A training study to establish the cause of children’s ‘defaulting errors’ in Spanish
Martin, J., Bannard, C. & Pine, J. M. (2017). A training study to establish the cause of children’s ‘defaulting errors’ in Spanish. Paper presented at the 14th Congress of the International Association for the Study of Child Language (IASCL), Lyon, France.
Support for constructivist models of morphological development comes from the many studies that show a relationship between patterns of errors in child speech and the frequencies of forms in child-directed speech. Such findings, however, are only correlational. We introduce an experimental approach designed to establish a causal link between linguistic experience and language development.
We test the claim that Spanish children’s present-tense errors - specifically their use of the third person singular (3SG) form in non-3SG contexts – reflect ‘defaulting’ to the most frequent verb form. 45 adults and 34 children (aged 9-10) were shown animations involving different characters (including an animated version of themselves). These were described in Spanish using different person/number forms, which participants were asked to imitate. After 40-80 such training trials participants were shown another sequence of animations and asked to describe them without guidance. The distribution of inflections at training was either skewed toward the 3SG form as in natural Spanish, or was uniform. This was repeated on 3 separate days.
Both adults and children showed greater accuracy on 3SG forms but reduced accuracy on other forms in the skewed relative to the uniform condition, and 3SG substitutions accounted for a significantly greater proportion of errors in the skewed condition. However, 3SG substitutions were also common in the uniform condition, despite there being no frequency bias at training.
These results suggest that the frequency of 3SG forms does affect learning and is likely to play a role in children’s ‘defaulting’. However, it is apparently not the only cause, with semantic or phonological factors also playing a role. We are currently exploring an extension of our method in which we decouple phonology and semantics at training in order to distinguish between these two possibilities.