How cognition affects eye movements during reading

Based on recent data (Yang & McConkie, 2001), we argue against direct cognitive control of most saccades made during reading. Rather, cognitive influences appear to be more indirect. We propose that saccades are executed at random times, with a pattern over the fixation period that is described by a hazard curve that is low initially, rises quickly to an asymptote and then typically drops more slowly or remains fairly constant. Cognition influences saccade onset times (which produce the fixation durations) in three ways. First, processing difficulties can inhibit and delay saccades. Second, parameters can be adjusted in response to text or task conditions. Third, given enough time, cognition can control saccades more directly. The latter appears to occur quite infrequently. Regressions are also not necessarily cognitively controlled.