Learning Count Nouns and Adjectives: Understanding the Contributions of Lexical Form Class and Social-Pragmatic Cues

Abstract In two experiments, one hundred ninety-two 3-year-olds, 4-year-olds, and adults heard a novel word for a target object and then were asked to extend the label to one of two test objects, one matching in shape-based object category (the shape match) and the other matching in a property other than shape (the property match). We independently manipulated the lexical form class cues (count noun, adjective) and social-pragmatic cues (point actions, property-highlighting actions) accompanying the label. The impact of these two types of cue on extension differed markedly across age groups. Adults and 4-year-olds extended the word to the property match significantly more often when the term was modeled as an adjective and when it was presented with property-highlighting actions; but adults extended both adjectives and count nouns systematically to the property match when the speaker highlighted the non-shape property, whereas 4-year-olds systematically extended only adjectives to the property match under these conditions. Three-year-olds extended the word to the property match significantly more often when the label was modeled as an adjective but were not significantly affected by the social-pragmatic cues; and they failed to extend either adjectives or count nouns systematically to the property match when the speaker highlighted the non-shape property. We discuss the results in terms of the proposal that word learning draws on cues from multiple sources and the nature of the "shape bias" in lexical development. ACKNOWLEDGEMENTS We thank Rachel Moser and Mijke Rhemtulla for their input and assistance. We are grateful to two anonymous reviewers for helpful comments. Support for this research came from a grant from the Social Sciences and Humanities Research Council of Canada. Notes ∗Significantly different from chance. N = 12 per condition. 1All adults and all 4-year-olds made a selection on all trials. However, six 3-year-olds failed to choose on one trial for either visible or non-visible properties. The missing data were handled as follows: For the proportion of property-match choice analyses, the proportions were calculated as numbers out of two, not three. For the pattern analyses, the missing data points were counted as being neither property-match nor shape-match choices when we determined property-match choosers and shape-match choosers. 2We conducted two further analyses. One focused on the familiarity of the 18 stimulus items. We asked all preschoolers, after they completed the task, to name each object they had seen. Three- and 4-year-olds provided an appropriate count noun on 92.8% and 93.7% of trials, respectively. These findings indicate that the stimuli were familiar, in that children knew a count noun for them. A second analysis focused on the performance of the individual stimulus triads. We calculated the mean proportion of property-match choices for each of the six triads across the 144 participants. Using one-way repeated-measure ANOVAs, we found no significant difference across either the three visible or the three non-visible property triads. In other words, the triads within each property type performed similarly across participants. 3Samuelson, Horst, Schutte, and Dobbertin (Citation2008; see also Samuelson & Smith, Citation2000) have recently reported that 3-year-olds but not 4-year-olds extended a novel object word by shape rather than material, even when the speaker highlighted a non-visible deformable property (squishy or bendable) when introducing the label, and regardless of whether the word was presented as a count noun (e.g., "This is a kiv") or as a mass noun (e.g., "This is some kiv"). The authors suggested that the result reflects an overgeneralization of the shape bias among 3-year-olds, given that 1) most count nouns children know at this age name rigid objects in categories organized by shape and 2) the deformable objects that children know how to label at this age are like rigid things in that they are both solid and named by count nouns (e.g., a napkin, a towel). As a result, the authors argued that 3-year-olds tended to generalize words for demonstrably deformable objects by shape. This result is consistent with our finding on the duck trial that 3-year-olds failed to extend a word on the basis of squishiness, even when they saw an action highlighting the property. 4We note, however, that extension to the shape match does not necessarily indicate a failure to interpret the novel word as a property term. Although it is possible that choosing the shape match reflects a shape-based category interpretation, it is also possible that such a choice reflects a shape-property interpretation. ∗Significantly different from chance. N = 12 per condition. 5As in Experiment 1, we conducted two further analyses. One focused on the unfamiliarity of the 18 stimulus items. We presented the 18 (unfamiliar) stimulus items from Experiment 2 along with the 18 (familiar) stimulus items from Experiment 1 to a separate group of four 4-year-olds. The unfamiliar and familiar objects were intermixed and presented in one of two random orders. Children were asked to name an object if they knew a name, and to say, "I don't know," if they did not. The experimenter first gave them a pair of practice objects. One was familiar (a pencil). All children provided the appropriate count noun and received praised. The other practice object was unfamiliar (a novel elastic object). The experimenter encouraged children to say, "I don't know," for this object, telling children that she had never seen anything like it before. If children did not say, "I don't know," (e.g., tried to label or describe it), the experimenter asked if they were sure and repeated that she had never seen it before. The experimenter continued in this manner, until she elicited an "I don't know" response, which she praised. The experimenter then showed the 36 stimulus items, one at a time. Consistent with the results of Experiment 1, children provided an appropriate count noun for 94.5% of the familiar objects. In contrast, children provided a count noun for only 34.8% of the unfamiliar objects. This 34.8% includes any word that could have been a count noun, including those that were ambiguous (e.g., "stone") or inappropriate (e.g., "carpet" for the furry cylindrical object). Despite using this liberal criterion, we found that the difference between sets of stimuli was highly significant, F(1, 34) = 59.42, p < .0001, reassuring us that we had succeeded in manipulating the familiarity of the objects between experiments. As in Experiment 1, we also calculated the mean proportion of property-match choices for each of the six triads across the 48 participants. Using one-way repeated-measures ANOVAs, we found no significant difference across either the three visible properties or the three non-visible properties triads. Thus, the triads within each property type behaved similarly across participants. 6We have proposed that the property-highlighting actions used in these experiments affected word extension by providing learners with a marker of the intent of the speaker to refer to some feature of an object, but it is possible that these actions operated simply by making the property more salient. If the latter explanation is correct, then we would expect to find that highlighting the property through non-intentional means (e.g., producing the same action accidentally upon the object) would result in a similar effect on word extension.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.