Accounting for regressive eye-movements in models of sentence processing: a reappraisal of the selective reanalysis hypothesis

When people read temporarily ambiguous sentences, there is often an increased prevalence of regressive eye-movements launched from the word that resolves the ambiguity. Traditionally,such regressions have been interpreted at least in part as reflecting readers’ efforts to re-read and reconfigure earlier material, as exemplified by the Selective Reanalysis hypothesis [Frazier, L., & Rayner, K. (1982). Making and correcting errors during sentencecomprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14, 178–210]. Within such frameworks it is assumed that the selection of saccadic landing-sites is linguistically supervised. As an alternative to this proposal, we considerthe possibility (dubbed the Time Out hypothesis) that regression control is partly decoupled from linguistic operations and that landing-sites are instead selected on the basis of low-level spatial properties such as their proximity to the point from which the regressive saccade was launched. Two eye-tracking experiments were conducted to compare the explanatory potential of these two accounts. Experiment 1 manipulated the formatting of linguistically identical sentences and showed, contrary to purely linguistic supervision, that the landing site of the first regression from a critical word was reliably influenced by the physical layout of the text. Experiment 2 used a fixed physical format but manipulated the position in the display at which reanalysis-relevant material was located. Here theresults showed a highly reliable linguistic influence on the overall distribution of regression landing sites (though with few effects being apparent on the very first regression). These results are interpreted as reflecting mutually exclusive forms of regression control with fixationsequences being influenced both by spatially constrained, partially decoupled supervision systems as well as by some kind of linguistic guidance. The findings are discussedin relation to existing computational models of eye-movements in reading.