Poster group
Details of individual items:
poster
It is well established that young children and adults prefer tofix a novel label to a novel object in word learning situations. However,in recent work, we have observed younger infants (14-15 months) behavingin ways inconsistent with this pattern. The goals of the present studyaimed to quantify the developmental progression of this word learningstrategy from infancy through adulthood, and to motivate the possibilityof developmental continuity given seemingly discontinuous performance. Both the onset and the psychometrics of the novel label/novelobject strategy have been difficult to study without a sensitive measureof on-line processing. The current paradigm incorporates positive aspectsof earlier efforts in assessing word comprehension and yields continuousreal-time data. Participants sit between two computer monitors. Duringeach trial, two different objects appear simultaneously, one on eachscreen; preferential looking is measured prior to onset of a label, 'Lookat that [CAR]' which correctly labels only one object. Comprehension ismeasured as increased looking time to the labeled object above preference.Known/Known trials, where both objects have known labels, illustrate theeffectiveness of this paradigm for word comprehension and serve as acomparison for Known/Novel trials. It is these Known/Novel trials, wherethe novel object is given a novel label, which elicit the novellabel/novel object strategy. Participants were 20 15-month-olds, 30 17-month-olds, 20 3-and4-year-olds, and 20 adults. Choice of known and novel objects and thespeed of presentation were set independently for each age group. Allsubjects saw a random combination of 12 Known/Known trials and 12Known/Novel trials. On six of the Known/Novel trials the novel object waslabeled. On the other six trials the known object was labeled. All age groups successfully paired known object labels with theirreferents. As shown in Figure 1, looking time to the known objectsincreased significantly above baseline preference; all p < 0.001.However, only the preschoolers and adults paired novel objects with theirlabels (see Figure 2); p < 0.005. 15-month-olds significantly decreasedtheir looking to the novel object, p < 0.03, and 17-month-olds show nopreference. Thus, there is an apparent discontinuity between 15-monthsand preschool. Both adults and preschoolers increase looking to the novelobject [DAX] when given a novel label, 'Look at that DAX'. In contrast,17-month-olds show no change in looking and, as Figure 2 representspercent of total looking time, 15-month-olds increase looking to the KNOWNobject [CAR]. That is, given the label, 'Look at that DAX,' 15-month-oldsbehave counter to the novel label/novel object strategy and increaselooking to the [CAR]. Given this apparent discontinuity, can we describethe change that occurs or, if not, is there some interpretation ofcontinuity which can embrace this difference? One possibility forcontinuity is that subjects at all ages must first decide that 'DAX'cannot be another label for [CAR] before settling on the novel object asthe referent. Under this hypothesis, increased looking to the [CAR] mayshow that subjects must first decide that 'DAX' cannot be a label for the[CAR]. One source of data which may support this continuity claim comesfrom adult patterns of switch in gaze. When labeling of the known object'CAR' occurs while adults are already looking at that object, they rarelyswitch their gaze. However, when labeling of the novel object 'DAX'occurs while adults are looking at that object, they increase switches ingaze to check the known object [CAR] (X2 91.88, p < 0.001). Suchswitches are consistent with the continuity proposal that all subjectsmust first decide that 'DAX' cannot be a label for [CAR].
poster
In approximately eight months, children go from a limited vocabulary ofone or two words to a productive corpus of 300+ items, with some estimatesindicating that children are learning in excess of nine new words a day.Unfortunately, the mechanisms whereby such a dramatic shift occurs are notyet well understood. Explanations range from social/pragmatic accountswhich highlight the role of the parent-child interaction, toconstraints/principles accounts which emphasize the importance of cognitiveheuristics in language acquisition, to associationistic theories whichemphasize domain-general learning mechanisms of attention. In addition,there are also interactionist models that attempt to combine aspects of allof these positions. Because any of these accounts might seem plausible on the surface, thereare numerous reasons why they benefit from computational implementation.First, the level of specificity involved in creating a computational modelis necessarily greater than that utilized in theory building. Thus, theeffect of a larger array of details, absent in the theoretical models can beexamined in the computational ones. Second, computational models can makepredictions, sometimes unexpected, about the behavior of the populationunder study. As a result, one can use the model to design future empiricalexplorations, which may help distinguish critical predictions made by eachof the prevailing theories. Finally, computational models shed light uponthe potential neural mechanisms involved in the behavior in question. Thus,considerations about the biological plausibility of the system may furtherconstrain theorizing and inform upon subsequent explanations. In this poster, four models for word learning, based on each of thepredominant theories are implemented. Each addresses the fundamentalmechanistic assumptions of the four approaches as well as their plausibilityof implementation within the human brain. In addition, these modelsexplore, at a high level of detail, the behavioral predictions made bythese. While the first three models simulate the associationistic,constraints/principles and social-pragmatic views, respectively, the finalmodel simulates the interactionist perspective. Taken together, thesemodels provide quantitative, testable, hypotheses of behavioral differencesbetween current theories of word learning. In the final section of theposter, empirical data is presented which tests each of these hypotheses,concluding that only the interactionist perspective can account for theresults seen.
poster
A major question concerning the acquisition of nominals concernsthe type of information (i.e., syntactic or semantic) children use tocategorize novel nouns. In particular, the relative weight given to syntaxand semantics in the categorization of novel nouns has been a question thathas fueled many research endeavors. While previous research has focused onthe count/mass distinction, other domains are equally relevant. Thepresent research examines toddlers' categorization of novel NP-type nouns.NP-type nouns (e.g., school , church) are nouns that can be used both ascount nouns (e.g., She is at a school ) and as noun phrases (e.g., She isat school). Soja (1994) has demonstrated that as soon as young childrenare reliably producing determiners with count and mass nouns, they are alsousing both sentence constructions of an NP-type noun. Soja (1994) has argued that all NP-type nouns must satisfy twocriteria in order to be so designated: 1) all NP-type nouns must refer toinstitutions associated with a ritualized, habitual activity;, and 2) eachNP-type noun must belong to one of many conflation classes where allmembers from a specific class refer to a cultural institution of the samekind. Some of these classes are education, religion, and meals. In aseries of experiments, Burns and Soja (in press) demonstrated that adultsand young children are able to categorize a novel nominal as an NP-typenoun if the novel noun satisfied the second criterion, regardless of thesyntactic information presented. In addition, they also demonstrated thatthree (Burns, 1997) and five year old children will ignore syntacticinformation if it is inconsistent with the semantic information presentedalong with the novel nominal. The present research extends these findings by testing the firstsemantic criterion. Ninety-eight children were told stories containing anovel nominal (e.g., 'blicket'). The stories manipulated whether the novelnoun satisfied the first criterion or not. In addition, syntacticinformation was also varied such that the novel noun could be presentedsyntactically as either a count or an NP-type noun. This was presented ina manner either consistent or inconsistent with the semantic informationpresent in the stories. The children were then asked to make a forcedchoice between members of a pair of test sentences that varied only by thepresence or absence of an 'a' before the novel nominal (recall thatsingular count nouns require a determiner but NP-type nouns do not). Itwas found that children were able to categorize the new nouns solely on thebasis of semantic information. Syntactic support was not necessary. This research supports the first semantic criterion for inclusioninto the NP-type noun category. In addition, it also supports the claimthat semantics play a dominant role in the categorization of novel nouns,and that while syntactic information is useful for such categorizations, itis not necessary. These findings are discussed in light of Bloom (1990)and Gordon's (1988) semantic accounts of the count/mass distinction.
poster
Words cannot explain unless a person already knows what words mean. Yet learning what words mean is what the infant 'means' to learn. Caregivers have resolved this dilemma by linking speech and ongoing events with gestures when communicating with their infants. These dyads engaged in cycles of perceiving and acting that promoted the comprehension of speech. They sensitively monitored, modified, and 'upped the perceptual ante' as their infants developed.Extending these findings to infant production, this research examined how the interactional environment of the infant educates attention to perceive referring actions. We investigated whether the target/topic of caregiver messages changes as infants develop and relates to their infants' early lexicon. Specifically, does the complexity of caregiver messages increase with development? Do these shifts in complexity inform infants' later production? If so, the target of mutual attention in caregivers' speech should be a step or two ahead of the infant's expressive level. These caregiver methods may be characteristic of one class, a particular educational level, language community, or culture. On the other hand, these caregiver practices may be ubiquitous or range over some continuum (Schieffelin & Ochs, 1986; Watson-Gegeo & Gegeo, 1986). We examined monthly video data from a naturalistic, longitudinal study of five European-American and six Latino families in the Western USA from the infant's sixth month through the one-word period. During attention-directing interactions caregivers directed infants to notice specific elements, relations, or events. This collection of interactions included all instances of perceptual imperatives uttered by caregivers, such as look!, listen!, touch!, smell!, taste!, etc., their accompanying gestures, and the gestures alone. Targets of attention encompass non-dynamic person/object, agent-action-object-recipient sequences, or possession, instrument, location, and so on. The five attention-directing gestures include embody, show, demonstrate, point, and look.Multivariate frequency analysis of the observed frequencies determined the interdependence of five variables (infant expressive level, target of attention, gesture, language, and subject). These findings suggest that caregivers' referring actions accompany and mirror an ever-expanding and increasingly challenging array of activities that may facilitate learning to talk. In both language communities, similar targets of attention expressed in caregiver messages occur a step or two ahead of the expressive level of the infant. Attentive caregiving practices are not limited to Euro-American middle class, highly educated families. These methods occur as well among Latino working class families with little formal education. However, the messages of Latino caregivers displayed a greater frequency and elaboration of gestures than those of European-Americans. Recent studies of linguistic typology may account for observed variations in gestural style observed among Spanish- and English-speaking families (Slobin, 1997; McNeill, 1997, Talmy, 1985). Variations in the expression of motion events related to linguistic typology may explain these differences.
poster
The extraordinary ability of children to acquire vocabulary so quickly andseemingly easily has motivated researchers to determine the types ofinformation that guide children in disambiguating word reference.Researchers have recently demonstrated that young children are sensitiveto a number of socio-pragmatic cues during language acquisition, and inparticular, when asked to map a novel word onto an object in scenarioscontaining more than one nameless object. These socio-pragmatic cuesinclude: attention or gaze direction (e.g., Baldwin, 1993), affect(including facial expression; e.g., Tomasello & Barton, 1994), discoursenovelty (e.g., Akhtar et al., 1996), and speaker's intentionality (Baldwinet al., 1996). In the present studies,twenty-two-to-twenty-six-month-olds' use of different attentional cues tomap a new word onto a new object in the presence of two nameless objectswas examined. Three types of attentional cues (i.e., general, specific,and affective) were directly contrasted. Productive vocabulary size wasalso examined using the MacArthur Communicative Development Inventory. InStudy 1, the experimenter directed 'like' (affective) attention toward oneobject (e.g., 'See, I like this one.') and general attention toward theother object (e.g., See, look at this one.'). It was found that toddlersresponded randomly when asked for the referent of a novel word. In Study2, the experimenter demonstrated a property (specific attention) of oneobject (e.g., 'See, you can squeeze it.') and directed general attentionto the other object. Again, toddlers responded randomly when asked forthe referent of a novel word. Finally, in Study 3, the experimenterdirected 'like' attention toward one object and demonstrated a property ofthe other object. Toddlers chose objects given a property at above chancelevels. As well, a significant positive correlation was found betweenproductive vocabulary size and the percentage of objects directed 'like'attention that were chosen, and a marginally significant (p .062)negative correlation, was found between productive vocabulary size and thepercentage of objects directed general attention that were chosen. Therewas no significant correlation between productive vocabulary size andpercentage of objects given a property that were chosen. Overall, thesefindings suggest that toddlers can use specific attentional cues to helpthem in word learning, and that as productive vocabulary increases, theyare more likely to use affective cues and less likely to use general cues.
poster
Recent research (Stager and Werker, 1997) demonstrated that14-month-old infants could rapidly learn to associate words with objectswhen the words were phonetically dissimilar ('lif' and 'neem'), but theycould not do so when the words were phonetically similar ('bih' and 'dih').However, acoustic similarity alone cannot account for the impairedperformance on the word-learning task, as 14-month-old infants couldsuccessfully discriminate these similar sounding words when tested using aspeech discrimination task. Rather, Stager and Werker suggested that theseresults could be explained by the challenging computational demands of thetask for new word learners, as it requires the infant not only todiscriminate the two similar sounds, but also to link these sounds toobjects. The present research was designed to investigate this hypothesisby testing older infants who are better able to deal with the computationalcomplexities of the word-learning task, and thus may be better able toattend to fine phonetic detail. Three groups of infants were tested using the switch procedure ofStager and Werker, with two modifications: (1) the trial duration andhabituation criteria were changed to make them appropriate for olderinfants; and (2) the visual stimuli were altered to make the two objectsmore visually distinct. The first study was conducted to ensure that themodified design did not change the nature of the task in any significantway. Sixteen 14-month-olds were tested. Results showed that the infantsdid not succeed on the task (Xsame 9.29s, Xswitch 9.20s, t (15) 0.044, p 0.966), thus replicating the Stager and Werker findings. In thesecond study, sixteen 20-month-olds were tested. Since infants at this ageare typically past the vocabulary spurt and are more proficient wordlearners, we expected them to be better able to attend to the phoneticdetail in this task. Indeed, infants did look significantly longer to theswitch trial (Xswitch 13.52s) than to the same trial (Xsame 9.64s),demonstrating that they were able to associate the phonetically similarwords with the two objects (t (15) -2.494, p 0.025). A third studywas conducted with 17-month-olds, who are generally on the threshold of thenaming explosion. To date, 10 out of 16 infants have been tested and theresults show longer looking times to the switch trial (Xswitch 12.52s)than to the same trial (Xsame 9.36s, t (9) -2.584, p 0.027). Takentogether these results suggest that the ability to access fine phoneticdetail in a word-object association task comes on line between 14- and17-months of age, and is well consolidated in older infants. Ongoingresearch is examining the role of computational complexity and vocabularysize in explaining this change.
poster
Previous research with children has demonstrated that vocabulary size isrelated to fine phonetic discrimination in early word learners(Charles-Luce & Luce, 1995; Walley, 1987). It has been hypothesized thatan infant with a small lexicon would have little pressure to pay attentionto fine phonetic detail because most words would be quite phoneticallydissimilar. However, as the lexicon space becomes more crowded and theamount of phonetically similar words increases, it would become necessaryto attend to detail (Walley, 1987, 1993).Stager and Werker (1997) recently investigated the discrimination of finephonetic detail in a word-object associative task. They discovered that14-month-old infants did not notice when object labels that differed byonly one phoneme ('bih' vs. 'dih') were switched on a test trial, althoughthey could do the task when the labels were phonetically dissimilar ('lif'vs. 'neem'). These results were attributed to age, since the infant of 14months is just beginning the word learning experience. Vocabulary size wasnot examined as a factor.The current study explored the relationship between lexicon size andsuccess on the task used in Stager and Werker (1997). Based on theprevious research with young children, we hypothesized that largercomprehension vocabulary size would correlate with success on theobject-label associative task in infants.We obtained comprehension vocabulary scores for two groups of infants(14-month olds and 17 month-olds) and tested them using the same design andaudio stimuli as in Stager & Werker (1997). As in the previous study, the14 month-olds did not notice the label switch. The 17-month-old group didlook longer at the switched trial, demonstrating that they were able todiscriminate fine phonetic detail in a word-learning task. To obtain ameasure of the infant's lexicon, we utilized raw scores on the VocabularyComprehension section of the MacArthur Communicative Development Inventory- Infant. We divided the infants into two groups, those who looked longerat the switch and those who did not, and correlated this dichotomousvariable with lexicon size. The 14 month-olds demonstrated a significantpositive correlation (i.e. larger vocabulary related to success on task)and the 17-month-olds did not. When the two groups were pooled, thecorrelation was found to be similar to the 14 months with an even highersignificance.The results from the 14-month group clearly demonstrate that a largerlexicon size is related to success at attending to fine phonetic detail inan object-label associative task. The fact that the 17-month correlationwas not significant can be attributed to a ceiling effect. Most of theseinfants successfully noticed the switch and had high vocabulary scores.The strengthening of the positive correlation when their data is added tothe 14-month-olds demonstrates that the trend is present, even at the olderage.
poster
Previous research has suggested that toddlers selectively attend to certainperceptual object cues, such as shape, (Jones et al 1991) or parts(Poulin-Dubois et al, 1995), when learning and generalising newlyencountered object labels. Furthermore, it has been suggested that thisselective attention to shape and/or parts is a product of early wordlearning: it is an acquired strategy that constrains word meanings and hencefacilitates the rapid acquisition of object labels, particularlypost-vocabulary spurt. However, categories of objects vary in terms of their relative homogeneity,or perceptual similarity. For example, natural kinds may share manyexternal features (e.g. birds), whereas artifacts may not (e.g. musicalinstruments). Much research on object label extension has made use ofexperimenter-created artifacts (e.g. shape-bias studies by Landau andcolleagues), limiting the generalizability of findings. Whererepresentations of real natural kinds and artifacts have been used, analysisof potential differences across semantic domains is rarely conducted. Thecurrent study will compare object label learning and extension across thesedifferent domains. In addition, researchers dispute whether object namingderives from non-strategic (e.g. Smith et al, 1996) or strategic (e.g.Gelman & Medin, 1993) control of attention. The current study examined the relationship between toddlers' vocabularysize and the acquisition of new object labels. After training sessions,during which a picture of the object was shown to the children and labelled,multiple-choice comprehension tests were administered. In these tests thetarget object was pitted against same-category foils. Two types of targetobjects (4 of each type) were taught to the children: (i) objects rated byadults as having a salient part (e.g. pelican/beak), and (ii) objects ratedby adults as having no salient parts (e.g. thrush). In addition, theobjects were from differing semantic domains - 4 living things (e.g.birds/insects) and 4 artifacts (e.g. instruments/tools). If performance in the comprehension tests is dependent upon experience withobject category terms, then larger vocabulary size will be related to betterlearning of object labels. It was found that for natural kind terms,performance was related to vocabulary size but this was not the case forartifacts. In a subsequent generalization task, the extent to which toddlers treatedsalient parts as characteristic or defining was investigated. For bothnatural kinds and artifacts, children preferred a part-removed exemplarrather than a part-added non-referent.These findings are discussed in the light of competing theories concerningthe role of strategic and non-strategic control of attention in early objectword acquisition.
poster
As adults, we routinely use links between syntax and meaning to understandsentences. In 'John hits Bill,' for example, the placement of John andBill in the sentence tells us which is the guilty party and which thevictim. The chief claim of the syntactic bootstrapping theory is thatprecursors of these links between structure and meaning guide children inlearning the meanings of verbs. This claim is supported by many studiesshowing that children between about 2 and 5 years of age take novel verbsin different sentence structures to have different meanings.How might this process begin? That is, what aspects of sentence structuremight be available to children early in the process of learning theirnative language, and how might they influence the interpretation of earlyverbs?The present research explores a possible structural cue to sentenceinterpretation--word order. Prior research has shown that 17-month-oldsinterpret word order appropriately in sentences with known verbs: in apreferential looking task, infants looked longer at a video in which thegrammatical subject of their test sentence was the agent rather than therecipient of the familiar action (e.g., tickling, feeding). If childrenare to use word order to guide interpretation of new verbs, however, theirunderstanding of the significance of word order must be abstract enough totranscend the meaning of a particular verb.To explore the effect of word order in verb learning, two experiments useda revised preferential looking procedure. In Experiment 1, a group of26-month-olds learned a single new transitive verb. In the critical testtrial, two versions of the same causal action appeared simultaneously onside-by-side video screens. The two versions differed only in theassignment of characters to roles: In one, a bunny rotated a duck on aswivel stool by pulling on a scarf wrapped around the duck's waist. In theother, the duck did the same to the bunny. Children heard one of two testsentences: 'The bunny is blicking the duck around' or 'The duck isblicking the bunny around.' The question of interest was whether thechildren would look longer at the event in which the subject of their testsentence played the agent's role. The test trial was preceded by a seriesof trials designed to familiarize the children with the two characters andtest events.Children used the word order in the sentences to interpret the novel verbs:In the test trial, a clear effect of word order could be seen. The26-month-olds looked significantly longer at the matching screen than thenon-matching screen.Another group of 26-month-olds participated in a second experiment in whichthey saw the same events. Their test sentences, however, were 'The bunnyand the duck are blicking around' or 'The duck and the bunny are blickingaround.' These sentences contain the same two nouns, and the same novelverb, as those in Experiment 1. But the order of the two nouns should nothave the same significance: the first noun need not refer to the agent ofthe event. In this case, the children did not look significantly longer atthis 'matching' screen as they heard the test sentence.These results document robust understanding of word order by 26 months ofage. Though the verb was unknown, children interpreted the subject of atransitive sentence as the agent of the action, and were not fooled by asimilar sequence of nouns in an intransitive sentence. These findings, andongoing experiments varying sentence and event types, are helping to revealhow children use the very beginnings of syntactic knowledge to guidesentence interpretation.
poster
The development of the lexicon has been characterized as one of nonlinearchange. Children initially establish their lexicons slowly. Learning speedsup dramatically following knowledge of a critical mass of words. Thispattern of development is usually interpreted as indicating that twodifferent mechanisms are required for vocabulary development -one thatallows children initially to slowly map words to their meanings and one thatemerges later underlies the period of rapid growth. A plausible alternativeis that nonlinear growth patterns result from a single nonlinear learningequation that operates according to the same rules throughout vocabularydevelopment. Curve-fitting models of longitudinal data will be presented toassess the plausibility of this alternative.Parents of 28 children (17 boys and 11 girls) completed the MacArthurCommunicative Development Inventory (CDI)) monthly beginning when childrenwere 8 months old and continuing until they were 30 months old. Childrenalso participated in laboratory measures of lexical production andcomprehension. The growth-curve modeling is based on the parent report data,but the shape of growth with development correlates significantly with thelaboratory measures for both production and comprehension. Althoughtremendous variability exists in the timing and size of vocabularyacquisition across children, most show a burst in production with anearlier, less pronounced burst in comprehension.The results of a curve-fitting procedure suggest that non-linear developmentcan result without the onset of a new learning mechanism. Rather, the samemechanisms that let children slowly learn their first words enable a burstin learning. A nonlinear dynamic equation (i.e., one in which the rate ofchange can vary over time), dy/dt ay**2 + by + c, was used to fit thecurves generated by the children's vocabulary growth. dy/dt represents therate of change in y per unit of change in t. y represents the number ofwords, and t represents time. a, b, and c are parameters that are estimatedto fit the growth observed for each child. The curves using this equationwere generated by calculating a deviation score (the average number of wordsseparating the predicted vs. observed curve at each time point) as a measureof error. Results of curve-fitting the production data is complete and werepresented at the International Congress for the Study of Child Language lastJuly (a substantially more limited audience than attends ISIS). Acrosschildren, deviation scores averaged 8.85, ranging from a low of 1.71 to ahigh of 19.94. Curve-fitting the comprehension data is ongoing. Both setsof curve-fitting data and a discussion of their implications forpsychological theories of word learning will be presented in the poster. Thecurve-fitting model makes plausible that a single growth equation appliedcontinuously over time may explain nonlinearities in vocabulary acquisition.Methods that pinpoint the factors that interact to determine thepsychological mechanisms of vocabulary acquisition are proposed.
poster
Several studies have shown that young children acquiring new verbs arefairly conservative in their productions. For example, one study suggeststhat young children have difficulty producing new verbs with appropriateinflectional endings unless they have heard these particular forms in theambient speech (Perez-Periera, 1989). This difficulty exists even forinflections that are highly regular and common. At around the same age,children also have difficulty producing new verbs in sentence contextsthat they have not explicitly encountered in the environment (Akhtar &Tomasello, 1997; Olguin & Tomasello, 1993). Again, this inability togeneralize is found even for highly common sentence types (e.g., the SVOsentence). Thus, before the age of 3, children's understanding of aparticular verb appears to be relatively isolated from other verbs, andchildren show little ability to generalize from one verb to another(Tomasello, 1992). The present study examines whether Chilean Spanish-speaking children areable to extend their knowledge of inflections to a new verb, or whetherthey are conservative in their verb use. Sixteen 2 1/2- and 16 3-year-oldchildren heard two familiar verbs (bailar or 'to dance' and manejar or 'todrive/ride') and were taught two novel verbs (ponzar or 'to stand on yourhead' and megar or 'to push an object over'). For both verb types, onewas heard in the third-person singular form and one in the third-personplural form. An adult then asked for the other, unmodelled form in anelicited production task. Only 4 of 16 younger children and 6 of 16 olderchildren produced an unmodelled form of a novel verb. Children were moresuccessful in producing familiar verbs. They were also somewhat moresuccessful in a comprehension task than they were in the production task. These findings are consistent with previous analyses of children'sspontaneous speech (e.g., Pizzuto & Caselli for Italian). The findingssuggest that children's mastery of verbal paradigms is initially piecemealand lexically based, and provide little evidence to support parametersetting accounts (e.g., Hyams, 1992, 1994).
poster
Many different accounts on the process of word-learning abound in the fieldof language acquisition. The accounts can be separated into three generaltypes, each with explanations for the ease with which children pick up novellabels with endless referents. These three main approaches can be found inspecific theoretical accounts. One account postulates that children come toword learning with explicit knowledge of word-to-world mapping (e.g.,Woodward). Another (e.g., Tomasello, Baldwin) postulates that 'smart' wordlearning is due to knowledge of the pragmatics of everyday communication. Alast account (e.g., Smith) states that general cognitive processes, such ascontext shifts and their effect on memory and attention, control smart wordlearning. The present study aims to shed light on the above accounts byexamining the role of emotion in the word-learning scenario. The subjects are studied in a word-learning situation where three differentemotions(positive/happy, negative/sad, and surprise) are present. The inclusion ofsurprise is in response to the claim that emotion affects the retention oflabels by changing the perceived intentionality of the novel label. Byincluding positive, negative, and surprise emotions we aim to separate andexamine the effects of intentionality and emotionality. During labeling ofthe novel object each emotion is presented to the subject in a scriptcontaining the labels. The only variable between the trials is thediffering randomized emotion. Specifically, 24 two-year-olds are presented with three pairs of toys.Each toy in a pair(matched for salience) is handled equally during thetraining session. The training session is done by a trained experimenter ineither a happy/positive, sad/negative, or surprise emotion tone of voice andfacial expression. The training session consists of a three to five minuteinteraction. One of the objects is labeled nine times with a novel word(i.e., 'dax'), while the other object is handled equally but labeled withthe empty label 'toy.' This is done three times, once for each pair.Multiple choice tasks (conducted by an independent experimenter) are runbetween each condition and after all three conditions. An ANOVA (conditionx sex) will be conducted. Data are currently being collected and coded. The results of this study will help to further hone the debate between theabove threeaccounts. If none of the emotions affect labeling, support is lent to thetheory that children have knowledge about word-to-world mapping that islargely uninfluenced by omnipresent social factors (like emotional states).If there is a main effect of any emotion on the labeling process, then thosewho posit a domain-specific knowledge of word learning must account for thatinteraction. An effect found only in the surprise condition points todeveloping pragmatics dictating word learning. If main effects are foundfor multiple emotions, then it is possible that emotion impacts infantsgeneral cognitive abilities during the encoding of the novel label,supporting the view that general cognitive processes are responsible forsmart word learning.
poster
Thirteen to 18-month-olds can learn a new object name after minimalexposure to the word and its referent, and they show comprehension of thenew word after a 24-hour delay in favorable testing situations (Schafer &Plunkett, 1998; Werker, Cohen, Llyod, & Casasola, 1998; Woodward, Markman,& Fitzsimmons, 1994). Although the ability to quickly form new word-objectmappings was originally thought to be specific to the task of wordlearning, recent findings with older children have shown that 2- to4-year-olds can also quickly learn and remember arbitrary non-lexicalinformation about novel objects (Markson & Bloom, 1997). This suggeststhat children have a broader ability to quickly acquire new, arbitraryinformation. It is possible, however, that such rapid learning abilitiesare initially specific to language and later are generalized to other typesof information. We have begun to investigate this possibility by studyingrapid learning in children closer to the start of word learning.Sixteen 18-month-olds were introduced to two novel objects, and wereencouraged to play freely with them for several minutes. One of the objects(the target) possessed a hidden, arbitrary feature (e.g., it made a noisewhen shaken), which the child 'discovered' while interacting with theobject, whereas the other object (the distracter) did not. Theexperimenter made sure that comparable attention was paid to both of theobjects by the child during this initial 'training phase'. The childreturned to the lab the next day for the testing session, whichinvestigated whether he/she remembered the hidden feature of the targetobject after a 24-hour delay. Memory was assessed using a violation ofexpectation object manipulation procedure (Baldwin, Markman, & Melartin,1993) in which the child explored a replica of the target object thatlacked the original hidden feature (e.g., it no longer made a noise whenshaken). We observed how many target actions (e.g., shaking the object)the child performed on the target object compared to both the actions ofthe same child on a different object, and to the actions of differentchildren who had not previously been exposed to the target object's hiddenfeature, on the same object.The results support the thesis that rapid learning about objects is notrestricted to language contexts, even for children at an early point inlanguage acquisition. Thirteen out of 16 children performed at least onetarget action on the target object, compared to only 3 out of 16 childrenperforming a target action on the object which served as the distracter[t(15) 5, p < 0.001]. In addition, 12 of these 16 children alsosuccessfully learned which of two different novel objects was the referentof a new label (e.g., jippi) using a multiple choice test procedure [t(15) 2.24, p < 0.05]. Further studies are currently underway with13-month-olds and with longer delay intervals between training and testing.The present findings reveal that children are not only proficient atlearning word-object relations, but readily acquire other arbitraryrelations as well. The results move us one step closer to understandingwhether children's capacity for word learning depends on a mechanismdedicated specifically to language, or results from a broader mechanism forlearning and retaining arbitrary information in multiple domains ofknowledge.