The Effect of Input on Children’s Cross-Categorical Use of Polysemous Noun-Verb Pairs

Abstract

Using an observational task followed by an experimental task with an Intermodal Preferential Looking Paradigm, we examined the effect of input on children’s acquisition of class extension rules by investigating the relationship between the amount of polysemous noun-verb pairs in French-speaking 2-year-olds’ input and both their spontaneous production of these words and their comprehension of novel instances of these words. Study 1 demonstrated that the number of words children used cross-categorically was related to the number of words their mothers used cross-categorically. Children also used object-denoting words cross-categorically more often than nonobject- and action-denoting words. Study 2 demonstrated that only children whose mothers frequently used noun-verb pairs cross-categorically were able to understand the cross-categorical use of the novel object-denoting words in the experimental task. This suggests that semantic and distributional cues associated with object-denoting noun-verb pairs in the input play an important role in children’s acquisition of class extension rules.

Notes

¹ “Logical relatedness of meanings was determined on the basis of Putejovsky’s (1995) analyses of nominalization and artifact kind denoting nominals. For example, a noun and a verb were related if their lexical entries had identical event, qualia, and argument structures (with the exception of category specific information like event headedness for verbs). Other relevant relations involve associations in the lexical conceptual paradigm (p. 133), and the specification of a given verb (e.g., swing) in a related nominal’s telic quale, which may be understood as representing the function of the object denoted by the nominal (see Moravcsik 1998). Otherwise stated, noun/verb pairs were considered to have logically related meanings if the action denoted by a verb form (e.g., “Drink your juice”) took on a noun form that denoted the same action (e.g., “Have a drink of juice”) or a physical object or stuff associated with the action (e.g., “Finish your drink”)” (Oshima-Takane et al. 2001:1157).

² A one-tailed test was used when there was a priori prediction about the directionality of the difference, whereas a two-tailed test was used when there was no a priori prediction about the directionality of the difference (Ferguson & Takane 1989).

³ The exclusion rates are relatively high because this was our first time using an eye tracker, and several adjustments had to be made in order to obtain the best tracking possible. However, most of these children met the criteria of the observational study. Thus, we report the data from all the children who were included in the observational study, including those who did not meet the criteria of the experimental study, as long as they met the inclusion criteria for the observational study.

⁴ The GazeTracker program fills gaps up to .035 seconds long, by assuming the participant continued looking at the last gazepoint until the next gazepoint arrives. Thus these gaps do not count toward our calculation of inattentiveness. Gaps larger than .035 seconds are considered as gaps in the data, and no data are recorded for that point. These gaps count toward our calculation of inattentiveness. We determined the 25% cutoff by looking at a sample of pilot data to see what was the lowest cutoff we could use that would include as many participants in the final sample as possible, while keeping a similar trend in the data. We calculated inattentiveness by taking the total amount of time the children did not look at either of the test stimuli and dividing that by the total time of the two test clips (12 seconds). The average looking time for each 6-second test clip was 4.52 seconds (72.24%), with a standard deviation of 1.5 seconds. One participant looked 0 seconds at one for the 6-second test trials. For this child, the average score for their age group was used as their score for that test video as in previous studies with an Intermodal Preferential Looking Paradigm (e.g., Naigles, Bavin & Smith 2005).

⁵ We excluded children who did not learn the parent nouns, as Lippeveld & Oshima-Takane (2013) demonstrated that children’s ability to understand novel derived words relies on their ability to learn the parent words.

⁶ For both the independent and the paired t-tests, we performed the arc sine transformation on the proportion scores in order to stabilize variances. Thus all t values reported for these tests were obtained from the arc sine transformed data. However, we reported the original means and standard errors in the text and figures because a monotonic transformation does not change what is originally measured by the dependent variables. Hence we can draw conclusions on the original measures (Ferguson & Takane 1989).