PII: S0010-0277(02)00013-6

Children’s command of quantiﬁcation

Jeffrey Lidz

a,1

, Julien Musolino

b,1,

Department of Linguistics, Northwestern University, 2016 Sheridan Rd., Evanston, IL 60208-4090, USA

Department of Speech and Hearing Sciences, Indiana University, 200 S. Jordan Avenue,

Bloomington, IN 47405-7002, USA

Received 16 December 2001; received in revised form 2 January 2002; accepted 8 February 2002

Abstract

In this article we present data from two sets of experiments designed to investigate how children

and adult speakers of English and Kannada (Dravidian) interpret scopally ambiguous sentences

containing numerally quantiﬁed noun phrases and negation (e.g. Donald didn’t ﬁnd two guys).

We use this kind of sentence as a way to ﬁnd evidence in children’s linguistic representations for

the hierarchical structure and the abstract relations deﬁned over these structures (in particular, the

relation of c-command) that linguists take to be at the core of grammatical knowledge. Speciﬁcally,

we uncover the existence of systematic differences in the way that children and adult speakers

resolve these ambiguities, independent of the language they speak. That is, while adults can easily

access either scope interpretation, 4-year-old children display a strong preference for the scopal

interpretation of the quantiﬁed elements which corresponds to their surface syntactic position.

Crucially, however, we show that children’s interpretations are constrained by the surface hierarch-

ical relations (i.e. the c-command relations) between these elements and not by their linear order.

Children’s non-adult interpretations are therefore informative about the nature of the syntactic

representations they entertain and the rules they use to determine the meaning of a sentence from

Keywords: Language acquisition; Cross-linguistic; Ambiguity resolution; Quantiﬁer scope; Negation; C-

command; Kannada

1. Introduction

In this paper we investigate children’s knowledge of the linguistic principles governing

the interpretation of the quantiﬁcational expressions of their language (e.g. no man, two

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 113

Cognition 84 (2002) 113–154

www.elsevier.com/locate/cognit

PII: S0010-0277(02)00013-6

COGNITION

* Corresponding author. Fax: 11-812-855-5531.

E-mail address: [email protected] (J. Musolino).

The authors are engaged in a continuing collaboration in which the order of names alternates from one paper

to the next. The authors contributed equally to the work reported here.

women). The study of natural language quantiﬁers belongs to a long tradition of philoso-

phical inquiry concerned with the logical analysis of language that can be traced back to

antiquity. A well-known problem raised by quantiﬁcational expressions is the inherently

non-referential nature of their meaning; a fact that Lewis Carroll understood all too well:

“Who did you pass on the road?” the King went on, holding out his hand to the

Messenger for some more hay.

“Nobody” said the Messenger.

“Quite right”, said the King: “this young lady saw him too. So of course Nobody walks

slower than you”.

“I do my best”, the Messenger said in a sullen tone. “I’m sure that nobody walks much

faster than I do!”

“He can’t do that”, said the King, “or else he’d have been here ﬁrst…”.

– Lewis Carroll, Through the Looking-Glass and What Alice Found There

The King’s error in the passage above was to treat nobody as though it referred to an

individual whereas in fact, quantiﬁed expressions express generalizations about sets of

individuals. So, the sentence Nobody passed the Messenger expresses the idea that the

intersection of the set of individuals and the set of entities having the property of passing

the Messenger is the null set. Another important property of quantiﬁers, guiding the King’s

reasoning in Lewis Carroll’s example, is the rich set of inferences to which they give rise.

In addition to vexing philosophers, logicians and kings, quantiﬁed expressions also raise

serious psychological issues, in particular in the domain of conceptual development and

language acquisition. Consider the problem of a learner trying to determine the meanings

of words. The learner needs to avoid the problem faced by the King in our example above.

That is, learners must avoid treating quantiﬁers as referring to individuals. But this should

only complicate their task: after all, once learners have realized that quantiﬁers are non-

referential, it is unclear what kinds of perceptual regularities they could possibly associate

with such expressions. To complicate matters further, the conceptual generalizations

expressed by quantiﬁed expressions are abstract, and so the learner’s ability to map

these concepts onto phonological form depends on cognitive development having reached

the point at which such concepts can be mastered and put to use.

In the psychological literature, children’s comprehension of quantiﬁers has traditionally

been examined as a means of assessing their logical competence. To be sure, understand-

ing the development of reasoning necessitates a proper appreciation of the development of

the logical vocabulary available in natural language and the kinds of representations that it

is associated with. In their classic studies on the development of logical competence,

Inhelder and Piaget (1958, 1964) examined children’s understanding of the universal

quantiﬁer in a number of class inclusion tasks. What they found was that children do

not understand sentences containing the universal quantiﬁer until the age of 7 or 8 which,

in the Piagetian framework, corresponds to the onset of concrete operational reasoning.

Inhelder and Piaget attributed children’s misunderstanding of the universal quantiﬁer to a

lack of ability to divide groups of objects into separate classes.

In recent years, children’s comprehension of sentences containing quantiﬁed expres-

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154114

sions has received a good deal of attention from investigators in the ﬁeld of linguistics (see

Crain, 2000 for a review). This renewed interest coincides with recent theoretical devel-

opments in the study of formal properties of quantiﬁers (Barwise & Cooper, 1981; Keenan

& Stavi, 1986; van Benthem, 1986) and was motivated in part by a desire to begin to map

out the development of children’s knowledge of the principles underlying natural language

semantics. The observation that emerges from these studies is that preschool children often

fail to interpret statements or questions containing quantiﬁed expressions the way adult

speakers do. For example, Philip and Aurelio (1991) found that when presented with a

picture showing three mice each in a cup and one extra cup without a mouse and asked of

that picture “Is every mouse in a cup?” 3–5-year-old preschool children often answered

‘no’ and pointed to the empty cup. Similar ﬁndings, reminiscent of Piaget’s original

observations, have been reported by a number of investigators (Bucci, 1978; Donaldson

& Lloyd, 1974; Inhelder & Piaget, 1964) and have appeared under various labels such as

“quantiﬁer spreading”, “symmetrical reading” or “conversion” (see Philip, 1996). More-

over, this phenomenon has been studied in a variety of languages including English,

French, Chinese, Japanese, Dutch and Catalan (for a review, see Philip, 1995).

While Piaget and his collaborators viewed children’s non-adult interpretations of quan-

tiﬁed expressions as reﬂecting a problem at the conceptual level, the studies cited above

typically regard children’s difﬁculties with quantiﬁers as stemming from their immature

linguistic representations. For example, Bucci (1978) proposed that statements of the form

‘All As are Bs’ are often parsed by children and by adults as simple strings of unordered

words lacking any syntactic structure. Philip (1995) offered a semantic explanation,

arguing that the “symmetrical interpretation” observed in preschool children corresponds

to a linguistic representation in which the child over-generalizes a tendency to quantify

over individual events rather than individual objects. This latter position is rejected by

Crain et al. (1996) who argued that children have full adult linguistic competence in the

domain of universal quantiﬁcation. According to these authors, earlier observations of

children’s non-adult behavior in tasks involving the comprehension of universally quan-

tiﬁed propositions are pragmatic in nature; they are due to violations of the felicity

conditions associated with the use of such expressions.

As noted by Brooks and Braine (1996), the vast majority of studies on children’s

comprehension of quantiﬁed expressions have focused almost exclusively on the distri-

butional reading of universal quantiﬁers. However, the set of natural language quantiﬁers

is not limited to elements such as all and every. In order to gain a broader understanding

of children’s knowledge of the complex array of linguistic principles underlying the

interpretation of quantiﬁcational elements it is therefore necessary to look beyond chil-

dren’s comprehension of the universal quantiﬁer. This is what we propose to do in the

present article. We present here data from two sets of experiments designed to investi-

gate how children and adult speakers of English and Kannada (Dravidian) interpret

ambiguous sentences containing numerally quantiﬁed noun phrases (e.g. two horses)

and negation. In addition to producing a more detailed picture of development, our

goal is twofold: for adults, our aim is to determine whether the intuitions of linguists

about the possible interpretations of these sentences can be veriﬁed experimentally. To

the extent that this is the case, we then ask when children become aware of the complex

mapping between syntax and semantics required for the interpretation of these linguistic

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 115

expressions. Anticipating the results, we uncover in this domain as well the existence of

systematic differences in the way that preschool children and adults interpret sentences

containing these expressions. We suggest that children’s non-adult interpretations have

psychological implications for the abstract format of the linguistic representations they

entertain. We show that these non-adult interpretations are informative about children’s

syntactic representations and the rules they use to determine the meaning of a sentence

from its structure.

To preface the experimental sections of our study, we lay out some important facts

concerning fundamental properties of language and we introduce the technical vocabulary

later used to describe the linguistic phenomena we shall be concerned with. We then

review a number of pertinent ﬁndings on children’s and adults’ comprehension of quanti-

ﬁcational expressions. Finally, we present a set of experiments in which children and adult

speakers of English and Kannada were asked to interpret sentences containing multiple

quantiﬁcational elements.

2. Linguistic background

There are two facts about language which seem to us to be uncontroversial. First,

sentences have a hierarchical structure. The words in a sentence are not strung together

in a list, but rather are composed into nested constituents above the word level. Second,

languages exhibit a phenomenon which, following Chomsky (1995), we refer to as displa-

cement: certain elements in a sentence are not interpreted entirely in the positions in which

they appear.

2.1. Hierarchy

Two of the central conclusions of the foundational work by Chomsky (1957) in syntax

are that linguistic representations are hierarchical and that the rules of syntax make

reference to this hierarchy. A sentence like (1) is not just a string of words, but rather

can be represented as a nested tree structure in which words combine to form phrases and

these phrases combine to form larger phrases, as in (2).

(1) The clown juggled three balls

(2)

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154116

Not only are linguistic expressions structured, but the rules and constraints that apply to

these expressions also make reference to this structure. Consider the following contrasts:

(3) a. Hillary

thinks she

won the election

b. * She

thinks that Hillary

won the election

c. After Bill embarrassed her

, Hillary

won the election

The contrast between (3a) and (3b) illustrates that there are constraints determining when a

pronoun, like she, and a referential expression, like Hillary, can refer to the same entity.

This constraint is not based on precedence, as can be seen from examining (3c) in which

the pronoun precedes the referential expression. The appropriate generalization becomes

clear when we examine the structure underlying (3a) and (3b).

(4)

What rules out (3b) is a constraint barring a pronoun from being interpreted as identical in

reference to an expression that is contained in the smallest nontrivial constituent contain-

ing the pronoun itself. Here, the smallest nontrivial constituent containing the matrix

subject is the entire sentence and so a pronoun in that position cannot be interpreted as

coreferential with anything else in the sentence. In the structure underlying (3a), the

smallest nontrivial constituent containing the pronoun is the embedded S. Since Hillary

is not contained within this constituent, coreference with the pronoun is allowed.

This generalization is supported by the observation that if we put the pronoun inside a

branching constituent in subject position, coreference is allowed. Here, the smallest

constituent containing the pronoun is the subject NP. Since there is no coreferential

expression in this constituent, the pronoun is licit.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 117

Indices are used to indicate that two NPs are coreferential, i.e. that they refer to the same individual. Within

this notation, the asterisk on (3b) indicates that an interpretation of that sentence on which the pronoun she and the

NP Hillary refer to the same individual is not available.

(5)

The relation expressed in the previous discussion has been formalized under the notions of

c-command and binding (Chomsky, 1981; Reinhart, 1976).

(6) x

c-commands y iff

a. the ﬁrst branching node dominating x also dominates y

b. x does not dominate y

c. x – y

(7) x

binds y iff

a. x c-commands y

b. x and y are coreferential

The constraint on coreference can now be stated as (8) (Chomsky, 1981; cf. Lasnik, 1976):

(8) A referring expression cannot be bound

Because the referring expression Hillary is c-commanded by and coreferential with the

pronoun in (3b), coreference is blocked by the constraint (8). The crucial observation for

our purposes is that the explanatory predicate “bind” is based on a relation deﬁned over

hierarchical representations, namely c-command. It is only in deﬁning relations over

hierarchical representations that we ﬁnd an explanation for the observed phenomenon,

and so we have evidence that these representations exist.

2.2. Displacement

As rock-bottom a phenomenon in language as hierarchy is “displacement”: certain

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154118

expressions are not interpreted in the positions in which they appear. Consider the obser-

vation that the direct object must occur adjacent to the verb in English:

(9) a. Rudy ate the pizza

b. * Rudy ate quickly the pizza

When the direct object is questioned, as in (10), this adjacency requirement is relaxed:

(10) What did Rudy eat __?

In its role as a sentential operator, the wh-phrase occurs at the beginning of the sentence.

This position is in accord with its interrogative interpretation:

(11) [what x] [Rudy ate x]

Nonetheless, the wh-phrase is interpreted as the object of the verb and, hence, as though it

is in object position.

Intuitively, displacement phenomena illustrate that different kinds of

semantic information must be represented in a single structure. When the requirements of

different kinds of information are in conﬂict, only one can be satisﬁed.

Displacement phenomena come in two varieties. As we have seen, an element that does

not occur in an argument position can still be interpreted as an argument. In such cases, it

appears that the element has “moved” out of its argument position. Conversely, elements

occurring in an argument position are sometimes interpreted as sentential operators, as in

(12):

(12) a. Who ate what

b. wh kx,yl [x ate y]

Here, the wh-phrase what is interpreted as part of a sentential operator asking for pairs,

despite the fact that it has not moved out of its argument position. That is, the appropriate

response to (12a) is a set of pairs such that the ﬁrst element ate the second element (Rudy

ate pizza, Bill ate a Big Mac, Hillary ate a sandwich, etc.). What expressions involving

multiple wh-phrases demonstrate is that displacement does not always involve the appear-

ance of “movement” out of an argument position.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 119

This characterization of the phenomenon is independent of the mechanisms that capture it. That is, every

theory of interrogation in English must account for the fact that wh-expressions are interpreted simultaneously as

sentence-level operators and arguments of the predicate, whether this is done by traces (Fiengo, 1977), copies

(Chomsky, 1993), slash-categories (Pollard & Sag, 1994) or other such mechanisms.

Notice that the postulation of traces as real entities represents the simplest of theories in this domain since it

allows the displaced element to actually occur in both positions, eliminating a need for statements concerning the

resolution of the conﬂict between argument structure and quantiﬁcational structure.

2.3. Hierarchy, displacement and scope ambiguity

This kind of “covert” displacement is also exhibited by ambiguities involving the

quantiﬁcational expressions discussed above, as illustrated in (13a). In an expression

like (13a), there are two possible interpretations, given as (13b) and (13c):

(13) a. Everyone didn’t smile

b. ;x[: [x smiled]]

c. : [;x [x smiled]]

The interpretation represented by (13b) states that everyone has the property that he didn’t

smile; in other words, no-one smiled. On this interpretation we say that the universal

quantiﬁer “;”

takes scope over negation. The interpretation in (13c) states that it is not

the case that everyone smiled; perhaps some people did and some people didn’t. Here,

negation takes scope over the universal quantiﬁer. The interpretation in (13b) is straight-

forwardly mapped from the surface syntax of (13a), given in (14):

(14)

In (14) the NP everyone c-commands the negative clitic n’t. Similarly, the semantic

counterpart of this expression takes scope over negation in (13b). However, in (13c),

negation is interpreted as taking scope over the entire sentence, illustrating again that

certain elements are interpreted in positions different from those they occur in in the

syntax.

It is important to observe at this point that the representations of both the syntax and the

semantics are hierarchical. That is, the representations in (13b) and (13c) are equivalent to

the nested tree-structure representations in (15a) and (15b):

(15)

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154120

Thus, the scope of the quantiﬁcational elements “ : ” and “;” is determined by the c-

command relations in the semantic representation. That is, given two quantiﬁcational

elements a and b, a takes scope over b if a c-commands b in the semantic representa-

tion.

Now, the c-command relations holding between two quantiﬁcational elements in the

surface syntax do not map directly to the c-command relations between these elements in

the semantics. This is the covert displacement phenomenon in evidence. As we have seen,

the quantiﬁcational expression everyone c-commands negation in the surface syntax, but

either c-command relation is possible in the semantics.

We can conclude, then, that the surface syntactic structure and the semantic structure

are not isomorphic. Data concerning displacement and quantiﬁcational ambiguity demon-

strate that there is a complex mapping from one kind of structure (surface syntax) to

another kind (semantics). It is standard practice in linguistics and philosophy to assume

that the meaning of a sentence is a function of the meaning of its parts and the way that

they compose (Frege, 1892/1990). So, in order to account for mismatches between the

surface syntactic structure of a sentence and its meaning, we can either complicate the

rules of syntactic composition or the rules of semantic composition.

Taking the ﬁrst option, we could treat the quantiﬁcational expressions on a par with

multiple questions, positing that there are syntactic displacement operations that reposi-

tion the quantiﬁcational elements in the syntax without phonological consequence

(Huang, 1982; May, 1977; McCawley, 1970). This approach simpliﬁes the mapping to

the semantics by complicating the syntax. On this view, the syntactic structure directly

determines the relevant aspects of interpretation; c-command in the syntax determines

scope in the semantics. Taking the second option, we could take the surface syntactic

representation to be the input to semantics and develop complex mapping operations to

determine the scope of a quantiﬁer independent of its syntactic position (Cooper, 1983;

cf. Steedman, 1987). On this view, the syntax and semantics are homomorphic and so the

grammar must specify where the two representations diverge. The choice between these

approaches may have important theoretical consequences; however, for our purposes it is

sufﬁcient to observe that the surface syntax and the semantic structure are not

isomorphic, whether this is achieved by a syntactic or a semantic operation inverting

the scope of the two operators.

3. Psycholinguistic background

Following our description of the formal aspect of quantiﬁcational interactions, we now

consider children’s knowledge of quantiﬁcation. As we said earlier, we will focus here

on scope phenomena involving the interaction between numerally quantiﬁed NPs (e.g.

two birds) and negation. Our motivation for doing so stems from the fact that the vast

majority of existing studies have focused on children’s knowledge of universal quanti-

ﬁers, i.e. all and every, in constructions typically involving a universally and an existen-

tially quantiﬁed NP, e.g.

every boy saw a squirrel. Another reason to consider

interactions between quantiﬁed NPs and negation is that there are reasons to believe

that scope ambiguities involving these elements employ different formal mechanisms

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 121

than those involving two quantiﬁcational NPs (Hornstein, 1984; Reinhart, 1995, 1997).

To the extent that this is the case, studying children’s comprehension of sentences

involving quantiﬁed NPs and negation will produce a map of the development of seman-

tic knowledge which is both broader and ﬁner than that produced by previous work.

Finally, we focus our attention on the interaction between numerally quantiﬁed NPs and

negation in order to ensure that the two scopal orderings can be distinguished. The

interpretation of a universal quantiﬁer outside the scope of negation (13b) entails the

interpretation in which negation takes wider scope (13c). That is, every situation that

makes (13b) true also makes (13c) true. Hence, we can only test one of these readings

directly. With numerally quantiﬁed NPs, however, this entailment problem does not

arise, as we will see below.

In spite of the growing number of studies on children’s comprehension of quantiﬁca-

tional interactions, relatively few have examined interactions involving quantiﬁers and

negation (but see de Boysson-Bardies & Bacri, 1977; Drozd & Philip, 1993; Kra

mer,

2000; O’Leary & Crain, 1994; Roeper & Matthei, 1975; Thornton, 1995). To our knowl-

edge, only one study, by Musolino, Crain, and Thornton (2000), presents a systematic

investigation of children’s comprehension of quantiﬁer–negation interactions. This study

therefore constitutes a natural starting point for our investigation.

Musolino et al. (2000) begin with the observation that sentences containing quantiﬁers

and negation often permit readings which do not directly follow from an isomorphic

mapping of surface form to semantic interpretation (see discussion in Section 2 and also

ring, 1997; Horn, 1989; Jackendoff, 1972 among others). Consider the examples below.

(16) Every horse didn’t jump over the fence

a. ;x [horse (x) !:jump over the fence (x)]

b. : ;x [horse (x) ! jump over the fence (x)]

(17) The detective didn’t ﬁnd some guys

’x [guy(x) ^ : ﬁnd(detective, x)]

As we pointed out earlier, sentences like (16) are scopally ambiguous. On one interpreta-

tion, (16) can be paraphrased as meaning that every horse is such that it didn’t jump over

the fence. In other words, none of the horses jumped over the fence. Here, the universally

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154122

For example, when a universal quantiﬁer occurs in object position, it may not take scope over negation:

(i) John didn’t borrow every book from the library.

(ii) John didn’t borrow two books from the library.

The sentence in (i) does not have an inverse scope interpretation, i.e. the interpretation that John borrowed no

books from the library. However, the sentence in (ii) clearly does allow the inverse scope interpretation in which

there are two books that John didn’t borrow. This differs from the situation in which there are two quantiﬁcational

NPs in a clause. Here, inverse scope is always allowed:

(iii) Some student borrowed every book from my ofﬁce.

(iv) Some student borrowed two books from my ofﬁce.

quantiﬁed subject takes scope over negation (abbreviated every . not), as illustrated by

the logical representation given in (16a). We call this an

isomorphic interpretation because

in this case the scope relation between the universally quantiﬁed NP and negation coin-

cides with their surface position. Another possible interpretation of (16) is that not all of

the horses jumped over the fence. In this case, negation takes scope over the quantiﬁed

subject (abbreviated not . every), as shown in the logical representation given in (16b).

We call this a

non-isomorphic interpretation because here negation takes scope over the

whole sentence, illustrating again that certain elements are interpreted in positions differ-

ent from the ones they occupy in surface syntax. Whereas (16) is scopally ambiguous, (17)

seems to require a non-isomorphic interpretation expressing the idea that there are some

guys that the detective didn’t ﬁnd.

In order to determine whether preschool children are aware of the complex mappings

between form and meaning involved in sentences like (16) and (17), Musolino et al. (2000)

tested children’s comprehension of such sentences using the Truth Value Judgment Task

(TVJT) methodology (Crain & Thornton, 1998). The main ﬁnding from Musolino et al.’s

study is that while adults can easily access the non-isomorphic interpretations of sentences

like (16) and (17), 4-year-olds systematically assign such sentences an isomorphic inter-

pretation only.

For example, one of the stories used to test children’s comprehension of (16) involved

three horses trying to jump over a fence. Two of the horses successfully jump over the

fence but the third one doesn’t. At the end of the story, a puppet describes the situation as

follows: “Every horse didn’t jump over the fence”. Note that this statement is true on the

non-isomorphic (i.e. ‘not all’) reading of the sentence since it is true that not all of the

horses jumped over the fence. However, the puppet’s statement is false on the ‘none’,

isomorphic, reading, since two horses did jump over the fence. Musolino et al. found that

in this case, adults always accepted the puppet’s statement and therefore showed that they

could easily access the non-isomorphic interpretation. Children, on the other hand,

systematically rejected the puppet’s statements. When asked to justify their answers,

children typically said that the puppet was wrong because two of the horses did jump

over the fence. Children, therefore, seem to only be able to access the isomorphic reading

of the sentence.

To determine whether children’s difﬁculty is linguistic or conceptual in nature, Muso-

lino et al. used sentences like (18) as a control.

(18) The Smurf didn’t buy every orange

The example in (18) was used in the context of a story in which a Smurf had bought only

one of three oranges, and therefore not all of them. Musolino et al. found that in this case,

children had no difﬁculty accepting the puppet’s statements that “the Smurf didn’t buy

every orange”. Notice that in this case the ‘not all’ reading corresponds to an isomorphic

interpretation, unlike the example in (16). The ﬁnding that children systematically assign

examples like (16–18) isomorphic interpretations is what Musolino et al. call the ‘Obser-

vation of Isomorphism’.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 123

The Observation of Isomorphism: (Musolino et al., 2000, p. 14)

Unlike adults, young children systematically interpret negation and quantiﬁed NPs on

the basis of their position in overt syntax.

Musolino et al.’s ﬁndings raise a number of interesting questions.

First, these results do not tell us the nature of the constraint underlying children’s resis-

tance to non-isomorphic interpretations. One possibility is that children’s overly isomorphic

interpretations reﬂect the linear arrangement between the quantiﬁed NPs and negation. This

wouldn’t be surprising since it has been suggested in the psycholinguistic literature that for

adults, linear order is one of the factors inﬂuencing the resolution of quantiﬁer scope

ambiguities on-line (Kurtzman & MacDonald, 1993). Alternatively, children’s interpreta-

tions may be constrained by the surface c-command relations holding between these

elements. This would also make sense since c-command is the linguistic determinant of

scope. It is unclear whether isomorphism obtains as a consequence of the linear position of

the quantiﬁed NPs and negation or whether it follows from the c-command relations

between these elements because these two notions are systematically confounded in the

material used by Musolino et al. As can be seen in the tree diagrams below, the subject

position always precedes and falls outside the c-command domain of negation while the

object position always follows and falls within the c-command domain of negation.

Teasing apart linear order from hierarchical structure is important because the answer to this

question can give us insight into the nature of children’s linguistic representations. In

particular, we can ﬁnd out whether the theoretical constructs of syntactic theory play an

explanatory role not only in models of adult grammatical competence but also in our under-

standing of children’s non-adult behavior. Simply put, do children’s linguistic representa-

tions have the same character as those of the adults they eventually become? We will call

this the structural question.

Another question raised by Musolino et al.’s ﬁndings is whether the observation of

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154124

isomorphism in children is bound to the lexical properties of the quantiﬁed NPs tested by

these authors, i.e. every N and some N. This is what we call the lexical question. This

question grows out of the observation that the availability of non-isomorphic interpreta-

tions sometimes depends in the adult grammar on the lexical nature of the quantiﬁed NPs

involved. For example, the sentence The detective didn’t ﬁnd some guys, as we discussed

earlier, seems to require a non-isomorphic interpretation (i.e. there are some guys that the

detective didn’t ﬁnd). However, replace the NP some guys by any guys and the sentence in

(19) no longer allows a non-isomorphic interpretation. That is, (19) must mean that the

detective found no-one, as shown by the logical representation in (19b). Now replace any

guys by one guy, as in (20), and the sentence becomes ambiguous again between an

isomorphic and a non-isomorphic interpretation, as shown in (20a) and (20b), respectively.

(19) The detective didn’t ﬁnd any guys

a. * ’x [guy(x) ^ : ﬁnd(detective, x)]

b. : ’x [guy(x) ^ ﬁnd(detective, x)]

(20) The detective didn’t ﬁnd one guy

a. : ’x [guy(x) ^ ﬁnd(detective, x)]

b. ’x [guy(x) ^ : ﬁnd(detective, x)]

A third important question, related to the scope of the phenomenon observed by Muso-

lino et al., is whether isomorphism can be observed in the acquisition of languages other

than English, provided of course that they manifest the same kind of scope ambiguities

with respect to QNPs and negation. We call this the cross-linguistic question.

Finally, an important question concerns the empirical basis for the observation of

isomorphism. Musolino et al. showed that children rejected statements like (16) and

(17) when they were true on a non-isomorphic interpretation. The children they tested

justiﬁed their negative answers by invoking the fact that the statements were false on the

isomorphic interpretation. In the case of (16) for example, in a situation in which two out

of three horses had jumped over a fence, children who answered ‘no’ to the puppet’s

statement that Every horse didn’t jump over the fence said that the puppet was wrong

because two of the horses had jumped over the fence. Musolino et al. therefore inferred

from children’s justiﬁcations that they were accessing isomorphic interpretations. That is,

they reasoned that children’s answers only made sense if they had taken (16) to mean that

none of the horses had jumped over the fence.

However, Musolino et al. did not directly show that children (or adults) would accept

the same statements if the isomorphic reading were true. This step would have provided a

clear demonstration that sentences like (16) are indeed ambiguous for adults (who

accepted them on the non-isomorphic interpretation). In addition, it would have provided

a direct empirical validation of the inference made by these authors based on children’s

rejections of non-isomorphic interpretations. That is, since children rejected the non-

isomorphic reading of sentences like (16) by invoking the fact that the isomorphic reading

was false, one would predict that they should accept the same sentences when the

isomorphic reading is true and the non-isomorphic reading is false.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 125

However, as mentioned above, every situation which satisﬁes the truth conditions of the

isomorphic (i.e. ‘none’) reading also satisﬁes the truth conditions of the non-isomorphic

(i.e. ‘not all’) reading. Coming back to our example, we can see that if it is true that none of

the horses jumped over the fence (the isomorphic interpretation), it is also necessarily true

that not all of the horses jumped over the fence (the non-isomorphic interpretation). Notice

that a similar entailment pattern obtains for sentences like (17). In this case, if it is true that

the detective didn’t ﬁnd anyone (the isomorphic interpretation) it is also true that there are

some individuals that the detective didn’t ﬁnd (the non-isomorphic reading). These entail-

ment patterns are shown in (21).

(21) a. ;x[ : P(x)] !:[;x [P(x)]] none ! not all

b. : [;x[P(x)]] K ;x[ : P(x)] not all K none

The unfortunate consequence of these entailment patterns for experimental design is that it

is impossible to directly test children’s (and adults) knowledge of the isomorphic readings

of sentences like (16) and (17) because scenarios in which the isomorphic reading is true

and the non-isomorphic reading is false are impossible to construct (but see Musolino &

Lidz, 2002). We refer to this as the entailment problem.

In Section 4 and 5, we present a series of experiments on English and Kannada,

designed to address these questions. The ﬁrst experiment addresses the lexical question

and the entailment problem. The second experiment addresses these questions as well as

the structural and cross-linguistic questions.

4. Experiments on English

We tested children’s and adults’ interpretations of ambiguous sentences like (22)

containing negation and a numerally quantiﬁed object.

(22) The detective didn’t ﬁnd two guys

On one reading, given in (23a), (22) can be paraphrased as ‘It is not the case that the

detective found two guys’. In this case, the QNP two guys is interpreted in the scope of

negation, giving rise to an Isomorphic interpretation. The other reading of (22), given in

(23b), can be paraphrased as ‘There are two guys that the detective didn’t ﬁnd’. Here, the

QNP two guys takes scope over negation, giving rise to a Non-Isomorphic interpretation.

(23) a. : ’

x [guy(x) ^ ﬁnd(detective, x)]

b. ’

x [guy(x) ^ : ﬁnd(detective, x)]

An important feature of sentences like (22) compared to the sentences used by Musolino

et al. is that while they are scopally ambiguous, there is no entailment relation between the

two scope readings. The entailment problem is solved for this kind of sentence. This

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154126

means that it is possible to conceive of a situation in which either scope reading of (22) can

be true while the other reading is false. Take for example a situation in which the narrow

scope reading of (22) is true. Imagine for instance that the detective only found one out of

two guys that he was looking for. In this case, the narrow scope reading of the numeral in

(22) is true since it is not the case that the detective found two guys (he only found one of

them). However, the wide scope reading of the quantiﬁcational NP is false because in this

situation it is not true that there are two guys that the detective didn’t ﬁnd. In fact there is

only one such person. Let us now consider a situation in which the wide scope reading of

the quantiﬁcational NP in (22) is true. Suppose that the detective only found two out of

four guys that he was looking for and therefore that there are two guys that he didn’t ﬁnd.

In this case, the wide scope reading of the quantiﬁcational NP in (22) is true. This situation

also falsiﬁes the narrow scope reading of the quantiﬁcational NP in (22). According to this

reading, it should not be the case that the detective found two guys. Yet in this situation,

the detective found exactly two guys, falsifying the narrow scope reading of the quanti-

ﬁcational NP. Using sentences like (22), we therefore circumvent the entailment problem

discussed earlier and use materials that allow us to directly test the subject’s interpretation

of both isomorphic and non-isomorphic scope readings.

One part of our research question is to determine experimentally whether our intuitions

about the ambiguity of sentences like (22) can be conﬁrmed by the judgments of adult

speakers of English. Recall that the Musolino et al. experiments did not explicitly test

adults’ (or children’s) interpretation of both of the scope readings of ambiguous sentences

containing quantiﬁed NPs and negation. A second part of our research question is to reﬁne

our understanding of children’s non-adult interpretations of quantiﬁcational expressions.

Testing young children’s interpretation of sentences like (22) provides a clear proving

ground for the observation of isomorphism by addressing the lexical question. This

experiment will determine whether or not the observation of isomorphism is restricted

to the kinds of quantiﬁed NPs tested by Musolino et al. (2000).

4.1. Method

4.1.1. Subjects

We tested 24 English-speaking children (12 boys and 12 girls) between the ages of 3;11

and 4;11 (mean 4;4). Four subjects were replaced in the design because they were too shy

or because they could not provide any justiﬁcations to their answers. The children were

recruited from daycare centers at the University of Maryland and in the Philadelphia area.

We chose 4-year-olds because previous studies (i.e. Musolino et al., 2000) showed that it is

at that age that children give non-adult responses to ambiguous sentences containing

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 127

Notice that solving the entailment problem is vital to making the Truth Value Judgment Task methodology

work properly. The reasoning behind this method is that subjects will always assent when the experimenter says at

least one thing that is true (Crain & Thornton, 1998). In other words, the method relies on listeners giving

speakers the beneﬁt of the doubt. Hence, if anything that the speaker says is true, then subjects respond by saying

that the speaker did in fact speak truthfully. Thus, when we present sentences that are true on one reading but false

on another and the subjects reject the statement as false, we conclude that the subjects did not access the true

reading. Experimental situations in which both readings are true do not allow us to see whether speakers have

access to both readings since an assent to the truth of the statement could be due to either reading being true.

quantiﬁed NPs and negation. It is also important to note that choosing the TVJT as our

experimental methodology imposed limits on the age of our child subjects since the task

usually proves to be too complex for children under the age of 4. In addition to children,

we also tested 24 adult native speakers of English, all undergraduate students at the

University of Pennsylvania.

4.1.2. Procedure

We tested children’s and adults’ interpretations of sentences like (22) using the TVJT

methodology (Crain & Thornton, 1998; Musolino et al., 2000). The TVJT involves two

experimenters. The ﬁrst experimenter acts out short stories in front of the subjects using

small toys and props. The second experimenter plays the role of a puppet (in this case

Elmo) who watches the stories alongside the subjects. At the end of the story, the puppet

makes a statement about what he thinks happened in the story. The subjects’ role is to

determine whether the puppet’s statement is ‘right’, in which case the puppet gets a reward

(usually a piece of candy), or whether the puppet was ‘wrong’ in which case he gets

punished and has to do some exercise (usually some push-ups). Finally, the subjects are

asked to justify their answers by explaining why they think that the puppet was right or

wrong. For a more detailed description of the TVJT, see Crain and Thornton (1998).

The children were ﬁrst introduced to the task as a group (during circle time) and then

tested individually in a quiet room away from the class. Each child received two pretest

stories (one in which the puppet’s statement was true and one in which it was false) and if

the child could answer those appropriately (including appropriate justiﬁcations for their

answers), they would then hear seven more stories: four test stories involving sentences of

type (22) and three control stories, administered in a pseudo-random order.

Adult subjects were shown a videotaped version of the stories witnessed by the children,

including the pretest stories. They were given a score sheet and were instructed to indicate,

for each story, whether the puppet was right or wrong. They were also asked to provide a

brief justiﬁcation for their answers. Adult subjects were tested either individually or in

groups of up to three.

4.1.3. Materials

We placed subjects in an experimental situation in which both scope readings of

sentences like (22) are relevant in the context of the stories. The stories were constructed

in such a way as to make one of the readings true and the other false. Answers of YES or

NO to the puppet’s statements

(along with appropriate justiﬁcations) were therefore taken

as a measure of subjects’ ability to access one reading or the other.

Two versions of each story were constructed. In the ﬁrst one, the wide scope reading of

the numeral in sentences like (22) was true (abbreviated, Wt) and the narrow scope reading

of the numeral was false (abbreviated, Nf). In the second version, the wide scope reading

of the numeral was false (abbreviated, Wf) and the narrow scope reading of the numeral

was true (abbreviated, Nt). Recall that what we call here the narrow scope reading of the

numeral (abbreviated not . two) corresponds to an isomorphic interpretation, since nega-

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154128

Answers in which the subject said that the puppet spoke truthfully are coded as YES and answers in which the

subject said that the puppet made a mistake are coded as NO.

tion both precedes and c-commands the numerically quantiﬁed NP, whereas the wide

scope reading of the numeral (abbreviated two . not) corresponds to a non-isomorphic

interpretation. Thus, if subjects accept the Wf/Nt version of the puppet’s statement, that

would indicate that they are able to access the isomorphic interpretation; if they accept the

Wt/Nf version of the puppet’s statement, that would indicate that they are capable of

accessing the non-isomorphic interpretation.

In the Wt/Nf version of the story (corresponding to the example in (22)), Donald and

four of his friends are playing hide-and-seek. Donald ﬁnds two of his friends but fails to

ﬁnd the other two. At the end of the story, the puppet says: “I know what happened:

Donald didn’t ﬁnd two guys, am I right?” In this case, the wide scope reading of the

numeral in (22) is true because there are indeed two guys that Donald didn’t ﬁnd. On the

other hand, the narrow scope (isomorphic) reading of the numeral, asserting that it is not

the case that Donald found two guys, is false since Donald found exactly two guys. In this

case therefore, a YES response (along with appropriate justiﬁcation) indicates a preference

for the wide scope interpretation while a NO response (along with appropriate justiﬁca-

tion) indicates a preference for the narrow scope reading. Fig. 1, which corresponds to the

scene at the end of one of the stories we used, illustrates this condition.

In the Wf/Nt version, Donald is playing hide and seek with two of his friends. He ﬁnds

one of them but fails to ﬁnd the other. At the end of the story, therefore, the wide scope

reading of the numeral is false because it is not true that there are two guys that Donald

didn’t ﬁnd. In fact there is only one such person. On the other hand, the narrow scope

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 129

Fig. 1. ‘Donald didn’t ﬁnd two guys’, WtNf Condition.

reading of the numeral is true because it is indeed not the case that Donald found two guys.

He only found one. Fig. 2 illustrates this condition.

The statements made by the puppet on each of the four test trials are given in Table 1

(see Appendix A for the plot). When making these statements, the experimenter holding

the puppet was instructed to say the sentences in a way that is the most naturally compa-

tible with the appropriate reading on which the sentence was a true description of what had

happened in the story. This step was taken to ensure that if there are any prosodic cues

associated with the different scope readings, they would be provided to the child subjects

by the experimenter holding the puppet.

In addition to the four test stories, each subject also witnessed three control stories.

Unlike the test items, the statements made by the puppet on the control stories were not

ambiguous. The purpose of these stories was to control for children’s knowledge of the

meaning of the separate linguistic elements involved in the scope ambiguities discussed

above, i.e. negation and NPs of the form two N. The experimenter holding the puppet had a

choice between two different statements for each of the control stories. One statement was

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154130

Fig. 2. ‘Donald didn’t ﬁnd two guys’, WfNt Condition.

Table 1

Puppet’s statements on test stories for both Wt/Nf and Wf/Nt conditions

Test story 1 ‘Cookie Monster didn’t eat two slices of pizza’

Test story 2 ‘The caveman didn’t ride two horses’

Test story 3 ‘Donald didn’t ﬁnd two guys’

Test story 4 ‘The boy didn’t pet two animals’

true in the context of the story and the other was false. If the child had answered YES to a

given test story, the experimenter holding the puppet was instructed to pick the statement

for the following control story corresponding to a NO answer, and vice-versa. This ensured

that the number of YES and NO answers was balanced. Another precaution that was taken

to ensure that children knew the meaning of two was to have each subject count the

number of toys or characters in each of the stories as they were being laid out on the

table. The list of statements made by the puppet on the control stories is given in Tables 2

and 3 for each of the two conditions.

Finally, the 48 subjects (24 adults and 24 children) were randomly assigned to each

condition (i.e. WtNf and WfNt) thus giving rise to the following 2 £ 2designwithscope

condition and age as between subject factors and 12 subjects per cell (Table 4). The group of

12 children assigned to the WtNf condition ranged in age between 3;11 and 4;11 (mean 4;4,

median 4;4) and the range for the group assigned to the WfNt condition was 4;1 to 4;9 (mean

4;5, median 4;5). The mean ages for these two groups do not differ signiﬁcantly (P ¼ 0:46).

Since the puppet’s statements on critical trials are ambiguous, we chose to treat scope

condition (i.e. WtNf and WfNt) as a between subject factor, instead of a within subject

factor, in order to avoid potential contaminating effects between the two possible readings.

That is, once children become aware of one of the possible interpretations for the ambig-

uous statements, they may ﬁnd it difﬁcult to later assign a similar statement a different

interpretation. In other words, the initial interpretation that children assign to statements of

the form NP didn’t V two N may inﬂuence the way they interpret subsequent statements of

the same form.

4.2. Results

In the analysis below, our dependent measure was the proportion of YES responses to

the puppet’s statements. Beginning with subjects’ responses to the test items (Fig. 3), we

found no signiﬁcant difference in adult’s acceptance rates in the WtNf and WfNt condi-

tions (93 and 97%, respectively). However, we found that children accepted the puppet’s

statements reliably more often in the WfNt condition, as compared to the WtNf condition

(81 vs. 33%, respectively, tð22Þ¼3:62, P ¼ 0:001).

The proportions of YES responses

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 131

One way to avoid this problem in a within subject design is to use a ‘block design’ in which subjects are

presented with all statements of one kind, say WtNf, followed by all statements of the other kind, i.e. WfNt. By

counterbalancing order of presentation (half of the subjects hear WtNf statements ﬁrst and the other half hears

WfNt ﬁrst) one can control for the existence of the kind of effects mentioned above.

On the assumption that chance performance on this task corresponds to a 50% acceptance rate, children’s

performance in the WtNf condition (i.e. 33% acceptance rate) did not signiﬁcantly differ from chance (P ¼ 0:16).

In the WfNt condition children’s acceptance rate, i.e. 81%, was signiﬁcantly above what would be expected by

chance (P , 0:001). However, these estimates should be treated with caution as it is unclear what chance

performance on this task would actually correspond to. In order to get a more accurate estimate, one would

need to quantify potential biases that children may display in favor of one kind of response or another. For

example, children who are confused may generally opt for a YES answer, thus giving rise to a well-known YES

bias (see Crain & Thornton, 1998). In such a case, we would also expect children not to be able to provide

adequate justiﬁcations for their responses. Notice that this is not what we found here. Whether they accepted or

rejected the puppet statements, children consistently provided appropriate justiﬁcations. Finally, note that the

experiments presented here were not designed to assess children’s behavior against expected chance behavior but

rather to compare their behavior in different experimental conditions (i.e. WtNf, WfNt).

were entered into an analysis of variance (ANOVA) with two factors: age (adults, 4-year-

olds) and condition (WtNf, WfNt). The analysis revealed a main effect of age

(Fð1; 44Þ¼27:23, P , 0:0001), a main effect of condition (Fð1; 44Þ¼12:43,

P , 0:001) and a reliable interaction between age and condition (Fð1; 44Þ¼8:77,

P , 0:01).

On the control items, children gave correct answers 97% of the time in the WfNt

condition and 94% of the time in the WtNf condition (recall that the puppet’s statements

in this case were not ambiguous). Adults gave correct answers to the control items 100%

of the time, in both conditions (Fig. 4). An ANOVA with two factors, age and condition,

was performed on the proportion of correct responses to control items. We found no

reliable effect of age (Fð1; 44Þ¼1:8, P ¼ 0:18), no reliable effect of condition

(Fð1; 44Þ¼0:19, P ¼ 0:66) and no interaction between age and condition

(Fð1; 44Þ¼0:19, P ¼ 0:66).

In the WtNf condition, in which two objects/characters were eaten/ridden/found/petted

by the main character and two were not, children rejected the puppet’s statements, i.e.

sentences of the form ‘NP didn’t V two N’, 66% of the time (i.e. on 32 out of 48 trials).

When asked why the puppet was wrong, children typically answered that the main char-

acter did indeed ride/eat/ﬁnd/pet two of the objects/characters. Fig. 5 shows some exam-

ples of children’s justiﬁcations.

Responses of this type, making direct reference to the fact that the character in question

had indeed ridden/eaten/found/petted two of the objects or other characters, accounted for

90% of children’s negative responses. That is, children almost always rejected the

puppet’s statements invoking the isomorphic interpretation of the target sentences.

In the WfNt condition, in which one object/character was eaten/ridden/found/petted by

the main character and one was not, children accepted the puppet’s statements, that is,

sentences of the form ‘NP didn’t V two N’, 81% of the time (i.e. on 39 out of 48 trials).

When asked why the puppet was right, children typically answered that the puppet was

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154132

Table 2

Puppet’s statements on control stories, WfNt condition

Control 1 ‘Donald didn’t ﬁnd all the rings’ (FALSE)

‘Donald found all the rings’ (TRUE)

Control 2 ‘Minnie didn’t buy any of the balloons’ (FALSE)

‘Minnie only bought two of the balloons’ (TRUE)

Control 3 ‘Smurfette didn’t buy the bird’ (FALSE)

‘Smurfette bought a bird and a cat’ (TRUE)

Table 3

Puppet’s statements on control stories, WtNf condition

Control 1 ‘The old man hurt two guys’ (TRUE)

‘The old man didn’t hurt any of the guys’ (FALSE)

Control 2 ‘The troll only bought one pumpkin’ (TRUE)

‘The troll bought all the pumpkins’ (FALSE)

Control 3 ‘The cheetah only ate the hotdog’ (TRUE)

‘The cheetah ate all the food’ (FALSE)

right because the main character only ate/rode/found/petted one of the objects/characters.

Fig. 6 shows some examples of children’s justiﬁcations.

Responses of this kind, making direct reference to the fact that the main character had

performed the relevant action with respect to only one of the objects/characters, accounted

for 91% of children’s justiﬁcations. This clearly shows that children correctly accepted the

puppet’s statements because the isomorphic reading of the test sentences was true, i.e. it

was not the case that the main character ate/rode/found/petted two objects/characters.

Adults always gave appropriate justiﬁcations to the puppet’s statements, in both condi-

tions.

4.3. Discussion

The results presented above show that our intuitions about the scope ambiguities in

sentences like Donald didn’t ﬁnd two guys are correct. Adult speakers of English can

easily assign either scope interpretation, as indicated by the proportions of YES responses

in each condition (93% for WtNf and 97% for WfNt). These results also underscore an

important property of the TVJT, namely the fact that subjects seem to prefer to give the

puppet credit for speaking truthfully, if that is possible. Notice that this could in principle

have been otherwise since the statements that subjects were asked to evaluate in each

condition could have legitimately been interpreted as either true or false, depending on

which scope reading one judges to be relevant (see Crain & Thornton, 1998 for further

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 133

Table 4

Wide True/Narrow False Wide False/Narrow True

4-year-olds (n)12 12

Adults (n)12 12

Fig. 3. Proportion of YES responses to test trials for children and adults in each of the two conditions.

discussion). Second, children’s almost perfect performance on control stories (i.e. 95%

correct answers on average) indicates that they didn’t experience any difﬁculty with the

task. More importantly, this shows that neither negation, nor quantiﬁed NPs of the type

two N, when considered separately, posed any problems to children. This also shows that

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154134

Fig. 4. Proportion of correct responses to control trials for children and adults in each of the two conditions.

Fig. 5.

children were able to say YES and NO when appropriate since the control stories were

designed to evoke both types of answer.

Finally, these results strengthen the ﬁndings of Musolino et al. (2000) by showing with

numerally quantiﬁed NPs that 4-year-olds have a reliable preference for the narrow scope,

isomorphic reading of quantiﬁed objects (i.e. 81 vs. 33% acceptance rate, respectively,

P , 0:001, along with appropriate justiﬁcations).

The observation of isomorphism, now extended to children’s interpretation of negation

and numerally quantiﬁed NPs, leaves open a number of important questions. As always,

when differences in the linguistic behavior of children and adults are uncovered, one needs

to try to explain what factors, linguistic or other, give rise to such differences.

Our observation also raises a question discussed at length in the psycholinguistic litera-

ture on ambiguity resolution (Kurtzman & MacDonald, 1993; others). When the rules of

grammar allow more than one interpretation to be assigned to a sentence, it is standard

practice for psycholinguists to regard preferences for one interpretation or the other as

arising from the operation of comprehension/processing principles (Frazier & Fodor,

1978). In the case of quantiﬁer scope ambiguities two main principles have been invoked

to explain such preferences. The ﬁrst is based on the notion of linear order. According to

this principle, the preferred scope reading for quantiﬁed phrases corresponds to the left-to-

right ordering of the phrases in the surface form of the sentence (Bunt, 1985; Fodor, 1982;

Johnson-Laird, 1969; Kroch, 1974; among others). The other principle relies on the

hierarchical relations between the constituents of a sentence. One version of this principle,

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 135

Fig. 6.

for example, seeks to account for scope preferences in terms of the c-command relation

between the quantiﬁed phrases (May, 1985; Reinhart, 1983; also see Frazier, 2000; Kurtz-

man & MacDonald, 1993 for discussion of this principle). Although a good deal is known

about the parsing mechanisms used by adult speakers to resolve ambiguous sentences (see

Tanenhaus & Trueswell, 1995), this question has only rarely been investigated in the case

of children (but see Crain & Thornton, 1998; Trueswell, Sekerina, Hill, & Logrip, 1999).

Our results so far suggest that children’s interpretation of ambiguous sentences contain-

ing negation and quantiﬁed NPs is constrained by the position of these elements in the

surface string. However, as we have noted, it is unclear at this point whether isomorphism

obtains as a result of a constraint imposed by the linear order between negation and the

quantiﬁed NP or by the more abstract notion of c-command deﬁned over the hierarchical

relations between these elements. This is because in the examples that we have tested so

far, the two notions are confounded (this is also true of the examples used by Musolino et

al., 2000). As can be seen in (24), the quantiﬁed object falls within the c-command domain

of negation. Hierarchical relations therefore map to a narrow scope reading. The quantiﬁed

object also follows negation. The linear relation between negation and the object QNP

therefore also maps to a narrow scope interpretation. Both notions are therefore compa-

tible with children’s interpretations.

(24)

Finally, we return to the cross-linguistic question. Is the observation of isomorphism an

observation about the acquisition of quantiﬁcational structure in English alone or is it a more

widespread property that can be observed in children’s acquisition of quantiﬁcational struc-

ture in other languages as well? In Section 5, we present the results of an experiment on child

and adult speakers of Kannada, a Dravidian language which offers the advantage of allowing

us to simultaneously address the cross-linguistic and the structural questions. We discuss the

questions regarding the possible causes of children’s difﬁculty with non-isomorphic inter-

pretations and the transition to the adult system of interpretation in Section 6.

5. Experiments on Kannada

5.1. Scope ambiguities and the structure of Kannada

Kannada is a Dravidian language spoken by approximately 40 million people in the

state of Karnataka in south-western India. The canonical word order in Kannada is

Subject-Object-Verb (SOV) and Kannada displays the same kind of scope ambiguities

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154136

as English with respect to negation and quantiﬁed NPs (Lidz, 1999). These properties are

illustrated in (25), which can be interpreted as meaning that it is not the case that I read two

books, a narrow scope interpretation of the numeral, or that there are two books that I

didn’t read, a wide scope interpretation of the numeral.

(25) naanu eraDu pustaka ood-al-illa

I-nom two book read-inf-neg

‘I didn’t read two books’

The crucial difference between Kannada and English for our purposes is that in Kannada,

linear order and c-command relations are not confounded. Consider the tree diagrams given

in (26). In English, negation both precedes and c-commands the object position, as

discussed earlier. In Kannada, however, negation c-commands the object but does NOT

precede it. This means that in Kannada, a c-command account of isomorphism would

predict a narrow scope reading of the object with respect to negation whereas a linear

account of isomorphism would predict a wide scope reading of the object (Table 5).

(26)

Kannada therefore provides the ideal paradigm to tease apart the contribution of linear

and hierarchical relations in children’s interpretation of scope relations. To the extent that

Kannada-speaking children are restricted to one of the two possible interpretations of

sentences like (25) in the way that English-speaking children are, linear order and hier-

archical relations make opposite predictions:

Linear prediction

If children’s interpretations of scope relations are constrained by linear order, then

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 137

Table 5

Predictions of the two accounts of isomorphism

Kannada English

C-command Narrow scope reading of object NP Narrow scope reading of object NP

Precedence Wide scope reading of object NP Narrow scope reading of object NP

children will display a preference for the wide scope reading of the object with respect to

negation.

Hierarchical prediction

If children’s interpretations of scope relations are constrained by c-command relations

between negation and the quantiﬁed object, then children will display a preference for the

narrow scope reading of the object with respect to negation.

5.2. Method

5.2.1. Subjects

In order to test native speakers of Kannada, we went to Mysore, India where we tested

24 Kannada-speaking 4-year-olds (11 girls and 13 boys) between the ages of 4;0 and 4;11

(mean 4;5) from Pushkarini and Swami Vivekanada preschools and 24 Kannada-speaking

adults from Mysore University. An additional six children were tested but eliminated from

the study either because they always gave the same answer or because they could not

justify their answers.

5.2.2. Procedure

As in the experiment on English, speakers of Kannada were tested using the TVJT

methodology. Both experimenters – the one telling the stories and the one playing the part

of the puppet – were native speakers of Kannada. Unlike their English counterparts,

Kannada-speaking adults did not see a videotaped version of the stories. Instead, the

stories witnessed by Kannada-speaking adults were also acted out in front of them.

Subjects received a pretest consisting of two stories followed by a series of seven stories

(four test stories and three ﬁller stories) administered in a pseudo-random order. The 48

subjects (24 adults and 24 children) were randomly assigned to each of the two conditions

(i.e. WtNf and WfNt) giving rise to the same 2 £ 2 design used in the experiment on

English. The children were tested individually in a quiet area outside the classroom. The

adults were tested individually or in groups of up to ten.

5.2.3. Materials

As before, two versions of the stories were constructed. In one version, the wide

scope reading of the numeral was true (abbreviated, Wt) and the narrow scope reading

was false (abbreviated, Nf). To accomplish this, the stories typically involved a main

character and four objects on which an action was to be performed. In the stories, the

character performed the action on two of the four objects but not on the other two. In

the end therefore, a statement like NP didn’t V two N was true on a wide scope

interpretation of the numeral since there were two objects on which the action hadn’t

been performed. On the other hand, the narrow scope reading of the numeral was false

because the number of objects on which the action had been performed was exactly

two.

In the other version of the stories, the wide scope reading of the numeral was false

(abbreviated, Wf) and the narrow scope reading was true (abbreviated, Nt). This was

achieved by having the main character attempt to perform the action with respect to

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154138

two objects instead of four. In the stories, the character would perform the action on one of

the objects but not on the second one. At the end of the stories, therefore, the wide scope

reading of sentences like NP didn’t V two N was false since the character failed to perform

the action with respect to only one of the objects. On the other hand, the narrow scope

reading was true because it was indeed not the case that the character performed the action

with respect to two objects. The statements made by the puppet on each of the four test

trials in both conditions are given in Table 6. The list of statements made by the puppet on

control stories is given in Table 7.

5.3. Results

As before, our dependent measure was the proportion of YES responses to the

puppet’s statements. Beginning with subjects’ responses to test items, we found that

Kannada-speaking adults gave YES responses to the wide scope reading 87.5% of the

time in the WtNf condition and that they gave YES response to the narrow scope reading

85.4% of the time in the WfNt condition. No signiﬁcant differences were found between

these two acceptance rates. However, we found that 4-year-olds accepted the puppet’s

statements reliably more often in the WfNt condition than in the WtNf condition (i.e.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 139

Table 6

Puppet’s statements on test stories for both Wt/Nf and WF/Nt conditions

Test story 1 Anoop eradu kaaru toley-al-illa

‘Anoop didn’t wash two cars’

Test story 2 Avanu eradu biskit tinn-al-illa

‘He didn’t eat two cookies’

Test story 3 Rashmi kekku-ge eradu moTTe hak-al-illa

‘Rashmi didn’t put two eggs into the bowl’

Test story 4 Huduganu eradu ungra karedis-al-illa

‘Ernie didn’t buy two rings’

Table 7

Puppet’s statements on control stories

Control story 1 Simha muru haavu hiD-i-tu (FALSE)

‘The lion found three snakes’

Simha eradu haavu hiD-i-tu (TRUE)

‘The lion found two snakes’

Control story 2 Ha manushya ella baNDe ett-id-a (FALSE)

‘That man lifted all the rocks’

Ha manushya eradu baNDe ett-id-a (TRUE)

‘That man lifted two rocks’

Control story 3 Ella kappe maney-a meele haar-i-tu (FALSE)

‘All the frogs jumped over the house’

Eradu kappe maney-a meele haar-i-tu (TRUE)

‘Two frogs jumped over the house’

75% of the time vs. 22.9% of the time, tð22Þ¼3:49, P ¼ 0:002).

The proportions of

YES responses to the puppet’s statements were entered into an ANOVA with two

factors: age (adults, 4-year-olds) and condition (WtNf, WfNt). The analysis revealed a

main effect of age (Fð1; 44Þ¼20:19, P , 0:0001), a main effect of condition

(Fð1; 44Þ¼8:97, P , 0:01) and a reliable interaction between age and condition

(Fð1; 44Þ¼10:53, P , 0:01) (Fig. 7).

On control items, we found that adults gave correct answers 94% of the time in both

conditions. Children gave correct responses to control items 97% of the time in the WtNf

condition and 100% of the time in the WfNt condition. The proportions of correct

responses to control items were entered into a ANOVA with two factors, age and condi-

tion, and we found no reliable effect of age (Fð1; 44Þ¼1:94, P ¼ 0:17), no reliable effect

of condition (Fð1; 44Þ¼0:21, P ¼ 0:64) and no reliable interaction between age and

condition (Fð1; 44Þ¼0:21, P ¼ 0:64) (Fig. 8).

Finally, the justiﬁcations given by the Kannada children and adults in each of the two

conditions matched those reported for English. That is, in the WtNf condition, children

rejected the puppet statements by invoking the fact that the main character had indeed

performed the relevant action with respect to two objects/characters. In the WfNt condition,

they accepted the puppet statements and correctly pointed out that the main character had only

performed the relevant action with respect to one of the objects/characters (Figs. 5 and 6).

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154140

Fig. 7. Proportion of YES responses to test trials for children and adults in each of the two conditions.

Since we did so for English, we compare here children’s performance on each of the two critical conditions to

chance performance. Of course, these estimates deserve the same qualiﬁcations as the ones discussed in the case

of English-speaking children. We found that Kannada-speaking children’s acceptance rate in the WtNf condition,

i.e. 22.9%, was reliably lower than what would be expected by chance (P ¼ 0:01), suggesting that children were

reliably selecting the narrow scope interpretation. Similarly, in the WfNt condition, children’s acceptance rate,

i.e. 75%, was signiﬁcantly above what would be expected by chance (P , 0:05), suggesting that children were

reliably selecting the narrow scope interpretation.

5.4. Discussion

Our results on Kannada conﬁrm the fact that sentences containing negation and numer-

ally quantiﬁed objects are ambiguous in that language too, as witnessed by the fact that

adult speakers of Kannada can easily assign either scope interpretation to these sentences

(i.e. 87.5% acceptance rate in the WtNf condition and 85.4% in the WfNt condition). For

children, the results on Kannada also replicate the pattern observed for English, namely the

fact that children, unlike adults, display a reliable preference for one of the two scope

readings. In that regard, recall the predictions that were made previously regarding

Kannada-speaking children’s interpretation of such ambiguous sentences.

Linear prediction

Ifchildren’sinterpretations ofscope relations areconstrained by linear order, then children

will display a preference for the wide scope reading of the object with respect to negation.

Hierarchical prediction

If children’s interpretations of scope relations are constrained by c-command relations

between negation and the quantiﬁed object, then children will display a preference for the

narrow scope reading of the object with respect to negation.

Our results show that children have a reliable preference for the narrow scope reading of

the quantiﬁed objects with respect to negation in favor of the wide scope reading (i.e. 75 vs.

22.9% acceptance rate, respectively). We conclude that the Isomorphism effect is a conse-

quence of hierarchical structure rather than linear order. Children’s interpretations of scope

relations are limited in a way that falls out directly from building syntactic structure, since in

the surface syntax, negation c-commands the quantiﬁcational expression. Although prece-

dence relations can be determined directly from the speech signal, children do not use these

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 141

Fig. 8. Proportion of correct responses to control trials for children and adults in each of the two conditions.

relations in computing semantic scope. Rather, the syntactic structure imposed by a speaker

on the speech signal determines the scope of quantiﬁcational elements.

6. General discussion

Let us summarize our ﬁndings. In the experiments presented here, we asked how speakers

of English and Kannada resolve scope ambiguities involving the interaction of negation and

numerally quantiﬁed object NPs. Our main ﬁnding is that in both languages adults readily

assign either interpretation, whereas 4-year-olds display a strong preference for the narrow

scope reading of the object. Thus, regardless of language, children’s preferences are char-

acterized by an over-reliance on the surface properties of linguistic structures. This result

extends the Musolino et al. (2000) Observation of Isomorphism to the case of numerally

quantiﬁed NPs. More importantly though, our results on Kannada show that children

compute scope relations on the basis of the surface c-command relations between negation

and the object, not on the basis of linear order. This demonstrates that young children, like

adults, represent sentences as hierarchical constructs and not mere strings of words, lending

support, in children, to one of the fundamental discoveries of modern linguistics. Where

children differ from adults is in their readiness to apply the covert displacement operations

discussed at the outset of this paper. That is, children take the surface syntactic representa-

tion to encode the same hierarchical relations as the semantic representation.

Returning to a question that was raised earlier, we may now ask why children differ

from adults in the way that we observed. In other words, why do children initially fail to

assign a wide scope interpretation to quantiﬁed objects? As always in such cases, two

types of explanation are possible. The ﬁrst one is that until a certain stage in grammatical

development, children do not have implicit knowledge of the fact that such interpretations

can be generated by the grammar of their language. This kind of account is usually known

as a ‘competence’ or ‘grammatical’ account in the linguistic literature (Chomsky, 1965,

1986). On such an account, children reject non-isomorphic interpretations because their

grammars do not include the covert displacement operation (see discussion in Section 2)

that is required to generate these readings. Kra

mer (2000) proposes a semantic account

along these lines on the basis of results from Dutch-speaking children. Speciﬁcally, she

suggests that these children may initially lack knowledge of the fact that indeﬁnite NPs can

receive what she calls a free variable interpretation, which is equivalent to what we have

been calling a wide scope interpretation. This, in essence, amounts to the claim that

children lack knowledge of one of the grammatical operations responsible for the inter-

pretation of indeﬁnite NPs.

Another possibility is that child interpretations result from some limitation on the

computational resources that children deploy during language comprehension. That is,

although children’s grammars may generate both possible readings, they may not be able

to access the non-isomorphic interpretations because they are computationally too taxing

(see Frazier, 2000 for a speciﬁc implementation of this view). On this view, often referred

to as a ‘performance’ or ‘processing’ account in the linguistic literature (Grodzinsky &

Reinhart, 1993), the child and the adult are assumed to share the same underlying gram-

matical knowledge but to differ in the way they implement that knowledge in the time-

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154142

course of language comprehension. One possibility suggested by recent ﬁndings on chil-

dren’s ability to resolve ambiguities on-line is that children may experience a garden-path

effect from which they cannot recover (Trueswell et al., 1999). In the case at hand,

children may initially access the isomorphic reading and end up being stuck with it for

lack of the ability to revise that initial interpretation. This view makes the interesting

prediction that although adults are capable of accessing either interpretation, they may

nevertheless experience a measurable garden-path effect in processing sentences with non-

isomorphic readings, albeit one from which they can quickly recover. An interesting way

to test this prediction would be to present adults with contexts in which both scope read-

ings are true and examine the justiﬁcations for their necessarily positive answers. We will

undertake this task in future research.

Although the data presented here do not allow us to distinguish between a competence or

a performance account, it is worth mentioning that in an attempt to tease apart the two views,

we presented children with a context designed to enhance the availability of the wide scope

reading of sentences like ‘The Smurf didn’t catch two birds’ (Musolino & Lidz, 2002). In

order to achieve this, we presented children with a story in which a Smurf was trying to catch

three cats and four birds. In the end, the Smurf manages to catch all the cats but only catches

two of the four birds. The puppet’s statement in this case was “The Smurf caught all the cats

but he didn’t catch two birds”. We tested ten 4-year-olds on this condition, using the same

method as described earlier and found that children still rejected the puppet’s statement 75%

of the time. Note that in a similar attempt, Musolino (in press) compared children’s perfor-

mance on their comprehension of sentences like “Every horse jumped over the fence” vs.

“Every horse jumped over the log but every horse didn’t jump over the fence”. Although

previous work had demonstrated that children never allow a wide scope reading of negation

over the universally quantiﬁed subject in sentences like “Every horse didn’t jump over the

fence”, their performance dramatically improved when the sentence was introduced with an

overt contrast. In sum, contextual factors do allow us to see that a performance account of

children’s reliance on isomorphic readings in sentences involving negation and a univer-

sally quantiﬁed subject NP is likely to be correct. However, these same factors were unable

to shift children’s preferences in sentences with a numerally quantiﬁed object NP. Further

investigation is therefore required to determine whether isomorphism effects are due to

performance or competence factors across the board.

Another question raised earlier in the discussion is how children eventually manage to

converge on the adult system of interpretation. The answer to this question depends on how

we answer the previous question regarding the nature of children’s difﬁculty with non-

isomorphic interpretations. Suppose for the sake of argument that we are dealing with a

competence problem and that children’s grammars are initially incapable of generating non-

isomorphic interpretations. At some point, children may hear a sentence such as You didn’t

put away two toys uttered in a situation that veriﬁes the truth conditions of a non-isomorphic

interpretation. That is, such a sentence could be uttered in a situation in which there are two

speciﬁc toys that the child didn’t put away but lots of toys s/he did put away. The pairing

between a sentence of this kind and the appropriate situation may provide children with the

relevant positive evidence necessary to postulate the existence of the syntactic operations

that give rise to non-isomorphic interpretations. On the other hand, if it turns out that

children are able to generate non-isomorphic interpretations, albeit with a large computa-

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 143

tional cost, then answering the convergence question amounts to explaining how children’s

preferences shift over time. One possibility is that as children grow older their computa-

tional resources become more adult-like and they eventually manage to access non-

isomorphic interpretations more reliably. In this respect, it is interesting to note that a

group of 12 5-year-old children, tested on the WtNf condition, accepted the puppet’s state-

ments, and therefore accessed the wide scope, non-isomorphic interpretation signiﬁcantly

more often than our group of 4-year-olds, i.e. 79% of the time vs. 33% of the time, respec-

tively (tð22Þ¼3:09, P , 0:01). Statistical analysis also revealed that 5-year-olds didn’t

differ signiﬁcantly from our control group of adults in their acceptance rate of the puppet

statements (79 vs. 93%, respectively, tð22Þ¼1:26, P ¼ 0:21). Thus, by 5 years of age,

children are able to assign sentences of the form NP didn’t V two N non-isomorphic inter-

pretations in an adult-like fashion (Musolino & Lidz, 2002). In future research we are

planning to address the question of what factors inﬂuence the development of the ability

to access non-isomorphic interpretations.

Although it is at present unclear why young children experience difﬁculty with the non-

isomorphic interpretation of sentences like NP didn’t V two N, it is worth pointing out the

existence of an interesting parallel between our results and recent ﬁndings from the

developmental literature. These ﬁndings come from a set of studies designed to investigate

children’s sensitivity to a class of conversational inferences known as scalar implicatures

(Carston, 1998; Gazdar, 1979; Grice, 1989; Harnish, 1976; Hirschberg, 1985; Horn, 1972;

Levinson, 2000; Sperber & Wilson, 1995; among many others). Scalar implicatures arise

in examples like Some students can write well where the speaker’s use of some typically

indicates that s/he had reasons not to use a more informative term, e.g. all. The sentence,

Some students can write well, therefore invites the inference – i.e. gives rise to the

implicature – that not all students can write well. Recent work on the development of

pragmatics suggests that preschool children are often insensitive to such implicatures

when they interpret scalar terms (Chierchia, Crain, Guasti, Gualmini, & Meroni, 2001;

Musolino & Lidz, in press; Noveck, 2001). For example, using a modal reasoning

scenario, Noveck (2001) investigated children’s and adults’ interpretations of statements

expressing x might be y in contexts in which the stronger statements x must be y were true.

As Noveck points out, x might be y can be interpreted logically (as compatible with must)

or pragmatically (as exclusive to must). What Noveck found is that 7–9-year-old children

treated x might be y logically (i.e. as compatible with x must be y) much more often than

adults. Noveck also reports that 8–10-year-old children typically treated French certains

(‘some’) as compatible with tous (‘all’) whereas adults in the same task were equivocal

(for a similar observation with English some, see Smith, 1980 and for similar observations

with the disjunction operator, or, see Braine & Rumain, 1981; Chierchia et al., 2001;

Gualmini, Crain, Meroni, Chierchia, & Guasti, 2001; Paris, 1973). Such behavior has led

Noveck to conclude that “younger, albeit competent reasoners, initially treat a relatively

weak term logically before becoming aware of its pragmatic potential”, and that, in this

respect, “children are more logical than adults” (Noveck, 2001, p. 165).

The parallel between children’s interpretation of sentences of the form NP didn’t V two

N and their interpretation of sentences containing scalar terms can now be stated: in both

cases, two interpretations are available (logical vs. pragmatic in the scalar experiments and

isomorphic vs. non-isomorphic in the experiments reported on here) though young chil-

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154144

dren only seem to be able to access one (the logical interpretation of scalar terms and the

isomorphic interpretation of quantiﬁed statements). This similarity is sharpened when we

take into account the fact that number terms such as two N are typically regarded as

belonging to the class of scalar terms (Horn, 1972). Is it possible then that children’s

difﬁculty with sentences like NP didn’t V two N and their reluctance to interpret scalar

terms pragmatically have a common origin? We believe that in spite of the similarities

noted above, the case for a uniform treatment of the two phenomena faces a number of

difﬁculties. First, it is important to observe that the phenomenon of scalar implicature

typically relies on the existence of entailment relations between the alternative interpreta-

tions of scalar terms (e.g. some/all, and/or, might/must, etc.). So for example, All students

can write well entails that some students can write well, but not vice-versa. The use of all

in All students can write well therefore gives rise to a more informative statement (i.e. one

that is true in a narrower set of circumstances) than the one expressed by the proposition

Some students can write well. The implicature associated with Some students can write

well is therefore typically regarded as arising from a violation of the Gricean maxim of

Quantity and in particular the submaxim of informativess given in (i).

(27) Quantity maxim (Grice, 1989)

(i) Make your contribution as informative as required

(ii) Do not make your contribution more informative than is required

An important difference between statements like Some students can write well and

Donald didn’t ﬁnd two guys is that in the latter case, as already discussed in the context

of Experiment 1, no entailment relation holds between the two scope readings (i.e. not .

two and two . not). It is therefore unclear how the relative informative strength of the two

scope readings could be deﬁned and therefore how a scalar implicature could arise. The

parallelism discussed above further breaks down when we consider the developmental

pattern of the respective phenomena. Studies on the development of pragmatic abilities

indicate that children as old as 10 still experience difﬁculty with scalar implicatures (see

Noveck, 2001 and references mentioned therein). By contrast, we found that by the age of

5, children are adult-like in their interpretation of sentences like NP didn’t V two N. Also

note that using the same task as the one we used here, i.e. the TVJT of Crain and Thornton

(1998), Chierchia et al. (2001) found that 5-year-olds performed poorly on tasks involving

the derivation of scalar implicatures (for similar ﬁndings see Musolino & Lidz, in press).

When combined, these observations suggest that the apparent similarity between chil-

dren’s difﬁculty with scalar implicature and scope ambiguity (in the case of statements like

NP didn’t V two N) is only superﬁcial.

7. Conclusion

In the ﬁeld of psycholinguistics, the study of ambiguity resolution has traditionally been

used as a tool to investigate the operation of the basic principles underlying the human

comprehension system (Fodor, Bever, & Garrett, 1974; Frazier & Fodor, 1978; Frazier &

Rayner, 1982; among many others). Here, we have extended this approachto the study of child

language and shown that the way young children interpret scopally ambiguous sentences has

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 145

psychological implications for the format of the linguistic representations they entertain and

the rules they use to determine the meaning of a sentence from its structure.

More importantly, however, we have used this kind of sentence as a way to ﬁnd

evidence for the hierarchical structure that linguists take to be at the core of grammatical

knowledge. While this paper leaves open many important questions concerning the details

of children’s errors and the path from childhood to linguistic maturity, we have recon-

ﬁrmed experimentally what every linguist wouldn’t doubt: namely, that sentences are built

by nesting structures inside of other structures and that relations deﬁned over these struc-

tures play an explanatory role not only in the characterization of adult knowledge but also

in the characterization of children’s knowledge (however incomplete). While children’s

representations may not always be identical to their adult counterparts, we can be sure that

their representations do not differ in kind from the abstract hierarchical structures that best

characterize adult representations.

Acknowledgements

We would like to thank the children and the staff of the following daycare centers

without whom none of the research reported here would have been possible: the Center

for Young Children at the University of Maryland, the Mulberry Daycare Center and the

Schoolhouse Learning Center in the Philadelphia area and the Pushkarini and the Swami

Vivekananda preschools in Mysore, India. Special thanks to Susie Metrick and Kirsten

Thorpe for assistance with running subjects. We would also like to thank the Central

Institute for Indian Languages, especially Omkar Koul, P.P. Giridhar and A.Y. Vijayku-

mar, for logistical and experimental assistance in Mysore. In addition, we would like to

thank audiences at the University of Chicago, the University of Delaware, the University

of Massachusetts, Northwestern University, Yale University, the Central Institute of

Indian Languages (Mysore), and the Central Institute for English and Foreign Languages

(Hyderabad), as well as the audience of the 2000 Boston University Conference on

Language Development, for helpful discussions on parts of the material reported here.

Finally, our thanks go to our colleagues at the University of Pennsylvania and North-

western University for their helpful comments and suggestions throughout the elaboration

of this work. This work was supported in part by a NIH National Research Service Award

Postdoctoral Fellowship to Lidz.

Appendix A

A.1. Test stories for English

A.1.1. Wide True, Narrow False Condition

A.1.1.1. Story 1

Plot: This story features Cookie Monster and four giant slices of pizza. Cookie Monster

is very hungry. He eats the ﬁrst slice of pizza then he eats the second one but is barely able

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154146

to ﬁnish it. Cookie Monster then says that the two remaining slices look good but he’s full

and can’t have another bite to eat.

Puppet’s statement: ‘Cookie Monster didn’t eat two slices of pizza, am I right?’

A.1.1.2. Story 2

Plot: This story features a brave caveman and four wild horses. The caveman decides

that he wants to ride on the horses’ backs. He gets on the ﬁrst horse, rides it and gets

thrown to the ground. The same thing happens with the second horse. The caveman then

walks to the third and the fourth horses, tells them that he would like to ride on their back

but that he won’t because he doesn’t want to fall again and hurt himself.

Puppet’s statement: ‘The caveman didn’t ride two horses, am I right?’

A.1.1.3. Story 3

Plot: This takes place at the zoo and features a boy, a giraffe, a kangaroo, a tiger and a

cheetah. The boy has always wanted to go to the zoo to pet the animals there. He ﬁrst

approaches the giraffe and pets it. He does the same with the Kangaroo. He then walks to

the tiger to try to pet it but the tiger starts growling at the boy and he can’t even get close to

it. The same thing happens with the Cheetah.

Puppet’s statement: ‘The boy didn’t pet two animals, am I right?’

A.1.1.4. Story 4

Plot: see Section 4

Puppet’s statement: ‘Donald didn’t ﬁnd two guys, am I right?’

A.1.2. Wide False, Narrow True Condition

A.1.2.1. Story 1

Plot: This story features Cookie Monster and two giant slices of pizza. Cookie Monster

is very hungry. He starts eating the ﬁrst slice of pizza but is barely able to ﬁnish it. Cookie

Monster then says that the remaining slice looks good but he’s full and can’t have another

bite to eat.

Puppet’s statement: ‘Cookie Monster didn’t eat two slices of pizza, am I right?’

A.1.2.2. Story 2

Plot: This story features a caveman and two wild horses. The caveman decides that he

wants to ride on the horses’ backs. He gets on the ﬁrst horse, rides it and gets thrown to the

ground. The caveman then walks to the second horse and tells it that he would like to ride

on its back but that he won’t because he doesn’t want to fall again and hurt himself.

Puppet’s statement: ‘The caveman didn’t ride two horses, am I right?’

A.1.2.3. Story 3

Plot: This story takes place at the zoo and features a boy, a giraffe and a dinosaur. The

boy has always wanted to go to the zoo to pet the animals there. He ﬁrst approaches the

giraffe and pets it. He then walks to the dinosaur to try to pet it but the dinosaur starts

growling at the boy and he can’t even get close to it.

Puppet’s statement: ‘The boy didn’t pet two animals, am I right?’

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 147

A.1.2.4. Story 4

Plot: In this story, Donald is playing hide and seek with two of his friends. Donald ﬁnds

the ﬁrst friend but he fails to ﬁnd the second one.

Puppet’s statement: ‘Donald didn’t ﬁnd two guys, am I right?’

A.2. Control stories for English

A.2.1. Story 1

Plot: This story features an old man and three of his friends. After a nice meal at the old

man’s house, his three friends decide to take a nap on the lawn. Meanwhile, the old man,

who isn’t tired, decides to mow his lawn. Although he tries to be really careful, the old man

hurts two of his friends with his lawnmower.

Puppet’s statements: ‘The old man hurt two guys, am I right?’ (TRUE) or

‘The old man didn’t hurt any of the guys, am I right?’ (FALSE)

A.2.2. Story 2

Plot: This story features a troll at a Halloween store. The troll is looking at various

Halloween items including three pumpkins. He ends up buying one of them but not the two

others.

Puppet’s statements: ‘The troll only bought one pumpkin, am I right?’ (TRUE) or

‘The troll bought all the pumpkins, am I right?’ (FALSE)

A.2.3. Story 3

Plot: In this story, a hungry cheetah asks his human friend to cook him dinner. The

cheetah’s friend serves him a hotdog and some peas. The cheetah complains about the peas

and says that he won’t eat them. He ends up eating only the hotdog.

Puppet’s statements: ‘The cheetah only ate the hotdog, am I right?’ (TRUE)

‘The cheetah ate all the food, am I right?’ (FALSE)

A.2.4. Story 4

Plot: In this story, Donald is looking for three rings that were hidden by one of his

friends. Donald ends up ﬁnding all three rings.

Puppet’s statements: ‘Donald didn’t ﬁnd all the rings, am I right?’ (FALSE)

‘Donald found all the rings, am I right?’ (TRUE)

A.2.5. Story 5

Plot: In this story, Minnie is looking at three rings and three balloons that she’s

considering buying. She ends up buying the rings and two of the three balloons.

Puppet’s statements: ‘Minnie didn’t buy any of the balloons, am I right?’ (FALSE)

‘Minnie only bought two of the balloons, am I right?’ (TRUE)

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154148

A.2.6. Story 6

Plot: In this story, Smurfette is at the pet store. There is a bird, a cat and a dog.

Smurfette ends up buying the bird and the cat.

Puppet’s statements: ‘Smurfette didn’t buy the bird, am I right?’ (FALSE)

‘Smurfette bought a bird and a cat, am I right?’ (TRUE)

A.3. Test stories for Kannada

A.3.1. Wide True, Narrow False Condition

A.3.1.1. Story 1

Plot: This story features a boy named Anoop and his four new cars. Anoop wants to

wash his cars because he knows it is important to take care of things you like. Today, he

washed his ﬁrst car and then his second car. When he came to the third car, it was already

clean because he hadn’t driven it yet. The fourth car is a convertible and Anoop is worried

about getting water on the inside so he decides not to wash it now.

Puppet’s statement: Anoop eradu kaaru toley-al-illa

anoop two car wash-inf-neg

‘Anoop didn’t wash two cars.’

A.3.1.2. Story 2

Plot: This story features Cookie Monster and four cookies. Cookie Monster is very

hungry. He eats the ﬁrst two cookies. When he comes to the third cookie, he sees that it is

shaped like a heart, not a cookie and so even though it has frosting that looks good to eat,

he decides not to eat it. The fourth cookie is shaped like a person and he considers eating it

but he’s not sure if its a cookie or a person so he decides to play it safe and not eat it.

Puppet’s statement: avanu eradu biskit tinn-al-illa

he two cookie eat-inf-neg

‘He didn’t eat two cookies.’

A.3.1.3. Story 3

Plot: This story features a girl named Rashmi who is planning to make a cake. She has

all the ingredients: four eggs, ﬂour, sugar and milk. She puts in the ﬂour, sugar and milk

and then she starts to put in the eggs. She puts in two but the other two have black spots on

them. They might be rotten so she decides not to use them.

Puppet’s statement: Rashmi keeku-ge eradu moTTe haak-al-illa

Rashmi cake-dat two egg put-inf-neg

‘Rashmi didn’t put two eggs into the cake.’

A.3.1.4. Story 4

Plot: This story features a boy who goes into a store to buy a present for his friend. The

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 149

shopkeeper has 4 rings and the boy has 4 rupees. The ﬁrst two rings cost 1 rupee each and

so he buys them. The second two cost 3 rupees each and so he doesn’t have enough money

to buy either of them.

Puppet’s statement: huduganu eradu ungra karedis-al-illa

boy two ring buy-inf-neg

‘The boy didn’t buy two rings.’

A.3.2. Wide False, Narrow True Condition

A.3.2.1. Story 1

Plot: This story features a boy named Anoop and his two new cars. Anoop decides to

wash his cars because he knows it is important to take care of things you like. Today, he

washes the ﬁrst car. The second car is a convertible and Anoop is worried about getting

water on the inside so he decides not to wash it now.

Puppet’s statement: Anoop eradu kaaru toley-al-illa

anoop two car wash-inf-neg

‘Anoop didn’t wash two cars.’

A.3.2.2. Story 2

Plot: This story features Cookie Monster and two cookies. Cookie Monster is very

hungry. He eats the ﬁrst cookie. When he comes to the second cookie, he sees that it is

shaped like a heart, not a cookie and so even though it has frosting that looks good to eat,

he decides not to eat it.

Puppet’s statement: avanu eradu biskit tinn-al-illa

he two cookie eat-inf-neg

‘He didn’t eat two cookies.’

A.3.2.3. Story 3

Plot: This story features a girl named Rashmi who is planning to make a cake. She has

all the ingredients: two eggs, ﬂour, sugar and milk. She puts in the ﬂour, sugar and milk

and then she starts to put in the eggs. She puts in one but the other one has a black spot on

it. It might be rotten so she decides not to use it.

Puppet’s statement: Rashmi keeku-ge eradu moTTe haak-al-illa

Rashmi cake-dat two egg put-inf-neg

‘Rashmi didn’t put two eggs into the cake.’

A.3.2.4. Story 4

Plot: This story features a boy who goes into a store to buy a present for his friend. The

shopkeeper has 2 rings and the boy has 2 rupees. The ﬁrst ring costs 1 rupee and so he buys

it. The second ring costs 3 rupees and so he doesn’t have enough money to buy it.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154150

Puppet’s statement: huduganu eradu ungra karedis-al-illa

boy two ring buy-inf-neg

‘The boy didn’t buy two rings.’

A.4. Control stories for Kannada subjects

A.4.1. Story 1

Plot: This story features two lions playing hide and seek. One lion hides in the woods

and the other lion turns his head to count. After counting to ten, the lion starts to look for

his friend. First he looks behind a tree but there is a snake there. Then he looks behind

another tree and there is a snake there too. He decides that there are too many snakes in the

woods and so it is too dangerous to play there.

Puppet’s statements: simha muru haavu hiD-i-tu (FALSE)

lion three snake ﬁnd-pst-3sn

‘The lion found three snakes.’

simha eradu haavu hiD-i-tu (TRUE)

lion two snake ﬁnd-pst-3sn

‘The lion found two snakes.’

A.4.2. Story 2

Plot: This story features a very strong man who likes to lift things. He sees a pile of

boulders sitting next to a platform and decides to lift the rocks. He lifts the smallest rock

ﬁrst and puts it on the platform. Then he lifts the next bigger rock and puts it on the

platform. He has trouble lifting this one because it is very heavy and he is a little tired from

lifting the ﬁrst rock. He tries to lift the third rock but it is much bigger than the others. It

won’t budge so he gives up.

Puppet’s statements: ha manushya ella baNDe ett-id-a (FALSE)

that man all rock lift-pst-3sm

‘That man lifted all the rocks.’

ha manushya eradu baNDe ett-id-a (TRUE)

that man two rock lift-pst-3sm

‘That man lifted two rocks.’

A.4.3. Story 3

Plot: This story features three frogs hopping through the woods. They come upon a

house and decide to try to jump over it. The ﬁrst frog jumps over the house. The second

frog also jumps over the house. The third frog has webbed feet and so he says he is not

good at jumping over high things. He decides not to try.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 151

Puppet’s statements: ella kappe maney-a meele haar-i-tu (FALSE)

all frog house-gen over jump-pst-3sn

‘All of the frogs jumped over the house.’

eradu kappe maney-a meele haar-i-tu (TRUE)

two frog house-gen over jump-pst-3sn

‘Two frogs jumped over the house.’

References

Barwise, J., & Cooper, R. (1981). Generalized quantiﬁers and natural language. Linguistics and Philosophy, 75,

87–106.

Braine, M., & Rumain, B. (1981). Children’s comprehension of “or”: evidence for a sequence of competencies.

Journal of Experimental Child Psychology, 31,46–70.

Brooks, P., & Braine, M. (1996). What do children know about the universal quantiﬁers all and each? Cognition,

60, 235–268.

Bucci, W. (1978). The interpretation of universal afﬁrmative propositions: a developmental study. Cognition, 6,

55–77.

Bunt, H. (1985). Mass terms and model-theoretic semantics. Cambridge: Cambridge University Press.

ring, D. (1997). The great scope inversion conspiracy. Linguistics and Philosophy, 20, 175–194.

Carston, R. (1998). Informativeness, relevance, and scalar implicature. In R. Carston & S. Uchida (Eds.),

Relevance theory: applications and implications (pp. 179–236). Amsterdam: Benjamins.

Chierchia, G., Crain, S., Guasti, M. T., Gualmini, A., & Meroni, L. (2001). The acquisition of disjunction:

evidence for a grammatical view of scalar implicatures. In A. H. J. Do, L. Dominguez & A. Johansen

(Eds.), Proceedings of the 25th annual Boston University conference on language development (pp. 157–

168). Sommerville, MA: Cascadilla Press.

Chomsky, N. (1957). Syntactic structures. The Hague: Mouton.

Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.

Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris.

Chomsky, N. (1986). Knowledge of language: its nature, origin, and use. New York: Praeger.

Chomsky, N. (1993). A minimalist program for linguistic theory. In K. Hale & S. J. Keyser (Eds.), The view from

building 20 (pp. 1–52). Cambridge, MA: MIT Press.

Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press.

Cooper, R. (1983). Quantiﬁcation and syntactic theory. Dordrecht: Reidel.

Crain, S. (2000). Sense and sense ability in child language. In S. C. Howell, S. A. Fish & T. Keith-Lucas (Eds.),

Proceedings of the 24th annual Boston University conference on language development. Somerville, MA:

Cascadilla Press, pp. 22–44.

Crain, S., & Thornton, R. (1998). Investigations in universal grammar: a guide to research on the acquisition of

syntax and semantics. MIT Press: Cambridge, MA.

Crain, S., Thornton, R., Boster, C., Conway, L., Lillo-Martin, D., & Woodams, E. (1996). Quantiﬁcation without

qualiﬁcation. Language Acquisition, 5 (2), 83–153.

de Boysson-Bardies, B., & Bacri, N. (1977). The interpretation of negative sentences. International Journal of

Psycholinguistics, 4,73–81.

Donaldson, M., & Lloyd, P. (1974). Sentences and situations: children’s judgments of match and mismatch. In F.

Bresson (Ed.), Problemes actuels en psycholinguistique. Paris: Centre National de Recherche Scientiﬁque.

Drozd, K., & Philip, W. (1993). Event quantiﬁcation in preschoolers’ comprehension of negation. In E. V. Clark

(Ed.), Proceedings of the 24th annual child language research forum. Stanford, CA: Center for the Study of

Language and Information, pp. 72–86.

Fiengo, R. (1977). On trace theory. Linguistic Inquiry, 8,35–61.

Fodor, J. A., Bever, T., & Garrett, M. (1974). The psychology of language. New York: McGraw-Hill.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154152

Fodor, J. D. (1982). The mental representation of quantiﬁers. In S. Peters & E. Saarinen (Eds.), Processes, beliefs,

and questions (pp. 129–164). Dordrecht: Reidel.

Frazier, L. (2000). On interpretation: minimal lowering. In M. W. Crocker, M. Pickering & C. Clifton Jr. (Eds.),

Architectures and mechanisms for language processing (pp. 303–323). Cambridge: Cambridge University

Press.

Frazier, L., & Fodor, J. D. (1978). The sausage-machine: a new two-stage parsing model. Cognition, 6, 291–328.

Frazier, L., & Rayner, K. (1982). Making and correcting errors during sentence comprehension: eye movements

in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14, 178–210.

Frege, G. (1892). On sense and nominatum. Reprinted in A. P. Martinich (Ed.), The philosophy of language.

Oxford: Oxford University Press (1990).

Gazdar, G. (1979). Pragmatics. New York: Academic Press.

Grice, P. (1989). Studies in the way of words. Cambridge, MA: Harvard University Press.

Grodzinsky, Y., & Reinhart, T. (1993). The innateness of binding and coreference. Linguistic Inquiry, 24 (1), 69–

101.

Gualmini, A., Crain, S., Meroni, L., Chierchia, G., & Guasti, M. T. (2001). At the semantics/pragmatics interface

in child language. In R. Hastings, B. Jackson & Z. Zvolenszky (Eds.), Proceedings of SALT XIIthaca, NY:

Cornell University.

Harnish, R. (1976). Logical form and implicature. In T. Bever, J. Katz & T. Langedoen (Eds.), An integrated

theory of linguistic ability (pp. 313–391). New York: Crowell.

Hirschberg, J. (1985). A theory of scalar implicature. Doctoral dissertation, University of Pennsylvania, Phila-

delphia.

Horn, L. (1972). On the semantic properties of the logical operators in English. Doctoral dissertation, UCLA

(Distributed by IULC, Indiana University, Bloomington).

Horn, L. R. (1989). A natural history of negation. Chicago, IL: University of Chicago Press.

Hornstein, N. (1984). Logic as grammar. Cambridge, MA: MIT Press.

Huang, C.-T. J. (1982). Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT,

Cambridge, MA.

Inhelder, B., & Piaget, J. (1958). The growth of logical thinking from childhood to adolescence, New York: Basic

Books.

Inhelder, B., & Piaget, J. (1964). The early growth of logic in the child. London: Routledge.

Jackendoff, R. (1972). Semantic interpretation in generative grammar. Cambridge, MA: MIT Press.

Johnson-Laird, P. (1969). On understanding logically complex sentences. Quarterly Journal of Experimental

Psychology, 21,1–13.

Keenan, E., & Stavi, J. (1986). A semantic characterization of natural language determiners. Linguistics and

Philosophy, 9, 253–326.

Kra

mer, I. (2000). Interpreting indeﬁnites: an experimental study of children’s language comprehension.

Doctoral dissertation, MPI Series in Psycholinguistics.

Kroch, A. (1974). The semantics of scope in English. PhD dissertation, MIT, Cambridge, MA.

Kurtzman, H., & MacDonald, M. (1993). Resolution of quantiﬁer scope ambiguities. Cognition, 48, 243–279.

Lasnik, H. (1976). Remarks on Coreference. Linguistic Analysis, 2,1–21.

Levinson, S. (2000). Presumptive meanings. Cambridge, MA: MIT Press.

Lidz, J. (1999). The morphosemantics of object case in Kannada. In S. Bird, A. Carnie, J. D. Haugen & P.

Norquest (Eds.), Proceedings of the 18th west coast conference on formal linguistics (pp. 325–336). Somer-

ville, MA: Cascadilla Press.

May, R. (1977). The grammar of quantiﬁcation. Doctoral dissertation, MIT, Cambridge, MA.

May, R. (1985). Logical form. Cambridge, MA: MIT Press.

McCawley, J. (1970). Where do noun phrases come from? In R. Jacobs & P. Rosenbaum (Eds.), Readings in

English transformational grammar. Waltham, MA: Ginn.

Musolino, J. (in press). Structure and meaning in the acquistion of scope. In L. Frazier, T. Roeper & K. Wexler

(Eds.), Studies in Theoretical Psycholinguistics Acquisition meets Semantics. Kluwer: Academic Press.

Musolino, J., Crain, S., & Thornton, R. (2000). Navigating negative quantiﬁcational space. Linguistics, 38 (1), 1–

32.

Musolino, J., & Lidz, J. (2002). Why children aren’t universally sucessful with quantiﬁcation, Indiana University

and Northwestern University, in preparation.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154 153

Musolino, J., & Lidz, J. (in press). Preschool logic: truth and felicity in the acquisition of quantiﬁcation. In B.

Skarabela, S. Fish, & A. H. J. Do (Eds.), Proceedings of 26th annual Boston University conference on

language development. Sommerville, MA: Cascadilla Press.

Noveck, I. (2001). When children are more logical than adults: experimental investigations of scalar implicature.

Cognition, 78, 165–188.

O’Leary, C., & Crain, S. (1994). Negative polarity items (a positive result), positive polarity items (a negative

result). Paper presented at the 1994 Boston University conference on language development, Boston, MA.

Paris, S. (1973). Comprehension of language connectives and propositional logical relationships. Journal of

Experimental Child Psychology, 16, 278–291.

Philip, W. (1995). Event quantiﬁcation in the acquisition of universal quantiﬁcation. Doctoral dissertation,

University of Massachusetts, Amherst.

Philip, W. (1996). The event quantiﬁcational account of symmetrical interpretation and a denial of implausible

infelicity. In A. Stringfellow, D. Cahana-Amitay, E. Hughes & A. Zukowski (Eds.), Proceedings of the 20th

annual Boston University conference on language development (pp. 564–575). Somerville, MA: Cascadilla

Press.

Philip, W., & Aurelio, S. (1991). Quantiﬁer spreading: pilot study of preschooler’s “every”. In T. Maxﬁeld & B.

Plunkett (Eds.), Papers in the acquisition of WH: proceedings of the Umass Roundtable, May 1990. Amherst,

MA: Graduate Linguistics Students Association, Linguistics Department, UMASS, pp. 267–281.

Pollard, C., & Sag, I. (1994). Head driven phrase structure grammar. Chicago, IL: University of Chicago Press.

Reinhart, T. (1976). The syntactic domain of anaphora. Doctoral dissertation, MIT, Cambridge, MA.

Reinhart, T. (1983). Anaphora and Semantic Interpretation, London. Croon Helm.

Reinhart, T. (1995). Interface strategies (OTS working papers, TL-95-002), Utrecht University.

Reinhart, T. (1997). Quantiﬁer scope: how labor is divided between QR and choice functions. Linguistics and

Philosophy, 20, 335–397.

Roeper, T., & Matthei, E. (1975). On the acquisition of all and some. Papers and reports on child language

development, 9. Stanford, CA: Stanford University Press.

Smith, C. L. (1980). Quantiﬁers and question answering in young children. Journal of Experimental Child

Psychology, 30, 191–205.

Sperber, D., & Wilson, D. (1995). Relevance: communication and cognition. (2nd ed.) Cambridge, MA: Harvard

University Press.

Steedman, M. (1987). Combinatory grammars and parasitic gaps. Natural Language and Linguistic Theory, 5,

403–440.

Tanenhaus, M. K., & Trueswell, J. C. (1995). Sentence comprehension. In J. L. Miller & P. D. Eimas (Eds.),

Handbook of perception and cognition: speech, language, and communication (2nd ed.). (pp. 217–262). San

Diego, CA: Academic Press.

Thornton, R. (1995). Children’s negative questions: a production/comprehension asymmetry. In J. Fuller, H. Han

& D. Parkinson (Eds.), Proceedings of the Eastern States Conference on Linguistics. Ithaca, NY: Cornell

University Press.

Trueswell, J. C., Sekerina, I., Hill, N., & Logrip, M. (1999). The kindergarten-path effect: studying on-line

sentence processing in young children. Cognition, 73,89–134.

van Benthem, J. (1986). Essays in logical semantics. Dordrecht: D. Reidel.

J. Lidz, J. Musolino / Cognition 84 (2002) 113–154154