Perceptual Development

Cognitive Psychology

Cognitive Psychology 47 (2003) 241–275

www.elsevier.com/locate/cogpsych

Use of the mutual exclusivity assumption by young word learners

Ellen M. Markman,* Judith L. Wasow, and Mikkel B. Hansen

Department of Psychology, Stanford University, Jordan Hall, Bldg. 420, Main Quad, Stanford,

CA 94305-2130, USA

Accepted 6 January 2003

Abstract

A critical question about early word learning is whether word learning constraints such as

mutual exclusivity exist and foster early language acquisition. It is well established that chil-

dren will map a novel label to a novel rather than a familiar object. Evidence for the role

of mutual exclusivity in such indirect word learning has been questioned because: (1) it comes

mostly from 2 and 3-year-olds and (2) the findings might be accounted for, not by children

avoiding second labels, but by the novel object which creates a lexical gap children are moti-

vated to fill. Three studies addressed these concerns by having only a familiar object visible.

Fifteen to seventeen and 18–20-month-olds were selected to straddle the vocabulary spurt.

In Study 1, babies saw a familiar object and an opaque bucket as a location to search. Study

2 handed babies the familiar object to play with. Study 3 eliminated an obvious location to

search. On the whole, babies at both ages resisted second labels for objects and, with some

qualifications, tended to search for a better referent for the novel label. Thus mutual exclusiv-

ity is in place before the onset of the naming explosion. The findings demonstrate that lexical

constraints enable babies to learn words even under non-optimal conditions—when speakers

are not clear and referents are not visible. The results are discussed in relation to an alternative

social-pragmatic account.

� 2003 Elsevier Science (USA). All rights reserved.

* Corresponding author. Fax: 1-650-725-5699.

E-mail address: [email protected] (E.M. Markman).

0010-0285/$ - see front matter � 2003 Elsevier Science (USA). All rights reserved.

doi:10.1016/S0010-0285(03)00034-3

242 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

1. Introduction

One of the major accomplishments of early childhood is the acquisition of lan-

guage including the learning of thousands of words (Anglin, 1993). Once children be-

come reasonably proficient speakers, there are several sources of information they can use to narrow down the meaning of a novel term. For children who have ac-

quired enough syntax, grammatical form class cues can limit the hypotheses they

consider, with count nouns more likely to refer to objects, adjectives more likely

to refer to properties, and verbs more likely to refer to actions (Brown, 1957). In fact,

distinctions within syntactic frames lead to more refined hypotheses about a verb�s likely meaning (Fisher, Gleitman, & Gleitman, 1991; Fisher, Hall, Rakowitz, &

Gleitman, 1994). For older children, morphology becomes an important means of

expanding their lexicon (Anglin, 1993). Contrast within a lexical domain (e.g., ‘‘that�s not red it�s maroon’’) could provide further specification of a term�s meaning,

narrowing it down from any property, to a more specific domain, say color (Au,

1990; Au & Laframboise, 1990; Carey & Bartlett, 1978; Heibeck & Markman,

1987). Social-pragmatic cues which reveal information about the communicative

intent of the speaker are another source of information that, along with the non-

linguistic context, can elevate some hypotheses about a word�s meaning over others

(Baldwin, 1995; Tomasello & Akhtar, 1995). Given the importance of word learning,

it makes sense that there would be multiple and often redundant sources of informa- tion that children can use to interpret a novel term (Markman, 1992; Woodward &

Markman, 1998). But consider very young children who may find themselves in sit-

uations where these sources of information are uninformative, ambiguous, or even

non-existent. To account for the success of these young children, several investiga-

tors have postulated that children rely on some constraints on word meanings to

shape their initial hypotheses about the meaning of novel words (e.g., Bloom,

1994; Clark, 1987, 1988; Markman, 1992, 1994; Markman & Hutchinson, 1984;

Markman & Wachtel, 1988; Mervis, 1987; Waxman & Gelman, 1986; Waxman & Kosowski, 1990). These word learning biases serve as default assumptions or good

first guesses as to a word�s meaning (Merriman & Bowman, 1989; Woodward &

Markman, 1991). They help the child solve this inductive problem by narrowing

the hypothesis space which, especially in absence of other sources of information

about a word�s meaning, would be overwhelmingly large.

The focus of this paper is on one of the postulated constraints: mutual exclusivity,

which leads children to prefer that each object have only one category label (Au &

Glusman, 1990; Liittschwager & Markman, 1994; Markman, 1989, 1992; Merriman & Bowman, 1989; Merriman & Schuster, 1991). Mutual exclusivity enables children

to avoid redundant hypotheses (Liittschwager & Markman, 1994), helps children

narrow overextension of words (Clark, 1983, 1987; Merriman & Bowman, 1989),

and allows children to overcome the whole-object assumption thus freeing them to

acquire words for parts, substances, and other properties (Markman & Wachtel,

1988; Soja, Carey, & Spelke, 1991; Woodward, 1992).

Another function of mutual exclusivity and other lexical constraints is that they

enable children to successfully infer the referents of novel terms even under less than

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 243

optimal conditions such as the absence of direct referential cues from the speaker.

That is, there are circumstances in which children can determine what object a novel

word maps onto without the speaker pointing to or otherwise directing the child�s attention to the object. Suppose, for example, a child sees two objects, one of which

is familiar, say a ball, and another novel, say a whisk, and hears someone say ‘‘Can you hand me the whisk?’’ According to mutual exclusivity a child should reject a sec-

ond label for ball and thus infer that ‘‘whisk’’ might refer to the whisk given it is the

only other object around. In fact there are now several powerful demonstrations that

young children will treat a novel term as referring to the novel rather than the famil-

iar object (Golinkoff, Hirsh-Pasek, Bailey, & Wenger, 1992; Markman & Wachtel,

1988; Merriman & Bowman, 1989: Merriman & Schuster, 1991). The role mutual

exclusivity plays in such indirect word learning is the main focus of this paper.

Despite the evidence for the various roles that mutual exclusivity plays in young children�s word learning, a number of concerns have been raised about the argument

that mutual exclusivity serves as a fundamental constraint on word learning. One

widespread objection is that there are exceptions to this principle found in children�s vocabularies (Banigan & Mervis, 1988; Gathercole, 1987, 1989; Mervis, 1987; Mer-

vis, Golinkoff, & Bertrand, 1994a; Nelson, 1988). Violations of the constraint found

in a child�s lexicon, however, are not necessarily evidence against the existence of the

constraint. To claim that children are biased to treat object labels as mutually exclu-

sive is not to claim that they can never learn more than one label for the same object. To determine whether a child uses mutual exclusivity requires examining the initial

hypotheses children consider. Children may well ultimately be successful at overrid-

ing the mutual exclusivity assumption yet nevertheless rely on mutual exclusivity as

an initial guide to interpreting a novel label (see Woodward & Markman, 1991, and

Merriman & Bowman, 1989, for elaborations of this point).

A second objection to the earlier work on mutual exclusivity is that it was consis-

tently demonstrated only for children over 2 1/2 years old yet children become capa-

ble of rapid acquisition of words by roughly 18–19 months of age (Bloom, Lifter, & Broughton, 1985; Corrigan, 1983; Dromi, 1987; Goldfield & Reznick, 1990; Halli-

day, 1975; McShane, 1979; Nelson, 1973). Paralleling the doubt that young children

use mutual exclusivity, is the claim that mutual exclusivity is simply a learned heu-

ristic that older children might find useful, rather than a constraint that is essential

for early word learning (MacWhinney, 1989). If mutual exclusivity is to play a role in

the early rapid acquisition of vocabulary, it needs to be available to children well be-

fore 2 1/2 years. Several investigators have expressed skepticism that mutual exclu-

sivity is available to younger children (Banigan & Mervis, 1988; Evey & Merriman, 1998; MacWhinney, 1989; Merriman & Bowman, 1989; Mervis & Ber-

trand, 1994; Mervis et al., 1994a; Mervis, Johnson, & Mervis, 1994b; Nelson,

1988), basing their doubt in part on diary data and experimental findings that chil-

dren are capable of acquiring second labels for things.

The issues of second label learning and the age at which mutual exclusivity first

occurs were both addressed in a series of studies by Liittschwager and Markman

(1994). The studies were designed to monitor children�s acquisition of new object

labels at the time of the first exposure to the novel words. As a test of whether

244 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

children prefer words to be mutually exclusive, children�s success at learning second

labels for objects was compared to their success at learning first labels. In the first

study, 24-month-olds were taught a novel term either for an object that they could

already name or for one that they could not already name. Contrary to the predic-

tions from mutual exclusivity, these 24-month-olds learned second labels as well as they learned first labels. These findings were surprising not just because they went

counter to the predictions but because they failed to replicate the Banigan and

Mervis (1988) finding that simple labeling of an object with a known first label

was not sufficient for children to acquire the new label. Liittschwager and Markman

explain these unexpected results by suggesting that depending on the information

processing requirements of the task, 24-month-olds may be capable of overriding

the mutual exclusivity assumption. In Banigan and Mervis� (1988) study, children were taught several new terms, while in Liittschwager and Markman�s (1994) study, they were taught only one. This hypothesis was tested in a second study in which 24-

month-olds were taught two rather than one new word—either two new first labels

for two different objects or two new second labels for two different objects. Increas-

ing the information processing demands of the task should make it harder for chil-

dren to override mutual exclusivity and thus harder for them to learn second

compared to first labels. This was exactly what was found. The 24-month-olds read-

ily learned two new first labels for two different objects, but in marked contrast

failed to acquire the second labels. Building on these findings, Liittschwager and Markman reasoned that younger babies with more limited information processing

abilities should be less able to override mutual exclusivity even when acquiring a sin-

gle new word. They tested this in a third study where 16-month-olds were taught a

single new word—either a first label for an object or a second label for an object

whose name they already knew. As predicted, 16-month-olds successfully acquired

first but not second labels.

Thus, mutual exclusivity is used by babies as young as 16 months of age to reject

second labels for things and thus avoid redundant hypotheses. It remains an open question, however, whether babies this young can benefit from the other functions

of mutual exclusivity. One goal of the present work is to address this question with

respect to indirect word learning.

Golinkoff, Mervis, and Hirsh-Pasek (1994) argue that such indirect word learning

will not be achieved by children before the time of the vocabulary spurt. Their frame-

work postulates a novel name-nameless category (N3C) principle, which holds that

novel labels map on to novel objects and makes similar predictions as mutual exclu-

sivity in many situations. On their account N3C is a second-tier principle that be- comes available to children at the time of the vocabulary spurt and not as part of

the first-tier principles available to children at the very beginning of language learn-

ing. To test this, Mervis and Bertrand (1994) presented 16–20-month-olds with dis-

plays of four familiar objects and one unfamiliar object and provided a novel label to

see if they would map the novel label to the unfamiliar object. Following each task

the experimenter provided the children with feedback by directly labeling the unfa-

miliar objects three times. Note that after the first exposure where children infer the

referent, this task becomes a direct teaching task.

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 245

Their results were that children who correctly mapped the label showed evidence

of having entered the vocabulary spurt, producing an average number of 95 words

according to parental reports. Non-mappers produced an average of 45 words. Ac-

cording to Mervis and Bertrand these results indicate that the ability to use N3C in

indirect word learning becomes available only at the time of or after the child has entered the vocabulary spurt. However, this procedure likely yields a conservative

estimate of children�s ability to disambiguate novel labels. As noted by Evey and

Merriman (1998) and Graham, Poulin-Dubois, and Baker (1998), children may fail

at this fast mapping task for a variety of reasons such as lack of processing capacity

to manipulate four familiar and one unfamiliar objects. This explanation gains fur-

ther credibility from another of Mervis and Bertrand�s findings. After the first expo-

sure to the items where children had to infer the referent, children were given

ostensive cues and feedback as to which novel object the label referred. Nevertheless, children still failed to learn. On Mervis and Bertrand�s logic then, we would have to

conclude that children cannot learn words at all prior to the vocabulary spurt. Given

the explicit teaching of the words, not only the fast-mappers but all children should

have learned that the novel words refer to the novel objects.

To reduce the information processing load, Graham et al. (1998) presented chil-

dren in the same age range with an indirect word learning task with three objects

rather than five. They too found an effect of vocabulary size, but in contrast to Mer-

vis and Bertrand�s findings, non-mappers had a mean productive vocabulary of 82 words and fast-mappers a mean productive vocabulary of 237 words. But Graham

et al.�s finding that non-mappers produce an average of 82 words is not compatible

with the generally accepted cut-off of 50 words for the vocabulary spurt. Further-

more, a vocabulary of 82 words in Graham and colleagues� non-mappers is very

close to the vocabulary of 95 words in children considered fast-mappers in Mervis

and Bertrand�s study. Taken together the studies of Mervis and Bertrand and

Graham et al. do not provide a coherent picture of children�s success at indirect word learning at different levels of vocabulary development and varying levels of task difficulty. The current studies attempt to resolve this issue.

Again, the earlier results are overwhelmingly that from about 2 1/2 years on, chil-

dren consistently treat the novel word as referring to the novel object on indirect

word learning tasks. Although this is consistent with the predictions from mutual

exclusivity, there are three other explanations for children�s performance on this

particular task.

Merriman and Bowman (1989) provided an important alternative and very differ-

ent explanation for these results. They point out that these results could be obtained if children were predisposed to fill lexical gaps. One critical difference between the

two hypotheses is that in this context mutual exclusivity and other lexical constraints

provide a backup strategy for children to help them learn a new word even when no

obvious candidate referent is around. In marked contrast, the lexical gap hypothesis

is dependent on the presence of a candidate object to label. The lexical gap hypoth-

esis concerning the indirect word learning situation is derived from a more general

principle that states that in the presence of an object that as yet has no known label,

children are motivated to discover its name (cf. Clark, 1983, 1987). The assumption

246 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

that things have names is related to the development of the ‘‘nominal insight’’ that

McShane (1979) and others have postulated. Many children early in the course of

language acquisition actively request labels for things. Thus, in the indirect word

learning tests of mutual exclusivity children could map the novel word to the novel

object because they have a novel object that they want to find a name for. If children had no reluctance to have second labels for things but simply were motivated to find

first labels for things, then the results of these studies would be the same. Thus the

lexical gap hypothesis might explain the results of the indirect word learning studies

that have been interpreted as evidence in favor of mutual exclusivity. Whether chil-

dren are in fact using mutual exclusivity in these cases is an open question.

One main goal of the present studies was to test whether the lexical gap hypothesis

is a viable alternative explanation for children�s success on indirect word learning

tasks. To distinguish between mutual exclusivity and the lexical gap hypothesis, the indirect word learning procedure was modified such that babies heard a novel

label in the presence of a familiar object but no novel object was visible. If younger

children are guided by mutual exclusivity to reject second labels for objects then they

should avoid selecting the visible familiar object in the presence of a novel label. The

lexical gap explanation is that upon seeing a novel object, children are motivated to

find out what it is called. It could not explain a rejection of second labels in this sit-

uation because no novel object is visible to create a gap in the child�s lexicon. Mutual exclusivity works in conjunction with the whole-object and the taxonomic

assumptions, which state that words refer to whole objects of like kind. Implicit in

these assumptions is the notion that children take words to refer. In the lexical prin-

ciples framework (e.g., Golinkoff et al., 1994) this notion has been made explicit as

the fundamental first-tier principle of reference. If children avoid mapping a second

label to the visible object, they are left with having heard a novel label being used to

refer to something but without an appropriate referent in sight. This, in turn, would

lead children to search for a referent for the label.

In other words, by blocking the immediately obvious choice, mutual exclusivity has left the target of a referential label unfilled and should therefore contribute, in

conjunction with something like Golinkoff et al.�s principle of reference, to children�s active seeking out of appropriate referents. The lexical gap principle is triggered by

the presence of a novel object that creates a gap. Without the object to create the

gap, lexical gap filling does not guide children�s behavior. On the lexical gap hypoth-

esis, children require a nameless object to initiate their mapping of the word. On the

mutual exclusivity account, children do not need an external incentive to map the

novel word—they are motivated by the lexical constraints, in particular mutual exclu- sivity and the principle of reference, to find an appropriate referent.

In the same way, mutual exclusivity appears different from the N3C principle. Ac-

cording to the N3C principle, children map novel labels onto objects for which they

do not yet have a name. The presence of the nameless object is required for the novel

name–nameless category principle to operate. Situations where children hear a novel

label in the presence of nothing but previously named objects are explicitly outside

the scope of N3C (Golinkoff et al., 1992, 1994; Mervis et al., 1994a). N3C further-

more explicitly allows children to make the initial assumption that objects can have

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 247

more than one label (Golinkoff et al., 1992). The modification of the indirect word

learning task in the present studies thus serves to clarify the distinction between mu-

tual exclusivity and the N3C principle.

The third alternative explanation for previous findings from indirect word learn-

ing tasks is that children may be able to determine a speaker�s referential intent through their knowledge of social practices involved in communication and through

pragmatic cues (e.g., Bloom, 1998, 2000; Clark, 1997; Diesendruck & Markson,

2001; Tomasello, 2001). In the case of indirect word learning, children base their rea-

soning on assumptions about speakers’ intent to refer rather than assumptions about

how words refer. In part, children may reason along the lines of, ‘‘If the speaker

meant to refer to the ball, she would have said �ball.’’� This social-pragmatic account

would predict the same pattern of results as mutual exclusivity. Although the present

studies were not designed to distinguish between this explanation and mutual exclu- sivity, we will have more to say about this issue in the discussion.

In sum, the present studies were designed to assess the role of mutual exclusivity

in fostering indirect word learning. In particular we examined word learning that

takes place without explicit referential cues from the speaker and without a novel ob-

ject to serve as a ready referent for the new word. The first goal of this work was to

determine whether this positive function of mutual exclusivity is available to babies

before the time of the naming explosion or whether, as others have argued, it is a

later development. The second goal of this work was to determine whether the use of mutual exclusivity can be differentiated from a propensity to fill lexical gaps. Spe-

cifically this work addressed whether mutual exclusivity leads 15–20-month-olds to

reject second labels for objects and whether it thereby helps motivate them to search

for another referent.

2. Study 1

The purpose of Study 1 was to determine whether babies between 15 and 19

months of age are capable of using mutual exclusivity unaided by a bias to fill lexical

gaps. If babies attempt to preserve the mutual exclusivity of object labels, then that

should motivate them to reject a second label for a known object and help lead them

to search for another possible referent of the novel term.

The procedure was to first briefly familiarize babies with a bucket as a location in

which to search for unseen objects. Then on each trial a baby viewed a bucket and a

familiar object whose label was known to that baby. On Familiar Label trials babies were asked to find the familiar object. On Novel Label trials babies were asked to

find the ‘‘press’’ or some other novel object. On baseline No Label trials, children

were asked to find ‘‘one.’’ Mutual exclusivity should lead children to indicate the fa-

miliar object less and to search more in the Novel Label than in the Familiar Label

condition. Children could be searching less when hearing a familiar label because

their understanding of the familiar label causes them to inhibit their baseline level

of searching. Thus it is important to compare the rate at which children select the

familiar object and search for another object when they hear a novel label relative

248 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

to their baserate of searching without hearing a label. In sum, the predictions for ob-

ject selection are that hearing a novel label should motivate children to reject second

labels for the familiar object and thus indicate the familiar object less in the Novel

Label than in the Familiar Label and No Label conditions. Similarly, the predictions

for search are that babies should search more in the Novel Label condition than in the Familiar Label and No Label conditions.

2.1. Methods

2.1.1. Participants

Thirty-two babies participated in this study. Their parents had responded to ads

in Palo Alto and San Francisco Bay Area community newspapers and parent-di-

rected publications inviting them to participate in the study. Parents were offered a small gift to thank them for their participation. All of the babies were full-term

healthy infants from largely middle-class monolingual English-speaking families.

The babies were divided into younger and older age groups designed to span 18

months—the average age of the vocabulary spurt. Sixteen babies were in the younger

age group, 8 males and 8 females. They ranged in age from 15 months, 10 days to 17

months, 18 days, with a mean age of 16 months, 6 days. Sixteen babies were in the

older age group, 8 males and 8 females. They ranged in age from 18 months, 2 days

to 19 months, 28 days, with a mean age of 18 months, 28 days. Two additional ba- bies were run but dropped from the study, one because of fussiness, and one because

of equipment failure.

2.1.2. Materials

The materials consisted of an opaque plastic bucket, five novel objects and six fa-

miliar objects, plus a ball and a shoe which were used in the familiarization phase.

The six familiar objects for each baby were selected from a possible set of 10 includ-

ing: baby, banana, bottle, car, cat, cup, dog, keys, phone, and spoon. Which familiar toys were selected for a given child depended on which words the child�s parents were most confident that the child understood. As parents were contacted to be scheduled,

the experimenter asked them to judge which of the 10 object labels their baby knew

and to rate their confidence in this judgment as ‘‘sure,’’ ‘‘probably,’’ ‘‘maybe,’’

‘‘probably not,’’ and ‘‘no.’’ From this list, six objects were selected to be included

in the experimental phase of the study. When possible, only object labels that the

parent was sure his or her baby knew were selected. When this was not possible, ob-

ject labels that the parent judged the baby probably knew were included to bring the total to 6. Two of the five novel objects were used only in the familiarization phase: a

flat disk surrounded with suction cups and a half-sphere-shaped noisemaker. The

three novel experimental items were a wonton maker (called a ‘‘crimp’’), a garlic

press (called a ‘‘press’’), and a corn butterer (called a ‘‘gadget’’).

2.1.3. Procedure

The baby was brought to the experimental room by his or her parent. While the

experimenter explained the procedure to the parent, the baby was free to crawl

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 249

around the room and play. Once the experimenter and parent judged that the baby

was comfortable, the baby was placed in a sassy seat at a small table. The parent sat

next to the baby and the experimenter sat across from the baby. Parents were asked

to refrain from initiating any interactions with their babies and to avoid pointing at

or discussing any of the toys. Familiarization trials. To familiarize the baby with the bucket as a possible loca-

tion for objects, each baby was asked to find something in the bucket four times.

None of the toys used in this familiarization phase were used in the experiment prop-

er. For two of the trials the experimenter visibly placed a familiar toy in the bucket,

say a ball, shook the bucket, and asked the child ‘‘Can you find the ball? Where is the

ball?’’ If the child reached for the bucket, the experimenter emptied the toy out of the

bucket, allowing the child to play with it for a few seconds. If the child did not reach

for the bucket, the experimenter shook it again and repeated the question. If the child still did not reach for the bucket, the experimenter placed the bucket in front

of the child and emptied the object out onto the table and allowed the baby to play

with it. For two of the trials, this procedure was repeated but with an object for

which the baby knew no label. For example, the experimenter visibly placed a flat

disk ringed with suction cups in the bucket, shook the bucket and asked the child

‘‘Can you find the gripper? Where is the gripper?’’ Whether the first object in the

bucket was familiar or novel was counterbalanced across children. On the following

trials, the order of familiar and novel objects alternated. On two trials the bucket was placed on the left and on two trials on the right side of the table. Thus, these four

trials provided equal opportunity of seeing familiar and novel objects in the bucket.

Experimental trials. After the brief familiarization, the experimental trials began.

On each of the nine experimental trials, the experimenter placed a visible familiar ob-

ject and a bucket on the table in one of two fixed positions out of reach of the baby—

77 cm from the baby, 47 cm apart (aligned with the arms of the sassy seat). The left/

right position of the bucket was randomly determined with the constraint that the

bucket could be in the same position for not more than five trials out of the nine, and for not more than three consecutive trials. Out of sight of the child, an object

had been placed in the bucket. When the bucket was brought into view, the exper-

imenter shook it to let the child know something was there. The visible object and

the bucket were placed on the table in a stereotyped way: both of them marching

from the edge of the table to the fixed positions. This was done to draw the child�s attention to both the toy and the bucket equally. Once the objects were placed on the

table, the experimenter called the child�s name and tapped the table to get the child�s attention. After the experimenter established eye-contact with the child, she asked one of the three experimental questions. During this time and until the baby made

a selection, the experimenter avoided looking at the toy or bucket and instead looked

at the baby.

In the Familiar Label condition, the experimenter asked the child to find the vis-

ible toy (which was always familiar) saying, for example, ‘‘Can you find the bottle?

Where is the bottle?’’ The familiar objects were called by the label the parent had re-

ported was most commonly used with the baby. So the doll, for example, might be

called ‘‘doll,’’ ‘‘dolly,’’ ‘‘baby doll,’’ or ‘‘baby,’’ depending on what a given child�s

250 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

parent would have said. The experimenter requested the object until the baby clearly

indicated the visible familiar toy or the bucket, or showed some other obvious signs

of searching such as looking on the floor, or was non-responsive after five repeti-

tions. The experimenter then gave the baby both toys to play with. That is, the visible

familiar toy was pushed within reach of the baby and the bucket was emptied in front of the child. The baby was allowed to play with the toys for about 10 or

15 s, then the child was asked to hand the toys over to a puppet. The next trial then

began.

For the baseline No Label condition, the experimenter said ‘‘Can you find one?

Where is one?’’ Again once the baby indicated either the bucket or the visible famil-

iar toy, or seemed to look around for something else, the visible familiar toy was

brought within reach of the baby and the bucket emptied in front of the baby. There

were three familiar trials and three baseline no label trials. For three of these six trials the bucket toy was a familiar toy for which the baby knew a label, and for three it

was an object without a known label.

In the Novel Label condition, the experimenter used a novel term in asking the

child to find an object, for example, ‘‘Can you find the crimp? Where is the crimp?’’

As in the other two conditions, once the baby indicated the visible familiar toy or the

bucket, the visible familiar toy and the bucket toy were placed in front of the baby to

play with. For these novel label trials, the bucket contained a different novel object

for each trial. There were three trials of each of the three conditions. These nine questions were

divided into three blocks of three such that each block contained one question from

each of the three conditions. Within a block of three, the order of the questions was

determined randomly for each participant. The sessions were videotaped with the

baby and the objects in full view.

At the conclusion of the session the parent was instructed on how to complete the

MacArthur Communicative Development Inventory (Fenson et al., 1991), a lan-

guage inventory which assesses early vocabulary acquisition. Parents were requested to mail us the completed inventory.

2.1.4. Coding

The videotapes were coded with the coder blind as to which questions babies had

been asked. This was accomplished by constructing coding tapes that: (1) had all of

the audio information removed, (2) had the segment of each trial that began with eye

contact between the experimenter and the baby and ended just before the experi-

menter revealed the bucket toy, and (3) each coding tape had only one trial from a given baby so that no comparative information across trials was available to infer

what condition was being coded for a given baby.

The coder judged whether the baby indicated the visible toy (by reaching or point-

ing to it), whether the baby indicated the bucket toy (by reaching or pointing to-

wards the bucket), and whether the baby searched the environment (by looking on

the floor, glancing from side to side, scanning the room, etc.).

To assess reliability of the coding categories, a second coder independently coded

one of the coding tapes which comprised 1/9 of the data. The coding was highly

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 251

reliable with agreement across categories ranging from 94 to 97% and averaging 95%

overall.

2.2. Results

2.2.1. Vocabulary measures

Our goal in this study was to study children just around the age of the naming

explosion. We expected many of the 15–17-month-olds to be just on the verge of

the naming explosion while many of the 18–19-month-olds should be in the midst

of it. The data from the MacArthur vocabulary questionnaires confirmed that the

two age groups represented pre- and post-spurt populations. Questionnaires were

returned for 13 children at each age group. The 15–17-month-olds� reported produc-

tive vocabulary ranged from 5 to 160 words with a mean of 46. The 18–19-month- olds� productive vocabulary ranged from 23 to 297 with a mean of 127. The older

babies had significantly larger productive vocabularies than the younger ones,

F ð1; 24Þ ¼ 7:35; p < :05. The mean productive vocabulary scores for the older ba-

bies was significantly greater than 50, a conventional cut-off point for the vocabulary

explosion, tð12Þ ¼ 2:93; p < :05, but the younger babies� mean vocabulary was not

significantly less than 50, tð12Þ < 1. On the other hand, the distribution of vocabu-

lary scores established that many of the younger babies had not quite reached the

naming explosion, while many of the older ones had. Nine of the 13 older babies had vocabularies greater than 50, while only four of the younger babies did,

v2ð1;N ¼ 26Þ ¼ 3:85; p < :05. The first set of analyses was conducted on the number of times that babies indi-

cated the visible familiar toy in the three conditions. Table 1 presents these data con-

verted to mean percent of trials on which babies selected the visible familiar toy. The

data were analyzed by an Age (Old vs. Young)�Condition (Familiar Label, Novel

Label, No Label) ANOVA, with condition as a within-subjects variable. There was a

highly significant effect of condition, F ð2; 60Þ ¼ 18:56; p < :001. There was no main effect of age, F ð1; 30Þ ¼ :2, but there was an Age�Condition interaction,

F ð2; 60Þ ¼ 3:52; p < :05. This interaction is that the older babies picked the visible

familiar toy slightly less than the younger babies in the Novel Label condition and

in the No Label baseline condition but more than the younger babies in the Familiar

Label condition. Planned comparisons were used to test whether babies showed

Table 1

Mean percent of trials in Study 1 on which the visible familiar toy was selected according to age

Age

(months) Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

16 73 33 44 34 73 30

19 90 20 35 31 54 38

252 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

evidence of using mutual exclusivity in this situation. The analyses were conducted

separately for each age group.

One clear prediction from the hypothesis that very young children use mutual ex-

clusivity to narrow their interpretations of novel labels is that children should be less

inclined to indicate a familiar toy as a referent of a novel label compared to the known label. This was borne out for both age groups. The older babies indicated

the familiar toy 2.69 times out of a possible 3 (90%) when asked for the familiar

toy and only 1.06 times out of 3 (35%) when they heard a novel label, paired-

tð15Þ ¼ 7:34; p < :001. Similarly, the younger babies indicated the familiar toy

2.19 times out of 3 (73%) when asked to find familiar toys and only 1.31 times

(44%) when they heard the novel labels, paired-tð15Þ ¼ 2:41; p < :05. A sign test

confirmed these findings. Twenty-five babies picked the visible familiar toy more

in the Familiar Label than in the Novel Label condition compared to only two ba- bies who did the reverse, p < :005.

A second prediction from the hypothesis that very young children use mutual ex-

clusivity is that upon hearing a novel label children should select a familiar object as

its potential referent less often than their baseline level of selecting familiar objects.

This was again true of both age groups, though more clearly for the younger babies.

The older babies indicated the familiar toy 1.06 times (35%) in the Novel Label con-

dition and 1.63 times (54%) in the baseline condition, tð15Þ ¼ 1:95; p ¼ :07. The younger babies selected the familiar toy 1.31 times out of 3 (44%) in the Novel Label condition and 2.19 times (73%) in the No Label baseline condition, paired-

tð15Þ ¼ 3:05; p < :01. These differences were significant by a sign test as well. Overall

14 babies selected the visible familiar toy more in the No Label baseline condition

than in the Novel Label condition and only two babies selected the reverse, p < :005. In sum, both predictions were supported: while looking at a bottle, for example,

and asked to find, e.g., a ‘‘crimp,’’ babies pick the bottle less often than when asked

for a bottle and less often than their baseline level when asked to find ‘‘one.’’ Thus,

as predicted, upon hearing a novel label, 15–19-month-old babies are led to reject a familiar toy as a possible referent of the novel term.

Mutual exclusivity helps babies avoid mapping errors and it can also help lead

them to search for a novel object as a potential referent for the novel label. To de-

termine if this function of mutual exclusivity is available to babies we examined the

number of times babies either indicated the bucket or searched in some other way,

such as scanning the room or leaning over to look on the floor. Although this anal-

ysis is not independent of the first, it is not entirely redundant. That is, babies might

fail to pick the familiar toy but also fail to search for another object. They could just be confused or non-responsive on a given trial and not be credited with either re-

sponse. Conversely babies might sometimes simultaneously indicate both the visible

toy and the bucket.

The mean percent of times babies appeared to search for an object other than the

visible familiar object is presented in Table 2. The number of times babies appeared

to search for an object was analyzed by an Age (Older vs. Younger)�Condition

(Familiar Label, Novel Label, No Label) ANOVA, with condition being a within-

subjects variable. There was a significant main effect of condition, F ð2; 60Þ ¼

Table 2

Mean percent of trials in Study 1 on which babies searched according to age

Age (months) Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

16 21 32 52 32 27 30

19 10 20 60 33 44 40

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 253

18:75; p < :001. There was no main effect of age, F ð1; 30Þ ¼ :35, nor did age interact

with condition, F ð2; 60Þ ¼ 2:18; p ¼ :12. Even though there was no interaction with

age, we analyzed the results separately for each age group to double check whether

the findings were reliable at each age.

The first prediction from the mutual exclusivity hypothesis is that babies should

search for an alternative object as a possible referent more in the Novel Label than

in the Familiar Label condition. This prediction was supported for both the 18–19- month-olds and the 15–17-month-olds. The older babies showed signs of searching

1.81 times out of 3 (60%) in the Novel Label condition and only .31 times (10%)

in the Familiar Label condition, paired-tð15Þ ¼ 6:71; p < :001. The younger babies

showed signs of search 1.56 times (52%) in the Novel Label condition compared to

.62 times (21%) in the Familiar Label condition, paired-tð15Þ ¼ 2:91; p < :05. A sign

test confirmed these findings with 27 babies searching more in the Novel Label than

in the Familiar Label condition with only one baby showing the reverse pattern,

p < :002. The second prediction is that babies should show more signs of searching for a

novel object in the Novel Label condition than in the No Label baseline condition.

This prediction was supported for the younger babies but did not reach significance

for the older. The weaker finding for the older babies is caused not by any drop in

their searching in the Novel Label condition but by their higher baserate preference

for looking in the bucket even in the baseline No Label condition. The 18–19-month-

olds searched a mean of 1.81 times out of 3 (60%) in the Novel Label condition com-

pared to 1.31 (44%) in the No Label condition, paired-tð15Þ ¼ 1:65; p ¼ :12. The 15–17-month-olds searched 1.56 times out of a possible 3 (52%) in the Novel Label

condition compared to .81 times (27%) in the No Label condition, paired-

tð15Þ ¼ 2:82; p < :05. Non-parametric analyses confirmed the prediction: Overall

12 babies searched more in the Novel Label than in the No Label condition, with

four showing the reverse pattern, p < :05, by a sign test.

Although data on vocabulary size was not reported for all the babies, the avail-

able data suggest that the effects hold up both for babies who have not yet entered

the vocabulary spurt and for those who are in the midst of it. Table 3 presents the percentage of trials on which babies selected the visible toy with a known label in

each condition. The babies are divided into two groups: Those with productive vo-

cabularies of less than 50 words and those with vocabularies of 50 or more words,

regardless of age. These data were analyzed by a Vocabulary Level (Less than 50,

50 or more)�Condition (Familiar, Novel, and No Label) ANOVA. As expected

Table 3

Mean percent of trials in Study 1 on which the visible familiar toy was selected according to babies� vo- cabulary level

Vocabulary level Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

Less than 50 words (N ¼ 12) 81 22 47 33 58 32

50 or more words (N ¼ 14) 83 31 29 29 67 37

254 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

condition was highly significant, F ð2; 48Þ ¼ 14:50; p < :0001. Moreover, there was no main effect for vocabulary level, F ð1; 24Þ ¼ :11, nor did vocabulary level interact

with condition, F ð2; 48Þ ¼ 1:44; p ¼ :27. Thus the overall effect of condition can be

generalized to both pre- and post-spurt babies. Individual t tests were used to exam-

ine if the two predictions from mutual exclusivity held up for both pre- and post-

spurt babies. The first prediction compares babies� selection of a familiar toy as a

referent of a novel label compared to the known label for a toy. Pre-spurt babies in-

dicated the familiar toy 2.42 times of a possible 3 (81%) when asked for the familiar

toy and 1.42 times out of 3 (47%) when they heard a novel label, paired- tð11Þ ¼ 3:63; p < :005. Similarly, post-spurt babies indicated the familiar toy 2.50

times (83%) in the Familiar Label condition and .85 times (29%) in the Novel Label

condition, paired-tð13Þ ¼ 3:97; p < :005. Thus, this prediction was confirmed for

both vocabulary groups for these selection data. The second prediction compares ba-

bies� selection of a familiar toy when they hear a novel label with a baseline condition

of selecting a familiar toy when no label is heard. The pre-spurt babies selected the

familiar toy 1.42 times (47%) when they heard a novel label and 1.75 times (58%)

when no label was heard, paired-tð13Þ ¼ 1:17; p ¼ :26. The post-spurt babies se- lected the familiar toy .86 times (29%) when they heard a novel label and two times

(67%) out of three in the No Label condition, paired-tð13Þ ¼ 3:17; p < :01. Thus the predicted difference against baseline was confirmed for the post-spurt babies but not

the pre-spurt babies.

An analogous analysis was carried out on the search measure. Table 4 presents the

percentage of trials on which pre- and post-spurt babies searched, either in the buck-

et or some other location, for each condition. These data were analyzed by a Vocab-

ulary (Less than 50, 50 or more)�Condition (Familiar, Novel, and No Label) ANOVA. Again condition was highly significant, F ð2; 48Þ ¼ 13:75; p < :0001.

Table 4

Mean percent of trials in Study 1 on which babies searched according to vocabulary level

Vocabulary level Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

Less than 50 words (N ¼ 12) 14 22 50 30 42 32

50 or more words (N ¼ 14) 17 31 64 33 31 38

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 255

And again there was no main effect of vocabulary level, F ð1; 24Þ ¼ :07 and no inter-

action of vocabulary level with condition, F ð2; 48Þ ¼ 1:20; p ¼ :31. The pattern of

the individual comparisons by paired t tests paralleled the significance levels reported

above for the data from the selection of the visible toy.

Whether the babies in the study were grouped by age or vocabulary, the predicted difference between the Familiar Label and the Novel Label conditions was upheld

for all groups. The predicted differences against baseline, however, was confirmed

only for younger babies and did not quite reach significance for the older babies. This

contrasts with the results when babies were grouped by vocabulary, where the pre-

diction was upheld for the high vocabulary group but did not reach significance

for the low vocabulary group. At first sight, these findings might suggest that only

young and only post-spurt babies use mutual exclusivity to guide search—an odd

conclusion given that age and vocabulary status are positively, not negatively, correlated. A closer inspection of the data reveals that the data in the Novel Label

condition remains essentially unchanged whether grouped by age or vocabulary

whereas the data in the No Label baseline condition changes direction as a function

of age. Examining babies� behavior by age in the No Label baseline condition, Table

2 shows that older babies as a group prefer to search more in the bucket (44%) than

younger babies (27%). Consequently the older babies, as shown in Table 1, tend to

select the visible toy less (54%) than the younger babies (73%). Recall that the vocab-

ulary data showed that four older children had productive vocabularies of less than 50 words and four younger children had vocabularies of 50 words or more. As the

four older children who are inclined to search are moved to the low vocabulary

group and the five younger children who are less inclined to search are moved to

the high vocabulary group, the baseline levels for searching and selecting the visible

toy in the vocabulary data are pulled in the opposite direction of those in the age

data. In other words, when asked to ‘‘show me one’’ older babies are more interested

in seeing what is in the bucket. We are arguing that this reflects their age and explor-

atory behavior, not their vocabulary acquisition. The primacy of age over vocabu- lary level on babies� search behavior in the No Label baseline condition was

confirmed with a regression analysis. The percentage of trials on which babies

searched on the baseline trials was entered as the dependent variable and vocabulary

and age as independent dichotomous variables. Age significantly predicted search-

ing, b ¼ :46; p < :05, while vocabulary level did not, b ¼ �:30; p ¼ :14. This was

also the case when babies� selection of the visible toy was entered as the dependent

variable. Age was significantly negatively related to selection, b ¼ �:50; p < :05, while vocabulary was unrelated, b ¼ :28; p ¼ :16. Regressions were also performed for the other two dependent variables. Neither age nor vocabulary were significant

predictors in the Familiar label or the Novel Label conditions. Thus, while it may

appear as if the predicted difference between search in the Novel Label condition rel-

ative to baseline is dependent on the size of babies� vocabulary, the data show that it

likely reflects age differences in babies� search behavior in the baseline condition

rather than inherent differences in their use of mutual exclusivity according to vocab-

ulary size. This interpretation will be given further support by a related finding in

Study 3.

256 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

To summarize, this study was designed to determine whether babies around the

average age of the naming explosion are capable of using mutual exclusivity to guide

their interpretation of novel labels. According to mutual exclusivity babies should

reject second labels for objects. In order to avoid having children use an alternative

strategy of filling lexical gaps, they were prevented from seeing any novel objects at the time the label was heard. Babies in this study learned that a bucket was a possible

location in which to find objects. They then viewed a familiar object and an opaque

bucket whose contents were unknown. To take one concrete example, a baby might

see a toy dog and the bucket placed on the table. They were then asked to find an

object in one of three ways, either using the familiar label, ‘‘Can you find the

dog?’’, a novel label, ‘‘Can you find the crimp?’’, or no label ‘‘Can you find one?’’

As predicted under the assumption that babies are rejecting second labels for objects,

babies indicated the visible familiar object less when hearing a novel label than when hearing the object�s familiar label, and less than their baserate of picking the familiar

object. Conversely they searched for something else as a potential referent more

when hearing the novel label than when hearing the familiar label and at least the

younger babies more than their baseline level of searching. Thus babies as young

as 15 months old were found to honor the mutual exclusivity assumption, unaided

by a bias to fill lexical gaps.

Study 2 was designed to probe the limits of when mutual exclusivity will lead ba-

bies to reject second labels for objects and search for other possible referents. In Study 1, a familiar object and a bucket as a potential location for objects were both

placed out of reach of children as the experimenter asked them to find an object. In

Study 2, babies were given a familiar object to play with while the bucket remained

out of reach. This was done to make the familiar object more compelling and more

salient to the babies at the time the novel label was heard. The question was whether

mutual exclusivity is powerful enough (1) to lead children to reject a second label for

an object they are playing with and (2) to interrupt their play to search for a novel

object as a potential referent of the novel label.

3. Study 2

3.1. Methods

3.1.1. Participants

Thirty-two babies participated in this study. Their parents had responded to ads in Palo Alto and San Francisco Bay Area newspapers inviting them to participate in

the study. Parents were given a small gift for their participation. All of the babies

were full term healthy infants from monolingual English-speaking families. The ba-

bies were divided into younger and older age groups designed to span 18 months—

the average age of the vocabulary spurt. Sixteen babies were in the younger age

group, 8 males and 8 females. They ranged in age from 15 months, 5 days to 17

months, 25 days, with a mean age of 16 months, 3 days. Sixteen babies were in

the older age group, 8 males and 8 females. They ranged in age from 18 months,

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 257

1 day to 20 months, 15 days, with a mean age of 19 months, 5 days. Three additional

babies were run but dropped from the study, two because of fussiness, and one be-

cause of equipment failure.

3.1.2. Materials

The materials were identical to that of Study 1, including the familiar and novel

objects and an opaque bucket.

3.1.3. Procedure

The procedure was identical to that of Study 1 with one exception. In Study 1,

both the visible familiar toy and the opaque bucket were out of reach of the baby

at the time the experimenter asked the baby to, e.g., ‘‘find the dog,’’ ‘‘find the crimp,’’

or ‘‘find one.’’ In Study 2, the visible familiar object was given to the child to play with but the bucket was kept out of reach. So when the child was asked to find

an object, they were already engaged in play with the visible familiar toy.

3.1.4. Coding

As in Study 1, coders were blind as to what question a baby had been asked. This

was accomplished by constructing coding tapes according to the procedure used in

Study 1.

Coders judged whether the baby indicated the visible toy, the bucket toy, or other- wise searched the environment (such as looking around the room, on the floor, etc.).

As in Study 1, babies indicated the bucket toy by reaching for it or pointing to it. But

because the babies were playing with the visible toy, they could also hold it up or

touch it as well as pointing to it. Because the possible responses were somewhat more

varied, we had the coder rate her confidence in her decision.

Again, the coding categories were highly reliable. One indication that the coding

was straightforward was the primary coder�s confidence in the judgments made. On a

scale from 1 to 5 where 5 is most confident, the average confidence across categories was 4.7. In particular, average confidence was high (4.6) for judgments about

whether or not babies indicated the toy they were playing with.

A second coder independently coded one of the coding tapes or 1/9 of the data.

Agreement across categories ranged from 94 to 100%, and averaged 97%.

3.2. Results

3.2.1. Vocabulary measures

We received MacArthur vocabulary questionnaires for 12 of the younger babies

and 13 of the older ones. The range in reported productive vocabulary was 6–196

for the 15–17-month-olds, with a mean of 67 words. The range in reported produc-

tive vocabulary was 18–20 to 406 for the 19-month-olds with a mean of 162 words.

Although the mean vocabulary scores were different for the two age groups,

tð17Þ ¼ 2:19; p < :05, the distributions were not different when 50 words was taken

as the cut-off point. Six of the 12 15–17-month-olds had vocabularies of less than 50

words, while four of the 13 18–20-month-olds did.

258 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

The first set of analyses was conducted on the number of times that babies indi-

cated the visible familiar toy in the three conditions. Table 5 presents these data. The

data were analyzed by an Age (Old vs. Young)�Condition (Familiar Label, Novel

Label, No Label) ANOVA, with condition as a within-subjects variable. There was a

highly significant effect of condition, F ð2; 60Þ ¼ 13:12; p < :001. There was no main effect of age, F ð1; 30Þ ¼ 1:82; p ¼ :19, nor did age interact with condition,

F ð2; 60Þ ¼ 1:31; p ¼ :26. Although there was no interaction with age, the results

were analyzed separately for each age group to determine whether the effects were

reliable at each age.

The first analyses tested the prediction from mutual exclusivity that children

should indicate the familiar toy they were playing with less when asked to find some-

thing with a novel label than when the familiar label was used. This was confirmed

for both age groups. The older babies indicated the familiar object 2.75 times (92%) in the Familiar Label condition compared to 1.56 (52%) times in the Novel Label

condition, paired-tð15Þ ¼ 3:88; p < :005. The younger babies indicated the familiar

object 2.69 (90%) times in the Familiar Label condition compared to 2.06 (69%)

times in the Novel Label condition, paired-tð15Þ ¼ 3:10; p < :01. Overall, 23 of

the babies selected the familiar toy more in the Familiar Label condition than the

Novel Label condition, while four showed the reverse pattern, p < :005, by a sign

test.

A second set of analyses examined the prediction that babies should indicate the familiar toy as a referent less in the Novel Label condition than in the No Label

baseline condition. This prediction was confirmed for both age groups combined.

The results were not significant for both ages taken separately, however. The older

babies indicated the familiar toy they had in hand 1.56 (52%) times in the Novel La-

bel condition compared to 2.25 (75%) times in the No Label condition, paired-

tð15Þ ¼ 2:03; p ¼ :06. The younger babies indicated the familiar toy 2.06 (69%)

times in the Novel Label condition and 2.38 (79%) times in the No Label condition,

paired-tð15Þ ¼ 1:43; p ¼ :17. The prediction was confirmed with both age groups combined: babies picked the familiar toy 1.81 (60%) times in the Novel Label con-

dition versus 2.31 (77%) times in the No Label condition, paired-

tð31Þ ¼ 2:49; p < :05, and as reported earlier there was no interaction with age.

Non-parametric analyses confirmed this finding: Overall, 16 babies picked the famil-

iar toy less in the No Label condition than the Novel Label condition, while six

showed the reverse pattern, p ¼ :052 by a sign test.

Table 5

Mean percent of trials in Study 2 on which the visible familiar toy was indicated according to age

Age (months) Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

16 90 16 69 23 79 27

19 92 15 52 32 75 23

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 259

Thus the mutual exclusivity assumption helped 15–20-month-old babies reject a

familiar object as a referent of the novel term even though they were holding or play-

ing with the object at the time the novel label was heard. They indicated the familiar

object less in the Novel Label condition than in either the Familiar Label or the base-

line Novel Label condition. Under these more taxing circumstances, the results were not always as striking as those of Study 1.

Analyses were also conducted on the number of times babies showed signs of

searching for another object, either by reaching to see what was in the bucket or

by looking around the room, etc. As in Study 1, the search data are not independent

of the data on reaching for the visible toy, but they are not entirely redundant as on a

given trial, a child could exhibit both behaviors or neither behavior. Table 6 presents

the percent of time babies searched in each condition.

An Age (Old vs. Young)�Condition (Familiar Label, Novel Label, No Label) ANOVA was used to analyze the search data, with condition as a within-subjects

variable. There was a significant effect of condition, F ð2; 60Þ ¼ 8:58; p < :001. Al-

though there was no significant effect of age, F ð1; 30Þ ¼ 2:35; p ¼ :14, age did inter-

act with condition, F ð2; 60Þ ¼ 10:99; p < :001. As will be verified in the analyses to

be reported next, older babies searched more in the Novel Label condition than in

the other two conditions as predicted, but the younger babies did not.

The first set of planned comparisons examined the prediction from mutual exclu-

sivity, that babies should search more in the Novel Label condition than in the Fa- miliar Label condition. Although this prediction was very clearly supported for the

older babies, it was not for the younger babies. The 18–20-month-olds searched a

mean of .44 (15%) times in the Familiar Label condition but a full 2.06 (69%) times

in the Novel Label condition, paired-tð15Þ ¼ 7:34; p < :001. The younger babies

showed very little explicit search in this study, searching a mean of .81 (27%) times

in the Familiar Label condition and .75 (25%) times in the Novel Label condition,

paired-tð15Þ < 1. Apparently the 15–17-month-olds were not willing to disrupt their

play with the toy they had in hand to search for something else. Non-parametric analyses confirmed this pattern for the older babies: Fifteen of the older babies

searched more in the Novel Label condition than in the Familiar Label condition,

with no babies showing the reverse pattern, p < :005 by a sign test.

The second set of analyses examined the prediction that babies should search

more in the Novel Label condition than in the No Label condition. Here again,

the prediction was clearly supported for the older but not the younger babies. The

younger babies searched .75 (25%) times in the Novel Label condition and 1.06

Table 6

Mean percent of trials in Study 2 on which babies searched according to age

Age (months) Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

16 27 28 25 31 35 31

19 15 17 69 28 35 31

260 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

(35%) times in the No Label baseline condition, paired-tð15Þ ¼ 1:04; p ¼ :32. Again

the babies may have been unwilling to stop playing with the toy in hand long enough

to search. In contrast, the older babies searched 2.06 (69%) times in the Novel Label

condition compared to 1.00 (33%) times in the No Label condition, paired-

tð15Þ ¼ 3:78; p < :005. Eleven of the older babies searched more in the Novel Label than in the No Label condition with only one baby showing the reverse pattern,

p < :05, by a sign test.

Another way of analyzing the data is to look at children�s performance as a func-

tion of their vocabulary level instead of age. As in Study 1, babies were divided into

two groups—those with productive vocabularies of less than 50 words (n ¼ 11) and

those with 50 or more words (n ¼ 14). The percent of times babies selected the visible

toy is presented in Table 7 as a function of vocabulary level and condition. The data

were analyzed by a Vocabulary (Less than 50, 50 or more)�Condition (Familiar, Novel, and No Label) ANOVA, with condition a within-subjects variable. There

was a strong effect of condition, F ð2; 46Þ ¼ 11:14; p < :001, but no effect of vocab-

ulary level, F ð1; 23Þ ¼ :03, nor any interaction of vocabulary with condition,

F ð2; 14Þ ¼ :30. Thus the overall condition differences found generalized across both

high and low vocabulary levels.

The search data were also analyzed dividing the babies by vocabulary level instead

of age. Table 8 shows how often (converted to percentages) babies searched as a

function of their vocabulary level and condition. A Vocabulary level (Less than 50, 50 or more)�Condition (Familiar, Novel, and No Label) ANOVA with condi-

tion as a within-subjects variable was used to analyze the mean search scores. There

was a significant effect of condition, F ð2; 46Þ ¼ 5:59; p < :01, but no effect of vocab-

ulary level, F ð1; 23Þ ¼ :002, nor any interaction of vocabulary and condition,

Table 7

Mean percent of trials in Study 2 on which babies indicated the visible familiar toy according to vocabu-

lary level

Vocabulary level Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

Less than 50 words (N ¼ 11) 77 17 61 20 77 22

50 or more words (N ¼ 14) 93 14 55 33 74 30

Table 8

Mean percent of trials in Study 2 on which babies searched according to vocabulary level

Vocabulary level Condition

Familiar Label Novel Label No Label

Mean SD Mean SD Mean SD

Less than 50 words (N ¼ 11) 27 20 49 38 39 20

50 or more words (N ¼ 14) 19 17 52 36 43 38

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 261

F ð2; 46Þ ¼ :32. Thus the overall effect of condition generalized across babies of both

high and low vocabularies. We do not present further analyses of the vocabulary

data past the omnibus ANOVA. Evidence from Study 1 showed that grouping the

data by vocabulary level is influenced by babies� tendency to search in the bucket

at different age levels. Babies� behavior in Study 2 is further influenced by their will- ingness to ignore the visible object in their hands. Since age and vocabulary are im-

perfectly correlated, these factors make it difficult to discern the influence of age vs.

vocabulary when the data is grouped by vocabulary. The number of participants in

the study does not allow dividing a given age group into subgroups of high and low

vocabulary without loss of statistical power.

The failure of the younger babies to search in this study may be a function of the

task demands of inhibition as well as their lexical principles. In Study 1 with both the

visible toy and the bucket toy out of reach, the younger babies requested the bucket toy as predicted on the mutual exclusivity hypothesis. Hearing a novel label caused

these 15–17-month-olds to reject a familiar toy as an appropriate referent and search

for another potentially more appropriate referent for the new word. When babies

were given the visible toy to play with in Study 2, only the 18–20-month-olds

searched as predicted upon hearing a novel label relative to both a familiar label

and no label. The 15–17-month-olds did not search for another referent in either

of those cases.

To summarize, both younger and older babies tended to perform as predicted when their acceptance of a label was measured by how often they indicated the vis-

ible toy upon hearing a novel label relative to a familiar label. The predicted differ-

ence relative to the baseline No Label condition was obtained only for the older

children. Thus even in this more difficult situation where the salience of the toy with

a known label was heightened by allowing children to play with it, children used mu-

tual exclusivity to guide their interpretation of a novel term to some extent. But in

this situation only the 18–20-month-olds searched for another referent. One likely

explanation for the discrepancy between the selection and search measures for the 15–17-month-olds is that to search for another potential referent would have re-

quired a more sustained interruption or inhibition of their play and, although the

older babies accomplished this, the younger babies were not sufficiently motivated

or able to do so.

4. Study 3

Study 3 further explored the limits of mutual exclusivity to motivate search for a

referent by removing the obvious place for babies to search. Studies 1 and 2 have

established that 15–20-month-olds can use mutual exclusivity to reject second labels

for objects and, at least under some circumstances, to search for an appropriate ref-

erent for a novel word. In both of these studies a bucket was made available as a

potential container for another object. In naturally occurring situations sometimes

there will be such an obvious location to search, but sometimes not. In this next

study, we explored whether babies around the age of the naming explosion would

262 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

be motivated by mutual exclusivity to reject a second label for an object and search

for a referent for a novel term even when there was no particularly obvious place to

search. The procedure was similar to that of Studies 1 and 2 except that no bucket

was used and no novel objects were ever presented to the children. The main ques-

tion was whether babies would show spontaneous search of the room.

4.1. Methods

4.1.1. Participants

The participants in this study were 36 healthy, full-term, and monolingual

English-speaking babies whose parents responded to ads in local newspapers and

parent-directed publications. Parents were offered a small gift for their participation.

The younger age group consisted of 18 babies ranging in age from 15 months 7 days to 15 months 28 days with a mean of 15 months 18 days. The older group consisted

of 18 babies ranging in age from 18 months 1 day to 18 months 25 days with a mean

of 18 months 13 days.

4.1.2. Materials

The materials consisted of 12 objects that are among the most likely objects to

have known names by very young children: baby, dog, cat bottle, duck, car, keys,

book, spoon, shoe, and banana. As in Studies 1 and 2, before the start of the study, each parent was asked to rate whether their baby understood each of these words.

For each baby, the experimenter selected three objects representing three of the

words that the baby was most familiar with. An attempt was made to vary the items

as much as possible across babies.

4.1.3. Procedure

Throughout the procedure babies were seated in a sassy seat at the table and their

parent was seated off to the side. The purpose of this study is to determine if babies will search for an unseen object

as the referent of a novel term to avoid accepting a second label for an object with a

well-known name. Once a baby scans the room, it will become obvious that there are

no other candidate objects around as potential referents for a novel term which may

influence their scanning on subsequent trials. For this reason, we had to severely lim-

it the number of trials each baby received. As in the earlier studies we had three kinds

of questions: Familiar Label, Novel Label, and No Label baseline, but in this study

babies heard only one question of each type. For each of the questions the experimenter first gave the baby a familiar object

with a well-known label and allowed the baby to play with it for a brief time. The

experimenter then retrieved the toy (sometimes having the baby give it to a puppet)

and placed it on the table in a fixed position out of reach of the baby. The experi-

menter then tapped on the table, called the baby�s name, and once she established

eye-contact with the baby she asked one of the three questions.

In the Familiar Label condition, the experimenter used the correct label for the

toy, repeating the question several times. If a spoon was placed on the table, the

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 263

experimenter would say ‘‘Can you show me the spoon? Where�s the spoon? Find the

spoon.’’

In the Novel Label condition a novel term was used in the request, but otherwise

the procedure was identical to that in the Familiar Label condition. For example, a

baby might see a shoe placed on the table out of reach and then be asked, ‘‘Can you show me the toma? Where�s the toma? Find the toma.’’

In the No Label condition, designed to assess baseline levels of searching, the re-

quest was made without using any labels. Otherwise the condition was identical to

the other two conditions. A baby might see a banana on the table and be asked

‘‘Can you show me one? Where is one? Find one.’’

The order of presentation of the Familiar Label, Novel Label, and No Label ques-

tions was counterbalanced across children such that each question occurred in the

first, second, and third position equally often. As in Studies 1 and 2, parents were asked to complete the MacArthur Communi-

cative Development Inventory (Fenson et al., 1991) and mail it back to us within the

next few days.

4.1.4. Coding

The videotaped responses of babies were coded without the soundtrack so that the

coder had no knowledge of what had been requested. Coding began from the time

the experimenter requested the object (which was determined in a preliminary coding of the tapes with the soundtrack on). The coder judged whether the baby indicated

the visible toy (by reaching or pointing), whether the baby searched the environment

(by looking around the room, looking on the floor, looking from side to side, etc.), or

whether the baby made some other response such as prolonged looking at the exper-

imenter. After each judgment the coder was asked to rate her confidence on a 5-point

scale where 5 is most confident.

The coding system was highly reliable. First, the primary coder was very confident

of the judgments made. Overall categories the mean confidence in the judgments was 4.87 on a 5-point scale. For the search category alone the mean confidence was 4.89

indicating that babies made quite identifiable search responses that were readily

coded. A second coder independently judged 33% of the data. The coders agreed

on 95% of the judgments.

4.2. Results

4.2.1. Vocabulary measures

As expected the 15 and 18-month-olds straddled both sides of the naming explo-

sion. Fifteen of the 18 parents of 15-month-olds completed the MacArthur vocabu-

lary questionnaire. The 15-month-olds productive vocabulary ranged from 0 to 62

words, with a mean of 24.1 which was significantly less than 50, tð14Þ ¼ 5:19; p < :001. Fourteen parents of the 18-month-olds completed the questionnaires.

The reported productive vocabulary of the 18-month-olds ranged from 20 to 385

with a mean of 121.4 which is significantly greater than 50, tð13Þ ¼ 2:30; p < :05. The 15-month-olds had a significantly smaller vocabulary than the 18-month-olds,

264 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

tð13Þ ¼ 3:07; p < :01. The distribution of vocabulary scores less than 50 versus 50 or

more differed significantly by age. Thirteen of the 15 15-month-olds had vocabularies

of less than 50, while only four of the 14 18-month-olds did, v2ð1;N ¼ 29Þ ¼ 10:07; p < :01.

In this study, the only object in view at the time babies were asked to retrieve something was a familiar object with a well known name. In particular, there was

no bucket to serve as a potential location for a novel object. At one point, the exper-

imenter used a novel label in making the request, e.g., ‘‘Can you show me the toma?’’

The main question the study addressed was whether the preference for keeping cat-

egory terms mutually exclusive would motivate babies to search their environment

for a candidate object to label.

The two predictions from the hypothesis that babies are using mutual exclusivity

here is that: (1) they should search the environment for a novel object more when they hear a novel word used in the request than when they hear the familiar label

for the visible object and (2) more than when they are simply asked to find ‘‘one.’’

The percentages of children at each age who searched in each of the conditions is

presented in Table 9. Because there was only one binary response per condition

for each child, the data were analyzed with a categorical repeated measures ANOVA

using the SAS CATMOD procedure followed by non-parametric planned compari-

sons. An Age (Older vs. Younger)�Condition (Familiar Label, Novel Label, and

No Label) categorical repeated measures ANOVA with condition as a repeated mea- sure was conducted to examine whether there were differences in the frequency with

which babies searched across conditions. There was a significant main effect of con-

dition, v2ð2;N ¼ 36Þ ¼ 21:77; p < :0001. There was no main effect of age, v2ð1;N ¼ 36Þ ¼ :13; p ¼ :72, nor did age interact with condition, v2ð2;N ¼ 36Þ ¼ 2:62; p ¼ :27. We also analyzed the results separately for each of the predictions from

mutual exclusivity.

The first prediction from the mutual exclusivity hypothesis is that babies should

search for an alternative object as a possible referent more in the Novel Label than in the Familiar Label condition. This prediction was supported for both the 18-

month-olds and the 15-month-olds. In the older group, 15 babies (83%) showed signs

of searching in the Novel Label condition compared with 9 babies (50%) in Familiar

Label condition. The means of each condition differed significantly by a Cochran

test, Q with 1 df ¼ 6:00; p < :05. In the younger group, 16 babies (89%) showed

signs of searching in the Novel Label condition compared to 10 babies (56%) in

the Familiar Label condition, Q with 1 df ¼ 6:00; p < :05.

Table 9

Percentages of babies in Study 3 who searched according to age

Age (months) Condition

Familiar Label Novel Label No Label

15 56 89 50

18 50 83 72

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 265

Babies could be searching more in the Novel Label condition than in the Familiar

Label condition because they are rejecting second labels for the objects with well-

known labels and searching for another possible referent. Alternatively, they could

be searching less in the Familiar condition because the presence of the familiar label

causes them to inhibit their baseline level of searching. Thus, the prediction is that babies should search more in the Novel Label condition than their baseline level

of searching in the No Label baseline condition. This prediction was clearly sup-

ported for the younger babies but not the older babies. Of the 18-month-olds, 15 ba-

bies (83%) searched in the Novel Label condition compared to 13 babies (72%) in the

No Label condition, Q with 1 df ¼ :67; p ¼ :41. Of the 15-month-olds, 16 babies

(89%) searched in the Novel Label condition compared to nine babies (50%) in

the No Label condition, Q with 1 df ¼ 7:00; p < :01. Vocabulary data were reported for 29 babies and indicated that the predictions

were generally held up both for babies who had not yet entered the vocabulary spurt

and those who were in the middle of it. The babies were divided into two groups

based on their productive vocabulary, with 50 words as the cut-off. Table 10 shows

the percentages of children at each vocabulary level who searched in each of the con-

ditions. Whether babies searched was analyzed with a Vocabulary (Less than 50, 50

or more)�Condition (Familiar Label, Novel Label, and No Label) categorical re-

peated measures ANOVA with condition as a repeated measure. There was a signif-

icant main effect of condition, v2ð2;N ¼ 29Þ ¼ 18:93; p < :0001, with no main effect of vocabulary, v2ð1;N ¼ 29Þ ¼ :74; p ¼ :39, and no interaction between vocabulary

and condition, v2ð2;N ¼ 29Þ ¼ :54; p ¼ :76. Planned comparisons showed that the

predicted difference for the Familiar and Novel Label conditions was found at both

vocabulary levels. For the post-spurt babies, 10 babies (83%) searched in Novel La-

bel condition compared with 5 babies (42%) in the Familiar label condition, Q with 1

df ¼ 6:00; p < :05. For the pre-spurt babies, 16 babies (94%) searched in the Novel

Label condition compared with 10 babies (59%) in the Familiar Label condition, Q

with 1 df ¼ 6:00; p < :05. The predicted difference between the Novel Label and baseline No Label conditions was found for the pre-spurt but not the post-spurt ba-

bies. For the post-spurt babies, 10 babies (83%) searched in the Novel Label condi-

tion compared with 8 babies (67%) in the No Label condition, Q with 1

df ¼ :67; p ¼ :41. For the pre-spurt babies, 16 babies (94%) searched in the Novel

Label condition compared with 11 babies (65%) in the No Label condition, Q with

1 df ¼ 5:00; p < :05. The individual comparisons when the data were divided by vo-

cabulary thus paralleled those when the data were divided by age.

Table 10

Percentages of babies in Study 3 who searched according to vocabulary level

Vocabulary level Condition

Familiar Label Novel Label No Label

Less than 50 words (N ¼ 12) 59 94 65

50 or more words (N ¼ 17) 42 83 67

266 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

Once again the prediction that babies viewing a familiar object should search

more when hearing a novel label than a familiar label was confirmed for both 15-

month-olds and 18-month-olds regardless of vocabulary level. The prediction that

babies should search more when hearing a novel label relative to no label, however,

was confirmed only for the younger babies, and only for the pre-spurt babies. As in Study 1, the weaker finding for the older babies is not caused by a drop in their

searching in the Novel Label condition but by their higher baserate preference for

looking at something else than the object before them in the No Label condition.

It seems unlikely that 15 but not 18-month-olds use mutual exclusivity in this situa-

tion. A more plausible explanation is that 18-month-olds take advantage of the op-

portunity to explore their surroundings when the baseline condition does not make a

specific demand of them.

When young children hear a novel label while attending to an object whose label is known, they tend to reject the term as a possible second label and instead search

for another more appropriate referent. Even when only one candidate object is in

view and when there is no obvious place to search, babies as young as 15 months

spontaneously looked around the room as if searching for an appropriate referent

for the novel word.

5. Discussion

By 2 1/2 years of age, children are highly consistent in mapping a novel object label

to an object whose name is as yet unknown rather than to an object whose name is al-

ready known. When shown, for example, a spoon and a whisk and asked to hand the

speaker the whisk, young preschoolers almost always hand over the whisk. This task

was devised byMarkman andWachtel (1988) to reveal children�s use of mutual exclu-

sivity. On their account, mutual exclusivity should lead children to reject a second label

for an objectwith a known label and thus lead them to infer that that the term applies to the object on display whose label is not yet known. Although this finding is highly rep-

licable (e.g., Golinkoff et al., 1992;Merriman&Bowman, 1989;Merriman&Schuster,

1991), its interpretation and its significance has been called into question. This work

addressed two of the main challenges about this role of mutual exclusivity.

The first concern raised about the earlier findings is that the children studied were

old enough that they could have noticed thatmany terms in their vocabulary happened

to be mutually exclusive and thus used mutual exclusivity as a convenient heuristic for

acquiring some subsequent vocabulary. Yet mutual exclusivity and other word learn- ing constraints have been postulated as being essential for helping children narrow

down the otherwise intractably large hypothesis space for mapping words to concepts.

The hypothesis tested here is whether mutual exclusivity is available to children before

the naming explosion or vocabulary spurt which is seen on the average at 18–19

months. Vocabulary data were available for a majority of the children and the data

were analyzed according to vocabulary level in addition to age.

Second is the question of whether children�s consistent performance in this indi-

rect learning situation reflects mutual exclusivity per se or whether it can be

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 267

accounted for by children�s attempt to fill lexical gaps as Merriman and Bowman

(1989) proposed. On their hypothesis, a novel object must be present to create a lex-

ical gap and motivate the word learning. Mutual exclusivity instead can function

even in the absence of an obvious candidate object and can motivate search for a ref-

erent that is out of sight. A lexical gap is created when babies come across something they cannot label. If

there is no unfamiliar object around, then no lexical gap is created. The key proce-

dural change in these studies was to eliminate the novel object from the display in

front of the babies to rule out the possibility that children could be attempting to fill

a lexical gap. In Study 1 an opaque bucket and an object with a well-known label

(e.g., a spoon) were placed on the table in front of the baby. The experimenter shook

the bucket so the baby could hear that something was inside it, but could not see

what it was. In Study 2 the familiar object was handed to the baby to play with. In Study 3 the bucket as an obvious location to search was removed.

Across the three studies, two kinds of measures and two kinds of comparisons

generated several different predictions from mutual exclusivity. In Studies 1 and 2,

the two measures were: (1) children�s selection of the visible object and (2) their ten-

dency to search. We examined children�s selections of objects and search when they

viewed, e.g., a spoon and a bucket on the table, and asked them either, ‘‘Can you

show me a spoon?’’, ‘‘Can you show me a gadget?’’, or ‘‘Can you show me one?’’

The two comparisons looked at children�s selection or search in (1) the Novel Label condition relative to the Familiar Label condition, and (2) the Novel Label condition

relative to the No Label condition.

The first prediction from mutual exclusivity is that if children are avoiding second

labels for objects, they should indicate the spoon less when asked for a ‘‘gadget’’ than

when asked for ‘‘a spoon.’’ In both Studies 1 and 2, young and old babies and pre-

and post-spurt babies alike consistently avoided picking the spoon in the Novel La-

bel condition relative to the Familiar label condition. This difference is predicted by

mutual exclusivity and although it demonstrates children�s resistance to second la- bels, it also could be obtained if children preferred to see what was in the bucket

and suppressed that tendency when asked for a well-known object. To evaluate this,

we compared children�s tendency to pick, e.g., the spoon when asked for a ‘‘gadget’’

relative to when asked for ‘‘one.’’ In Study 1 both old and young babies selected the

spoon less when asked for ‘‘a gadget’’ than when asked for ‘‘one,’’ but the effect was

slightly stronger for the younger babies. Study 2 was designed to push the limits of

children�s use of mutual exclusivity by handing them the toy to play with which made

the toy more salient and interesting and thereby added strong inhibitory require- ments to revealing mutual exclusivity. Although there was no interaction with age

and both age groups combined indicated the spoon they were playing with less when

asked for ‘‘a gadget’’ compared to ‘‘one,’’ the results did not reach significance for

the age groups separately and appear stronger for the older children, who likely have

more inhibitory control. We take this as reflecting developmental differences in the

demands of the situation in Study 2 with inhibitory requirements affecting the re-

sults. So with this qualification, the selection data from Studies 1 and 2 support both

predictions from mutual exclusivity.

268 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

The second measure across all studies was whether babies searched for an ap-

propriate referent for the novel label. The argument here is that if babies have re-

jected a second label, they are then left with a word without a referent and,

according to the principle of reference, they might be motivated to find an appro-

priate object. Again we evaluated babies� search both when hearing a familiar label and when hearing no label. As predicted, in Study 1 babies at both ages and both

vocabulary levels searched more when asked for ‘‘a gadget’’ than when asked for

‘‘a spoon.’’ In Study 2, while actually playing with the spoon, only the older babies

searched more, but younger babies in this study hardly searched at all in any con-

dition—they remained interested in the object in hand. Here the lexical constraints

are pitted against this inhibitory requirement and younger babies do not interrupt

their play to search for an appropriate referent. It is striking, however, that the

older babies do. In Study 3, when the bucket was removed and there was no ob- vious place to search, babies at both age groups and both vocabulary levels

searched more in the Novel label than in the Familiar label condition. So across

all three studies this search prediction was confirmed for all babies at both ages

and vocabulary levels, except for in Study 2 when the younger babies playing with

a toy hardly search at all. The second prediction about babies� search behavior is

that babies should search more when asked for ‘‘a gadget’’ than when asked ‘‘for

one.’’ In Study 2, as just mentioned, younger babies do not interrupt their play to

search, but the older babies did and confirmed this prediction. In Studies 1 and 3 it is the younger babies that confirm the second prediction more strongly. This is par-

ticularly striking in Study 3 where younger babies� spontaneously search the room

despite there being an obvious object in front of them and no obvious location for

another object. The reason younger babies show the effect more than older babies

is that the baseline No Label condition revealed developmental differences in chil-

dren�s exploratory behavior. When asked for ‘‘one’’ and thus free to do what they

liked, younger babies were more likely to reach for the salient visible object

whereas older babies were more interested in exploring the bucket or searching the room. In sum, although the effects were modulated somewhat by the explor-

atory and inhibitory demands of the tasks, across all three studies there is clear ev-

idence that both younger and older babies search more when hearing a novel label

than in the baseline No Label condition.

Taken together, the results rule out the lexical gap hypothesis and also contradict

the findings of Mervis and Bertrand (1994) and Graham et al. (1998) who found that

1 1/2 year-old babies with limited vocabularies were not able to solve similar indirect

word learning tasks. Their tasks most resemble the main selection prediction from mutual exclusivity that babies will avoid selecting the familiar object upon hearing

a novel label relative to hearing a familiar label. Study 1 showed without any qual-

ifications that babies from 15 to 19 months of age used this version of mutual exclu-

sivity regardless of age and vocabulary level. One key difference between the studies

is the number of objects that babies can choose as referents for the novel label. The

current studies involved a two-choice task compared to the five-choice task of Mervis

and Bertrand (1994) and three-choice task of Graham et al. (1998). This suggests

that their tasks may have been too demanding for children, a notion that is

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 269

supported by the failure of Mervis and Bertrand�s babies to learn even from direct

labeling with corrective feedback on the same task.

Another lexical constraint, the N3C principle, has been proposed to account for

the findings from indirect word learning studies. The N3C principle is predicted to

become available to children only at or beyond the time of the vocabulary spurt and it requires that a novel object be present at the time where a novel label is heard.

Thus it would be hard for N3C to account for our findings. In Study 3 only a single

familiar object was present and babies still searched. Golinkoff et al.�s framework

specifies that children would either assume that the novel label is a second label

for the familiar object and not search, or rely on pragmatic cues of the type posited

by the social-pragmatic approach to provide a motivation for searching elsewhere

(Mervis et al., 1994a).

The social-pragmatic view presents an alternative account of word learning to that of the lexical constraints approach. On that view, lexical principles are not

needed for children to infer the referents of words. Instead, the findings from indirect

word learning studies can be explained by children�s motivation to figure out the in-

tended referent of the speaker and their expectation that the speaker will be a collab-

orative communicator. Children would reason that if the speaker had wanted the

ball, he or she would have asked for the ball by using its conventional name (e.g.,

Bloom, 1998, 2000; Clark, 1990, 1997; Diesendruck & Markson, 2001; Gathercole,

1989; Tomasello, 2001). Although children use a range of sources of information to infer a speaker�s intentions to refer to an object, it is notable that indirect word

learning studies render directly observable pragmatic cues such as eye gaze and

pointing uninformative. Indeed, this ambiguity is a motivating force behind studies

of mutual exclusivity. Thus, it is the child who must identify the relevant knowledge

that will disambiguate the referent of the label and use this to choose among the

possible referents. This knowledge may consist of lexical principles or knowledge

of conventional communicative practices as well as other types of knowledge. As

the current studies were not designed to distinguish between the mutual exclusivity and social-pragmatic accounts, both of these accounts remain theoretically possible.

However, we know of no empirical studies from a social-pragmatic point of view

that demonstrate that 15–18-month-old children are able to recruit knowledge of

conventional communicative practices in cases where direct cues are uninformative,

and we know of only one study with older children. Before turning to that study, we

review the available evidence from studies that address young children�s abilities to infer the intended referent when a speaker uses a novel label.

Well before word learning beings, infants are able to attribute intention to entities that have features in common with animate beings (Johnson, 2000; Woodward,

1998, 1999; Woodward & Sommerville, 2000). Later, in the months preceding the

vocabulary explosion, infants show understanding of the pragmatics involved when

a speaker labels objects. Baldwin (1991, 1993; Baldwin, Markman, Bill, Desjardin, &

Irwin, 1996) has explored whether 1–2-year-olds require some evidence of intent to

refer to an object before learning a novel word. The alternative possibility would be

that infants hear a label and associate it with whatever object that they are currently

engaged with. In a series of ‘‘discrepant labeling’’ tasks, a speaker uttered a novel

270 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

label as the child was looking at a novel toy. The speaker, however, was looking into

an opaque bucket and not at the visible toy (Baldwin, 1991). Even in the case where

there was not a joint focus of attention, 18-month-olds learned that the word re-

ferred to the object that had been in the bucket and even 16-months-olds avoided

making mapping errors. Thus before 18 months of age children are aware that conventional expressions of intent to refer must be present before mapping labels

to novel objects.

Not all interaction with an object signals intent to refer. For example, adults may

accidentally handle objects that they did not mean to refer to or may attempt to refer

to objects that are not accessible. Older infants, between the ages of 18 and 24

months, are able to track speakers� communicative intent in a range of situations

that seem to require a more subtle understanding of indicators of intentional states.

In a representative study, Tomasello and Barton (1994) presented 24-month-olds with a series of objects that were hidden in buckets. The speaker announced that

her intention was to find ‘‘a toma,’’ yet she rejected the first two objects that she

found with a scowl and then excitedly found the target object in a third bucket. In

a control condition the speaker found the target on the first try. Children in both

conditions learned the novel word for the target object, illustrating that children

are aware that the achievement of a goal to refer was a condition for attaching

the label to the target. Children at this age are also sensitive to frustrated goals when

naming objects. In a similar study (Akhtar & Tomasello, 1996), 24-month-olds were shown a series of objects which were then hidden in different locations. The speaker

announced that she wanted to find ‘‘a toma’’ and turned to a toy barn as her first

choice of a location to search. However, the barn doors turned out to be locked.

The experimenter expressed her disappointment and explained to the child that

she was unable to open the doors. She then progressed to examine and handle ob-

jects in the other locations. In a control condition, the experimenter immediately

found the target object behind unlocked doors. Follow-up comprehension tests

showed that children learned the intended referent of the novel label equally well in both conditions, even when the referent remained locked in the barn. These find-

ings and those of Tomasello and Barton (1994) have been replicated with 18-month-

olds (Tomasello, Strosberg, & Akhtar, 1996).

The results of these studies convincingly demonstrate the abilities of very young

children to infer communicative intent. Not only are children sensitive to pragmatic

cues such as pointing and handling but are also able to update such cues by paying

attention to expressions of disappointment or frustrated goals. It is notable, how-

ever, that none of the studies involve inferring the referent of a label under circum- stances where the directly observable pragmatic cues are uninformative. On the

contrary, for every object that the child might consider as a referent, the adult speak-

er gives a clear cue that she is now attending to it, and then marks whether to reject

or accept the object for further consideration with explicit statements about an error,

such as ‘‘whoops,’’ or emotional cues, such as excitement or disappointment, that

signal whether a goal has been achieved or not. In contrast, in the indirect word

learning studies possible referents are deliberately not marked by such action cues

(see Baldwin & Tomasello, 1998, for a related argument).

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 271

As far as we know, only one study of children�s social-pragmatic abilities presents

children with a task which resembles indirect word learning tasks in that the possible

referents are not directly acted upon by the speaker in some manner. Akhtar, Car-

penter, and Tomasello (1996) let 24-month-old children play with three novel objects

in the company of two experimenters and one of the child�s parents. Each of the ob- jects were singled out and described with excitement but never labeled. The parent

and one experimenter then left the room and in their absence another novel object

was introduced. All four objects were then placed in a row and the two adults re-

turned and excitedly exclaimed: ‘‘Look, I see a gazzer!’’ The adults carefully avoided

looking at any one particular object. In a control condition, the adults returned and

said ‘‘Look at that!’’ with equal excitement. Children�s comprehension was later

tested by asking the child to indicate the referent of the novel label. Only the children

in the novel label condition correctly indicated the target object. In this study, rather than having each object singled out by the speaker, children themselves must identify

that the relevant principle for choosing between the objects is the speaker�s emo-

tional reaction to an object introduced in her absence. Although not exactly parallel

to what would be required in the indirect word learning studies, this finding comes

closest to demonstrating that young children may be able to recruit the relevant so-

cial-pragmatic knowledge in the absence of supporting cues. Note that the children

in this study were 2 years old and that this has not yet been shown with 15–18-

month-olds. Suppose 15-month-olds are capable of such pragmatic reasoning. On the social-

pragmatic account, babies in Studies 1 and 2 would reason that speakers would nor-

mally use the conventional term for an object to refer to it. Since the speaker used

another term, she must have intended to refer to another object. As the babies know

that the bucket may hide this other object, they search in the bucket. Study 3, how-

ever, presents a single, familiar object in full view on the table without any other ob-

jects around and more strongly suggests that the single object in view is a likely

referent. Even if the speaker is not directly pointing to or looking at the object, the context alone is a strong pragmatic cue that the speaker intends to refer to that

object. Thus, babies find themselves in a situation where the pragmatic cue to the

speaker�s intention conflicts with the conclusion from his or her reasoning about

communicative conventions like ‘‘if she had meant that, she would have said that.’’

A stronger version of this type of conflict arises in Liittschwager and Markman�s (1994) Study 3 of second-label learning in 16-month-olds. Similar results with 18-

month-olds were found in an unpublished study by Nowinski and Markman. In

those studies, the speaker directly looks at and points to a familiar object while say- ing a novel label, yet the infants reject the object as a referent of the novel label. It is

not clear how the social-pragmatic account explains a situation where all the directly

observable pragmatic cues deviate from the conclusion that the infant would draw

based on other sources of knowledge about a speaker�s intentions. Under what cir-

cumstances would a speaker point to one object while meaning to refer to another

object, unless she also clearly indicated error, negative emotion, disappointment,

or possibly deception? And, in turn, how would a child of 16 months have come

to develop such strong assumptions about the existence of a communicative practice

272 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

where a speaker points to a different object than the one she intends to label without

any of the above markers? Even allowing for conflicts between pragmatic cues and

other sources of knowledge, it seems unlikely that 16-month-old babies� reasoning that ‘‘if she had meant that, she would have said that’’ alone would be a strong en-

ough motivator to override the pragmatic cues of pointing and eye gaze and cause babies to search elsewhere. The social-pragmatic explanation specifically does not in-

clude a mutual exclusivity assumption so there is nothing in principle to prevent the

baby from assuming that the speaker might use two labels to refer to the same object.

The social-pragmatic explanation of why 16-month-olds reject second labels for ob-

jects is a complicated story whereby ‘‘if she meant that, she would have said that’’

outweighs the fact that the speaker is obviously labeling the object. This seems much

less compelling than the straightforward explanation based on the mutual exclusivity

assumption where children would reason ‘‘that can�t be a crimp, it�s a spoon.’’ Can the success of 15-month-olds on these indirect word learning tasks be ex-

plained by social-pragmatic reasoning? Maybe—the jury is still out. But at this point,

related evidence supports the claim that mutual exclusivity is necessary to account

for our findings because: (1) there is explicit evidence that 16-month-olds avoid sec-

ond labels for objects as predicted by mutual exclusivity even in the face of clear, un-

ambiguous pragmatic cues that the speaker did intend to refer to the given object

with the second label, and (2) there is a lack of evidence that 15-month-olds have

the ability to do the necessary reasoning when explicit social-pragmatic cues are ab- sent or ambiguous.

In conclusion, our findings demonstrate both that mutual exclusivity is available

to children before the age of the vocabulary spurt and one of the purposes it serves is

to promote some kinds of indirect word learning. The results highlight the way in

which lexical constraints can serve as a resource for children when directly observa-

ble cues from the speaker or objects in the environment are lacking. By 15 months of

age babies reject second labels for objects and actively search for appropriate refer-

ents of a word rather than always relying on a speaker to direct their attention to a given referent or on having the mapping triggered by viewing a novel object as a can-

didate referent. Mutual exclusivity and the principle of reference provide children

with adequate grounds to actively search for appropriate referents of novel terms.

In this way, lexical constraints add a layer of resilience to babies� word learning, en-

abling them to form word-referent mappings even when speakers are not clear and

referents are not visible.

The findings are also significant in demonstrating that word learning constraints

play a role in the fast acquisition of words seen at around 19 months of age. Our findings do not, of course, address questions about the ultimate origins of this bias

(see Woodward & Markman, 1998). Whether and what kinds of exposure to linguis-

tic input are relevant to working out this assumption remains an open question. Our

findings do address, however, the question of whether word learning constraints are

necessary to account for young children�s ability to rapidly acquire vocabulary or

whether they are merely dispensable heuristics that children discover after having al-

ready become capable of figuring out the meanings of novel terms in some other

way. When the ability to recruit mutual exclusivity in this indirect word learning

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 273

situation was only found in 2-year-olds it was a fair criticism to argue that children

might be using a learned heuristic that developed as a consequence of having gone

through the naming explosion (MacWhinney, 1989). But the success of 15-month-

olds in the present study rules out this explanation. Instead of being a late developing

learned heuristic, mutual exclusivity is available to serve a vital role in the rapid learning of novel words for babies at the time of the naming explosion.

Acknowledgments

We are grateful to the many parents who generously volunteered to have their ba-

bies participate in these studies and to Bing Nursery School for providing research

facilities. We thank Colleen M. Fitzsimmons for her help in coding the data and the reviewers for their constructive suggestions. This work was supported in part

by NSF Grant BNS-91092 to Ellen M. Markman.

References

Akhtar, N., Carpenter, M., & Tomasello, M. (1996). The role of novelty discourse in early word learning.

Child Development, 67, 635–645.

Akhtar, N., & Tomasello, M. (1996). Two-year-olds learn words for absent objects and actions. British

Journal of Developmental Psychology, 14, 79–93.

Anglin, J. M. (1993). Vocabulary Development: A morphological analysis. Monographs of the Society for

Research in Child Development, 58(10, Serial No. 238).

Au, T. K. (1990). Children�s use of information in word learning. Journal of Child Language, 17, 393–

416.

Au, T. K., & Glusman, M. (1990). The principle of mutual exclusivity in word learning: To honor or not to

honor? Child Development, 61, 1474–1490.

Au, T. K., & Laframboise, D. E. (1990). Acquiring color names via linguistic contrast: The influence of

contrasting terms. Child Development, 61, 1808–1823.

Baldwin, D. A. (1991). Infants� contribution to the achievement of joint reference. Child Development, 62,

875–890.

Baldwin, D. A. (1993). Infants ability to consult the speaker for clues to word reference. Journal of Child

Language, 20, 395–418.

Baldwin, D. A. (1995). Understanding the link between joint attention and language. In C. Moore, & P.

Dunham (Eds.), Joint attention: Its origins and role in development (pp. 131–158). Hillsdale, NJ:

Erlbaum.

Baldwin, D. A., Markman, E. M., Bill, B., Desjardins, R. N., & Irwin, J. M. (1996). Infants� reliance on a

social criterion for establishing word–object relations. Child Development, 67, 3135–3153.

Baldwin, D. A., & Tomasello, M. (1998). Word learning: A window on early pragmatic understanding. In

E. V. Clark (Ed.), The proceedings of the twenty-ninth annual child language research forum. Chicago:

University of Chicago Press.

Banigan, R. L., & Mervis, C. B. (1988). Role of adult input in young children�s category evolution: 2. An

experimental study. Journal of Child Language, 15, 493–504.

Bloom, P. (1994). Possible names: The role of syntax–semantics mappings in the acquisition of nominals.

Lingua, 92, 297–329.

Bloom, L., 1998. Language acquisition in its developmental context. In W. Damon, D. Kuhn, & R. Siegler

(Eds.), Handbook of child psychology, Vol. 2. Cognition, perception, and language (pp. 309–370). New

York: Wiley.

274 E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275

Bloom, P. (2000). How children learn the meanings of words. Cambridge, MA: MIT Press.

Bloom, L., Lifter, K., & Broughton, J. (1985). The convergence of early cognition and language in the

second year of life: Problems in conceptualization and measurement. In M. Barrett (Ed.), Children�s one-word speech (pp. 149–180). New York: Wiley.

Brown, R. W. (1957). Linguistic determinism and the parts of speech. Journal of Abnormal and Social

Psychology, 55, 1–5.

Carey, S., & Bartlett, E. (1978). Acquiring a single new word. Papers and Reports on Child Language

Development, 15, 17–29.

Clark, E. V. (1983). Meanings and concepts. In J. H. Flavell, & E. M. Markman (Eds.), Handbook of child

psychology: Vol. 3. Cognitive development (pp. 787–840). New York: Wiley.

Clark, E. V. (1987). The principle of contrast: A constraint on acquisition. In B. MacWhinney (Ed.),

Mechanisms of language acquisition: The 20th annual carnegie symposium on cognition (pp. 1–34).

Hillsdale, NJ: Erlbaum.

Clark, E. V. (1988). On the logic of contrast. Journal of Child Language, 15, 317–335.

Clark, E. V. (1990). On the pragmatics of contrast. Journal of Child Language, 17, 417–431.

Clark, E. V. (1997). Conceptual perspective and lexical choice in acquisition. Cognition, 64, 1–37.

Corrigan, R. (1983). The development of representational skills. In K. W. Fischer (Ed.), Levels and

transitions in children�s development (pp. 51–65). San Francisco: Jossey-Bass.

Diesendruck, G., & Markson, L. (2001). Avoidance of lexical overlap: A pragmatic account.

Developmental Psychology, 37, 630–641.

Dromi, E. (1987). Early lexical development. Cambridge, England: Cambridge University Press.

Evey, J. A., & Merriman, W. E. (1998). The prevalence and the weakness of an early name mapping

preference. Journal of Child Language, 25, 121–147.

Fenson, L., Dale, P. S., Reznick, J. S., Thal, D., Bates, E., Reilly, J. S., & Hartung, J. P., 1991. The

MacArthur Communicative Development Inventory. San Diego, CA.

Fisher, C., Gleitman, H., & Gleitman, L. (1991). On the semantic content of subcategorization frames.

Cognitive Psychology, 23, 331–392.

Fisher, C., Hall, D. G., Rakowitz, S., & Gleitman, L. (1994). When it is better to receive than to give:

Syntactic and conceptual constraints on vocabulary growth. Lingua, 92, 333–375.

Gathercole, V. C. (1987). The contrastive hypothesis for the acquisition of word meaning: A

reconsideration of the theory. Journal of Child Language, 14, 493–532.

Gathercole, V. C. (1989). Contrast: A semantic constraint? Journal of Child Language, 16, 685–702.

Goldfield, B. A., & Reznick, J. S. (1990). Early lexical acquisition: Rate, content and the vocabulary spurt.

Journal of Child Language, 17, 171–183.

Golinkoff, R. M., Hirsh-Pasek, K., Bailey, L. M., & Wenger, N. R. (1992). Young children and adults use

lexical principles to learn new nouns. Developmental Psychology, 28, 99–108.

Golinkoff, R., Mervis, C. B., & Hirsh-Pasek, K. (1994). Early object labels: The case for a developmental

lexical principles framework. Journal of Child Language, 21, 125–155.

Graham, S. A., Poulin-Dubois, D., & Baker, R. K. (1998). Infants� disambiguation of novel object words.

First Language, 18, 149–164.

Halliday, M. A. K. (1975). Learning how to mean. In E. H. Lenneberg, & E. Lenneberg (Eds.),

Foundations of language development: A multidisciplinary approach: Vol. 1 (pp. 239–265). New York:

Academic Press.

Heibeck, T. H., & Markman, E. M. (1987). Word learning in children: An examination of fast mapping.

Child Development, 58, 1021–1034.

Johnson, S. C. (2000). The recognition of mentalistic agents in infancy. Trends in Cognitive Sciences, 4,

22–28.

Liittschwager, J. C., & Markman, E. M. (1994). Sixteen and 24-month-olds� use of mutual exclusivity as a

default assumption in second label learning. Developmental Psychology, 30, 955–968.

MacWhinney, B. (1989). Competition and lexical categorization. In R. Corrigan, F. Eckman, & M.

Noonan (Eds.), Linguistic categorization (pp. 195–241). Amsterdam, the Netherlands: John Benjamin.

Markman, E. M. (1989). Categorization and naming in children: Problems of induction. Cambridge, MA:

MIT Press.

E.M. Markman et al. / Cognitive Psychology 47 (2003) 241–275 275

Markman, E. M. (1992). Constraints on word learning: Speculations about their nature, origins and

domain specificity. In M. R. Gunnar, & M. P. Maratsos (Eds.), Modularity and constraints in language

and cognition: The Minnesota symposium on child sychology (pp. 59–101). Hillsdale, NJ: Erlbaum.

Markman, E. M. (1994). Constraints on word meaning in early language acquisition. Lingua, 92, 199–227.

Markman, E. M., & Hutchinson, J. (1984). Children�s sensitivity to constraints on word meaning:

Taxonomic versus thematic relations. Cognitive Psychology, 16, 1–27.

Markman, E. M., & Wachtel, G. F. (1988). Children�s use of mutual exclusivity to constrain the meanings

of words. Cognitive Psychology, 20, 121–157.

McShane, J. (1979). The development of naming. Linguistics, 17, 79–95.

Merriman, W. E., & Bowman, L. L. (1989). The mutual exclusivity bias in children�s word learning.

Monographs of the Society for Research in Child Development, 54(Serial No. 220).

Merriman, W. E., & Schuster, J. M. (1991). Young children�s disambiguation of object name reference.

Child Development, 62, 1288–1301.

Mervis, C. B. (1987). Child-basic object categories and early lexical development. In U. Neisser (Ed.),

Concepts and conceptual development: Ecological and intellectual factors in categorization (pp. 201–233).

Cambridge, MA: Cambridge University Press.

Mervis, C. B., & Bertrand, J. (1994). Acquisition of the novel name–nameless category (N3C) principle.

Child Development, 65, 1646–1662.

Mervis, C. B., Golinkoff, R. M., & Bertrand, J. (1994a). Two-year-olds readily learn multiple labels for the

same basic-level category. Child Development, 65, 1163–1177.

Mervis, C. B., Johnson, K. E., & Mervis, C. A. (1994b). Acquisition of subordinate categories by 3-year-

olds: The roles of attribute salience, linguistic input, and child characteristics. Cognitive Development,

9, 211–234.

Nelson, K. (1973). Structure and strategy in learning to talk. Monographs of the Society for Research in

Child Development, 38(Serial No. 149).

Nelson, K. (1988). Constraints on word learning? Cognitive Development, 3, 221–246.

Soja, N. N., Carey, S., & Spelke, E. S. (1991). Ontological categories guide young children�s inductions of word meaning: Object terms and substance terms. Cognition, 38, 179–211.

Tomasello, M. (2001). Perceiving intentions and learning words in the second year of life. In M.

Bowerman, & S. C. Levinson (Eds.), Language acquisition and conceptual development. New York:

Cambridge University Press.

Tomasello, M., & Akhtar, N. (1995). Two-year-olds use pragmatic cues to differentiate reference to objects

and actions. Cognitive Development, 10, 201–224.

Tomasello, M., & Barton, M. E. (1994). Learning words in nonostensive contexts. Developmental

Psychology, 30, 639–650.

Tomasello, M., Strosberg, R., & Akhtar, N. (1996). Eighteen-month-old children learn words in non-

ostensive contexts. Journal of Child Language, 23, 157–176.

Waxman, S. R., & Gelman, R. (1986). Preschoolers� use of superordinate relations in classification and

language. Cognitive Development, 1, 139–156.

Waxman, S. R., & Kosowski, T. D. (1990). Nouns mark category relations: Toddlers and preschoolers

word learning biases. Child Development, 61, 1461–1473.

Woodward, A. L., 1992. The role of the whole object assumption in early word learning. Doctoral

dissertation, Stanford University, Stanford, CA.

Woodward, A. L. (1998). Infants selectively encode the goal object of an actor�s reach. Cognition, 69, 1–34. Woodward, A. L. (1999). Infants� ability to distinguish between purposeful and non-purposeful behaviors.

Infant Behavior and Development, 22, 145–160.

Woodward, A. L., & Markman, E. M. (1991). Constraints on learning as default assumptions: Comments

on Merriman and Bowman�s the mutual exclusivity bias in children�s word learning. Development

Review, 11, 137–163.

Woodward,A.L.,&Markman,E.M.(1998).Earlywordlearning.InW.Damon,D.Kuhn,&R.Siegler(Eds.),

Handbookofchildpsychology,Vol. 2:Cognition,perception,and language (pp.371–420).NewYork:Wiley.

Woodward, A. L., & Sommerville, J. A. (2000). Twelve-month-old infants interpret action in context.

Psychological Science, 11, 73–76.