Word order in the input to Argentinian Spanish-learning children: How reliable as cue to clause transitivity?

AUDISIO, CYNTHIA PAMELA; QUIROGA, MACARENA SOL; RAMIREZ, MARÍA LAURA; ROSEMBERG, CELIA RENATA

Congreso; Society for Research in Child Development 2021 Biennial Meeting; 2021

Society for Research in Child Development

The acquisition of transitivity happens in the first 3 years of life (e.g., Pye, 1983; Berman, 1993) but remains challenging to younger children (Bavin & Growcott, 2000`; Hirsh-Pasek, Golinkoff, & Naigles, 1996), and even to 8-year-olds, who still commit transitivity errors (Brooks & Tomasello, 1999). According to constructivist approaches, input provides sufficient information to infer the grammatical patterns of any language (e.g., Lieven 2010). However, not all sorts of information are equally reliable: the Competition Model (MacWhinney, 1987) predicts that the order of acquisition of cues across development is determined by cue reliability. In some languages word order is a highly reliable cue to the transitive pattern. English, for instance, presents strict word order and does not allow null subjects. However, English-learning children younger than 4 use ungrammatical word orders as often as canonical ones in experiments with novel and low frequency verbs (Abbott-Smith, Lieven & Tomasello, 2001, Akhtar 1999, Franck, Millotte & Lassotta, 2011). Spanish-learning children still do this by the age of 4 (Aguado-orea, Witherstone, Bourgeois, & Baselga, 2019) and prefer local cues rather than word order (Kail & Charvillat, 1988). In Spanish the subject can be omitted and word order varies for pragmatic purposes.Although some studies have analysed child-directed speech (CDS) in some varieties of Spanish (Aguado-orea et al., 2019), other sources of variation in children?s input have been neglected. We aim to investigate the frequency and reliability of word order in the entire linguistic input to normally developing Argentinian Spanish-learning children that are socioeconomically diverse (N = 10 whose parents had, on average, 17 years education and N= 10 whose parents had 10 years of education). We have analysed a representative sample of the input to these children at 20 months drawn from a corpus that registers the naturalistic speech produced in their environment.Different word orders occur with varying frequency across SES groups (X2(6)=32.99, p-value<0.001) and addressees (X2(6)=42.6, p-value<0.001). Clauses without argument NPs (i.e., and thus a discernible order) are significantly more frequent in the input to low-SES children and CDS. The presence of postverbal argument NPs is significantly more frequent in the input to middle-SES children and in overheard speech (OHS). Clauses with both a preverbal and postverbal NP (e.g. AVO) are significantly more frequent in OHS. To assess word order reliability in the input to low and middle-SES children, two logistic regression models with clause transitivity as the response variable and word order as the explanatory variable were trained and tested. The models achieved significantly higher accuracy levels than the No Information Rate: 77.3% (95%CI: 69,78) and 78.8% (95%CI: 71,84) on the input to middle and low-SES children respectively. Preliminary evaluation of mismatches between predicted and actual classes revealed that clauses without argument NPs were most often misclassified (V>NP-V>V-NP>NP-V-NP/NP-NP-V/V-NP-NP). Classification errors were more often over OHS than CDS and over clauses produced by child-speakers rather than adult-speakers. Implications of these results are discussed.