Symposium
Chair: Susan A. Graham
Discussant: D. Geoffrey Hall
One of the most challenging tasks faced by the young child in acquiringher native language is the acquisition of a vocabulary. Studies havefound there are many sources of information available to the child thatcould be helpful during word learning including socio-pragmatic cues,attentional and conceptual cues, as well as linguistic cues (e.g.,Akhtar, Carpenter, & Tomasello, 1996; Baldwin, 1991, 1993; Hall, 1994;Hall & Graham, 1999; Tomasello & Akhtar, 1995). In this symposium, wepresent four papers that examine young children’s disambiguation ofnovel words. Each paper provides a diverse perspective on early wordlearning by focussing on a different type or types of cues children mayrely on when linking words with their referents. The first papercompares young children’s propensity to extend labels to new exemplarsto their treatment of other types of newly learned information, therebyproviding insight into young children’s understanding of the uniquenessof words. The second paper examines the relative influence of differenttypes of attentional cues (e.g., general, specific, and arbitrary cues )on children’s disambiguation of a novel object word when faced with morethan one nameless object. The third paper presents evidence thatchildren assess a speaker’s knowledge of the discourse context, anddisambiguate the referent of a new word on the basis of thatassessment. The final paper investigates the role of conceptual cues(i.e., animacy information) and linguistic cues (i.e., pronominalinformation) during verb acquisition. Taken together, these papersprovide evidence that children can rely on a diverse set of cues whenlearning new words, an issue of great interest in the field of wordlearning. Our discussant, with an extensive research program in earlyword learning, will provide an insightful critique of these papers.
Details of individual items:
paper
Recent studies by Markson and Bloom (1997) showed that 3-year-oldslearned a novel name and novel fact equally well, leading theseresearchers to conclude that word-learning was not a unique type oflearning. However, three studies by Behrend et al. (1999) showed that3-year-olds were more likely to extend novel names than novel facts toadditional exemplars, suggesting that these children already understandthat count nouns are uniquely and necessarily extendable while othertypes of facts are not. The two studies presented in this paperexamine the origins of this understanding by examining two-year-olds’extension patterns and by examining parental input while teaching novelnames and novel facts.Methods: In Study 1, 2 _- and 3-year-old children were taught a novellabel (e.g. 'This is a koba.') for one object and a novel fact (e.g.'this fell in the sink.')) for another object. After each teachingbout, children were than shown an array which included the originaltraining object, three additional exemplars of the training object(which differed perceptually from the original), and five distracters.Children were then asked to show the experimenter the koba/thing thatfell in the sink. After the children’s first choice, the experimenterprobed for additional choices by asking, 'Are there any otherkobas/things that fell in the sink here, or not?' The order of the nameand fact learning trials was counterbalanced. In Study 2, 16two-year-olds and their parents participated in a play session in whichparents were asked to teach their child a novel name for an unfamiliarobject and a novel fact for another unfamiliar object from and array of20 toys. Parental teaching behavior was analyzed in terms of numbersof exemplars used during teaching, linguistic content of teachingepisodes, and pragmatic contexts of label versus fact teaching.Results: In Study 1, the number of exemplars chosen by the children inthe name learning versus fact learning conditions was compared. Datafrom 19 children who have been tested show that children are extendingthe novel label significantly more frequently (M3.8, Maximum 4) toadditional exemplars than they are extending the novel fact (M 2.4) toadditional exemplars. No child failed to extend the label to at leastone additional exemplar, suggesting that 2-year-olds honor the necessaryextendibility of count nouns. Data analysis from the parent teachingsessions of Study 2 is in its preliminary stages. However, parents donot differ in the number of different exemplars they use when teachingtheir children a novel label or a novel fact, suggesting that children’smore frequent extension of labels than facts is not simply a reflectionof input frequency. However, we expect that other, more subtle aspectsof parental input during label teaching and fact teaching will makeimportant contributions to the emergence of young children’s propensityto extend labels more frequently than other types of newly learnedinformation.
paper
Researchers have recently demonstrated that young children can rely on anumber of external cues when asked to disambiguate the referent of anovel word in the presence of more than one nameless object. Forexample, several studies have documented that children’s word-mappingdecisions can be affected by a speaker’s gaze direction (e.g., Baldwin,1991, 1993), a speaker’s affective and/or behavioral cues (e.g.,Tomasello & Barton, 1994), and the relative novelty of objects in thediscourse context (e.g., Akhtar et al., 1996).In the present experiments, we examined the influence of different typesof attentional cues on 3-year-olds’ disambiguation of novel objectwords. Twenty-four children were tested in each of three experiments.The procedure was as follows: Across a series of trials, theexperimenter first presented the child with two unnamed objects, one ofwhich she picked up and described to the child. In Experiment 1, shesimply directed the child’s attention to the object (e.g., 'Look at thisone.'). In Experiment 2, the experimenter described an uninformativefact about the object (e.g., 'It’s on the table.'). In Experiment 3,she described a property of the object (e.g., 'You can squeeze thisone.'). No attention was directed to the second object in all threeexperiments. Once the object had been described, the experimenter askedthe children for the referent of a novel object word ('Show me a dax.').In Experiments 1 and 2, children chose randomly between the two namelessobjects when extending a novel label, indicating that providing childrenwith either an uninformative fact about an object or directing generalattention to one of the objects did not assist them in disambiguating anovel word’s referent. In contrast, in Experiment 3, children extendedthe novel object words to those objects which had not received anyattention significantly more than would be expected by chance (M 62.50%, p < .05). Thus, when children had been provided with a cue thatwas specific to the object (i.e., a property of the object wasdescribed), they extended the novel word to the object that hadinitially been ignored by the adult speaker. These findings suggestthat children may have treated the property as a way to describe orlabel the object and thus, mapped the novel label to the object that didnot yet have any label. Taken together, these studies indicate thatsimply providing any type of attentional cue does not assist children indisambiguating the referent of a novel object word. We are currentlyfollowing up these findings in a second set of studies which examine 22to 26-month-old children’s word-object mapping when different types ofattentional cues (general, specific, and affective-intentional) aredirectly contrasted on a given trial. Results to date indicate asignificant correlation between children’s reliance onaffective-intentional cues to disambiguate word-object reference andvocabulary size (r .50, p <. 03, n19).
paper
Researchers currently debate how young children infer the referent of anew name. Akhtar, Carpenter, and Tomasello (1996) argue that 2-year-oldsare capable of assessing a speaker’s knowledge of the discourse context,and infer a speaker’s intent based on such assessment. Samuelson andSmith (1998) offered an alternative account of Akhtar et al.’s findings,suggesting that all that is involved in children’s selection of thereferent of a new name is a child-centered, dumb mechanism driven byattention and memory demands. The present study was designed to testthese alternative accounts. Twenty-two 2-year-olds were tested in aprocedure identical to the one used by Samuelson and Smith (1998). Theexperimenter showed children three objects while sitting on the floor,and played with the objects by shaking them in a metallic container.The experimenter then invited children to move to a table, where shepresented them a fourth object and played with it by spinning it in abasket. The experimenter and the child then returned to the floor, theexperimenter put all four objects in a transparent box, and a puppetcame out of his house to play with them. In a Same-Speaker condition,which was identical to Samuelson and Smith’s (1998) study, theexperimenter looked into the box and said repeatedly 'There’s a tiga inhere'. In a Different-Speaker condition, the puppet, who did not knowwhat happened in the discourse context before his appearance, lookedinto the box and said 'There’s a Tiga in here'. The experimenter, thechild, and the puppet played with all four objects by shaking them inthe metallic container, and then depending on the child’s condition, theexperimenter or the puppet asked the child to give her/him the 'Tiga'.If children’s naming decisions are based on child-centered attentionalmechanisms sensitive to contextual variations, as Samuelson and Smithclaim, it should not matter who does the naming. Children in bothconditions should pick the fourth object because its presentationcontext stands out in memory and therefore in attention. Alternatively,if children’s naming decisions are guided by their pragmatic skills, inparticular their evaluation of a speaker’s knowledge and intent, it doesmatter who does the naming. Children in the Same-Speaker conditionshould pick the fourth object, because they infer that the experimenteris probably naming the object she treated in a different/special way.Children in the Different-Speaker condition should pick randomly,because the puppet did not present any of the objects, was absent whenthe objects were presented, and thus children have no clue as to hisreferential intent. So far we have found that 4 of the 10 children inthe Same-Speaker condition picked the fourth object, whereas none of the12 children in the Different-Speaker condition did so, X2 (1, N 22) 5.87, p < .05. This finding suggests that 2-year-olds’ interpretationof names is not simply guided by dumb attentional mechanisms, but ratheris the product of children’s assessments of a speaker’s knowledge of thediscourse context, and their inferences about a speaker’s referentialintent.
paper
Although there have been some investigations of children’s attention tovarious cues during noun learning, little is known about children’s useof similar cues when acquiring a new verb. The present studies testwhether 2 1/2- and 3-year-old children make use of two different kindsof cues, conceptual cues (i.e., animacy information) and linguistic cues(i.e., pronominal information) during verb acquisition. Animacyinformation may be useful during verb learning because many verbschildren hear refer to the actions of animate agents (e.g., climb, eat,drive). Moreover, Slobin (1981; 1985) has argued that an attention toprototypical (animate-inanimate) events may assist children in learninglanguage. Pronominal information may also be important to the youngchild because pronouns provide children with word-level cues to sentencestructure (e.g., she vs. her in English). In addition to examiningwhether children attend to these kinds of cues during verb learning, wewere also interested in examining the extent to which children werewilling to extend new verbs learned within particular animacy or lexicalcontexts to new situations. In Study 1, we taught 32 2 1/2- and 32 3-year-old children two novelverbs and then examined their understanding of these verbs using anenactment task. Children heard new verbs in sentences containing eithernouns or pronouns, and were also assigned to one of four animacytraining conditions (e.g., animate-animate events). We then testedchildren’s ability to extend the novel verbs to new lexical and animacycontexts. The younger children were extremely conservative in the verbcomprehension task. The older children were able to extend theirknowledge of the new verb to new sentences, yet also demonstrated aninfluence of animacy and word type. Specifically, 3-year-oldsdemonstrated the best understanding of a new verb when they initiallylearned the verb in sentences that included pronouns, particularly whenthose sentences referred to events with animate agents and inanimate oranimate patients (t (7) 4.7, p < .01 and t (7) 3.5, p < .01respectively).In Study 2, we were interested in further examining the role thatpronouns may play during verb acquisition. Sixteen 2 1/2- and 163-year-old children were taught two novel verbs. One verb was heardembedded in sentences that contained only nouns and the other insentences containing only pronouns (here, children heard examples acrossthe four animacy configurations). An analysis of the children’sproductions revealed a significant Creative (creative, noncreative) byVerb (novel, familiar) interaction, F (1, 31) 7.44, p< .02. Childrenat both ages were less likely to produce creative utterances with novelverbs as compared to familiar ones. The set of studies provides newevidence of children's attention to animacy and lexical information, aswell as the extent of their conservatism, during verb acquisition.