicated
Children are preschoolers (4-5 age range)
Experimental methodology_______________________________________________________
Truth Value Judgment Task
Crain and Thornton, (1998)
Universally quantified subjects_______________________________________________________
(1) Every horse didn’t jump over the fence.
Musolino, Crain and Thornton (2000) Linguistics
One of the stories_______________________________________________________
One of the stories_______________________________________________________
One of the stories_______________________________________________________
One of the stories_______________________________________________________
One of the stories_______________________________________________________
One of the stories_______________________________________________________
One of the stories_______________________________________________________
The end of the story “Every horse didn’t jump over the fence, am I right?”
Results (5-year-olds vs. adults)_______________________________________________________
0
0.2
0.4
0.6
0.8
1
5-year-olds Adults
Children’s justifications_______________________________________________________
“Every horse didn’t jump over the fence, am I right?”
Child:”You’re wrong because these two horses jumped over the fence!”
Numerally quantified objects_______________________________________________________
(2) The Smurf didn’t catch two birds
Lidz and Musolino (2002) Cognition
a. Not (caught 2)
b. 2 (not caught)
Isomorphic Condition_______________________________________________________
2 (not caught) = FALSE
Not (caught 2) = TRUE
Non-Isomorphic Condition_______________________________________________________
2 (not caught) = TRUE
Not (caught 2) = FALSE
NP didn't V two N: adult data
0
0.2
0.4
0.6
0.8
1
WtNf WfNt
Pro
port
ion
of Y
ES
res
pons
esResults: Adults_______________________________________________________
IsomorphicNon-Isomorphic
Results: Children_______________________________________________________
NP didn't V two N: child data
0
0.2
0.4
0.6
0.8
1
WtNf WfNt
Pro
po
rtio
n o
f Y
ES
re
sp
on
se
s
IsomorphicNon-Isomorphic
Children’s justifications (non-iso)_______________________________________________________
“The Smurf didn’t catch two birds, am I right?”
Child:”You’re wrong, she did catch two!”
Children’s justifications (iso)_______________________________________________________
“The Smurf didn’t catch two birds, am I right?”
Child:” You’re right! She only caught one”
The observation of Isomorphism_______________________________________________________
“Young children, unlike adults, have a strong tendency to interpret sentences containing quantified NPs and negation on the basis of the surface syntactic position of these elements”
Musolino, Crain and Thornton (2000) Linguistics
The observation of Isomorphism_______________________________________________________
Children Adults
Sentence type
Every horse didn’t jump over the fence
The Smurf didn’t catch two birds
2 22
Questions
Developmental question
Causal question
Structural question
The structural question _______________________________________________________
What underlies isomorphism?
Linear order ?
C-command ?
IP
SUBJECT I’
English (SVO)_______________________________________________________
INeg
VERB OBJECT
VP
Subj > Neg
Neg > Obj
Kannada
• Approximately 40 million speakers in Karnataka, south-western India.
Scope ambiguity in Kannada
naanu eraDu pustaka ood-al-illa
I-nom two books read-inf-neg
‘I didn't read two books.’
a. Not (read 2)b. 2 (not read)
S O V
IP
SUBJECT I’
Kannada (SOV)_______________________________________________________
INeg
OBJECT VERB
VP
Predictions for Kannada
To the extent that Kannada children display a preference for one of the two readings:
_______________________________________________________
C-command: same results as English
Linear order: opposite results from English
Results: Adults
'NP two Ns Ved not: adults"
0.87 0.85
0
0.2
0.4
0.6
0.8
1
Wide-true / Narrow-false Wide-false / Narrow-true
Pro
port
ion
of 'Y
es' r
espo
nses
2 (not caught) not (caught 2)
Results: English vs. Kannada
English KannadaNP didn't V two N: child data
0
0.2
0.4
0.6
0.8
1
WtNf WfNt
Prop
ortio
n of
YES
resp
onse
s
2 (not) Not (2)0
0.2
0.4
0.6
0.8
1
2 (not) not (2)2 (not) Not (2)
Conclusions _______________________________________________________
They differ in ways that are constrained by fundamental linguistic principles (i.e. c-command).
Children systematically differ from adults.
Children’s ‘errors’ tell us about the kinds of linguistic representations that they entertain.
Objections_______________________________________________________
Hint: complexity of displays, reverse linear order, selective focus
What else could account for the results on c-command?
The causal question_______________________________________________________
Lack of syntactic knowledge (Bucci, 1978; Roeper and deVilliers, 1991)
Lack of semantic knowledge (Philip 1995, Drozd & van Loosbroek, 1999)
Incomplete knowledge (Musolino, Crain and Thornton, 2000)
Lack of knowledge?_______________________________________________________
Children can be made to behave like adults
Adults can be made to behave like children
Musolino and Lidz (2003) Language Acquisition
Musolino and Lidz (2006) Linguistics
Turning children into adults_______________________________________________________
(2) Every horse jumped over the log but every horse didn’t jump over the fence
(1) Every horse didn’t jump over the fence.
One of the stories_______________________________________________________
The end of the story “Every horse jumped over the log but every horse didn’t jump over the fence, am I right?”
_______________________________________________________Results
0
0.2
0.4
0.6
0.8
1
5-year-olds Adults
Pro
po
rtio
ns
of
'YE
S' r
esp
on
ses
No contrast
Contrast
Conclusions_______________________________________________________
Context can be manipulated so as to boost children’s ability to access non-isomorphic interpretation
Renders ‘lack of knowledge’ account implausible
Turning adults into children_______________________________________________________
(2) NP didn’t V two N
What is the adult preference?
NP didn't V two N: adult data
0
0.2
0.4
0.6
0.8
1
WtNf WfNt
Pro
po
rtio
n o
f Y
ES
res
po
nse
s
Non-Isomorphic Isomorphic
Two > not = TRUE
Not > two = TRUE
Cookie Monster didn’t eat two slices of pizza
Same material as in Lidz and Musolino (2002)
Justifications_______________________________________________________
Narrow scope: “because he only ate one slice – not two”
Wide scope: “because there are two slices that he didn’t eat”
Unclear
Results (adults, n=20)_______________________________________________________
0.75
0.075
0.175
0
0.2
0.4
0.6
0.8
1
Narrow Wide Unclear
Turning adults into children_______________________________________________________
(2) NP didn’t V two N
What is the adult preference?
NP didn't V two N: adult data
0
0.2
0.4
0.6
0.8
1
WtNf WfNt
Pro
po
rtio
n o
f Y
ES
res
po
nse
s
Non-Isomorphic Isomorphic
Conclusions_______________________________________________________
In this case, children’s preference for isomorphic interpretations reflects an exaggerated preference also observable in adults
Inducing Isomorphism in adults_______________________________________________________
(2) Two frogs didn’t jump over the rock
There are two frogs that didn’t jump over the rock (two>not)
It is not the case that two frogs jumped over the rock (not>two)
Isomorphic condition_______________________________________________________
Two > not = TRUE
Not > two = FALSE
Non-Isomorphic condition_______________________________________________________
Two > not = FALSE
Not > two = TRUE
Results (adults, n=20)_______________________________________________________
0
0.2
0.4
0.6
0.8
1
Isomorphic Non-isomorphic
Helping adults_______________________________________________________
(1) Two frogs didn’t jump over the rock.
(2) Two frogs jumped over the fence but two frogs didn’t jump over the rock.
Two frogs jumped over the fence but two frogs didn’t jump over the rock.
Two > not = FALSE
Not > two = TRUE
Results (adults)_______________________________________________________
0
0.2
0.4
0.6
0.8
1
No-contrast Contrast
Conclusions_______________________________________________________
Remarkable continuity between the developing and the mature system
Children’s isomorphic behavior represents exaggerated preferences also observable in adults
The isomorphic effect can be induced in adults
The contextual factors that help children overcome their isomorphic tendencies have the same effects on adults
General conclusions_______________________________________________________
Systematic differences in the way children and adults interpret quantified statements.
Children’s errors can be used to uncover their grammatical knowledge.
Accounts based on ‘lack of knowledge’ are likely to be incorrect.
General conclusions_________________________________________________________
By the age of 5, children have adult-like knowledge of the grammar of quantification
However, children differ from adults in the way they implement their knowledge
Differences in processing resources deployed during language comprehension
The kindergartenpath effect (Trueswell et al. 1999)
SENTENCE
LF1 isomorphic LF2 non-isomorphic
Children & adults
Adults are better
The approach developed in part 1 and 2 naturally extends to the study of atypical development.
Part 1: Developmental PsychologyPart 2: Linguistic TheoryPart 3: Speech-language pathology
Focus on the case of Williams Syndrome (WS).
Musolino, Landau, and Chunyo (2007), In preparation
Issue
The status of grammatical knowledge in individuals with Williams Syndrome (WS)
Spared or impaired?
_______________________________________________________
Why should we care?
So WS can potentially to tell us a lot about the structure and development of the human mind.
WS suggests a potential dissociation between language and other aspects of cognition (e.g., spatial cognition).
_______________________________________________________
Main claim
Knowledge of core, abstract principles of syntactic and semantic computation is spared in WS.
Implications for the relevance of WS in the debate over modularity.
_______________________________________________________
Outline
Background on WS
Competing views
Apply the integrative approach
Experimental evidence
Implications
_______________________________________________________
Williams Syndrome_______________________________________________________
Williams Syndrome
Rare genetic disorder (1/15,000 live births) involving a micro-deletion on chromosome 7.
Physical anomalies along with mild to more serious mental retardation (IQ average 70).
Uneven cognitive profile with areas of strength (e.g., language) in the face of serious deficits in areas such as spatial cognition, motor planning, and number.
_______________________________________________________
Language in WS
“Verbal advantage over non-verbal intelligence” (Karmiloff-Smith et al., 1997)
Verbal IQ > Performance IQ
Different from individuals with similar levels of mental retardation (e.g., Down Syndrome)
_______________________________________________________
Interpretations
Is grammar spared in WS?
Yes (e.g., Pinker, 1999; Clahsen & Almazan, 1998; Brock, 2006; Mervis et al., 2003; Zukowski, 2006)
No (e.g., Karmiloff-Smith, 1998; Karmiloff-Smith et al., 1997; Thomas & Karmiloff-Smith, 2003 )
_______________________________________________________
Proponents of some version of modularity
Neuroconstructivist view
Knowledge of grammar is impaired or deviant in WS.
WS individuals learn language using different cognitive mechanisms.
Explicitly rejects modularity.
Emphasis on rote learning and inability to extract underlying regularities and form linguistic generalizations.
_______________________________________________________
Neuroconstructivist view
- Karmiloff and Karmiloff-Smith (2001)
“It has become increasingly clear, therefore, that the superficially impressive language skills of individuals with WS may be due to good auditory memory rather than an intact grammar module” (p.202-3).
“We argue that the language of WS people, although good given their level of mental retardation, will not turn out to be “intact””(p.247)
- Karmiloff et al. (1997)
_______________________________________________________
Neuroconstructivist view
“… they will tend to acquire a large number of words by rote and only weakly extract underlying regularities.” (p.257)
“This suggests that if WS children go about language acquisition differently from normal children … they will end up – as they indeed do – with large vocabularies but relatively poor system building” (p. 257)
- Karmiloff et al. (1997)
- Karmiloff et al. (1997)
_______________________________________________________
Goals
Focusing on knowledge of core syntactic and semantic principles.
Apply the integrative approach to test the predictions of these two opposing views.
Using an experimental technique (the TVJT) which has a proven track record in uncovering such knowledge.
_______________________________________________________
Looking at quantificational phenomena.
The phenomenon
The interaction of negation and disjunction
… NOT … OR …
_______________________________________________________
The phenomenon
(1) John had a beer or a glass of wine.
(2) * John had a beer and John had a glass of wine.
(3) John didn’t have a beer or a glass of wine.
(4) John didn’t have a beer and John didn’t have a glass of wine.
_______________________________________________________
De Morgan’s laws
(P Q) ( P) ( Q)
Not (beer or wine) (not beer) and (not wine)
“The negation of the disjunction of two propositions is logically equivalent to the conjunction of their negations”
_______________________________________________________
The phenomenon
The interpretation of negation and disjunction is governed by De Morgan’s laws of propositional logic
But only when disjunction occurs in the scope of negation
_______________________________________________________
Scope condition
C-command
Precede only
Inclusive (Neither)
Exclusive (Either)
(5) The man who got a pay raise didn’t buy a BMW or a Peugeot.
(6) The man who didn’t get a pay raise bought a BMW or a Peugeot.
_______________________________________________________
Knowledge to be tested
(a) Syntax: scope/c-command
(b) Semantics: entailment relations and De Morgan’s laws of propositional logic
(c) The relationship between (a) and (b)
… NOT … OR …
_______________________________________________________
Predictions
Spared grammar view: knowledge of scope, c-command, entailment relations and De Morgan’s laws should be spared.
Neuroconstructivist view: knowledge of scope, c-command, entailment relations and De Morgan’s laws should be impaired.
_______________________________________________________
Experiment
The idea
C-command
Precede
Neither
Either
The man who got a pay raise didn’t buy a BMW or a Peugeot.
The man who didn’t get a pay raise bought a BMW or a Peugeot.
_______________________________________________________
4 control conditions
The man [who got a pay raise] didn’t buy a BMW or a Peugeot.
Control 1
Control 2Control 3
Control 4
_______________________________________________________
The man who got a pay raise didn’t buy a BMW or a Peugeot
C-command
FALSE
The man who didn’t get a pay raise bought a BMW or a Peugeot
Precede
True
Participants (3 groups)
12 individuals with WS (M age = 16;4, Range = 11;10 to 21;11) (M IQ = 63; SD = 4.19)
12 Mental Age controls (M = 6;1, Range = 5;2 to 7;8) (M IQ = 118; SD = 2.44)
12 adults (all college undergraduates)
_______________________________________________________
Design
Conditions: 2 experimental and 4 control
- Precede vs. C-command (experimental)
3 (groups) x 6 (conditions)
Groups = WS, MA, Adults
- Or, negation, relative clauses, De Morgan’s law (controls)
_______________________________________________________
Methodology
Truth Value Judgment Task
- Short, animated vignettes presented on a computer monitor with pre-recorded narration.
- Statements about what happened are heard at the end of each vignette.
- Participants have to decide whether the statements are true or false.
_______________________________________________________
Results
Control conditions (proportion of correct responses)
0
0.2
0.4
0.6
0.8
1
Negation Disjunction Relative DeMorgan
WS
MA
- No main effect of group, condition, or interaction (All p values > .1)
Interim conclusions
Participants experienced no difficulty with the task.
WS and MA have knowledge of the meaning of components that make up the experimental items (or, negation, relative clauses, De Morgan).
WS do not differ from MA.
_______________________________________________________
Experimental conditions (proportion of correct responses)
0
0.2
0.4
0.6
0.8
1
C-command Precede
WS
MA
- Main effect of group (p < . 05), no main effect of condition, or interaction (both p values > . 4)
* *
Interim conclusions
Both WS and MA have knowledge of c-command and De Morgan’s laws.
MA are slightly better at implementing their knowledge.
_______________________________________________________
Implementation
No differences between the two groups on each of the components when considered in isolation.
Cumulative difficulty of these interacting components is more taxing for WS than MA.
Presumably due to differences in processing resources between the two groups.
_______________________________________________________
Implementation
Is level performance on experimental conditions related to level of performance on control conditions (correlational analysis)?
YES for both WS and MA (r ≥ .7)
_______________________________________________________
Implementation
What is the precise nature of these correlations (regression analysis)?
For both WS and MA, performance on negation was a significant predictor of overall performance (accounting for 46% and 64% of the variance, respectively)
Makes sense given what is known about processing difficulties associated with negation (e.g., Horn, 1989)
_______________________________________________________
Implementation
No qualitative difference in the way WS and MA implement their knowledge.
If we are on the right track:
Results comparing WS to 4-year-olds suggest that this is indeed the case.
WS performance should be similar to that of younger, typically developing children.
_______________________________________________________
WS vs. MA (proportion correct responses)
0
0.2
0.4
0.6
0.8
1
Precede C-command
Negation Disjunction Relative DeMorgan
WS
MA
WS vs. 4-year-olds (proportion correct responses)
0
0.2
0.4
0.6
0.8
1
prec
ede
c-co
mman
d
nega
tion
disjun
ction
relativ
es
demorg
an
WS
4s
Conclusions
Results are compatible with the spared grammar view.
Knowledge of core, and very abstract principles of syntactic and semantic computation is preserved, or intact, in the WS population (scope, c-command, entailment relations, De Morgan’s laws)
_______________________________________________________
Conclusions
Empirical challenge
Theoretical challenge
Our results directly contradict the claim that knowledge of grammar is not intact in WS.
How is knowledge of c-command, etc. acquired if the cognitive mechanisms involved are different from those used by typically developing children ?
Challenge for the neuroconstructivist view.
_______________________________________________________
Conclusions
The study of typical and atypical development must proceed hand and hand, with each area informing the other.
_______________________________________________________
If everything I said wasn’t clear, it is because I didn’t tell you
everything …