Citation

Barr DJ, Gann TM, Pierce RS. Acta Psychol. 2011; 137(2): 201-207.

Affiliation

Institute of Neuroscience and Psychology, University of Glasgow, 58 Hillhead St., Glasgow G12 8QB, United Kingdom.

Copyright

(Copyright © 2011, Elsevier Publishing)

DOI

10.1016/j.actpsy.2010.09.011

PMID

20961518

Abstract

Researchers often conduct visual world studies to investigate how listeners integrate linguistic information with prior context. Such studies are likely to generate anticipatory baseline effects (ABEs), differences in listeners' expectations about what a speaker might mention that exist before a critical speech stimulus is presented. ABEs show that listeners have attended to and accessed prior contextual information in time to influence the processing of the critical speech stimulus. However, further evidence is required to show that the information actually did influence subsequent processing. ABEs can compromise the validity of inferences about information integration if they are not appropriately controlled. We discuss four solutions: statistical estimation, experimental control, elimination of "on-target" trials, and neutral gaze. An experiment compares the performance of these solutions, and suggests that the elimination of on-target trials introduces bias in the direction of ABEs, due to the statistical phenomenon of regression toward the mean. We conclude that statistical estimation, possibly coupled with experimental control, offers the most valid and least biased solution.

Language: en