Effects of prediction error on episodic memory retrieval: evidence from sentence reading and word recognition

ABSTRACT

Prediction facilitates word processing in the moment, but the longer-term consequences of prediction remain unclear. We investigated whether prediction error during language encoding enhances memory for words later on. German-speaking participants read sentences in which the gender marking of the pre-nominal article was consistent or inconsistent with the predictable noun. During subsequent word recognition, we probed participants’ recognition memory for predictable and unpredictable nouns. Our results indicate that individuals who demonstrated early prediction error during sentence reading, showed enhanced recognition memory for nouns overall. Results from an exploratory step-wise regression showed that prenominal prediction error and general reading speed were the best proxies for recognition memory. Hence, prediction error may facilitate recognition by furnishing memory traces built during initial reading of the sentences. Results are discussed in the light of hypotheses positing that predictable words show a memory disadvantage because they are processed less thoroughly.

KEYWORDS:

• Prediction error
• sentence comprehension
• reading
• memory
• recognition

Acknowledgments

The authors wish to thank all students and RAs who helped with stimulus construction and data acquisition for this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 We included trial number in order to account for effects of fatigue or general customization with the experiment. Word length was included because our items were not controlled for along this variable (see “Materials” section).

2 Based on visual inspection of the data, trials with RTs longer than 6000 ms were removed, a procedure that retained 98% of all data points, a justifiable rate according to Ratcliff (1993).

3 Of note, we replicated all findings reported below when using inverse RTs and raw, untransformed RTs, as dependent variable.

4 A reviewer asked whether the condition by Pr score interaction on the third spill-over word might be caused by non-linearities in the PR score variable. We checked for this by means of residual plots, where Pearson's residuals are expected to be approximately uniform in the y direction if the model is correctly specified. The resulting plots did not raise concerns about a non-linear relationship for this variable.

5 A reviewer additionally suggested we use subject-wise accuracy averages from the self-paced reading comprehension questions as a predictor for attention. These analyses showed the same overall effects, in that the pre-nominal cost score for the third spill-over word (b = 0.25, SE = 0.12, t = 2.04, p = .05), but not the noun cost score (b = 0.2, SE = 0.12, t = 1.68, p = .1), significantly predicted recognition rates among subjects. Interestingly, comprehension accuracy did not significantly predict recognition accuracy (b = 0.1, SE = 0.12, t = 0.91, p = .4).

6 We would like to thank an anonymous reviewer, for suggesting this to us.

Additional information

Funding

This research was funded by a Saarland University start-up grant awarded to KIH and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation, Project 232722074 – SFB 1102 Information Density and Linguistic Encoding).