Prediction is facilitated by engaging the production system

Lelonkiewicz, J. , Pickering, M. & Rabagliati, H.

University of Edinburgh

Does the production system play an important role in how we predict linguistic content (Adank et al., 2010; Pickering & Garrod, 2007)? We tested this idea by varying how the production system was engaged whilst reading, and measured the effects on prediction.
Participants read sentence contexts word-by-word (e.g., the boat passed under the...). Contexts were either high or low cloze, and were length-matched. To manipulate the production system, we asked participants (within-subjects) to read context sentences silently or aloud.
In Experiment1, context sentences were followed by either existing words (bridge) or nonwords (bronge), and participants were asked to only read aloud existing words, staying silent for nonwords (50% of trials). Participants initiated speaking faster in high than low cloze sentences and, importantly, this prediction effect was stronger when they had read the sentence context aloud (604 vs 725ms) than silently (868 vs 918ms; B=-76.44, t=-2.56).
Experiment2 confirmed this finding. Participants named a picture that appeared after the context. Again, predictability had a greater effect on naming latency after reading the context aloud (aloud 323 vs 503ms; silently 494 vs 597ms; B=-69.19, t=-3.04).
We propose a model outlining how the production system affects prediction during sentence processing.