zum Inhalt springen

Prediction error and structural priming in L2 processing

Duygu F. Şafak & Holger Hopp

Technische Universität Braunschweig, Germany

This study explores the extent to which adult L2 learners recruit prediction error for learning in two priming experiments using visual world eye-tracking. To this end, we investigate structural priming in the L2 predictive processing of the dative alternation among adult L1-German–L2-English learners

Experiment 1 tests whether prediction error drives adaptation when ditransitive verbs biased to either the double-object (DO) or the prepositional-object (PO) structure occur in the other structure. Participants first read aloud prime sentences crossing Verb Bias (DO-bias/PO-bias) and Structure Type (DO/PO). Subsequently, they heard DO and PO target sentences – with non-biased verbs – while viewing visual scenes with corresponding referents (Figure 1). 

Results revealed PO-priming effects: L2 learners had more looks to the theme referent after hearing the target sentence verb following PO (vs. DO) prime sentences. PO-priming was modulated by surprisal effects, i.e., there were more looks to the theme referent when the prime structure mismatched prime verb bias (PO prime sentences with DO-bias verbs). No priming obtained for DO sentences.

In Experiment 2, we ask if prediction-error priming is constrained by grammatical restrictions on the dative alternation (Figure 2). Unlike in Experiment 1, PO target sentences contained (non-alternating) verbs that were grammatically restricted to PO. As in Experiment 1, PO-priming and surprisal effects were found in Experiment 2. Critically, the observed effects were smaller and more short-lived, suggesting that L2 learners apply verb-class restrictions to reign in priming.

These findings point to an error-driven prediction mechanism in L2 priming (Exp.1), as in natives [1], and further suggest that L2 learners immediately apply grammatical constraints to gate prediction-based priming (Exp.2).

Figure 1: An example of prime-target pairs in Experiment 1
Figure 1: An example of prime-target pairs in Experiment 2
  • [1] Chen et al. (2022). Error-based structure prediction in language comprehension: Evidence from verb bias effects in a visual-world structural priming paradigm for Mandarin Chinese. JoExpPsy:LMC, 48, 60-71.