Attention to Novel Objects during Verb Learningpsy2.fau.edu/~kersten/Papers/kersten+smith2002.pdf · learning may beneﬁt from prior learning of nouns. Noun learning per se is not

Child Development, January/February 2002, Volume 73, Number 1, Pages 93–109

Attention to Novel Objects during Verb Learning

Alan W. Kersten and Linda B. Smith

Three experiments provided evidence that 3.5- to 4-year-old English-speaking children (

N

�

72) attend to theappearances of novel objects, not only when they hear a novel noun, but also when they hear a novel verb.Children learning nouns in the context of novel, moving objects attended exclusively to the appearances of ob-jects, even though nouns were also related to the motions of those objects. Children learning verbs attendedequally to the appearances of objects and their motions. The latter result contrasted with the results fromadults (

N

�

20), who focused more strongly on motions than on the appearances of objects when learningverbs. When familiar objects were instead employed, child verb learners attended more to motions than to theappearances of objects. Children may attend to novel objects during verb learning because knowledge of anobject may be prerequisite to understanding what a verb means in the context of that object.

INTRODUCTION

Children who are learning a variety of different lan-guages have been found to learn nouns earlier thanverbs (Gentner, 1982). Although this noun advantagemay not be universal, children seem to learn nouns atleast concurrently with verbs even in “verb-centered”languages such as Korean and Mandarin, in whichnouns can be dropped from a sentence and verbsappear in the salient, sentence-final position (Au,Dapretto, & Song, 1994; Choi & Gopnik, 1995; Gopnik& Choi, 1995; Tardif, 1996; Tardif, Gelman, & Xu,1999). One explanation for the earlier acquisition ofnouns compared with verbs is Gentner’s (1982; Gent-ner & Boroditsky, 2001) natural partitions hypothesis.According to this hypothesis, the ability of humanperception to segregate objects from their surround-ing environment, coupled with a straightforwardmapping of those objects onto nouns, encourages theacquisition of object names. In contrast, although chil-dren are also quite good at perceiving motions, andmost of children’s earliest words refer to objects thatare capable of moving or being moved (Bloom, 1973;Nelson, 1973), the mapping of these motions ontoverbs is more difficult and arbitrary. For example,Talmy (1985) has noted that whereas the most com-mon verbs in English convey the manner of motion ofan object (e.g., run, skip), verbs in many other lan-guages more often convey information about the pathof an object (e.g., enter, exit). Thus, although percep-tual machinery is dedicated to motions as well as ob-jects, learning how a language conveys these per-ceived motions may be more difficult than learninghow it conveys information about perceived objects.

The natural partitions hypothesis suggests thatnouns are learned before verbs because verbs are moredifficult. An alternative explanation, however, is thatnouns are learned before verbs because verb learning

is dependent on knowledge of objects, which is typi-cally acquired during the process of noun learning.This view is suggested by a number of differentsources. First, Bloom (1981) has noted that whereaschildren frequently produce nouns during the one-word period, verb use is rare until children start pro-ducing multiword utterances. On the basis of this ob-servation, she proposed that verbs cannot be learnedin isolation, but rather must be learned in combina-tion with nouns. This suggests that to learn the mean-ing of a verb, children must rely on knowledge asso-ciated with nouns, typically knowledge of objects.

That verb learning is dependent on knowledge ofobjects is also suggested by Gentner’s (1981) theory ofrepresentational differences between nouns and verbs.According to this theory, object concepts are moreconceptually cohesive than are relational concepts suchas verbs. As a result, nouns are interpreted in muchthe same way regardless of their context. In contrast,representations of relational concepts interact morestrongly with representations of other concepts. As aresult, the interpretation of a verb may be dependenton the particular objects involved. For example, one’sinterpretation of the verb “running” may be very dif-ferent in the contexts of “man,” “horse,” “engine,” and“politician” (for evidence consistent with this sugges-tion, see Earles & Kersten, 2001). If this theory is cor-rect, one would expect that children would have dif-ficulty conceptualizing the motions of unfamiliarobjects, because the representations of those objectsare not yet established. In contrast, children shouldperform better at conceptualizing the motions of fa-miliar objects, because those objects form an estab-lished conceptual framework on which representa-tions of novel motions can be constructed.

© 2002 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/2002/7301-0007

94 Child Development

Finally, a role for object knowledge in verb learn-ing is also suggested by Kersten’s (1998a) division oflabor theory of motion representation in language. Ac-cording to this theory, object nouns are typically asso-ciated with a number of different manners of motion.For example, humans are capable of walking, run-ning, skipping, and sauntering, and children learn toassociate these motions with humans at an early age(Bertenthal, Proffitt, & Cutting, 1984). The role ofmanner of motion verbs, then, is to select which of thedifferent manners of motion associated with a noun isrelevant in the given instance. To accomplish this, amanner of motion verb is associated with a genericmeaning. For example, the verb “run” may be associ-ated with a meaning such as “to move rapidly.” Whenthis verb is used in conjunction with a noun such as“person,” the verb selects the manner of motion asso-ciated with “person” that is most consistent with thegeneric meaning of the verb. Because different man-ners of motion are associated with different nouns,this same verb may select different meanings in dif-ferent noun contexts (e.g., “person” versus “engine”).

The theories of Bloom (1981), Gentner (1981), andKersten (1998a) suggest that knowledge of the typesof motions that an object can perform (or be affectedby, when the object involved is a patient rather thanan agent) is necessary before a child can understandwhat a verb means in the context of that object. Thisknowledge would typically be acquired during theprocess of learning a noun for an object, and thus verblearning may benefit from prior learning of nouns.Noun learning per se is not absolutely necessary forthe learning of verbs, however. What is important isthat a child has knowledge of the category to which anobject belongs, in particular knowledge of the types ofmotions that are typical for objects in that category. Thisobject category knowledge may be acquired during theprocess of noun learning, but it may also be acquired in-dependently of a label (for evidence that motions can beassociated with object categories independently of anoun label, see Kersten, 1998b; Kersten & Billman, 1997).

The theories of Bloom (1981), Gentner (1981), andKersten (1998a) may be thought of as specific instan-tiations of a more general class of linguistic theoriesregarding the relation between the meaning of a verband its arguments. In particular, semantic structuretheorists (e.g., Gropen, Pinker, Hollander, & Gold-berg, 1991; Levin & Rappaport Horav, 1991; Pinker,1989) have proposed that the argument structure as-sociated with a verb is related to the meaning of theverb. Thus, if a verb appears in two different argu-ment structures, it has two different (albeit related)meanings. An example of this phenomenon comesfrom the locative alternation. The locative alternation

is exemplified in the sentences “The worker loadedbricks into the wagon,” and “The worker loaded thewagon with bricks.” According to semantic structuretheorists, the meaning of the verb “load” in these twocases is different. In particular, the first sentence merelyspecifies that the worker caused some bricks to moveinto the wagon, whereas the second sentence specifiesthat the worker caused the wagon to become full bymeans of moving bricks. This explains why it is ac-ceptable to say “The worker loaded bricks into thewagon but it was still only half full,” whereas it isstrange to say “The worker loaded the wagon withbricks but it was still only half full.” Thus, theories ofsemantic structure are consistent with the theoriesof Bloom (1981), Gentner (1981), and Kersten (1998a)in suggesting that the same verb can have differentmeanings when used along with different argumentsand argument structures. If these theories are correct,then attention to the number and nature of the argu-ments (i.e., nouns) associated with a verb may be usefulin determining the meaning of that verb (for evidencewith regard to argument structure, see Gleitman, 1990;Naigles, 1990; Naigles & Hoff-Ginsberg, 1995).

The theories of Bloom (1981), Gentner (1981), andKersten (1998a) make a number of predictions withregard to children’s performance in a verb-learningtask. First, if a novel verb were presented along with amotion performed by a novel object, children wouldhave to learn about the nature of that object beforefully comprehending the meaning of the verb. Thus,children may attend as much to a novel object as to itsmotion when learning a novel verb. Second, if a novelverb were presented along with a motion performedby a familiar object, children would not have to at-tend as strongly to that object to understand themeaning of the verb. In particular, they would al-ready have some understanding of the kinds of mo-tion that the object could participate in, and thuslearning a verb would require determining which ofthose motions was associated with the verb. Finally, averb learner with a more sophisticated understand-ing of the types of objects in the world and their inter-relations may be able to recognize a novel object as anexample of a previously encountered class of objects.Thus, an adult verb learner may attend less to novelobjects than does a child, because the adult may relyon more general knowledge of object classes when in-terpreting the motion of an unfamiliar object.

Attention to Novel Objects duringVerb Learning in Children

Evidence that children attend to novel objects dur-ing verb learning comes from the work of Behrend

Kersten and Smith 95

(1990). Behrend presented 3-year-olds, 5-year-olds,and adults with novel verbs along with events involv-ing novel instruments, novel actions, and novel re-sults. After each such event, he presented participantswith events that varied on one of these three at-tributes and asked them whether these were still ex-amples of the same verb. Although participants of allages were least likely to accept an event that involveda different result as an example of the same verb, 3-and 5-year-old children also rejected about 40% ofevents involving instrument changes. Thus, childrenattended to objects as well as actions when they werepresented with a novel verb. Adults were much morelikely to accept events involving instrument changesas examples of the same verb, indicating that this ten-dency to focus on objects decreases with age. Forbesand Farrar (1995) also found that 3-year-olds tendedto reject an event as an example of a verb because of achange in instrument, whereas adults were morelikely to accept such an event.

Further evidence that children attend to novel ob-jects when learning verbs comes from the classic studyof Brown (1957). Brown presented 3- to 5-year-old chil-dren with cards depicting novel objects, novel sub-stances, and novel actions. For example, one pictureinvolved a pair of hands performing a kneading actionon a mass of red confetti in a novel container. Each suchcard was labeled either by a novel count noun (e.g., “asib”), mass noun (e.g., “some sib”), or verb (e.g., “sib-bing”). Children were then shown three cards andasked to choose which was another example of thenovel word. Each card matched the original on oneproperty but mismatched on the other two. For exam-ple, the card that matched on the action involved anovel container with a mass of yellow confetti inside it.

The results revealed that children most often chosethe card that matched on the object (e.g., the con-tainer) in the count noun condition. In fact, out of 16children, only one child ever chose a card thatmatched on the action. This finding is consistent witha tendency to associate nouns with objects. Althoughthe majority of children’s choices in the verb condi-tion involved the card that matched on the action,33% of children’s choices still involved novel objectsand substances. Thus, there is evidence of a tendencyfor children to attend to novel objects not only whenthey are presented with nouns, but also when they arepresented with verbs.

Evidence that Adults Attend Primarilyto Motions When Learning Verbs

Unlike research with children, research with adultssuggests that they do not pay much attention to the

appearances of novel objects when they are presentedwith a novel verb, but rather focus their attention onthe actions of those objects. For example, Kersten(1998a) presented undergraduates with events in-volving interactions between two novel, buglike crea-tures. Each such event was accompanied by a sen-tence involving both a novel noun and verb (e.g.,“The zeebee is morping.”). Each noun and verb wasrelated to both an object attribute and an event at-tribute. In particular, each noun and verb always co-occurred with a particular body shape of one of thecharacters in the event as well as a particular pathalong which that character moved with respect to thesecond character. Participants were later tested ontheir knowledge of these relations by asking them tochoose which of two events was a better example of aparticular noun or verb. One event was entirely con-sistent with the word being tested, whereas the otherevent involved either a body shape or path that wasinconsistent with this word.

Participants were found to be nearly perfect atchoosing which of two events was the better exampleof a particular verb when those two events differedon the path of the moving character. In contrast, par-ticipants were at chance when asked to choose be-tween two events that differed on the body shape ofthe moving character. It was not the case that changesin the body shape of the moving character were toosubtle to be noticed, because participants were nearlyperfect at choosing between two events that differedon the body shape of the moving character whenknowledge of nouns rather than verbs was tested.These results suggest that adults focus their attentionon the motions rather than the appearance of an ob-ject when they are presented with a verb, even whenthe object is novel in appearance.

Research with children thus suggests that they at-tend strongly to novel objects during verb learning,whereas research with adults suggests that they donot attend as strongly to novel objects. What is re-sponsible for this difference? A possible explanationis that adults have a more sophisticated understand-ing of object categories and their interrelations. Inparticular, although the objects employed in the re-search of Kersten (1998a) were novel, participantsmay have treated them as examples of a more generalclass of objects, namely bugs. Consistent with thissuggestion, participants often spontaneously referredto these objects as “bugs” in postexperimental inter-views. Once participants identified these novel ob-jects as examples of bugs, they may not have attendedvery strongly to their individuating characteristics,but rather relied on their general knowledge of thetypes of actions that bugs are capable of to help them


figure out the meanings of the verbs. Children, on theother hand, may not always recognize the larger cat-egory to which an object belongs, and thus may treatnovel objects as truly novel and worthy of attention.

Overview of Experiments

The present experiments were designed to investi-gate the extent to which children and adults attend tothe appearances of novel objects in the same verb-learning task. In particular, they were designed to testwhether 3.5- to 4-year-old English-speaking children,like adults, focus their attention on actions in the pres-ence of a novel verb, or whether they also attend tothe appearances of novel objects. The experimentsalso tested whether a tendency to focus on objects isreduced when familiar rather than novel objects arepresent. The method for the experiments was adaptedfrom that of Kersten (1998a). Experiment 1 examinedwhether children and adults attend more strongly tothe appearance of a novel object or to the path of thatobject when learning a novel verb and when learninga novel noun. Experiment 2 examined attention tonovel objects in the context of a different type of motion,namely the manner of motion of an object. Finally, Ex-periment 3 tested children’s attention to an object andits path when familiar rather than unfamiliar objectswere involved in the events.

EXPERIMENT 1

Experiment 1 examined the learning of novel nounsand verbs in 3.5- to 4-year-old English-speaking chil-dren and in adults. As in Kersten (1998a), each nounor verb was related to the body shape of one characteras well as the path of that character with respect to asecond character. Participants were tested on theirknowledge of these relations by presenting them withevents involving a different body or path and askingthem if these test events were still examples of thewords heard during learning. On the basis of the re-sults of Kersten (1998a), adults were predicted to ac-cept an event as an example of a noun as long as thebody of the moving character was consistent withthat noun. That is, they were expected to pay little at-tention to motion when learning nouns. On the otherhand, they were expected to accept an event as an ex-ample of a verb as long as the path of the movingcharacter was consistent with that verb. In otherwords, they were expected to pay little attention tothe appearance of the object in motion when learningverbs. Children who were learning nouns were pre-dicted to perform similarly to adults, focusing stronglyon the body shape of the moving object. Children

who were learning verbs, however, were expected toperform quite differently from adults, attending notonly to the path of the object but also to the bodyshape of that object.

Method

Participants

Twenty children (10 boys and 10 girls) who rangedin age from 41 to 48 months (

M

�

44 months) partici-pated. These children were recruited from files of po-tential participants maintained by Indiana Universitydevelopmental psychology researchers. These filescontain names and ages of children found throughbirth announcements, advertisements in local news-papers and radio programs, or referrals from parentsof other participants. Although demographic datawere not collected on the children who participated inthis study, this population of children is predomi-nantly White and middle class, with English beingthe sole or primary language spoken in the home.Each child was tested in the laboratory and was givena small gift (e.g., a T-shirt or beach ball) at the end ofthe experiment. Five boys and 5 girls were randomlyassigned to the noun condition, whereas the remain-ing participants were assigned to the verb condition.Four additional children (2 boys and 2 girls) were dis-qualified because they failed to complete the experi-ment or said either “yes” or “no” to every test ques-tion. Of these children, 3 were in the verb conditionand 1 was in the noun condition.

Twenty adults (2 men and 18 women) who rangedin age from 21 to 40 years also participated in the ex-periment. These adults were recruited from an under-graduate course on personality at Florida AtlanticUniversity and received extra credit for their partici-pation. Ten adults were randomly assigned to the nouncondition, and 10 were assigned to the verb condition.

Stimuli

Pretraining events.

All stimuli in this experimentwere shown on a 19-inch (.48 m) color televisionmonitor, viewed at a distance of about 1.5 m. To famil-iarize participants with the task, they were firstshown pretraining events involving known objects.The first pretraining event depicted a teddy bearjumping up and down on a line representing theground. Participants in the noun condition were told“See that? That’s a bear!” Participants in the verb con-dition were told “See this one? He’s jumping!” Thisevent lasted 12 s. The second pretraining event de-picted a boy sleeping in a bed. Note that this eventvaried from the first pretraining event both on the ac-


tion and the object involved. Participants in the nouncondition were asked “Is this a bear?” Participants inthe verb condition were asked “Is this one jumping?”Participants who said “No” were given positive feed-back, whereas participants who responded incor-rectly or did not respond were told, “That’s not a bear,that’s a boy,” or “He’s not jumping, he’s sleeping.”

Learning Events.

Twenty-four events were pre-sented along with noun or verb labels to teach partic-ipants the meanings of those words. Each event in-volved two buglike creatures (see Figure 1). Onecreature, the agent, moved throughout each event,whereas a second creature, the reference object, re-mained stationary and served as a frame of referencefor the motion of the agent. Events varied on a num-ber of attributes. Some attributes involved the bodyparts of the two creatures. Both the agent and the ref-erence object were composed of three attributes: ahead, a body, and a pair of legs. Agent body had twopossible values, whereas all other parts of the crea-tures had four possible values. A further static at-tribute involved the environment in which the eventtook place, defined by background pictures in the twoupper corners of the screen. The four possible envi-ronments were a desert, a swamp, a mountainous ter-rain, and a rocky plain.

The motion of the agent also varied on two at-tributes. One attribute was agent path, or the direc-tion taken by the agent with respect to the referenceobject. Agents either moved toward or away from thereference object. A second attribute was the leg mo-tion of the agent. There were four possible values of

this attribute (for descriptions of the different legmotions, see Kersten, 1998a). An additional source ofvariation in the events came from the starting loca-tions of the two characters. The location of the refer-ence object was always in one of the two lower quad-rants of the screen. The location of the agent was tothe left of the reference object on half of the events andto the right on the other half. The agent started eachevent separated from the reference object by a dis-tance such that an agent moving toward the referenceobject would just come in contact with the referenceobject at the end of the event.

Each participant was presented with either twonouns (i.e., “zeebee” and “taygo”) or two verbs (i.e.,“morping” and “spogging”). In either case, the twowords could be distinguished by two different at-tributes, agent body and agent path. In particular, thewords “zeebee” and “morping” always correspondedto an agent with a rounded, orange body with redspots on it (see the object on the left in Figure 1) mov-ing toward the reference object. This event was pre-sented 15 times. In contrast, the words “taygo” and“spogging” always corresponded to an agent with arectangular, purple body with gray spots on it (see theobject on the left in Figure 2) moving away fromthe reference object. This event was presented onlynine times and was included primarily to give partici-pants negative feedback about the meanings of thewords “zeebee” and “morping.” All other attributesbesides agent body and path varied randomly acrossthe events. Note that this random variation includedaspects of the appearance of an object (e.g., the legs andhead of the agent), as well as aspects of the motion ofthat object (e.g., the leg motion of the agent). This ran-dom variation was included to simulate the variationin real-world examples of nouns and verbs. For exam-

Figure 1 Example event from Experiment 1. This eventwould have been labeled by the noun “zeebee” for partici-pants in the noun condition, whereas it would have been la-beled by the verb “morping” for participants in the verb con-dition. Examples of these words always involved an agent (onthe left in this figure) with a rounded, orange body that movedinto contact with the reference object (on the right in this fig-ure). Other attributes (i.e., the head, legs, and leg motion of theagent; the head, body, and legs of the reference object; and theenvironment) varied randomly across the different examplesof a word. Animated examples of the events used as stimuli inthis research are available on the World Wide Web at http://www.psy.fau.edu/akersten/home.html.

Figure 2 Second example event from Experiment 1. Thisevent would have been labeled by the noun “taygo” for partic-ipants in the noun condition, whereas it would have been la-beled by the verb “spogging” for participants in the verb con-dition. Examples of these words always involved an agent (onthe left in this figure) with a square, purple body that movedaway from the reference object (on the right in this figure).


ple, only certain aspects of an object (e.g., its shape)are relevant to the meaning of a noun, whereas others(e.g., its size) are irrelevant. Similarly, only certain as-pects of an event are relevant to the meaning of a par-ticular verb, and thus the examples of a verb oftenvary along a number of other aspects that are orthog-onal to the meaning of the verb (Huttenlocher & Lui,1979; Kersten & Billman, 1997).

Nouns were always presented in the present tense(e.g., “This is a zeebee.”). Verbs were presented in sev-eral different tenses because of findings by Tomaselloand Kruger (1992). This work demonstrated that chil-dren learned verbs better when they were presentedin either the future or past tense than in the presenttense. Thus, in addition to being labeled in the presenttense for each event (e.g., “This one is morping.”),verbs were often labeled in the future tense prior to anevent (e.g., “This one’s gonna morp.”), and in the pasttense after an event (e.g., “See that? He morped!”).

Test events.

Twenty-four events tested for knowl-edge of relations between the words accompanyingthe learning events and the attributes, agent body andagent path. The first test event came after the fifthlearning event, after which learning events and test

events were interspersed. All test events seen by aparticular participant tested for knowledge of oneword. In particular, participants in the noun condi-tion were asked, “Is this a zeebee?” Participants in theverb condition, on the other hand, were asked, “Is thisone morping?” There were four different types of testevents (see Figure 3). One type,

object

�

motion match

events, involved an agent body and path that wereboth consistent with the accompanying word. A sec-ond type,

no match

events, involved an agent bodyand path that were not consistent with the accompa-nying word, but rather were consistent with the otherword the participants had learned (i.e., “taygo” forparticipants in the noun condition or “spogging”for participants in the verb condition). A third test type,

object match

events, involved an agent body that wasconsistent with the accompanying word, but a paththat was inconsistent. A fourth test type,

motion match

events, involved a path that was consistent with theaccompanying word, but an agent body that was in-consistent. Each participant was presented with six ofeach of the four types of test events. All attributes thatvaried randomly during the learning events contin-ued to vary randomly in the test events. Thus, a par-

Figure 3 The four different types of test events in Experiment 1. Object � motion match events involved both an agent body anda path that were consistent with the tested word (“zeebee” in the noun condition and “morping” in the verb condition). Motionmatch events involved a path that was consistent with the tested word, but an agent body that was inconsistent. Object matchevents involved an agent body that was consistent with the tested word, but a path that was inconsistent. Finally, no match eventsinvolved both an agent body and a path that were inconsistent with the presented word.


ticipant could only use knowledge of agent body andpath when evaluating the test events.

Procedure

Children played with toys in the testing roomwhile their parents read and signed the consent form.Adults were simply led into the testing room andsigned the consent form. Adults were told that thetask was designed for children and that it would per-haps seem silly to them, but that it was very impor-tant for them to do their best. All participants werethen asked, “Do you want to watch some cartoons?”They were then shown the two pretraining events.Next, participants were introduced to a hand puppetcalled “Bobo.” Participants were told that Bobo was ababy dinosaur and needed to learn to speak dinosaurtalk. Participants were asked if they would like tohelp teach Bobo some new words. They were thenshown the learning and test events.

Design

There were three independent variables in this ex-periment. The first was the type of test event pre-sented to a participant (object

�

motion match versusobject match versus motion match versus no match),manipulated within participants. The second was thepart of speech of the word accompanying each testevent (noun versus verb), manipulated between par-ticipants. The third was the age group of the partici-pant (child versus adult), also manipulated betweenparticipants. The dependent variable was the propor-tion of “yes” responses to the test questions.

Results

Aggregated Results

The results of Experiment 1 are depicted in Figure4. An ANOVA was conducted on the proportion of“yes” responses to the four different types of testquestions. An

�

level of .05 was adopted for this andall other analyses. This analysis revealed a significantmain effect of test type,

F

(3, 108)

�

65.02,

p

�

.001,

MSE

�

.06. This main effect was moderated by sev-eral interactions, however. First, there was an interac-tion of test type and part of speech,

F

(3, 108)

�

32.79,

p

�

.001,

MSE

�

.06. Second, there was an interactionof test type and age group,

F

(3, 108)

�

3.33,

p

�

.05,

MSE

�

.06. Finally, there was a three-way interactionof test type, part of speech, and age group,

F

(3, 108)

�

4.22,

p

�

.01,

MSE

�

.06.Post hoc analyses were carried out to understand

this three-way interaction. Separate ANOVAs were

carried out for the noun condition and the verb con-dition. In the noun condition, there was a main effectof test type,

F

(3, 54)

�

89.40,

p

�

.001,

MSE

�

.04.Newman-Keuls tests revealed that the proportions of“yes” responses to the object

�

motion match eventsand to the object match events were not significantlydifferent, whereas these two proportions were signif-icantly greater than the proportions of “yes” re-sponses to the motion match and no match events.These last two proportions, in turn, were not signifi-cantly different. These results indicate that partici-pants were willing to accept an event as an exampleof the noun “zeebee” as long as it involved the correctobject, and that they ignored the motion of that objectwhen making their decisions. There was not a signif-icant interaction of test type and age group,

F

(3, 54)

�

1,

Figure 4 Aggregated results of Experiment 1. The y-axis rep-resents the proportion of “yes” responses to each of the fourtest types. Error bars reflect standard errors.


indicating that children and adults performed simi-larly in the noun condition.

In the verb condition, there was also a main effectof test type,

F

(3, 54)

�

27.50,

p

�

.001,

MSE

�

.08. Thismain effect was moderated, however, by a significantinteraction of test type and age group,

F

(3, 54)

�

5.64,

p

�

.01,

MSE

�

.08. Newman-Keuls tests revealedthat adults’ proportions of “yes” responses to theobject

�

motion match and motion match eventswere not significantly different, whereas these twoproportions were significantly greater than the propor-tions of “yes” responses to the object match and nomatch events. These last two proportions, in turn, werenot significantly different. These results indicate thatadults were willing to accept an event as an exampleof the verb “morping” as long as it involved the cor-rect motion, and that they ignored the appearance ofthe object in motion when making their decisions.

In contrast to the results of adults, Newman-Keulstests revealed that children made a significantly greaterproportion of “yes” responses to object

�

motionmatch events than to any of the other three test types.The proportions of “yes” responses to motion matchand object match events were not significantly differ-ent. The difference between each of these two propor-tions and the proportion of “yes” responses to no matchevents was also not significant, although there was atrend for children to make more “yes” responses to mo-tion match and object match events than to no matchevents. These results indicate that children were lesswilling to accept an event as an example of the verb“morping” if either the object or its motion werechanged, and they were even less likely to accept anevent if both the object and motion were changed.

Individual Differences

The aggregated results of children in the verb con-dition suggest that they attended equally to the ap-pearances of objects and the motions of those objects.Several different patterns of individual performancecould yield this result, however. First, the majority ofchildren could indeed focus equally on objects andmotions. Second, half of the children could focus pri-marily on objects and half could focus primarily onmotions. Finally, some combination of these two pat-terns could also produce this result.

To determine which of these patterns is evident,the results of individual participants are presented inFigure 5. In this figure, the proportion of “yes” re-sponses in the object match trials is represented onthe x-axis, whereas the proportion of “yes” responsesin the motion match trials is represented on the y-axis.The results of the adult participants are depicted in

the upper panel of Figure 5. Inspection of the opencircles reveals that all adults in the noun condition at-tended primarily to the appearances of objects. In-spection of the filled-in circles reveals that all but 2adults in the verb condition attended primarily tomotions. Of the remaining 2, 1 attended primarily

Figure 5 Performance of individual participants in Experi-ment 1. Each dot represents the performance of 1 participant.The x-axis represents the proportion of “yes” responses in ob-ject match events, which involved an agent body that was con-sistent with the tested word, but a path that was inconsistent.The y-axis represents the proportion of “yes” responses in mo-tion match events, which involved a path that was consistentwith the presented word, but an agent body that was inconsis-tent. Thus, participants who relied entirely on the appearanceof an object to make their choices would be represented in thelower right-hand corner, whereas participants who relied en-tirely on the path of an object would be represented in the up-per left-hand corner. A participant who relied equally on bothattributes would be represented along the central diagonal.


to objects (i.e., the one in the lower right-hand quad-rant), whereas the other attended equally to objectsand motions (i.e., the one along the diagonal).

The lower panel of Figure 5 depicts the results ofchildren. Inspection of the open circles reveals thatnearly all children in the noun condition focused pri-marily on the appearances of objects. All but 2 chil-dren are in the lower right-hand corner of the figure.Of the remaining 2, 1 was perfectly conservative (i.e.,the child said “no” to all of the object match and mo-tion match events), and 1 was perfectly liberal (i.e.,the child said “yes” to all of these events). Inspectionof the filled-in circles reveals several different pat-terns of performance among children in the verb con-dition. Some children focused primarily on the ap-pearances of objects. If this group is operationallydefined as those children who made at least 33% more“yes” responses to object match trials than to motionmatch trials, then 3 children fall into this group. Otherchildren in the verb condition focused primarily onmotions. If this group is operationally defined asthose children who made at least 33% more “yes” re-sponses to motion match trials than to object matchtrials, then 3 children fall into this group. Finally, 4children attended approximately equally to objectsand motions. Of these, 2 were relatively conservativeand 2 were relatively liberal. Thus, a combination ofthe two patterns of individual performance describedabove seems to be responsible for the overall patternof results. In particular, the results of some childrenresembled the group results in that they attendedequally to objects and motions. The remaining chil-dren preferred one type of information over the other,with half focusing primarily on objects and half fo-cusing primarily on motions.

Discussion

The results of Experiment 1 provide evidence thatchildren attend to the appearances of novel objectsnot only during noun learning but also during verblearning. English-speaking children were as likely toaccept an event involving a consistent object and aninconsistent motion as an example of a verb as theywere to accept an event involving an inconsistent ob-ject and a consistent motion. This suggests that theyattended as much to objects as they did to motionswhen they were learning the verb.

Children who were learning verbs attended moreto motions than did children who were learningnouns, who focused primarily on objects. This sug-gests that English-speaking children of this age havesome understanding that verbs are different fromnouns. Children’s performance was still quite differ-

ent from that of adults, however, who were willing toaccept an event as an example of a verb as long as themotion depicted in that event was consistent withprior examples of the verb. In contrast to the learningof verbs, the learning of nouns proceeded quite simi-larly in children and adults. Both children and adultswere willing to accept an event as an example of anoun as long as the appearance of the moving objectin the event was consistent with prior examples ofthat noun, regardless of the motion of that object.

Although the aggregated results of children in thisexperiment suggest that they attended equally to ob-jects and motions when learning verbs, analysis of in-dividual performance suggests that this overall pat-tern actually reflects a combination of three differentpatterns of performance. In particular, one group ofchildren attended primarily to motions, a second at-tended primarily to objects, and a third attended toboth objects and motions. One may be tempted toconclude that this third group of children was simplyguessing, accounting for their lack of preference forone type of information over the other in this task.Even these children, however, showed evidence ofsystematic responding. In particular, whereas thisthird group of children averaged 58% and 44% “yes”responses in the object match and motion matchevents, respectively, they averaged 83% “yes” re-sponses in the object

�

motion match events. The factthat they were less likely to accept an object matchevent than to accept an object

�

motion match eventas an example of “morping” suggests that they at-tended to motion when making their responses. Fur-thermore, the fact that they were less likely to accept amotion match event than to accept an object

�

motionmatch event suggests that they also attended to theappearances of objects when making their choices.Thus, the lack of preference for objects or motions onthe part of these children does not reflect unsystem-atic guessing, but rather a pattern of attending to bothobjects and motions.

Although the present results are consistent with thehypothesis that English-speaking children attend tonovel objects as well as to motions during verb learn-ing, there are some alternative hypotheses that mustbe considered. First, it is possible that children may havebeen confused by the presentation of verbs in severaldifferent tenses at several different points throughoutan event, accounting for their lack of preference formotion over the appearances of objects. In particular,a verb was presented sometimes in the future tense(e.g., “This one is gonna morp”) before an event be-gan, sometimes in the present progressive (e.g., “Thisone is morping”) while an event was taking place,and sometimes in the past tense (e.g., “See that? He


morped!”) after an event was over. The reason for thisvariation in morphology was to provide children withas much evidence as possible that the novel word theywere hearing was a verb. It is possible, however, thatthis variation may have been confusing to children.

To test this alternative hypothesis, an additionalexperiment was conducted that was identical to theverb condition of Experiment 1 except that verbs werealways presented in the present progressive tense whilean event was taking place. Ten children (6 boys and 4girls) ranging in age from 41 to 49 months (

M

�

44months) were recruited from the Karen Slattery ChildCare Center at Florida Atlantic University to partici-pate in this study. Children averaged 65% (

SD

�

35%)“yes” responses to object match events and 49%(

SD

�

33%) “yes” responses to motion match events.These percentages were not significantly different,

t

�

1.14,

p

�

.10, and were ordinally in the oppositedirection of the predictions of this alternative hypoth-esis. Thus, this study provided no support for the hy-pothesis that variation in the morphology associatedwith verbs was responsible for children’s lack of pref-erence for motion in a verb-learning task.

A second alternative hypothesis is that childrenwere attending to a different motion attribute thanthe one that was relevant to verb meanings in Exper-iment 1. According to Talmy (1985), the most commontype of verb in English conveys manner of motion in-formation. In contrast, the verbs in Experiment 1 weredistinguished on the basis of path information, a typeof motion that is commonly conveyed by verbs inmany of the world’s languages, but is not commonlyconveyed by English verbs. A type of manner of mo-tion was, in fact, present in the events seen by chil-dren in Experiment 1. In particular, the leg motion ofthe agent varied in different events. It is possible thatbecause of their English language-learning experi-ence, children focused on leg motion when they werepresented with a novel verb. Because leg motion var-ied randomly from event to event, however, childrenwould not have been able to discriminate between thetwo verbs. After discovering that they were unable todistinguish between the two verbs on the basis ofmanner of motion, children may have attended un-systematically to the other attributes of the events,yielding equal attention to objects and paths.

EXPERIMENT 2

Experiment 2 tested this alternative hypothesis forthe results of Experiment 1. In particular, it testedwhether English-speaking children focus more stronglyon motions than on the appearances of objects whenmanners of motion rather than paths are the relevant

type of motion. The design of Experiment 2 was iden-tical to that of Experiment 1 except that each nounand verb was always accompanied by a particularbody and leg motion of the agent rather than a partic-ular body and path of the agent. Path varied ran-domly from event to event, just as leg motion had var-ied in Experiment 1. If children focused specificallyon manner of motion in the presence of a novel verb,then they should have been better at associating verbswith leg motions in this experiment than they were atassociating verbs with paths in Experiment 1.

Method

Participants

Twenty children (10 boys and 10 girls) who rangedin age from 43 to 48 months (

M

�

45 months) partici-pated in this experiment. Children were recruited inthe same manner as in Experiment 1. None of the chil-dren in Experiment 2 had participated in Experiment1. Five boys and 5 girls were randomly assigned to thenoun condition, whereas the remaining participantswere assigned to the verb condition. Four additionalchildren (2 boys and 2 girls) were disqualified be-cause they failed to complete the experiment or saideither “yes” or “no” to every test question. Two ofthese children were in the verb condition and 2 werein the noun condition.

Stimuli

The learning events were identical to those of Ex-periment 1 except that each noun or verb was alwaysaccompanied by a particular leg motion of the agentrather than a particular path. The noun “zeebee” andthe verb “morping” were always accompanied by anagent that angled its legs forward and back as itmoved, with one leg moving forward as the othermoved back (see Figure 6). The noun “taygo” and theverb “spogging” were always accompanied by anagent that moved its legs up and down in tandemalong the length of its body. The test events wereidentical to those of Experiment 1 except that thematch or mismatch of verbs with leg motion ratherthan path was varied. In particular, object matchevents involved a body of the agent that was consis-tent with the presented word, but a leg motion thatwas inconsistent; whereas no match events involvedboth an incorrect body and an incorrect leg motion.

Procedure

The procedure was identical to that of Experiment 1.


Design

There were two independent variables in this ex-periment. One was the type of test event presented toa participant (object

�

motion match versus objectmatch versus motion match versus no match), manip-ulated within participants. The second variable wasthe part of speech of the word accompanying eachtest event (noun versus verb), manipulated betweenparticipants. The dependent variable was the propor-tion of “yes” responses to the test questions.

Results

Aggregated Results

The results of Experiment 2 are depicted in Figure7. An ANOVA was conducted on the proportion of“yes” responses to the four different types of testquestions. This analysis revealed a significant maineffect of test type,

F

(3, 54)

�

41.78,

p

�

.001,

MSE

�

.06. Newman-Keuls tests revealed that the propor-tions of “yes” responses to object

�

motion matchevents and to object match events were not signifi-cantly different, although there was a trend for chil-dren to make fewer “yes” responses to object matchevents. These two proportions, in turn, were signifi-cantly higher than the proportions of “yes” responsesto the motion match and no match events. There wasno significant difference between the proportions of“yes” responses to the motion match and no matchevents. These results indicate that children based their

decisions primarily on the appearances of the objectsin the events.

There was not a significant main effect of part ofspeech,

F

(1, 18)

�

1, nor did the interaction of partof speech and test type attain significance,

F

(3, 54)

�

2.09,

p

�

.11,

MSE

� .06. These results indicate thatchildren distributed their patterns of responses acrossthe four test types in a similar manner in the nouncondition and in the verb condition, making their de-cisions primarily on the basis of the appearance of theobject in motion for both nouns and verbs.


The results of individual participants are pre-sented in Figure 8. Inspection of the open circles re-veals that all but 1 child in the noun condition at-tended primarily to objects. The remaining child wasvery conservative, saying “no” to nearly all of the mo-tion match and object match events. Inspection of the

Figure 6 The meanings of the two different words heard byparticipants in Experiment 2. For participants in the noun con-dition, the noun “zeebee” always accompanied a creature witha rounded, orange body that angled its legs forward and backin alternation, giving the creature the appearance that it wasstepping forward. This creature is depicted on the left in thisfigure. The noun “taygo,” on the other hand, always accompa-nied a creature with a lumpy, gray body that moved its legsforward and back in tandem, giving the creature the appear-ance that it was pulling itself forward. This creature is de-picted on the right in this figure. For participants in the verbcondition, the verb “morping” always accompanied the crea-ture depicted on the left, whereas the verb “spogging” alwaysaccompanied the creature depicted on the right.



filled-in circles again reveals several different patternsof performance in the verb condition. Five participants(i.e., those in the lower right-hand corner) attendedprimarily to objects. Two participants (i.e., those inthe upper left-hand corner) attended primarily to mo-tions. The remaining 3 participants (i.e., those alongthe diagonal) attended equally to objects and mo-tions, with 2 participants being relatively liberal and 1participant being relatively conservative.

Discussion

The results of Experiment 2 again revealed thatEnglish-speaking children attend to the appearance ofa novel object when learning a verb. In fact, childrentended to focus more on objects than on the motions ofthose objects. Apparently, children’s failure to attendmore strongly to paths than to the appearances of objectsin Experiment 1 was not a result of a tendency to focuson leg motions. If this had been the case, one wouldhave expected that the children in Experiment 2 wouldhave been better at associating verbs with leg motions.

EXPERIMENT 3

The results of Experiments 1 and 2 indicate that 3.5- to4-year-old children attend just as strongly to the ap-

pearance of a novel object as to the motion of that ob-ject when learning verbs. The proposed explanationfor these results is that children use object categoryknowledge to help them figure out the meaning of averb. In particular, children’s knowledge of the typesof motions that are exhibited by a given type of objectmay help them to figure out what a verb may mean inthe context of that object. Thus, if a child is presentedwith a verb in the context of a novel object, childrenmay attend not only to the motion of that object butalso to the appearance of the object to figure outwhich types of motion a given type of object is capa-ble of. If this account is correct, children who werepresented with a verb in the context of a familiar ob-ject would not have to attend as strongly to the ap-pearance of that object, because they would alreadyknow what types of motion that object was capableof. Learning the verb would thus involve figuring outwhich of the types of motion associated with an objectcovaried with the presentation of the verb.

Experiment 3 was designed to test the predictionthat children learning verbs should not attend asstrongly to a familiar object as they do to a novel ob-ject. In particular, rather than presenting all childrenwith the motions of novel bugs, half of the chil-dren were presented with the motions of cars andtrucks. Because children were expected to already knowwhat types of motions these objects were capable of,they were predicted to attend more strongly to themotions of these objects than to the appearances ofthese objects, to determine which of these motions, infact, covaried with the presentation of a verb. The otherhalf of the children were again presented with the mo-tions of novel bugs. These children were expected toattend just as strongly to the appearances of these ob-jects as to their motions, as in Experiments 1 and 2.

Method

Participants

Thirty-two children (16 boys and 16 girls) whoranged in age from 41 to 48 months (M � 45 months)participated in this experiment. Sixteen children (8 boysand 8 girls) who ranged in age from 43 to 47 months(M � 45 months) were assigned to the novel objectscondition, whereas 16 children (8 boys and 8 girls)who ranged in age from 41 to 48 months (M � 45months) were assigned to the familiar objects condi-tion. Four additional children (2 boys and 2 girls) weredisqualified because they failed to complete the ex-periment or said either “yes” or “no” to every test ques-tion. Children were recruited in the same manner as inExperiments 1 and 2. None of the children in Experi-

Figure 8 The performance of individual participants in Ex-periment 2. Each dot represents the performance of 1 partici-pant. The x-axis represents the proportion of “yes” responsesin object match events, which involved an agent body that wasconsistent with the tested word, but a leg motion that was in-consistent. The y-axis represents the proportion of “yes” re-sponses in motion match events, which involved a leg motionthat was consistent with the presented word, but an agentbody that was inconsistent.


ment 3 had participated in either of the previous exper-iments. All children learned verbs in this experiment.

Stimuli

The stimuli for participants in the familiar objectscondition differed from those of Experiment 1 in thatthe events depicted the motions of cars and trucksrather than novel bugs. One event type (i.e., “morp-ing”) involved the motion of a car toward a van (seeFigure 9). The other event type (i.e., “spogging”) in-volved the motion of a truck away from the van (seeFigure 10). Thus, both the type of motion and the typeof object were related to verbs, as in the previous ex-periments. A line underneath the objects representedthe ground. The wheels of the cars and trucks rolledas those objects moved. The colors of the cars, trucks,and vans varied randomly, with four possible colorsof each object. As in Experiment 1, the moving objectappeared to the left of the reference object (i.e., thevan) on half of the trials and to the right of the refer-ence object on the other trials.

The stimuli for participants in the novel objectscondition were identical to those of the verb condi-tion of Experiment 1 except that there was no varia-tion on leg motion. Instead, each agent always movedby angling its legs forward and backward in tandem.

This was done to equate the number of motion at-tributes with that of the familiar objects condition,which involved no variation on manner of motion.Thus, in both conditions, path was the only motion at-tribute to vary.

Procedure

The procedure was identical to that of Experiment 1.

Design

There were two independent variables in this ex-periment. One was the type of test event presented toa participant (object � motion match versus objectmatch versus motion match versus no match), manip-ulated within participants. The other was the famil-iarity of the objects involved in the events (novel ver-sus familiar), manipulated between participants. Thedependent variable was the proportion of “yes” re-sponses to the test events.

Results and Discussion

Aggregated Results

The results of Experiment 3 are depicted in Figure11. An ANOVA was conducted on the proportion of“yes” responses to the four different types of testquestions. This analysis revealed a significant maineffect of test type, F(3, 90) � 22.64, p � .001, MSE �.08. This main effect, however, was moderated by an in-teraction of test type and object familiarity, F(3, 90) �3.26, p � .05, MSE � .08.

Figure 9 Example of the verb “morping” for participants inthe familiar objects condition of Experiment 3. The agent wasalways a car (on the left in this figure), which moved into con-tact with a van (on the right in this figure).

Figure 10 Example of the verb “spogging” for participants inthe familiar objects condition of Experiment 3. The agent wasalways a truck (on the left in this figure), which moved awayfrom the van (on the right in this figure).



Post hoc analyses were carried out to understandthis interaction. Separate ANOVAs were carried outfor the novel objects condition and the familiar ob-jects condition. In the novel objects condition, therewas a main effect of test type, F(3, 45) � 12.68, p �.001, MSE � .09. Newman-Keuls tests revealed that theproportion of “yes” responses to the object � motionmatch events was higher than to any of the other testtypes. The proportions of “yes” responses to the motionmatch and object match test events did not significantlydiffer, whereas the proportions of “yes” responses toeach of these two test types was significantly greaterthan the proportion of “yes” responses to the nomatch events. Thus, children in the novel objects con-dition were less likely to accept an event as an exam-ple of the verb “morping” if either the motion or theobject in motion was changed, and they were evenless likely to accept an event if both the object and itsmotion had changed.

There was also a significant main effect of test typein the familiar objects condition, F(3, 45) � 13.27, p �.001, MSE � .08. Unlike in the novel objects condition,however, Newman-Keuls tests revealed that the pro-portions of “yes” responses to the object � motionmatch events and to the motion match events werenot significantly different. These two proportions, inturn, were significantly higher than the proportionsof “yes” responses to object match and no matchevents. There was no significant difference betweenthe proportions of “yes” responses to the object matchevents and to the no match events. These results indi-cate that children in the familiar objects condition at-tended primarily to the motions of objects whenlearning verbs.


The results of individual participants are presented inFigure 12. As can be seen in the top panel of Figure 12,several different patterns of performance were againevident in children in the novel objects condition.Five children (i.e., those in the upper left-hand corner)attended primarily to motions. Nine children (i.e.,those in the lower right-hand corner) attended primar-ily to objects. The remaining 2 children (i.e., thosealong the diagonal) attended equally to objects andmotions. Of these, 1 was relatively liberal and 1showed no clear response pattern.

As can be seen in the lower panel of Figure 12, sev-eral different patterns of performance were also evi-dent in children in the familiar objects condition.Nine children (i.e., those in the upper left-hand cor-ner) attended primarily to motions. In marked con-trast to the novel objects condition, only 1 child (i.e.,

the one in the lower right-hand corner) attended pri-marily to objects. Finally, six-children (i.e., thosealong the diagonal) attended equally to objects andmotions. Of these, 2 were relatively conservative, 2were relatively liberal, and 2 showed no clear re-sponse pattern. Thus, individual patterns of perfor-mance, as well as the aggregated results, reveal thatchildren were more likely to attend to motions than tofamiliar objects when learning a verb.

Figure 12 The performance of individual participants in Ex-periment 3. Each dot represents the performance of 1 partici-pant. The x-axis represents the proportion of “yes” responsesin object match events, which involved an agent body that wasconsistent with the tested word, but a path that was inconsis-tent. The y-axis represents the proportion of “yes” responsesin motion match events, which involved a path that was con-sistent with the presented word, but an agent body that wasinconsistent.


GENERAL DISCUSSION

The results of the present research revealed that 3.5-to 4-year-old English-speaking children attend just asstrongly to novel objects as they do to motions whenpresented with a novel verb. Experiment 1 revealedthat children associated nouns primarily with objects,even though the objective relations of nouns to thepaths of those objects were equally strong. In contrast,children associated verbs just as strongly with objectsas with paths. These verb-learning results contrastedwith the results from adults, who associated verbsprimarily with paths. Experiment 2 revealed that chil-dren also attended to novel objects when learning adifferent kind of verb, namely manner of motionverbs. Finally, Experiment 3 employed well-knownreal-world objects and found that children attendedmore strongly to motions than to objects in the sameverb-learning task.

Implications for Theories of Verb Learning

These results have a number of implications fortheories of verb learning. First, by the time they are3.5 to 4 years of age, English-speaking children havesome knowledge that nouns and verbs convey differ-ent meanings. In particular, children were more likelyto attend to motions when learning a verb than whenlearning a noun. Different children varied in theirverb-learning strategies, however. Some children at-tended primarily to motions, some attended primar-ily to objects, and still others attended equally to ob-jects and motions. A possible explanation for thisvariation in verb-learning strategies is that childrenmay differ in their knowledge of object categories. Inparticular, some children may have recognized thenovel objects in Experiments 1 and 2 as examples ofthe superordinate category “bug,” whereas othersmay have treated these same objects as truly novel.Children who classified the two objects (i.e., “zee-bees” and “taygos”) into the same superordinate cat-egory may not have attended very strongly to theirindividuating characteristics, but rather may have re-lied on their knowledge of the types of actions thatare typically performed by bugs to help them deter-mine the meaning of a verb. Children who treated theobjects as truly novel, on the other hand, may have at-tempted to learn the types of actions performed byeach of the two categories of bugs, thus requiringthem to attend to objects as well as actions.

A second implication of the present results is thatthere is development in the tendency to focus onnovel objects when learning verbs. In particular, thepresent findings are consistent with the prior results

of Behrend (1990) and Forbes and Farrar (1995) inshowing that 3.5- to 4-year-old English-speaking chil-dren attend more strongly to novel objects than doadults when learning a verb. The objects involved inthese prior studies were instruments, whereas the ob-jects in the present study were agents. The findingthat both of these types of objects are given greater at-tention by children than by adults in a verb-learningsituation suggests that a tendency for children to fo-cus on novel objects during verb learning is very gen-eral in nature and not restricted to particular classesof objects. Adults may be less likely than children toattend to novel objects during verb learning becauseadults are more likely to infer the larger category towhich those objects belong. Adults may thus be ableto make use of existing category knowledge ratherthan having to learn what types of actions a giventype of object is capable of performing. Consistentwith this explanation, children in the familiar objectscondition of Experiment 3 performed quite similarlyto adults who were presented with novel objects inExperiment 1, suggesting that adults treated these novelobjects as examples of familiar categories.

A third implication of the present results is thatchildren are not as likely to attend to familiar objectsas they are to attend to novel objects when learning averb. This finding is consistent with Gentner’s (1981)theory of representational differences between nounsand verbs. According to this theory, representationsof object concepts are highly cohesive, with mutuallyfacilitatory connections among the various attributesof an object. Objects are thus interpreted in largely thesame way, regardless of the context in which they oc-cur, because the detection of even a subset of the at-tributes of an object is sufficient to activate the repre-sentations of all of the attributes of an object. Incontrast, representations of relational concepts aremuch less internally cohesive, with a smaller numberof mutually facilitatory connections among “core” at-tributes of a relation and a larger number of connec-tions to other concepts such as nouns. A verb maythus have different meanings in the context of differ-ent objects, because the external connections to therepresentations of these objects may result in differentsets of attributes becoming activated. This theory sug-gests that to learn the meaning of a verb in the contextof unfamiliar objects, children first have to establishrepresentations for those objects. Thus, children mayattend to objects as well as motions when those ob-jects are unfamiliar. In contrast, when learning themeaning of a verb in the context of familiar objects,representations of those objects are already estab-lished, and thus less attention to objects is required.

Greater attention to novel than to familiar objects is


also consistent with Kersten’s (1998a) division oflabor theory of motion representation in language.According to this theory, the role of English mannerof motion verbs is to select from a number of mannersof motion associated with an object category. Thus,knowledge of the different motions (e.g., strolling,sauntering) associated with a category (e.g., human)may be necessary before one can successfully inter-pret a verb used in the context of that object category.Learning which motions are associated with whichobject categories requires attention to both the ap-pearances of objects and the motions of those objects.Once an object is familiar, however, learning a verbmay only require figuring out which of the motionsassociated with that object covaries with the presenta-tion of the verb.

Although the division of labor theory predicts at-tention to novel objects during the learning of mannerof motion verbs, children in the present experimentsalso attended to novel objects when learning pathverbs. According to the division of labor theory,path verbs mean roughly the same thing regardless ofthe nature of the objects involved. For example, themotion of “colliding” can be represented reasonablyaccurately by the gesture of two fists moving into con-tact with one another, regardless of whether theobjects involved are football players, cars, or planets.Thus, the division of labor theory predicts that chil-dren would not have to attend to the nature of the ob-jects involved in order to learn path verbs. The factthat children did attend to novel objects when learn-ing path verbs may indicate that object categoryknowledge is useful not only when learning mannerof motion verbs but also when learning path verbs.Alternatively, children in the present experimentsmay simply have applied strategies that were usefulin their prior verb-learning experience. Because thechildren in the present experiments were Englishspeakers, this prior verb-learning experience in alllikelihood involved primarily manner of motionverbs. Thus, children may have applied strategiesthat were useful in the learning of manner of motionverbs, regardless of whether they were in fact learn-ing manner of motion or path verbs.

Relative Novelty of Actions and Objects

There is a possible alternative explanation for thepresent results. Given the evidence for a preferencefor novelty in verb learning (Golinkoff, Jacquet, Hirsh-Pasek, & Nandakumar, 1996; Merriman, Evey-Burkey,Marazita, & Jarvis, 1996; Merriman, Marazita, & Jarvis,1993), it is possible that children simply attended to themost novel characteristic of an event when they were

presented with a verb to be learned. Children maythus have attended to the buglike creatures in theseexperiments because they were the most novel char-acteristic of the events. In contrast, children may havebeen less likely to attend to the cars and trucks in Ex-periment 3 because these were less novel.

This hypothesis does not explain, however, whyadults attended less strongly to novel objects than didchildren. Adults as well as children show a preferencefor novelty when learning verbs (Kersten, Goldstone,& Schaffert, 1998), and yet adult verb learners in Ex-periment 1 focused primarily on motions even thoughthey were presented with the same novel objects aswere children. For the novelty hypothesis to be a via-ble explanation for the present results, additional ex-planatory mechanisms would be required to accountfor this developmental evidence. For example, onepossible explanation could be that adults are moresensitive than children to markers of syntactic cate-gory, and thus are able to ignore the novelty of objectsand focus on motions when they are presented witha novel word with verb morphology (e.g., “-ing,”“-ed”). Children, on the other hand, may be moredriven by visual stimulus characteristics, and thusmay focus on the most novel or interesting character-istic of an event, regardless of the type of word in-volved. Additional research is necessary to dissociatethis novelty hypothesis from the hypothesis that verblearning is dependent on knowledge of objects.

Conclusions

The present experiments provide evidence that 3.5-to 4-year-old English-speaking children attend notonly to motions but also to novel objects when theyare presented with a novel verb. This attention tonovel objects during verb learning is proposed to beuseful because the nature of an object places con-straints on what a verb may mean in the context ofthat object. This account also offers an explanation forwhy, in many languages, children acquire nouns be-fore verbs. In particular, both noun and verb learningare dependent on object category knowledge, butverb learning requires the further step of determiningwhich of the different motions associated with an ob-ject category is relevant to the meaning of a verb.Thus, provided that nouns are used with sufficientfrequency and saliency as labels for these object cate-gories, nouns will tend to be learned earlier than verbs.

ACKNOWLEDGMENTS

The authors would like to thank Stacy Aikins, Ca-milla Fancher, Heather Farmakis, Joycy Lacombe, Anel


Marachena, Nushka Remec, Marc Riccardi, JenniferSchaefer, Jessica Shuman, Blythe Silberman, and ScottTurner for their help in conducting the experiments.They would also like to thank Julie Earles for herhelpful comments on earlier versions of the article. Thisresearch was supported by NICHD Grant HD 28675.

ADDRESSES AND AFFILIATIONS

Corresponding author: Alan W. Kersten, Departmentof Psychology, Florida Atlantic University, Boca Raton,FL 33431-0991; e-mail: [email protected]. Linda B. Smithis at Indiana University, Bloomington, IN.

REFERENCES

Au, T. K., Dapretto, M., & Song, Y. K. (1994). Input vs. con-straints: Early word acquisition in Korean and English.Journal of Memory and Language, 33, 567–582.

Behrend, D. A. (1990). The development of verb concepts:Children’s use of verbs to label familiar and novelevents. Child Development, 61, 681–696.

Bertenthal, B. I., Proffitt, D. R., & Cutting, J. (1984). Infantsensitivity to figural coherence in biomechanical motions.Journal of Experimental Child Psychology, 37, 213–230.

Bloom, L. (1973). One word at a time: The use of single word ut-terances before syntax. The Hague: Mouton.

Bloom, L. (1981). The importance of language for languagedevelopment: Linguistic determinism in the 1980s. An-nals of the New York Academy of Sciences, 379, 160–171.

Brown, R. (1957). Linguistic determinism and the part ofspeech. Journal of Abnormal and Social Psychology, 55, 1–5.

Choi, S., & Gopnik, A. (1995). Early acquisition of verbs inKorean: A cross-linguistic study. Journal of Child Lan-guage, 22, 497–529.

Earles, J. L., & Kersten, A. W. (2001). Semantic context influ-ences memory for verbs more than memory for nouns.Manuscript submitted for publication.

Forbes, J. N., & Farrar, J. M. (1995). Learning to representword meaning: What initial training events reveal aboutchildren’s developing action verb concepts. CognitiveDevelopment, 10, 1–20.

Gentner, D. (1981). Some interesting differences betweenverbs and nouns. Cognition and Brain Theory, 4, 161–178.

Gentner, D. (1982). Why nouns are learned before verbs:Linguistic relativity versus natural partitioning. InS. Kuczaj II (Ed.), Language development: Vol. 2. Lan-guage, thought, and culture (pp. 301–334). Hillsdale, NJ:Erlbaum.

Gentner, D., & Boroditsky, L. (2001). Individuation, relativ-ity, and early word learning. In M. Bowerman & S.Levinson (Eds.), Language acquisition and conceptual devel-opment (pp. 215–256). Cambridge, U.K.: Cambridge Uni-versity Press.

Gleitman, L. (1990). The structural sources of verb mean-ings. Language Acquisition, 1, 3–55.

Golinkoff, R. M., Jacquet, R. C., Hirsh-Pasek, K., & Nanda-

kumar, R. (1996). Lexical principles may underlie thelearning of verbs. Child Development, 67, 3101–3119.

Gopnik, A., & Choi, S. (1995). Names, relational words, andcognitive development in English and Korean speakers:Nouns are not always learned before verbs. In M. Tomasello& W. E. Merriman (Eds.), Beyond names for things: Young chil-dren’s acquisition of verbs (pp. 63–80). Mahwah, NJ: Erlbaum.

Gropen, J., Pinker, S., Hollander, M., & Goldberg, R. (1991).Affectedness and direct objects: The role of lexical se-mantics in the acquisition of verb argument structure.Cognition, 41, 153–195.

Huttenlocher, J., & Lui, F. (1979). The semantic organizationof some simple nouns and verbs. Journal of Verbal Learn-ing and Verbal Behavior, 18, 141–162.

Kersten, A. W. (1998a). A division of labor between nounsand verbs in the representation of motion. Journal of Ex-perimental Psychology: General, 127, 34–54.

Kersten, A. W. (1998b). An examination of the distinctionbetween nouns and verbs: Associations with two differ-ent kinds of motion. Memory and Cognition, 26, 1214–1232.

Kersten, A. W., & Billman, D. (1997). Event category learn-ing. Journal of Experimental Psychology: Learning, Memory,and Cognition, 23, 638–658.

Kersten, A. W., Goldstone, R. L., & Schaffert, A. (1998). Twocompeting attentional mechanisms in category learning.Journal of Experimental Psychology: Learning, Memory, andCognition, 24, 1437–1458.

Levin, B., & Rappaport Horav, M. (1991). Wiping the slateclean: A lexical semantic exploration. Cognition, 41, 123–151.

Merriman, W. E., Evey-Burkey, J. A., Marazita, J. M., &Jarvis, L. H. (1996). Young two-year-olds’ tendency tomap novel verbs onto novel actions. Journal of Experimen-tal Child Psychology, 63, 466–498.

Merriman, W. E., Marazita, J., & Jarvis, L. H. (1993). Four-year-olds’ disambiguation of action and object word ref-erence. Journal of Experimental Child Psychology, 56, 412–430.

Naigles, L. (1990). Children use syntax to learn verb mean-ings. Journal of Child Language, 17, 357–374.

Naigles, L. R., & Hoff-Ginsberg, E. (1995). Input to verblearning: Evidence for the plausibility of syntactic boot-strapping. Developmental Psychology, 31, 827–837.

Nelson, K. (1973). Structure and strategy in learning to talk.Monographs of the Society for Research in Child Develop-ment, 38(1–2, Serial No. 149).

Pinker, S. (1989). Learnability and cognition. The acquisition ofargument structure. Cambridge, MA: MIT Press.

Talmy, L. (1985). Lexicalization patterns: Semantic structurein lexical forms. In T. Shopen (Ed.), Language typology andlinguistic description: Vol. 3. Grammatical categories and thelexicon (pp. 57–149). New York: Cambridge University Press.

Tardif, T. (1996). Nouns are not always learned before verbs:Evidence from Mandarin speakers’ early vocabularies.Developmental Psychology, 32, 492–504.

Tardif, T., Gelman, S., & Xu, F. (1999). Putting the noun biasin context: A comparison of English and Mandarin. ChildDevelopment, 70, 620–635.

Tomasello, M., & Kruger, A. C. (1992). Acquiring verbs inostensive and nonostensive contexts. Journal of ChildLanguage, 19, 311–333.

Documents

Attention to Novel Objects during Verb Learningpsy2.fau.edu/~kersten/Papers/kersten+smith2002.pdf · learning may beneﬁt from prior learning of nouns. Noun learning per se is not