Input Effects on the Acquisition of Finiteness

Understanding input effects is crucial for delineating the roles of Universal Grammar and learning in grammatical development. Unlike the input effects seen in vocabulary development (Huttenlocher, Haight, Bryk, Seltzer, & Lyons, 1991; Hart & Risley, 1995), effects on grammatical development have been more difficult to identify (Newport, Gleitman, & Gleitman, 1977; Hoff-Ginsberg, 1998; Huttenlocher, Vasilyeva, Cymerman, & Levine, 2002; Huttenlocher, Waterfall, Vasilyeva, Vevea, & Hedges, 2010; see also Valian, 1995 for review). One weakness of earlier studies has been their generality. For example, Newport et al. (1977) tried to uncover relationships between ten aspects of maternal speech, some structural and some interactional, and four measures of structure in child output, as well as the ubiquitous measure of mean length of utterance (MLU). Unfortunately, a methodological problem arises when a large number of correlations are tested; the multiplicity problem (Benjamini & Hochberg, 1995). As Valian (1995) has pointed out, studies in the mold of Newport et al. (1977) have turned up “no robust findings” (p. 512). In contrast, Hadley, Rispoli, Fitzgerald, and Bahnsen (2011) focused on a single input variable, input informativeness for tense, based on the theoretical work of Legate and Yang (2007). Hadley et al. found that proportionately richer signaling of English tense in parental input predicted individual differences in the rate of children’s acquisition of finiteness between 21 and 30 months of age. Another weakness of earlier studies has been a lack of control for differences in children’s developmental levels when parent input is sampled. Taking Newport et al. (1977) as an example, age and level of syntactic developmental varied in the children studied. The effects of age and prior knowledge had to be controlled statistically. In contrast, Hadley et al. (2011) identified an optimal point in time to explore input effects based on Rispoli, Hadley, and Holt’s (2009) quantitative model of growth in finiteness. Because the vast majority of children developing typically do not show emergence of finiteness morphemes at 21 months, but growth in the finiteness system is evident shortly thereafter, Hadley et al. measured input when children were 21 months old. Additionally, child vocabulary at 21 months was used to control for individual differences in development before measureable growth in finiteness began. Although the results of Hadley et al. (2011) supported Legate and Yang’s (2007) Variational Learning (VL) hypothesis, VL is incompatible with one empirical finding of the Rispoli et al. (2009) finiteness growth model. VL predicts variation at the onset of learning. Rispoli et al. (2009) found that finiteness productivity for the group was not significantly different from zero at the centering point (i.e., 21 months) nor were there significant individual differences among children at 21 months of age. Thus, a no-intercept model was the best fit to the data. In other words, children acquiring the finiteness system of English do not vary at the onset of learning as VL predicts. Rather, productivity of finiteness marking grows gradually from zero from during the third year of life. Additionally, the VL model does not attempt to explain differences in the learning of individual finiteness morphemes, but this almost certainly exists and must be part of the process of acquiring finiteness in English. On the other hand, the Gradual Morphosyntactic Learning hypothesis (GML; Rispoli & Hadley, 2011; Rispoli, Hadley, & Holt, 2012) is compatible with a no-intercept model for the acquisition of finiteness. GML also directly addresses how the learning of individual morphemes fits into the larger picture of acquiring