Transcript Document
Brass Tacks in Linguistic Theory
Stephen Crain, Andrea Gualmini and Paul Pietroski University of Maryland
and
Rosalind Thornton Gennaro Chierchia Luisa Meroni Teresa Guasti Graciela Tesan
The Challenge
“Certainly, humans are endowed with some sort of predisposition toward language learning. The substantive issue is whether a full description of that predisposition incorporates anything that entails
specific contingent facts
about natural languages.” [our emphasis] “We question whether children learn what transformational generative syntacticians think they learn.” (Pullum and Scholz 2002, TLR)
Specific Contingent Facts about Natural Languages
1. The ‘basic’ meaning of disjunction is inclusive-
or
2. The meanings of natural language Determiners are ‘conservative’ 3. The same structural notion (‘c-command’) ties together disparate linguistic phenomena: • Licensing negative polarity items • • Generating the ‘conjunctive’ reading of disjunction Disjoint reference between pronouns and names
…
Poverty of Stimulus Arguments
1. the linguistic knowledge children achieve, and the course of acquisition 2. the linguistic input children receive 3. the nonlinguistic capacities of children Nativists point to the disparity between (1) and (2), even given optimistic assumptions abut (3), so long as this consists of ‘domain general’ capacities to extract information and to form generalizations based on experience
Poverty of Stimulus Arguments
1. the linguistic knowledge children achieve, and the course of acquisition 2. the linguistic input children receive 3. the nonlinguistic capacities of children Nativists point to the disparity between (1) and (2), even given optimistic assumptions abut (3).
Three kinds of poverty-of-stimulus arguments are offered
Poverty of Stimulus Argument I
First
, children project beyond their experience in ways that the input does not even suggest: children learn partial linguistic generalizations. A constraint on interpretation: (a) * He said that the Troll jumped the best (b) The Troll said that he jumped the best.
(c) When he was jumping, the Troll ate sushi.
not the usual induction problem
Nativism is NOTestablished by the fact that children generalize: “ … projecting beyond experience is just one aspect of language acquisition. Children also fail to project beyond their experience in characteristic ways. It is
this
fact that most impresses nativists. The theoretical problem posed by human language learning is to explain why children project beyond their experience just so far and no further; the specific "angle" of projection seems arbitrary (and idiosyncratic to
linguistic
projection).” Crain and Peitroski (2001)
Poverty of Stimulus Argument II
Second
: children project beyond their experience in ways that the input does not even suggest. For example, some English-speaking children produce constructions that are not attested in the local language, but are found in other languages. The detailed patterns of children’s productions is difficult to explain using domain general capacities to form generalizations. By contrast, the patterns of children’s productions are consistent with the
Continuity Assumption
.
The continuity Assumption
“All principles and constructs of universal grammar are available at the outset and all child grammars will be ‘possible human grammars’, in the sense of falling within the patterns of adult grammars (either observed or permitted under the theory). The child’s grammar may, however, deviate from that of the language he will ultimately acquire.” (Goodluck 1991) Child language can differ from the language of adults in the same linguistic community, but only in ways that adult languages can differ from each other.
(Crain 1990)
…versus input matching
“... speakers acquire schemas or templates through exposure to actually-occurring expressions, and use those schemas to sanction new expressions.” (van Hoek 1995) “…perhaps the mental grammar actually consists of positive CONSTRUCTS that license sentences, perhaps ample positive evidence is available, and perhaps positive evidence can support learning. If so, then the negative evidence hallmark is irrelevant to questions about innateness ... “ (Rosen and Rosen 1994) “ such an approach relies on there being a strong relationship between the nature of the lexically-specific patterns in the child's speech and the frequency with which such patterns occur in the child's input.
” (Rowland and Pine 2000)
Medial Wh-questions in child English
Children project beyond their experience in ways that the input does not even suggest. For example, some English speaking children insert a ‘extra’ copy of a wh word in certain Wh-questions, but not in others.
Who do you think who that is
?
What do you think what is in the box
? The detailed pattern of the questions children do and do not ask is difficult to explain on an experience-dependent account of language development.
A Trace of Germanic in Child English
Wer i glaubst du wer i nach Hause geht?
‘Who do you think who goes home?’
*Wessen Buch i glaubst du wessen Buch i Hans liest? ‘
Whose book do you think whose book Hans is reading?’
#Which Smurf do you think which Smurf is wearing roller skates?
Cf. Which Smurf do you think is wearing roller skates?
Wen versucht Hans anzurufen? ‘
Whom is Hans trying to call?’
#Who is Hans trying who to call?
# Who do you want
WHO
to brush your hair
?
QuickTime™ and a Sorenson Video decompressor are needed to see this picture.
Poverty of Stimulus Argument III
Third
, children form ‘deep’ generalizations -- which tie facts together which could not be related by relying “entirely on generalization from experience by the ordinary methods that are used for learning other (nonlinguistic) things from experience” (Pullum and Scholtz 2002) 1. children know how to interpret statements with disjunction (e.g., ‘
not A or B’; ‘Every A or B’…
) 2. and children know the conditions that license expressions like ‘
any
’ and ‘
ever
’ (so-called Negative Polarity Items NPIs) 3. and children know when a pronoun and a name must have disjoint reference Linguistic theory suggests that these phenomena (and others) are governed by the same linguistic principles
Disjunction
The meaning of ‘ or ’ is twofaced: POSITIVE:
Snow-white will find the apple or the banana.
Snow-white will find the apple
or
Snow-white will find the banana
.
This sentence is true in circumstances associated with ‘exclusive-or’ NEGATIVE:
Snow-white will not find the apple or the banana
Snow-white will never find the apple the banana
and
Snow-white will never find
This sentence has the ‘ conjunctive ’ interpretation of ‘ or ’
Disjunction
• The conjunctive interpretation of
or
is derived only if it is assigned the truth conditions associated with inclusive-
or
• The input to children is almost entirely consistent with exclusive-
or
being the basic meaning of disjunction (see Appendix) • In limited positive contexts,
or
-statements may be understood as inclusive-
or
: E.g., situations of uncertainty (making a prediction or a bet)
More on Disjunction: De Morgan’s Laws
Across natural languages, negative statements with disjunction give rise to inferences that closely resemble the equivalences expressed by De Morgan ’ s Laws (Boster and Crain, 1993; Partee, ter Meulen and Wall, 1993).
(P Q) ( P Q) (P Q) ( P Q)
More Generally: Downward Entailment
a. Children left
before
we served zucchini or broccoli Children left before we served zucchini and children left before we served broccoli. b.
Every
child who likes zucchini or broccoli will eat soup Every child who likes zucchini will eat soup and every child who likes broccoli will eat soup.
c
.
None
of the children ate zucchini or broccoli None of the children ate zucchini and none of the children ate broccoli.
Minimal Pairs: Interpreting Disjunction
Every linguist or philosopher admires Chomsky.
Every psychologist admires Chomsky or Halle.
If a linguist or a philosopher goes to the gym, Geoff goes swimming.
If a linguist goes to the gym, Geoff goes swimming or golfing.
Geoff usually arrives at the gym before linguists or philosophers.
Geoff usually arrives at the gym after linguists or philosophers.
Geoff went to work without the car or the bike. Geoff went to work with the car or the bike.
Same Minimal Pairs - Licensing NPIs
Every psychologist who read any linguistics book admires Chomsky.
*Every psychologist admires any philosopher.
If any linguist goes to the gym, Geoff goes swimming.
*If a linguist goes swimming, Geoff goes to any gym.
Geoff usually arrives at the gym before any linguist.
*Geoff usually arrives at the gym after any linguist.
Geoff went to work without any means of transportation. *Geoff went to work with any means of transportation.
Descriptive Generalization
Linguistic environments that permit “any” also yield the conjunctive interpretation of “or”
Every linguist or philosopher admires Chomsky.
Every psychologist who read any linguistics book admires Chomsky.
Linguistic environments that permit do not “any” yield the exclusive-or interpretation
Every psychologist admires Chomsky or Halle.
*Every psychologist admires any philosopher.
A Deeper Generalization
Downward Entailing expressions permit “any” and yield the conjunctive interpretation of “or.” So, De Morgan’s Law is just the tip of the iceberg:
OP
DE
[A or B]
[OP
DE
A] and [OP
DE
B]
Do children know this, despite overwhelming evidence in the input -- that disjunction is exclusive-
or
?
Piecemeal Learning
(cf. Pullum and Scholz 2002)
The news that Noam had not won was a surprise to some/*any of the linguists.
The news that Noam had won was not a surprise to some/any of the linguists.
fact 1 a) … not+V+V+NP+P+some b) …V+not+NP+P+some/any The bear who never laughed expected to find some/*any dogs at the party.
The bear who laughed never expected to find any dogs at the party. fact 2 a) … never+V+ V+to+V+some b) … V+never+V+to+V+some/any
But, what do fact 1 and fact 2 (and so forth) have in common? And what do these facts have in common with the interpretation of disjunction?
QuickTime™ and a DV - NTSC decompressor are needed to see this picture.
Specific Contingent Facts about Natural Languages
1. The ‘basic’ meaning of disjunction is inclusive-
or Still to come:
2. The meanings of natural language Determiners are ‘conservative’ 3. The same structural notion (‘c-command’) ties together disparate linguistic phenomena: • • • Licensing negative polarity items Generating the ‘conjunctive’ reading of disjunction Disjoint reference between pronouns and names
…
Universal Quantification
The universal quantifier,
every,
is a Determiner. In natural languages, Determiner meanings are ‘conservative,’ (e.g., Barwise and Cooper 1981; Chierchia and McConnel-Ginet 2000) However, the findings of research on child language -- going back to Inhelder and Piaget -- have led many to conclude that: •
some children adopt a non-adult and non-conservative interpretation of the universal quantifier
Answer this Question
Is every boy riding an elephant?
The Symmetrical Response
No, not this one!
Is every boy riding an elephant?
The Event Quantification Account
Every boy is riding an elephant
Truth Conditions: Every Every NP [boy] VP [boy-is-riding-an-elephant] and NP [elephant] VP [boy-is-riding an elephant] Analysis: Every e NP [ boy e or elephant e ] VP [ boy-is-riding-an-elephant e ] ‘ every event in which there is a boy or an elephant is one in which the boy is riding the elephant ’
Conservativity: A Linguistic Universal
A determiner meaning is conservative iff: (DET NP)(VP) (DET NP)(NP VP) where DET is a two-place relation, with an internal argument = NP, and an external argument = VP The intuition:
Every boy is riding an elephant
is TRUE iff
every boy
is a subset of
boys who are riding an elephant
• Its truth can be decided without necessarily looking at every elephant in the domain of discourse
Conservativity
Conservativity of determiner meanings make these inferences valid: (DET NP)(VP) (DET NP)(NP VP) Few Americans smoke Few Americans are Americans who smoke Every Italian eats pasta Every Italian is an Italian who eats pasta No German drinks Bud
…
No German is a German who drinks Bud
The Event Quantification account
Philip (1995), cf. Drozd and van Loosbroek (1998, 2000)
Every boy is riding an elephant
Every NP [boy or elephant] VP [boy-is-riding-an-elephant]
• This analysis of
every
violates conservativity, because it blurs the distinction between the internal and the external argument • On this analysis, it is necessary to check every elephant to see that the boys and the elephants are equinumerous. • If all natural language Determiners are conservative, there cannot be a Determiner that means ‘equinumerous’
So, why else might children think there is a ‘missing’ boy?
“… children try to understand what people mean, not only what words mean ” (Freeman et al, 1982; p. 69). “ When a speaker asks a question about the presence of ‘ all the Xs ’ , he is implicitly requesting the hearer to carry out an exhaustive search to check that no X is missing. Asking someone such a question is only legitimate ‘ socio-dialogically ’ if there is at least the possibility that some X is (or some Xs are) in actuality missing.
” (Freeman et al., 1982; p. 64)
Satisfy Felicity Conditions
One way is by providing a different
possible outcome
. Adults are happy if there is an additional object that participants could have chosen.
QuickTime™ and a DV - NTSC decompressor are needed to see this picture.
Specific Contingent Facts about Natural Languages
1. The ‘basic’ meaning of disjunction is inclusive-
or
2. Natural language Determiners are ‘conservative’
Still to come:
3. The same structural notion (‘c-command’) ties together disparate linguistic phenomena: • • • Licensing negative polarity items Generating the ‘conjunctive’ reading of disjunction Disjoint reference between pronouns and names
…
Structure also Matters
1. For “any” to be permitted, it must be c-commanded by a downward entailing expression 2. For “or” to receive the conjunctive interpretation, it must be c-commanded by a downward entailing expression 3. If a pronoun c-commands a name, they must have disjoint reference
The operative structural relation in all these cases is “c-command”
OP DE
C-command
- NPI licensing - Interpreting Disjunction
NPI Or -conjunctive
OP DE Pronoun i
C-command
- NPI licensing - Interpreting Disjunction - Disjoint Reference
NPI Or -conjunctive Name j ≠ i
OP DE DO NOT TRESPASS
No C-command
- No NPI licensing - Change in Interpreting Disjunction *
NPI Or – ‘exclusive’
OP DE Pronoun i DO NOT TRESPASS
No C-command
- No NPI licensing - Change in Interpreting Disjunction - Coreference is possible *
NPI Or – ‘exclusive’ Name i
Children’s knowledge of c-command
Test sentences:
1. “The girl that stayed up late did not get a dime or a jewel” 2. “The girl that didn’t go to sleep got a dime or a jewel” 1. = the conjunctive-or reading 2. = the exclusive-or reading is preferred
A Typical Story
This is a story about two girls who lost their teeth.
“Well, I know I should go to sleep, but I really want to see what the Tooth Fairy looks like, so I am gonna stay up.”
The Tooth Fairy arrives! “Ok, I have some dimes and some jewels for good children who have lost their teeth!”
Merlin: “The girl that stayed up late will not get a dime or a jewel” Merlin: “The girl that didn’t go to sleep will get a dime or a jewel”
“This girl is asleep. I’ll give her a dime. Here’s a jewel too”
“Why are you awake? You know you are supposed to be asleep!”
“I know, I know! But I wanted to see what you look like. But, I lost a big tooth. You have to give me something!”
“Ok, ok, I’ll give you a jewel”
Merlin: “I said the girl that stayed up late would not get a dime or a jewel” Merlin: “I said the girl that didn’t go to sleep would get a dime or a jewel”
30 children (age from 3;11 to 5;9 – mean 5;0) participated in the both conditions of the experiment. Each subject was presented with two trials in each condition. First condition: Merlin: “I said the girl that stayed up late would not get a dime or a jewel.” Children rejected this statement 90% of the time.
The same children also had two trials in the second condition. Merlin: “I said that the girl that didn’t go to sleep would get a dime or a jewel.” The child subjects accepted this statement 87% of the time.
Summary: Specific Contingent Facts
1. The ‘basic’ meaning of disjunction is inclusive-
or
2. Natural language Determiners are ‘conservative’ 3. The same structural notion of ‘c-command’ ties together several apparently disparate linguistic phenomena:
•
Licensing negative polarity items
• •
Generating the ‘conjunctive’ reading of disjunction Disjoint reference between pronouns and names …
QuickTime™ and a Sorenson Video decompressor are needed to see this picture.