[PS-1.43] Is structural priming due to prediction error?

van Gompel, R. P.

University of Dundee

Theories by Chang et al. (2006) and Jaeger and Snider (2013) claim that structural priming is due to error-based learning. During comprehension of a prime, language users predict upcoming structure; if their prediction is incorrect, they are less likely to subsequently use the incorrectly predicted structure.

We tested error-based theories using PO/DO primes (1a-d) followed by a target (2). The first clause in the primes resulted in either a new-given or given-new order in the PO/DO. According to error-based accounts, comprehenders should frequently make prediction errors in primes (1a, 1c) because the structure does not have the preferred given-new order, whereas they should predict the given-new structures in (1b, 1d) correctly. Therefore, priming should be stronger in (1a-c).

After the politician was questioned, the activist handed (1a) it to the politician/(1b) him the petition.
After the petition was signed, the activist handed (1c) him the petition/(1d) it to the politician.

(2) The tourist handed ?

We observed clear priming, but it was no stronger after new-given than given-new primes. The same finding was obtained when the pronouns were replaced by definite noun phrases. These findings are more consistent with residual activation (e.g., Pickering & Branigan, 1998) than error-based learning theories.