Infants' Expectations about Object Label Reference
Graham, Susan A, Baker, Rachel K, Poulin-Dubois, Diane, Canadian Journal of Experimental Psychology
Abstract The aim of this research was to examine whether infants at the early stages of lexical development were sensitive to the word-category linkage. In Experiment 1, 16-to 19-month-old infants were requested to match a target with either a basic-level or a thematic match, with or without a novel label. Stimuli were presented using the preferential looking paradigm. Infants in the Novel Label condition looked significantly longer at the basic-level match than infants in the No Label condition. In Experiment 2, infants were presented with a target, followed by a basic-level match and a superordinate-level match with or without a novel label. Again, infants in the Novel Label condition looked significantly longer at the basic-level match than infants in the No Label condition. Taken together, these findings indicate that infants initially assume that novel words label basic-level categories and thereby do honour the word-category linkage.
Young children are amazingly adept word learners. Anglin (1993) recently estimated that between the ages of 18 months and 6 years, children acquire five to six new words a day. A large number of the words acquired during the infancy period refer to object categories (Benedict, 1979; Nelson, 1973). Most of these early object words refer to basic-level (e.g., dog) as opposed to superordinate-level (e.g., animal) or subordinate-level (e.g., collie) categories (Anglin, 1977; PoulinDubois & Graham, 1994). For the majority of infants, first words are acquired slowly, over the course of several months. Then around 18 months of age, many infants experience an accelerated period of lexical acquisition termed the vocabulary spurt (Goldfield & Reznick, 1990; Nelson, 1973; Reznick & Goldfield, 1992).
Infants' rapid acquisition of novel words is a remarkable feat considering the formidable demands of the word-learning task. Upon hearing a new word, infants must link the word to the appropriate referent and then generalize that word to other instances of the referent. This task is extremely complex when one considers the highly inductive nature of the word mapping task - even in ostensive labelling situations, there is a multitude of possible referents for any new word. Given children's success at word learning, many researchers have argued that lexical development is guided by biases or constraints that facilitate word learning (e.g., Golinkoff, Mervis, & Hirsh-Pasek, 1994; Markman, 1989,1992). These biases are proposed to simplify the process of word learning by limiting the number of possible referents a child must consider when confronted with a new word.
One word-learning bias that has received a great deal of empirical scrutiny in recent years is the taxonomic assumption or the word-category linkage (Markman, 1989; Waxman & Hall, 1993). According to this bias, children assume that a novel word refers to members of the same kind or taxonomic category, and not to objects that are related spatially or thematically (Markman,1992). That is, when taught a new label for an object (e.g., a dog), children will extend that new label to another object in the same taxonomic category (e.g., another dog or animal) rather than to an object that is spatio-temporally related to the original object (e.g., a doghouse). In the present experiments, we examined the nature of the word-category linkage at the early stages of lexical development. Specifically, we examined whether infants who are provided with a novel label will attend longer to basic-level matches than infants who do not hear a novel label. A second goal was to examine infants' initial assumptions about the categorical level of a novel word when faced with a choice between superordinate- and basic-level matches.
The notion of a linkage between nouns and taxonomic categories was first proposed by Markman and Hutchinson (1984). They presented 3- and 4-year-old children with a series of target objects (e.g., dog) followed by a thematic match (e.g., bone), and a taxonomic match (e.g., another dog) in a matchto-sample procedure. When the target object was labelled with a nonsense word, children chose the taxonomically related object significantly more often than would be expected by chance alone. When the target object was not labelled, the children selected the taxonomic match at chance levels. In subsequent experiments, the same taxonomic bias was produced by a novel label at the superordinate level and with artificial objects. Since these initial experiments, numerous other studies have demonstrated this linkage with preschoolage children (e.g., D'Entremont & Dunham, 1992; Golinkoff, Shuff-Bailey, Olguin, & Ruan ,1995; Hall, 1993; Waxman & Gelman, 1986; Waxman & Kosowski, 1990). Recent research also indicates that French- and Spanish-speaking preschoolers honour this bias (Waxman, Senghas, & Benveniste, 1997).
The child's assumption of an association between nouns and taxonomic categories is an efficient means of rapidly acquiring new words. However, whether this strategy arises out of word- learning experience or actually drives early lexical acquisition is difficult to ascertain. To date, the majority of studies examining the word-category linkage include preschool-aged children who are well beyond the period of rapid vocabulary expansion that occurs around 18 months of age. Yet, in view of the rapid and seemingly effortless acquisition of new words during infancy, it is reasonable to postulate that word-learning biases, particularly a sensitivity to a wordcategory linkage, may be operational during this developmental period.
The issue of a linkage between words and categories in infancy has been investigated in a number of ways. Some researchers have examined how linguistic input might facilitate young infants' categorization (e.g., Balaban & Waxman, 1997; Roberts, 1995; Waxman & Markow, 1995). Others have examined how a novel label might direct an infant's attention towards an object as opposed to some other aspect of the environment. For example, studies have shown that children as young as two years of age associate words with objects rather than with their motion, texture, colour, or substance (e.g., Echols, 1992; Landau, Smith, & Jones, 1988; Soja, Carey, & Spelke, 1991). Labels have also been found to increase the attention 10- to 14-month-old infants pay to objects (Baldwin & Markman, 1989). These studies are extremely informative in demonstrating that infants assume a word refers to a whole object or an object category versus some other aspect of the object itself or the environment. However, these results do not directly address the issue of how infants extend novel words when faced with two objects, one a taxonomic match and the other a thematic match. This issue is important given that infants have been found to extend words on the basis of different types of similarity, including thematic associations (e.g., Clark, 1983; Dromi, 1987). Indeed, Golinkoff et al. (1994) have argued that infants at the beginning of word learning are guided by a principle of extendibility, rather than an exclusive word-category linkage. This principle states that words can be extended to referents that bear some type of similarity, including any type of perceptual similarity or thematic association, to the original referent. With development, infants are expected to limit their word extensions to members of the same basic-level category (principle of categorical scope).
The two published studies to date that have examined the issue of taxonomic versus thematic word extension in infancy have yielded inconsistent results. Using a forced-choice match-to-sample task involving basic-level taxonomic and thematic matches, Bauer and Mandler (1989) found that 16and 20-month-old infants exhibited a high level of categorical responding which did not increase in the presence of a novel label. This finding suggests that novel nouns do not highlight categorical relations. However, two characteristics of the design may have contributed to these unexpected results. First, the experimenter systematically reinforced the children for taxonomic choices which may have masked the facilitative effect of the novel noun. Second, the side on which the taxonomic match appeared on each trial was systematically alternated, which may have provided children with an expectation about which position choice would be reinforced.
In a more recent study, Waxman and Hall (1993) presented 15- and 21-month-olds with object triads consisting of a target object, a taxonomic match, and a thematic match in a no word and novel word forced-choice match-to-sample task. Measures of expressive vocabulary were used to determine whether infants had achieved the vocabulary spurt. Children from both age groups selected the taxonomic match significantly more often in the novel word condition than in the no word condition, regardless of their vocabulary size. However, this condition effect was observed in the younger infants only after modifying the procedure to include demonstrations of the two types of relations. Furthermore, because most thematic choices represented animate objects whereas the target and taxonomic objects all represented inanimate objects, infants might have been biased toward choosing a match from the same ontological kind as the target. Although Waxman and Hall purposefully selected their taxonomic and thematic matches in this manner to highlight the distinction between the thematic and taxonomic relations, this ontological distinction may have led infants to extend the novel noun on the basis of inanimacy rather than taxonomic relations. In light of these methodological issues and the inconsistent results of the two studies, further research is needed to clarify the conditions under which children younger than three years honour the word-category linkage.
A related issue requiring empirical scrutiny is the hierarchical level at which infants first apply the word-category linkage. Research indicates that infants are able to categorize at a number of levels, including both the basic and superordinate (or global) levels, by the time they are two years of age (e.g., Balaban & Waxman, 1997; Behl-Chadha, 1996; Mandler & Bauer, 1988; Mandler, Bauer, & McDonough, 1991; PoulinDubois, Graham, & Sippola, 1995; Quinn, Eimas, & Rosenkrantz, 1993). Although some studies have found a pattern of global-level categories emerging before basic-level categories (e.g., Mandler & Bauer, 1988; Mandler et al., 1993; Poulin-Dubois et al., 1995), others have not (e.g., Waxman & Markow, 1995). Thus, on the basis of their categorization abilities, it is conceivable that infants could assume novel words label categories at either the basic or superordinate level. Indeed, Waxman and her colleagues have found that words can highlight both basic- and superordinate-level categories (e.g., Waxman & Hall, 1993; Waxman & Markow, 1995).
However, when one examines the nature of the input that infants receive and the types of words that they first produce, it is more likely that infants first assume that novel words label categories at the basic level. Several studies have found that parents tend to use basic-level terms when they first label objects for their children (e.g., Anglin, 1977; Blewitt, 1983; Callanan, 1985; Poulin-Dubois et al., 1995). Similarly, the majority of children's early object words are basic-level terms (e.g., Clark, 1983; Poulin-Dubois et al., 1995; see also Woodward & Markman, in press, for a discussion of these issues). Furthermore, studies with preschoolers indicate that children tend to first assume that a novel word labels basic-level categories rather than superordinate- or subordinate-level categories (e.g., Callanan, 1989; Hall, 1993; Taylor & Gelman, 1989; Waxman, 1990; Waxman & Senghas, 1992). To date, infants' first assumptions about the hierarchical level of referents of novel words, when presented with both basicand superordinate-level matches, have not been investigated.
The present research was designed to examine two issues regarding the nature of the word-category linkage in early lexical development: First, in light of inconsistent results in previous infancy studies, we sought to determine whether infants at the early stages of lexical development honour the word-category linkage when presented with taxonomic matches and thematic matches (Experiment 1). Second, we examined infants' initial expectations about the hierarchical level of a referent of a novel word when faced with a basiclevel and superordinate-level match (Experiment 2), a pairing that has not yet been empirically examined.
In the first experiment, we examined infants' attention to basic-level and thematic matches in the presence and absence of a novel label. We also examined whether attention to the basic-level match varied as a function of vocabulary size. Golinkoff et al. (1994) hypothesize that infants move from extending words on the basis of perceptual and thematic relations to limiting their extension to basic-level categories. If such is the case, those children with more words in their vocabulary should attend more to the basic-level match than those children with smaller productive vocabularies, in the presence of a novel word.
In this experiment, improvements were made to the typical tasks testing taxonomic versus thematic relations. First, the thematic choices used were familiar to infants within this age range. Although it has been established that infants form basic and superordinate (or global) level categories during the first two years of life (e.g., Mandler & Bauer, 1988; PoulinDubois & Graham, 1994), few studies have examined whether infants of this age are familiar with thematic relations between objects. It is possible that children chose taxonomically in previous studies because the novel label was effectively directing them to look for the relation with which they were familiar (e.g., Waxman & Hall, 1993). Second, the typical match-to-sample task was adapted for use within the preferential looking paradigm. The traditional match-to-sample task which requires infants to either point to objects or give them to the experimenter has been found to be too difficult for infants, necessitating some form of reinforcement or training (e.g., Bauer & Mandler, 1989; Waxman & Hall, 1993). In the preferential looking paradigm, the infant is presented with two displays and the amount of visual fixation to each of the screens is monitored as infants listen to auditory stimuli (e.g., Behrend, 1988; Golinkoff, Hirsh-Pasek, Cauley, & Gordon, 1987; Hirsh-Pasek & Golinkoff, 1996; Reznick, 1990). In our procedure, infants first saw a target item on both computer screens. This trial was followed by the presentation of a basiclevel match on one screen and a thematic match on the other screen, accompanied by instructions with or without a novel label. If infants in this age range do indeed exhibit a sensitivity to the linkage between words and categories, we expected greater attention to the basic-level match in the presence of the novel label than when no label is presented. Attention to a thematic match was not expected to be affected by the presence of a novel label.
Participants. The participants were 39 16- to 19-month-old infants for whom English was the sole or one of the predominant languages spoken at home. Six infants were tested but excluded from the sample for one of the following reasons: parents used real labels during testing (n = 2); English was not a predominant language in the home (n = 1); found to be statistical outliers (n = 3; see Data Reduction subsection). The final sample consisted of 33 infants (16 boys and 17 girls) ranging in age from 16.69 months to 19.62 months (M = 17.85 months, SD = .83).
Stimuli. The stimuli were five coloured picture triads composed of a target stimulus (e.g., a stroller), a picture of a different object from the same basic-level category (e.g., another stroller) and a picture of a thematically related object (e.g., a baby; see Table 1 for complete listing and Figure 1 for an example of a triad). Although the basic-level matches were members of the same basic-level category as the target, they were not identical to the target. In general, they differed from the target in both colour and orientation (see Figure 1). The five thematic matches were selected from a set tested with infants in the same age range (see following section). These pictures were digitized using the Abaton Scan 300/Colour Scanner for presentation using the Adobe Photoshop 2.01 software programme on computer monitors. The verbal instructions for the task were recorded (MacRecorder) and presented through speakers. The female experimenter used the higher pitch and intonational contours characteristic of the infant-directed speech register when recording the verbal instructions.
Apparatus. A three-sided testing chamber was designed for this task with a front panel display containing two 35-cm Macintosh colour monitors separated by 60 cm. Two audio speakers placed midway between the two monitors transmitted the verbal prompt. A 25-watt blue light situated above the speakers was used to direct the infant's attention to the centre of the apparatus between trials. The infant's state and parent's behaviour were monitored via a small television monitor. Trial length and synchronization of the auditory and visual stimuli were controlled by a Macintosh Centris microcomputer. All sessions were videotaped with an 8-mm camcorder, the lens of which was focused on the infant's face.
Selection of thematic matches. To identify thematic matches familiar to infants, 16 infants (mean age = 17.82 months, SD = .49 months, 7 boys, 9 girls), recruited from the same population as those tested in the word-category linkage task, were presented with a set of 20 picture triads consisting of a target (e.g., a pail), a thematically related match (e.g., a shovel), and an unrelated foil (e.g., a person). To control for possible salience effects, each thematic match was presented once as the thematic match and once as the unrelated foil on another trial. Infants were requested to choose the picture that belonged with the target (e.g., "Which one goes with the pail?").
The data were analyzed to determine what proportion of infants correctly chose the thematic match for each target. From the original 20 items, 5 were selected for use in the word-category linkage task. At least 64% of infants chose the thematic match on each of these triads, with the proportions ranging from 64.29% to 71.43% (M = 67.66%). This overall mean is significantly greater than chance level responding (50%), t(4) = 14.42, SEM = 1.23,p < .001. To ensure that the selection of these five thematic matches was not due to a general preference for these items regardless of their relationship to the target, the percentage of thematic items chosen when they were matches was compared to the percentage of these same items chosen when they were unrelated foils. Infants chose more thematic items when they were matches (M = 67.29%, SD = 21.05) than when foils (M = 30.83%, SD = 16.02), t(15) = 5.39, SEM = 6.77,p < .001. The percentage of infants who chose the thematic match when it was a match versus when it was an unrelated foil was also compared. On average, more infants selected these matches when they were the match (M = 67.66%, SD = 2.74) than when they were a foil (M = 36.31%, SD = 8.3), t(4) = 7.09, SEM = 4.42,p <.01. Thus, it seems clear that a salience effect cannot account for the selection of these five thematic matches.
Design and procedure. Infants were randomly assigned to one of the two experimental conditions: a No Label condition (n = 14) and a Novel Label condition (n = 19). These conditions were identical except for the instructions presented: In the No Label condition, the target picture was not labelled whereas in the Novel Label condition, the target was labelled with a novel noun. The No Label condition acted as a baseline condition in that it allowed examination of infants' visual preferences for the pictures in the absence of a label. These preferences were then compared to those of infants who heard the objects labelled with a novel word.
In both conditions, the infant was seated in a booster seat attached to a table in the laboratory with his or her parent seated directly behind or beside him or her. The infant faced the front panel of the testing chamber from a distance of 1.4 metres. Each trial began with both computer screens blank and the blue light blinking in order to draw the infant's attention to the centre of the apparatus. The target of each picture triad was then presented on both screens for 8 seconds with a digitized female voice giving the appropriate instructions. After a 3-second intertrial interval, the basic-level and thematic matches were presented on different screens and the same female voice gave the verbal instructions appropriate to the condition. The matches were presented for a total of 8 seconds.
In the No Label condition, the instructions for the presentation of the target were "Look! See this one? Look at this one. See this one." During the intertrial interval, the infant heard "Can you find another one?" The infant was then presented with the thematic and basic-level match with the following instructions: "Where's another one? Find the other one." In the Novel Label condition, the target was labelled with a one-syllable nonsense word (e.g., wug). The instructions were "Look! This is a wug. See the wug? Look at the wug." During the inter-trial interval, the infant heard "Can you find another wug? The matches then appeared on the screen and the infant was instructed as follows: " Where's the wug? Find the wug."
Infants were presented with 10 trials in their condition (target trial followed by matches trial for each of 5 triads). The side presentation of the thematic and basic-level matches was counterbalanced. The order of the triads was randomly determined and then standardized across subjects. Due to the random ordering, no pattern in the left-right presentation of the basic-level match was evident. The infant's parent was instructed not to label the pictures nor make any attempt to direct the infant toward a particular screen during testing. Parents were also asked to complete the MacArthur Communicative Development Inventory: Words and Sentences (Fenson et al., 1991) as a measure of their infants' expressive vocabularies.
Interobserver agreement and data reduction. The amount of time (in seconds) that the infant looked at each stimulus and offscreen was coded from the videotapes. Two trained coders, who were unaware of the order and position of the pictures, coded a random selection of 20% of the sessions. One coder was a graduate student and the other was a research assistant. Interobserver reliability was high, with a mean of r = .93 (SD = .08).
Three variables were coded for each trial: visual fixation time on the right screen, time on the left screen, and time off screen. The data of interest were the mean basic-level and thematic visual fixation times which were obtained by averaging across the five triads. Data were screened to assess for outliers, normality, and skewness. Three infants were eliminated from the analyses as they were significantly statistical outliers (t-scores +/- 3 on one of the variables), as indicated in the Participants section. In addition, data were omitted for any trial on which: (a) visual fixation to one of the stimuli was 100% of total fixation to the pair, (b) total fixation time was less than 2 seconds (1/4 of trial duration); (c) the infant did not fixate the target in the trial preceding the matches. These criteria are similar to those used in previous studies (e.g., Poulin-Dubois, Serbin, Kenyon, & Derbyshire, 1994; Reznick & Goldfield, 1992). Four trials in total (1%) were eliminated for these reasons. Finally, participants were excluded from the analyses if they exhibited a side preference across all trials (i.e., looking at one side more than 65% of the time across all trials) or if their total fixation time to both stimuli was less than 2 seconds (1/4 of trial length) for more than 30% of the trials. No eliminations were made on these bases.
Effect of label condition on attention to the matches. The mean duration of visual fixation for the thematic and basic-level matches in each condition are presented in Table 2. To assess the influence of a novel label on infants' attention to basiclevel versus thematic matches, we conducted a 2 (Label: Novel Label versus No Label) x 2 (Match Type: Basic-level versus Thematic) mixed factor analysis of variance (ANOVA). Label was the between-subjects factor and Match Type was the within-subjects factor. This analysis yielded a significant main effect of Match Type, F(1, 31) = 8.14, MSE = .55, p < .01, and a significant Label x Match Type interaction, F(1, 31) = 8.18, MSE = .55, p < .01. We then compared infants' visual fixation of the basic and thematic matches across conditions. Infants looked significantly longer at the basic-level match in the Novel Label condition than in the No Label condition, t(31) = 2.12, SEM = .24, p < .05. In contrast, infants looked longer at the thematic match in the No Label condition than in the Novel Label condition, t(31) = - 2.57, SEM = .22, p < .05.
Correlations between vocabulay size and attention to the matches. To assess the effect of vocabulary size on infants' responding to the basic-level and thematic match across the two conditions, we computed the correlations that are presented in Table 3. In order to control for any contribution of age, we computed partial correlation coefficients between infants' productive vocabulary size and the total amount of visual fixation to the basic-level and thematic match, with the effects of age partialled out. Infants' vocabularies ranged from 2 to 350 words (M = 91.09, SD = 82.20). In the Novel Label condition, there was a significant positive correlation between vocabulary size and visual fixation to the basic-level match, r = .58,p<.02. This correlation suggests that the larger the infants' vocabulary, the more likely they were to look at the basic-level match in the Novel Label condition. As expected, vocabulary size was not related to visual fixation of the thematic match in the Novel Label condition. In the No Label condition, there was a significant correlation between vocabulary size and visual fixation of the thematic match, r = .65, p < .02. No other correlations were significant.
The results of this experiment indicate differential patterns of attention to basic-level and thematic matches as a function of the presence or absence of a novel label. Infants who were instructed to look for the referent of a novel label looked significantly longer at basic-level matches than those who did not receive these instructions. An opposite pattern was found with the thematic matches: Infants in the No Label condition attended significantly longer to thematic matches than those in the Novel Label condition. These findings suggest that the presence of a Novel Label has differential effects on attention to basic-level and thematic matches.
A second notable finding in this study was that attention to the basic-level match in the presence of a novel label was related to vocabulary size. The larger the infants' vocabulary, the more likely they were to attend to the taxonomic match in the Novel Label condition. Because the possible confounding effects of age were controlled, this suggests a linguistic, rather than an age-related, phenomenon. It is worth noting that there was a shift in the correlation between vocabulary and preferences across the two conditions, suggesting a strong effect of labels on taxonomic matching. These findings are discussed in more detail in the Discussion section following Experiment 2.
In view of the findings from Experiment 1, we next sought to clarify the hierarchical level at which infants first apply the word-category linkage. Although there is evidence that words may highlight categories at multiple hierarchical levels, we wished to examine whether infants would assume first that words label basic-level objects. Using the preferential looking paradigm, infants were presented with a target (e.g., a banana), followed by a basic-level match (e.g., another banana) and a superordinate-level match (e.g., a pear) with or without a novel label.
Participants. The participants were 39 16- to 20-month-old infants recruited from the same population as those in Experiment 1. Seven infants were excluded from the sample for one of the following reasons: exhibited a side bias during the experiment (n = 4; see Data Reduction subsection); did not complete the task (n = 1); had reported hearing losses (n = 2). The final sample consisted of 32 infants (11 boys and 21 girls) ranging in age from 18.06 months to 20.10 months (M = 19.04 months, SD = .58).
Stimuli. The stimuli were 10 coloured picture triads composed of a target stimulus (e.g., a banana), a picture of an object from the same basic-level category (e.g., another banana), and a picture of an object from the same superordinate-level category (e.g., a pear) (see Table 1 for complete listing). These pictures were digitized and presented using the preferential looking apparatus described in Experiment 1.
Design and procedure. Infants were randomly assigned to one of the two experimental conditions: a No Label condition (n = 16) and a Nove/Label condition (n = 16). These conditions were identical except for the instructions presented with the No Label condition, which acted as a baseline condition. As in Experiment 1, the target of each picture triad was presented on both screens for 8 seconds with a recorded female voice giving the appropriate instructions. After a 3second intertrial interval, the basic and superordinate match were presented on different screens and the same female voice gave the verbal instructions appropriate to the condition. The matches were presented for a total of 8 seconds. The instructions in both the No Label condition (i.e., "Look! See this one?" followed by "Where's another one?") and Novel Label condition (i.e., "Look! This is a wug." followed by "Where's the wug?") were identical to those presented in Experiment 1.
Infants were presented with 21 trials in their condition (target trial followed by matches trial for each of 10 triads). After the presentation of a block of 10 trials, the infants were presented with an interest trial (a picture of a doll and a vehicle accompanied by a female voice saying, "Look at the toys!"). This trial was intended to keep infants' attention engaged in the task. The side of presentation of the basic-level and superordinate-level matches was counterbalanced across the protocol. As in Experiment 1, due to the random ordering of triads, no pattern in the left-right presentation of the taxonomic match was evident. The infant's parent was instructed not to label the pictures nor make any attempt to direct the infant toward a particular screen during testing. Parents were also asked to complete the MacArthur Communicative Development Inventory: Words and Sentences (Fenson et al., 1991).
Interobserver agreement and data reduction. A random selection of 20% of the sessions was coded twice by two trained coders, who were unaware of the order and position of the pictures. Interobserver reliability was high with a mean of r = .93 (SD = .03).
The data of interest were the mean basic-level and superordinate-level visual fixation times which were obtained by averaging across the 10 triads. Data were screened according to the criteria outlined in Experiment 1 and 21 trials in total (6%) were eliminated. As outlined in the Participants section, four infants were excluded from the analyses because they exhibited a side preference across all trials (i.e., looking at one side more than 65% of the time across all trials).
Effect of label condition on attention to the matches. The mean durations of visual fixation for the basic- and superordinatelevel matches in each condition are presented in Table 2. To assess the influence of a novel label on infants' attention to basic-level versus superordinate-level matches, we conducted a 2 (Label: Novel Label versus No Label) x 2 (Match Type: Basic-level versus Superordinate-level) mixed factor ANOVA. Novel Label was the between-subjects factor and Match Type was the within-subjects factor. This analysis yielded a significant main effect of Match Type, F(1, 30) = 7.45, MSE = .28, p < .01, and a significant Label x Match Type interaction, F(1, 30) = 6.65, MSE = .28, p < .02. We then compared infants' visual fixation of the basic and superordinate matches across conditions. Infants in the Novel Label condition looked significantly longer at the basic-level match than infants in the No Label condition, t(30) = 3.44, SEM = .15, p < .01. In contrast, infants attended to the superordinate-level match similar amounts of time in the No Label and Novel Label conditions, t(15) = -.73, SEM = .20, n.s.
Correlations between vocabulary site and attention to the matches. To assess the effect of vocabulary size on infants' responding to the basic-level and superordinate-level matches across the two conditions, we computed partial correlations between infants' productive vocabulary size and the total amount of visual fixation to the basic-level and superordinate-level match separately for infants in each condition, controlling for the effects of age. These correlations are presented in Table 3. Infants' vocabularies ranged from 8 to 367 words (M = 104.50, SD = 90.35). In the Novel Label condition, the correlation between vocabulary size and visual fixation to the superordinate-level match approached significance (r = .48, p < .10). In the No Label condition, there were no significant correlations.
Consistent with the results of Experiment 1, infants who were instructed to look for the referent of a novel label looked significantly longer at the basic-level match than those who did not receive these instructions. The two groups, however, did not differ in their attention to the superordinate matches. These findings indicate that infants do first assume that novel words label categories at the basic-level, a finding discussed further in the General Discussion.
In contrast to Experiment 1, there were no significant correlations between attention to the basic-level match in the Novel Label condition and vocabulary size. This lack of correlation might be attributable to the differing contrasts. between the experimental stimuli that infants were presented with in the two experiments: In Experiment 1, infants were presented with a contrast between a taxonomic match and an associatively or thematically related match whereas in Experiment 2, the contrast was between two taxonomic matches. We propose the following, admittedly speculative, account for the different patterns of correlations across the two studies. In Experiment 1, the correlation between attention to the basic-level match when labelled and vocabulary size suggests that 18-month-old infants with larger vocabularies are more sensitive to the linkage between taxonomic categories and words. The lack of correlation between vocabulary and differential attention to two taxonomic matches in Experiment 2 may suggest, however, that more word learning experience does not foster a specific link between labels and a particular type of taxonomic category at this age. This proposal is clearly speculative and needs to be evaluated in future studies.
We should note that there was one correlation in Experiment 2 that approached significance -- between vocabulary size and attention to the superordinate match in the Label condition. Although this correlation must be interpreted cautiously as it did not reach statistical significance, it does suggest that with more words in productive vocabulary, infants realize that words may also label categories at the superordinate level. This finding is in accord with a recent study which found that by the age of 24 months, children with a higher proportion of nouns in their vocabulary were more likely to use nonperceptual information about category membership when extending words (Poulin-Dubois, Frank, Graham, & Elkin, in press). In this same study, no correlations were found between attention to nontaxonomic shape matches and vocabulary size.
The present experiments were designed to examine the wordcategory linkage in early lexical development. The results of the first experiment provide some evidence that infants do honour the word-category linkage: The assignment of a novel label to an object produced more attention to an object from the same basic-level category relative to an object from the same thematic category. The results of the second experiment indicate that the presence of a novel label produces longer mean attention to an object from the same basic-level category, but not an object from the same superordinate-level category, as those infants in the Novel Label condition attended significantly longer to the basic-level match than those infants in the No Label condition. Taken together, these findings provide important insights into the nature of the word-category linkage during infancy.
First, consistent with both Waxman and Hall's (1993) findings with 15- to 20-month-old infants and research with preschool-age children (e.g., Golinkoff et al., 1995; Hall, 1993), the present studies indicate that infants do honour the word-category linkage. Infants attended differentially to a basic-level match as a function of the presence of a novel label. This finding supports the notion that the presence of language has unique effects on children's categorization behaviour. Second, the demonstration of the word-category linkage, during a developmental period where inconsistent effects have been found to date, offers additional support for the notion that this word-learning bias may play a pivotal role in the rapid lexical acquisition that occurs during the second year of life.
The results of Experiment 2 further clarify the nature of the word-category linkage in infancy: Infants first assume that novel words label basic-level categories, rather than superordinate categories. This finding is consistent with studies documenting the predominance of basic-level labels in early vocabulary (e.g., Anglin, 1977; Clark, 1983; PoulinDubois et al., 1995). The present findings do not rule out the possibility that infants may apply the word-category linkage at all categorical levels, depending on the context. They do indicate that infants initially assume that new words refer to the basic-level categories, in accord with the proposal made recently by Woodward and Markman (in press). These authors argue that children initially expect novel words to label basic-level objects and may require some degree of contextual support to interpret a novel label as referring to superordinate- or subordinate-level categories.
An initial expectation that new object words are basic-level labels is an efficient or productive word-learning strategy for infants for a number of reasons. First, as proposed by Rosch and colleagues (Mervis & Rosch, 1981; Rosch et al., 1976), basic-level categories are considered psychologically fundamental as there is maximal similarity within members of the category but minimal similarity with members of other categories. One can argue that members of basic-level categories actually share both perceptual and 'id" similarity. Thus, an initial assumption that words are basic-level terms would offer infants an easily accessible rule to extend novel words, as they could quickly determine whether objects belonged to the same basic-level category. Furthermore, by grouping objects together into basic-level categories with common labels, infants are likely to discover other commonalties amongst these objects, such as functional similarity. Research with preschool-age children suggests that perceptual similarity (which is high in basic-level categories) may indeed "bootstrap" other more abstract knowledge about objects (e.g., Gentner & Imai, 1994). Thus, this initial expectation may. pave the way for learning labels at other categorical levels. Finally, the assumption that words first label basic-level categories is consistent with the input that infants receive: Several studies have revealed that parents tend to use basiclevel terms when they first label objects for their children (e.g., Anglin, 1977; Blewitt, 1983; Callanan, 1985; Poulin-Dubois et al., 1995).
Infants' expectation that words first label basic-level categories is consistent with recent research indicating that perceptual similarity is a dominant basis for the word extension during infancy and the early preschool years (e.g., Baldwin, 1992; Imai, Gentner, & Uchida, 1994). In fact, given the nature of the contrasts presented to infants in both studies, it might be argued that any preference for the basiclevel match can be explained with reference to similarity alone rather than to taxonomic kinds. However, as noted above, it is difficult to dissociate perceptual similarity and "kind" similarity in basic-level categories. Other recent research does suggest that infants are not relying solely on perceptual similarity in word extension tasks. For example, PoulinDubois et al. (in press) found that 18- and 24-month-old infants were more likely to extend a novel label to a shape match (e.g., a telephone receiver) with no categorical relation to the target (e.g., a banana) than to a taxonomic match (e.g., a strawberry) which bore little similarity to the target. However, by 24 months of age, those children with a larger proportion of object names in their vocabulary were more likely to use taxonomic information when extending novel words. These results are consistent with the findings of other recent studies which suggest a shape-to-taxonomic shift over time might characterize children's assumptions about word meanings (e.g. Golinkoff et al., 1995; Imai et al., 1994).
We also examined the effects of differences in wordlearning experience, as measured by productive vocabulary, on the recognition of the linkage between word and categories. In Experiment 1, we found that the larger the infant's vocabulary, the more likely he or she was to attend to the basic-level match when the pictures were labelled. This may suggest that the acquisition of a larger vocabulary strengthens the linkage between words and taxonomic categories (as opposed to thematic relations). However, evidence for the proposal that such a linkage may exist with specific taxonomic categories (e.g., basic-level as opposed to superordinate-level) was not found in the present studies and awaits evaluation in future investigations.
In summary, early in lexical development, infants. narrow their lexical extensions to taxonomic categories, as opposed to thematic relations. Furthermore, they narrow these lexical extensions to taxonomic categories at the basic-level, as opposed to the superordinate-level.
This research was supported by grants to the first author from NSERC, the University of Calgary Research Grants Council and the Calgary Herald Young Innovator Award, and grants from NSERC and the Fonds FCAR awarded to the third author. The second author was supported by a predoctoral award from the Fonds FCAR. We would like to thank the infants and parents who enthusiastically participated in these studies. We would also like to thank Natasha Bergeron, Teressa Grosko, Mandy Steiman, Barbara Levine, Ruth Miller, Megan Wgs, and Joelene Huber for their assistance with these studies. Special thanks to Ilana Frank for her assistance in obtaining the thematic matches used in Experiment 1. Geoff Hall, Andrea Welder, and two anonymous reviewers provided us with invaluable comments on an earlier version of this manuscript. Portions of the data from Experiment 1 were presented at the International Conference on Infant Studies, Paris, France, June, 1994. Data from Experiment 2 were presented at the International Conference on Infant Studies, Atlanta, Georgia, April, 1998. Correspondence and requests for reprints should be directed to S. A. Graham, Department of Psychology, University of Calgary, 2500 University Dr. N.W., Calgary, AB T2N 1N4 (email: grahams(acs.ucalgary.ca).
Etant donne la facilite avec laquelle les enfants ajoutent de nouveaux mots a leur vocabulaire, de nombreux chercheurs affirment que l'acquisition des connaissances lexicales est guidee par des contraintes qui facilitent l'apprentissage de mots nouveaux (Golinkoff, Mervis & Hirsh-Pasek, 1994; Markman, 1989,1992). L'une de ces contraintes a recemment fait l'objet de nombreuses etudes empiriques: il s'agit de l'hypothese taxinomique ou, en d'autres termes, du lien motcategorie (Markman, 1989; Waxman et Hall, 1993). Du fait de cette contrainte, les enfants presument qu'un mot nouveau renvoie, non pas a des objets lies en fonction de l'espace ou d'un theme, mais plutot i des elements du meme type ou de la meme categorie taxinomique (Markman, 1991,1992). La presente etude visait a examiner deux aspects de la nature du lien mot-categorie au debut de l'acquisition des connaissances lexicales. Premierement, a la lumiere des resultats variables des etudes effectuees anterieurement aupres de jeunes enfants, nous avons cherche a determiner si, au debut de l'acquisition des connaissances lexicales, les enfants reconnaissent le lien mot-categorie lorsqu'on leur presente des possibilites de combinaisons taxinomiques et thematiques (experience 1). Deuxiemement, nous avons examine les attentes initiales des enfants quant au niveau hierarchique du referent d'un mot nouveau lorsque des possibilites de combinaisons de base et superordonnees leur sont presentees (experience 2). Dans l'experience 1, on a demande a des bebes iges de 16 a 19 mois de jumeler un objet (p. ex. une poussette) soit avec un element de la categorie de base (p. ex. une autre poussette), soit avec un element de la categorie thematique (p. ex. un bebe), en leur montrant ou non un nouveau mot. Des stimuli ont ete presentes selon le paradigme preferentiel. Les resultats indiquent que les bebes auxquels on a montre un nouveau mot ont examine beaucoup plus longtemps la combinaison de la categorie de base que ceux auxquels aucun nouveau mot n'a ete montre. Dans l'experience 2, on a presente a des bebes ages de 16 a 20 mois un objet (p. ex. une banane), puis un element de la categorie de base (p. ex. une autre banane) et un element de la categorie superordonnee (p. ex. une poire), en leur montrant ou non un nouveau mot. Comme dans l'experience 1, les bebes auxquels on a montre un nouveau mot ont examine beaucoup plus longtemps la combinaison de la categorie de base que ceux auxquels aucun nouveau mot n'a ete montre.
Examinees ensemble, ces conclusions nous eclakent grandement sur la nature du lien mot-categorie chez les jeunes enfants. En premier lieu, les resultats des presentes etudes indiquent que les jeunes enfants reconnaissent effectivement le lien mot-categorie. La presence d'un nouveau mot a modifie la facon dont ils ont traite un jumelage au niveau de base. Cette conclusion renforce l'idee que, chez les enfants, la presence du langage a des consequences particulieres sur la categorisation. En deuxieme lieu, la demonstration du lien mot-categorie pendant un stade de developpement ou des effets variables ont ete constates a ce jour, corrobore l'hypothese selon laquelle cette contrainte a l'apprentissage du vocabulaire peut servir de pivot dans l'acquisition rapide des connaissances lexicales qui survient durant la deuxieme annie de la vie. Enfin, les resultats de l'experience 2 precisent la nature du lien mot-categorie chez les jeunes enfants: ces derniers presument d'abord que les nouveaux mots designent des categories de base, plutot que des categories superordonnees. Cette conclusion est conforme a celles des etudes sur la predominance de mots des categories de base dans le vocabulaire des jeunes enfants (p. ex. Anglin, 1977; Clark, 1983; Poulin-Dubois et al., 1995).
Anglin,J. M. (1977). Word, object and conceptual development. New York: Norton.
Anglin, J. M. (1993). Vocabulary development A morphological analysis. Monographs of the Society for Research in Child Development, 58(10, Serial No. 238).
Balaban, M. T., & Waxman, S. R. (1997). Do words facilitate object categorization in 9-month-old infants? Journal of Experimental Child Psychology, 64, 3-26.
Baldwin, D. A. (1992). Clarifying the role of shape in children's taxonomic assumption. Journal of Experimental Child Pychology, 54, 392-416.
Baldwin, D. A., & Markman, E. (1989). Establishing word-object relations: A first step. Child Development, 60, 381-398.
Bauer, P. J., & Mandler, J. M. (1989). Taxonomies and triads: Conceptual organization in one- to two-year-olds. Cognitive Psychology, 21, 156-184.
Behrend, D. A. (1988). Overextensions in early language comprehension: Evidence from a signal detection approach. Journal of Child language, 15, 63-75.
Behl-Chadha, G. (1996). Basic-level and superordinate-like categorical representations in early infancy. Cognition, 60, 105-141. Benedict, H. (1979). Early lexical development: Comprehension and production. Journal of Child Language, 22, 89-106. Blewitt, P. (1983). Dog versus collie: Vocabulary in speech to young children. Developmental Pychology, 19, 602-609.
Callanan, M. A. (1985). How parents label objects for young children: The role of input in the acquisition of category hierarchies. Child Development, 56, 508-523.
Callanan, M. A. (1989). Development of object categories and inclusion relations: Preschoolers' hypotheses about word meanings. Developmental Psychology, 25, 207-216. Clark, E. V. (1983). Meanings and concepts. In J. H. Flavell & E. M. Markman (Eds.), Handbook of child p.cho/eg (Vol 3): Cognitive development, pp. 787-840. New York: John Wiley & Sons. D'Entremont, B., & Dunham, P. J. (1992). The noun-category bias phenomenon in 3-year-olds: Taxonomic constraint or translation? Cognitive Development, 7, 47-62.
Dromi, E. (1987). Early lexical development. Cambridge: Cambridge University Press.
Echols, E. H. (1992, May). Developmental changes in attention to labeled events during the transition to language. Poster presented at the International Conference on Infancy Studies, Miami Beach, FL. Fenson, L., Dale, P. S., Reznick, J. S., Thal, D., Bates, E., Hartung, J. P., Pethick, S., & Reilly, J. S. (1991). MacArthur Communicative Development Inventories. San Diego: San Diego State University.
Gentner, D., & Imai, M. (1994). A further examination of the shape bias in early word learning. In E. V. Clark (Ed.), Proceedings of the Child Language Research Forum. Stanford University. Goldfield, B. A., dc Reznick, J. S. (1990). Early lexical acquisition: Rate, content, and the vocabulary spurt. Journal of Child Language, 17, 171-183.
Golinkoff, R. M., Hirsh-Pasek, K., Cauley, K. M., & Gordon, L. (1987). The eyes have it: Lexical and syntactic comprehension in a new paradigm. Journal of Child Language, 14, 23-45. Golinkoff, R. M., Mervis, C. B., & Hirsh-Pasek, K. (1994). Early object labels: The case for lexical principles. Journal of Child Language, 21, 125-155.
Golinkoff, R. M., Shuff-Bailey, M., Olguin, R., & Ruan, W. (1995). Young children extend novel words at the basic level: Evidence for the principle of categorical scope. Developmental Psychology, 31, 494-507.
Hall, D. G. (1993). Basic-level individuals. Cognition, 48, 199-221. Hirsh-Pasek, K., & Golinkoff, R. M. (1996). The origins of grammar:
Evidence from ear# language comprehension. Cambridge: MIT Press. Imai, M., Gentner, D., & Uchida, N. (1994). Children's theories of word meaning: The role of shape similarity in early acquisition. Cognitive Development, 9, 45-75.
Landau, B., Smith, L. B., & Jones, S. S. (1988). The importance of
shape in early lexical learning. Cognitive Development, 3, 299-321. Mandler, J. M., & Bauer, P. J. (1988). The cradle of categorization: Is the basic level basic?. Cognitive Development, 3, 247-264. Mandler,J. M, Bauer, P. J., & McDonough, L. (1991). Separating the sheep from the goats: Differentiating global categories Cognitive Psychology, 23, 263-298.
Markman, E. M. (1989). Categorization and naming: Problems of induction. Cambridge: MIT Press.
Markman, E. M. (1991). The whole object, taxonomic, and mutual exclusivity assumptions as initial constraints on word meaning. In J. P. Byrnes & S. A. Gelman (Eds.), Perspectives on language and cognition: Interrelations in development, pp. 72-106. Cambridge: Cambridge University Press.
Markman, E. M. (1992). Constraints on word learning: Speculations about their nature, origins, and domain specificity. In M. R.
Gunnar & M. P. Maratsos (Eds.), Minnesota Symposium on Child Psychology, 25, pp. 55-101. Hillsdale, NJ: Lawrence Erlbaum Associates.
Markman, E. M., & Hutchinson, J. E. (1984). Children's sensitivity to constraints on word meaning Taxonomic versus thematic relations. Cognitive Pychology, 16, 1-27. Mervis, C. B., & Rosch, E. (1981). Categorization of natural objects. In M. R. Rosenzweig & L W. Porter (Eds.), Annual review of psychology (Vol. 32, pp. 89-115). Palo Alto, CA: Annual Reviews. Nelson, K. (1973). Structure and strategy in learning to talk. Monographs of the Society for Research in Child Development, 38 (Serial Number 149).
Poulin-Dubois, D., Frank, I., Graham, S. A., & Elkin, A. (in press). The role of shape in toddlers' lexical extensions. British Journal of Developmental Psychology.
Poulin-Dubois, D., & Graham, S. A. (1994). Infant categorization
and early object word meaning. In A. Vyt, H. Bloch & M. H. Bornstein (Eds.), Early child development in the French tradition, pp. 207-225. Hillsdale, NJ: Lawrence Erlbaum. Poulin-Dubois, D., Graham, S., & Sippola, L. (1995). Early lexical development: The contribution of parental labelling and infants' categorization abilities. Journal of Child Language, 22, 325-343. Poulin-Dubois, D., Serbin, L. A., Kenyon, B., dc Derbyshire, A. (1994). Infants' intermodal knowledge about gender. Developmental Psychology, 30, 436442.
Quinn, P. C., Eimas, P. D., dc Rosenkrantz, S. L. (1993). Evidence for representations of perceptually similar categories by 3-month-old and 4-month-old infants. Perception, 22, 463-475. Reznick, J. S. (1990). Visual preference as a test of infant word comprehension. Applied Pycholinguistics, 11, 145-166. Reznick, J. S., & Goldfield, B. A. (1992). Rapid change in lexical development in comprehension and production. Developmental Psychology, 28, 406-413.
Roberts, K. (1995). Categorical responding in 15-month-olds: Influence of the noun-category bias and the covariation between
visual fixation and auditory input. Cognitive Development, 10, 21-41. Rosch, E., Mervis, C. B., Gray, W. D., Johnson, D. M., & Boyes-Braem, P. (1976). Basic objects in natural categories. Cognitive Psychology, 8, 382-439.
Soja, N., Carey, S., & Spelke, E. (1991). Ontological categories guide young children's inductions of word meaning: Object terms and substance terms. Cognition, 38, 179-211. Taylor, M., & Gelman, S. A. (1989). Incorporating new words into the lexicon: Preliminary evidence for language hierarchies in two-year-old children. Child Development, 60, 625-636. Waxman, S. R. (1990). Linguistic biases and the establishment of conceptual hierarchies: Evidence from preschool children. Cognitive Development, 5, 123-150. Waxman, S. R., & Gelman, R. (1986). Preschoolers' use of superordinate relations in classification and language. Cognitive Development, 1, 139-156.
Waxman, S. R, & Hall, D. G. (1993). The development of a linkage between count nouns and object categories: Evidence from
fifteen- to twenty-one-month-old infants. Child Development, 64, 1224-1241.
Waxman, S. R., & Kosowski, T. D. (1990). Nouns mark category relations: Toddlers' and preschoolers' word-learning biases. Child Development, 61, 1461-1473.
Waxman, S. R., & Markow, D. B. (1995). Words as invitations to form categories: Evidence from 12- to 13-month-old infants. Cognitive Pychology, 29, 257-302. Waxman, S. R., & Senghas, A. (1992). Relations among word meanings in early lexical development. Developmental Pychology, 28, 862-873.
Waxman, S. R., Senghas, A., & Beneviste, S. (1997). A cross-linguistic examination of the noun-category bias: Its existence and specificity in French- and Spanish-speaking preschool-aged children. Cognitive Psychology, 32, 183-218.
Woodward, A. L., & Markman, E. M. (in press). Early word learning. In W. Damon, & R. Siegler, (Eds.), Handbook of Child Pychology,
SUSAN A. GRAHAM, University of Calgary
RACHEL K. BAKER and DIANE POULIN-DUBOIS, Concordia University…
Questia, a part of Gale, Cengage Learning. www.questia.com
Publication information: Article title: Infants' Expectations about Object Label Reference. Contributors: Graham, Susan A - Author, Baker, Rachel K - Author, Poulin-Dubois, Diane - Author. Journal title: Canadian Journal of Experimental Psychology. Volume: 52. Issue: 3 Publication date: September 1998. Page number: 103+. © Canadian Psychological Association Mar 2008. Provided by ProQuest LLC. All Rights Reserved.