Looking back on reading ahead: No evidence for lexical parafoveal-on-foveal effects

- Models of reading make different predictions about word processing in the parafovea.
- Four experiments (N = 244) show no evidence for lexical parafoveal-on-foveal effects.
- A Bayesian meta-analysis confirms the null effects predicted by serial models.
- These findings support E-Z Reader and cast doubt on correlational corpus methods.

Current models of eye movement control during reading make different predictions regarding the possibility of parafoveal-on-foveal effects - i.e. whether the lexical properties of upcoming, parafoveal words can affect reading time. To date, there have been contradictory findings from correlational corpus analyses and carefully controlled experimental studies regarding the existence of these effects. To address this controversy, we conducted four experimental studies (total N = 244) investigating the effects of parafoveal word frequency during natural reading. These experiments showed no evidence for parafoveal-on-foveal effects in an environment that should have been highly conducive to parallel lexical processing. In addition, a Bayes Factor meta-analysis of the prior experimental literature also provided clear support for the null hypothesis. These findings confirm the predictions of serial attention models such as E-Z Reader, and call into question the findings of previous correlational corpus studies.