2018 Friday Poster 6593

Friday, November 2, 2018 | Poster Session I, Metcalf Small | 3pm

How Children Learn to Disappear Causative Errors Using Positive Evidence
A. Irani

Problem: Children often extend causative forms to pure intransitive verbs like fall and disappear (e.g. Bowerman 1982), making errors like (1)-(4). This paper addresses the question of how retreat from the overgeneralization of causative forms takes place without negative evidence. Previous approaches have either denied the principled productivity behind causative errors or have proposed statistical preemption and entrenchment as a way of retreating from overgeneralization (Ambridge et al. 2008; Bowerman and Croft 2008; Pinker 1989). These approaches cannot predict the occurrence of causative errors with verbs of varying frequency (Table 1). Instead, we argue that children learn the causative form of verbs from positive evidence in the input. Adopting Yang’s (2016) Sufficiency Principle, we show that children initially overgeneralize when their vocabulary size is small, due to the sufficient majority of intransitive verbs that show the the causative alternation. As their vocabulary size increases, the generalization is no longer supported, causing children to retreat from their initial hypothesis.

Overgeneralization: We conducted a corpus study on data from two children in CHILDES who produce causative errors: Adam (Brown corpus) and Ross (MacWhinney corpus). We examined each unaccusative verb used by the children up to the point where the first causative error was produced, resulting in 49 verbs for Adam (until age 3;2) and 24 verbs for Ross (until age 3;0). Each verb was then marked for whether they were produced in a causative frame in CHILDES by any parent in order to estimate their feasible occurrence in the input data. Following Yang’s (2016) Sufficiency Principle, a rule is generalizable within a class of N members iff there are N – N/ln(N) positive members following the rule. Out of 49, 37 of the unaccusative verbs used by Adam occurred in a causative frame; the Sufficiency Principle needs at least 36. For Ross, 18 verbs occurred as causative when the Sufficiency Principle requires only 16. Thus, for these children, the causativization rule is productive, which predicts their causative errors.

Retreat from Overgeneralization: We also find that the causative alternation rule ceases to be productive over time with increase in vocabulary size. To estimate the learner’s growing vocabulary past the early stages of acquisition seen in CHILDES, we examined each unaccusative verb in Levin (1993) that occurred with a frequency of at least 1 per million in a corpus of adult speech (CELEX): a total of 261 verbs. Out of the 261 verbs, 65 only occur as intransitives, i.e., only 196 of the verbs have a causative form. According to the Sufficiency Principle, a class with 261 members requires at least N – N/ln(N) = 214 verbs to follow the rule. 196 is insufficient evidence for the learner to generalize, and therefore, they cease to apply the rule to verbs for which they have not encountered the causative form.

Conclusion: This analysis supports a view in which children overgeneralize the causative alternation, as well as stop making causative errors, using positive evidence in the linguistic input.

References

Ambridge, Ben, Julian M Pine, Caroline F Rowland, and Chris R Young (2008). “The effect of verb semantic class and verb frequency (entrenchment) on children’s and adults’ graded judgements of argument-structure overgeneralization errors”. In: Cognition 106.1, pp. 87– 129.

Bowerman, Melissa and William Croft (2008). “The acquisition of the English causative alternation”. In: Crosslinguistic perspectives on argument structure: Implications for learnability, pp. 279–307.

Levin, Beth. (1993). English verb classes and alternations: A preliminary investigation.

University of Chicago press.

Pinker, Steven. (1989). Learnability and cognition: The acquisition of argument structure. MIT Press.

Yang, Charles. (2016). The price of linguistic productivity. MIT Press.