The roles of entrenchment, preemption and semantics in the retreat from causative overgeneralisation errors: A combined production and judgment study.

Amy Bidgood, Ben Ambridge, Julian Pine, Caroline Rowland, Franklin Chang, and Daniel Freudenthal presented this poster at the Nijmegen Lectures, Nijmegen, Netherlands in 2013.

Abstract:

Errors in children’s language production can give us a unique insight into the language acquisition process. Productive language use depends on children first learning and then generalising grammatical rules. However, use of these rules must also be appropriately restricted if children are to avoid overgeneralisation errors, e.g. *Don’t giggle me. Causative overgeneralisation errors such as this occur when children hear pairs of sentences such as (a - The ball rolledà John rolled the ball) and assume that other verbs heard in intransitive sentences can also alternate between the two structures. While this productive rule enables creative language use (b - The plate brokeà  Sarah broke the plate), it can also lead to overgeneralisation errors (c: Sarah laughedà *John laughed Sarah (=made Sarah laugh).The present study tests three theories of how children retreat from causative overgeneralisation errors: semantics, entrenchment, and preemption. Previous experimental studies have offered differing levels of support for all three mechanisms, but they have mostly either included very few verbs or used only grammaticality judgments rather than a language production measure.The current study used a priming task to investigate children’s errors in production. Participants (aged 5-6, N=64) and an experimenter took turns to describe animations. The experimenter produced only grammatical, transitive sentences in order to prime the child to use the transitive construction. Half of the target verbs were alternating, the other half were intransitive-only. Thus, for intransitive-only verbs, an ungrammatical utterance would result if the child used the same construction as the experimenter. Regression analysis showed that the rate of causative overgeneralisation errors was significantly predicted by the entrenchment (p<0.001), preemption (p=0.018) and semantics (p=0.008) measures. These findings suggest the need for a model of language production that explains the relative contributions of semantics, entrenchment and preemption in the retreat from causative overgeneralisation errors.