Poster group
Details of individual items:
poster
The strategies that toddlers use when first learning words is a topiccentral to language acquisition. It is clear that children establishword meaning using numerous cues, including semantic (Bloom, 2000),pragmatic (Baldwin, 1994), and syntactic information (Hirsh-Pasek andGolinkoff, 1998). Most of the research on these cues has focussed oneach cue in isolation, however in a conversational context, these cuesare all present at the same time. Co-ordinating the use of these cuesis obviously a difficult task. We are interested in ways that theinput could be structured to facilitate this co-ordination. That is,caregiver's speech could highlight various aspects of word usage, andthus be a valuable tool for word learning. This possibility hasreceived little attention in the literature. The limited research onthis topic has demonstrated that caregivers adopt specific strategieswhen teaching novel nouns to their young toddlers. These caregiverstrategies accord well with previously demonstrated biases thattoddlers bring to the word learning task, such as mutual exclusivity(e.g., Markman, 1989). It thus seems likely that toddlers could findcaregiver input useful when learning all novel words, including thosefrom other lexical categories. However, no previous work hasinvestigated the strategies that caregivers follow when teachinglexical categories other than nouns, such as adjectives or propernames. The present research systematically examined how caregivers teachnovel words from three different lexical categories: Nouns,adjectives, and proper names. Forty-eight caregivers and their 2- to4-year-old toddlers were audio-taped in a story-book reading task. Inthis task, caregivers were presented with a book of ten black and whitedrawings, each depicting a familiar animal or object. The caregiverswere then asked to teach their toddlers novel words ('X') presentedwith the drawings. The novel words were presented in a sentence framewhich provided syntactic evidence that the novel words were eithernouns ('This is an X'), adjectives ('This is very X'), or proper names('This is named X'). The audio-tapes were then transcribed and codedfor content. It was found that caregivers adopted strikingly differentstrategies in how they taught the novel word as a function of thewords' lexical categories. That is, the input they provided for thesame picture differed dramatically if the word was interpreted as anadjective, a proper name, or a noun. Moreover, these distinctstrategies corresponded to the biases that toddlers of this age havebeen shown to possess when they interpret new words from thesecategories (e.g., Hall, 1994, 1996; Hall & Graham, 1999, Markman,1989). The results highlight the need to acknowledge the potentialrole of caregiver speech in directing toddlers' learning of new words. * A portion of this research was presented at the 1999 Brain,Behaviour, and Cognitive Science conference held in Alberta, Canada.
poster
During the course of early language development, children develop different word learning skills. They become capable of rapid word learning or forming an initial representation of a word after hearing or seeing it only a few times. In addition, they develop a novel mapping strategy that allows them to infer that a novel word refers to something for which they have no name. Deaf children of hearing parents tend to be significantly delayed in their vocabulary. We hypothesized that this delay may be related to the strategies that deaf children have available to them to learn new words. In a previous small-scale study, we found that the kind of strategies children use to 'bootstrap' their learning of new words correlates with their productive vocabulary. Specifically, we have shown that children with moderate expressive vocabularies learn new words effectively using the rapid word learning strategy while those with the large expressive vocabularies learn words via the rapid word learning and the novel mapping strategy. We also tested two children, twins with very low productive vocabularies, who were unable to learn novel words using either strategy. The goal of the current study was to extend the previous research to include more children with low language and children educated in a variety of language environments (Total Communication, Oral, and Bi-cultural-Bi-lingual). Eighty-one deaf children, ranging from 27 to 77 months, were administered two word learning tasks. The 'novel mapping task' was designed to test if children would map a novel word onto a novel object without ostensive labeling (i.e. infer that a novel word refers to a novel object rather than three objects for which they already have labels. The 'rapid word learning task' was designed to test if children can learn a new label for an unfamiliar object if it is presented to them only three times. Eleven of the children with the lowest vocabularies did not consistently learn new words using either strategy while 29 children with moderate vocabularies learned new words using the rapid word learning strategy, but not the novel mapping strategy. Thirty-nine children with the highest vocabularies learned new words consistently using both the rapid word learning strategy and the novel mapping strategy. The children's expressive vocabulary predicted word learning strategies children use above and beyond age, but age did not predict word learning strategy use above and beyond expressive vocabulary. These results suggest there are three 'levels' of word learning strategies that facilitate children's learning of new words. Furthermore, the word learning strategies available to children relate to their expressive vocabulary whether they are acquiring language through speech, sign, or both.
poster
Most words are categorical (i.e., 'animal' or 'ball'). However, many of the first words infants learn are proper nouns(i.e., 'Mommy'). At what point do children come to realize thatmost objects in the world are named by categories? Research withanimals has consistently shown that even pigeons and rats willgeneralize to an entire category of stimuli. Therefore, onemight assume that young infants would begin with a 'principle ofextendibility' -- an assumption that words label categories. Yettheories beginning with this assumption would then need toexplain why infants do not extend 'Mommy' to include all women. Possibly the nature of word-learning requires that infants beginby not generalizing, but rather treating new words as uniquelyassociated with a single object. How then do young infants learnhow (and when) to extend a word to include an entire category? Researchers have found categorization abilities beginning inearly infancy (Hall, 1991; Mandler, 1998; Waxman & Markow,1995). Even at 7-months of age, infants are capable ofseparating objects from two categories (Mandler, 1998). Furthermore, at 12-months of age, hearing a novel word whilebeing presented with several category members assists infants informing a category (Waxman & Markow, 1995). By two years of age,infants will extend a novel label to new category members (Hall,1991). In stark contrast, Smith (in press) suggests that younginfants need to build up, through multiple exposures, a principleof extendibility based on regularities in shape. How can theseapparently disparate findings be reconciled? One possibility isthat while infants can form categories and can learn words,combining these tasks creates a very difficult problem. Perhapsat first, infants are conservative -- using the word only for thespecific object that had been labelled. Only after seeing a wordextended to several category members will they be able togeneralize to the entire category. This would suggest thatinitially infants do not possess a mature principle ofextendibility, but rather that they develop this principlethrough experience. This study attempts to address when infants begin to use aprinciple of extendibility rather than assuming that objects arelabelled as proper nouns. 10-, 12-, 14-, and 19-month-olds weretested using the Interactive Intermodal Preferential LookingParadigm (IIPLP). Each infant was given three novel object toplay with, one at a time (exploration phase). Two of the objectswere identical (except for color) while the third was from adifferent category (the distractor). After the explorationphase, the infants were shown the two category members on a Faganboard (salience phase). The experimenter then picked up one ofthe objects and labelled it ten times while holding it andlooking at it (labelling phase). This was followed by three setsof test trials. In the first, the infants were shown thelabelled object and the distractor. This tested if the infantshad learned that the label applied to the original object(original test trials). Then, the infants were shown the twocategory members (i.e., the same object in two colors). Thistested if the infants would show a proper noun preference evenwhen the second object is in sight and available for rapidcomparisons (proper noun trials). Finally, the infants wereshown the unlabelled category member and the distractor, to testif the infants would extend the label when the orignal object wasnot present (extend trials). For half of the infants, the extendtrials preceded the proper noun trials. Overall, infants demonstrated an increasing ability to learn theword. While only 43% of the 10-month-olds demonstrated that theyhad attached the label during the original test trials, almosttwo-thirds of the 19-month-olds did so. Furthermore, infants atall ages were significantly biased to use the label as a propernoun. At all ages, whether or not the infants were willing toshow any evidence of extension depended on the order of the testtrials. If the infants did the proper noun trials first, thenthey were not willing to extend the label. In contrast, theinfants who did the extend trials before the proper noun trialsdemonstrated a willingness to extend the label. This findingsuggests that early abilities to extend novel words are quitefragile. When given any suggestion that the word might be aproper noun, the infants choose not to extend the label. Although infants might be able to generalize and categorizeoutside of word-learning situations, young infants do notnaturally extend words, but do indeed adopt a conservativeposture. This explains how children can have both proper nounsand common nouns. Proper nouns are the 'default' option, yetwith experience with multiple exemplars, infants quickly learn toextend the word to an extire category. This paper presents amid-line position between Smith and Mandler, Waxman, Markow, andHall. The infants do not appear to automatically link words andcategories, as might be suggested by the latter group. However,they might be able to learn to 'fast-map' a word to a category(in this study, the infants were only given a few seconds tolearn the word). In sum, infants do need experience to learn aprinciple of extendibility, but maybe not very much for thesystem to be able to work.References:Hall, D. G. (1991). Acquiring count nouns: Therestricted-unrestricted distinction. Dissertation AbstractsInternational.Mandler, J. (1998). Representation. In W. Damon (Ed.),Handbook of Child Psychology, Vol. 2: Cognition, perception, andlanguage. New York: Wiley.Smith, L. B. (in press). Learning how to learn words: Anassociative crane. In R. M. Golinkoff, K. Hirsh-Pasek, & G.Hollich (eds.), Breaking the word learning barrier. New York: Oxford.Waxman, S. R., & Markow, D. R. (1995). Words as invitations toform categories: Evidence from 12- to 13- month-old infants. Cognitive Psychology, 29, 257-302.
poster
Recent work has documented that 13-month-olds have begun to distinguish between novel words presented as count nouns vs. adjectives in fluent, infant-directed speech, and that their expectations for word meaning accord with this emerging sensitivity (Waxman, 1999; Waxman & Markow, 1995). Unfortunately, however, existing evidence for infants' expectations in word-learning is based entirely on novelty-preference tasks. In the experiment reported here, we advance this work, examining infants' expectations for novel count nouns and adjectives, using a new, word-extension paradigm. Thirty-six 14-month-olds were randomly assigned to a Novel Noun, Novel Adjective, or No Word condition. During a Familiarization phase, the ex introduced all infants to four different objects (e.g., 4 purple horses), all from the same category (e.g., horse) and instantiating the same property (e.g., purple). Objects were introduced either with a novel noun (e.g., 'These are blickets'), novel adjective (e.g., 'These are blick-ish'), or no novel words ('Look at these'). In the Test phase, infants were asked to extend the novel words beyond the familiarization objects. The Category choice (e.g., a blue horse) was from the same object category as the familiarization objects, but had a different property; the Property choice (e.g., a purple chair) was from a different category, but shared the same property as the familiarization objects. There were two test trials for each set. The probability that an infant would consistently select the Category choice (or Property choice) on both trials is .25. Each infant completed this procedure 4 times, with 4 different sets of objects.If infants expect that the novel word referred to the category of objects presented during familiarization (e.g., horses), they should prefer the Category choice at test. If they expect that the novel word referred to an object property (e.g., purple), they should prefer the Property choice. The results indicate that 14-month-olds have begun to distinguish between novel words presented as count nouns vs. adjectives, in fluent speech, and that these emerging distinctions are recruited by infants' in their extension of novel words. Infants in the Novel Noun condition (M : .43) consistently selected the Category choices at a rate that exceeded chance (.25) and exceeded performance in either the Novel Adjective (M :.22) or No Word (M :.29) conditions, all p's < .05. Infants in the Novel Adjective condition (M : .29) consistently selected the Property choice at a rate that exceeded performance in either the Novel Noun (M: .13) or No Word (M : .10) conditions, all p's < .05. When considered in conjunction with novelty-preference data, these results suggest that an expectation linking count nouns and object categories is established early in acquisition, and that this sets the stage for the emergence of other word-world mappings (e.g., adjective to object property). This line of research is important because it permits us to identify infants' initial expectations regarding word-to-world mappings, and to trace their evolution as a function of infants' experience with the structure of the native language under acquisition. Cross-linguistic and developmental implications will be discussed.
poster
Researchers have postulated word learning biases to explain infant's effortless and rapid acquisition of object words. One such bias, the whole object assumption, helps infants acquire new words by narrowing the referent of a novel label to an entire object, rather than to actions, spatial relations or parts of that object (Woodward & Markman, 1998). However, this hypothesis does not encompass the acquisition of action words. This is an important oversight because action words are well represented in infants' first words (Bloom, Tinker & Margulis 1993). Furthermore, infants are more responsive to objects in motion than to stationary ones (Slater, 1989), and they can detect objects in motion at greater distances (Burnham, 1987). Motion that provides information about an entity's status as animate or inanimate is particularly salient to young infants (Poulin-Dubois, 1999).The issue in the present experiment was whether the whole object assumption could be overridden when a novel label is presented in the context of a novel object displaying an animate or inanimate motion. Will the whole object assumption guide infants to ignore an object's motion path and focus on the identity of the object, or does the presence of distinct animate and inanimate motion paths cause infants to override this assumption?To address this issue, 15-month-old infants were presented with a novel object on a computer screen, engaged in an animate or inanimate motion, and they heard a novel label paired with this event. They were taught a label for an animal-like object jumping over an obstacle (e.g. gop) and another for a vehicle-like object bouncing off a wall (e.g. fep). On the test trials, infants were simultaneously presented with the event originally paired with the target label and the event paired with the other label. It was expected that infants would look longer at the event originally paired with the label, if they had learned this association. In the generalization trials, the original pairings presented in the training phase were switched, such that infants saw the animate object performing the inanimate motion and the inanimate object performing the animate motion. If infants associated the novel label with the object in the event, they were expected to ignore the motion path and to look longer at the screen with this object. However, if infants associated the novel label with the motion in the event, they were expected to look longer at the screen presenting the motion path initially paired with the label.Preliminary data with a small number of infants shows that in the test trials, they showed the ability to look longer at the event originally paired with the label. This suggests that 15-month-old infants can learn new words after only two exposures. They also preferred to look at the object in the generalization trials. These data suggest that despite the saliency of the animate and inanimate motion cues, the object bias is guiding infants' understanding of these labels. The current data will be discussed in the context of the recent word learning literature.