Abstract

The present study investigated whether semantic processing during sentence reading is automatic or mediated by top-down variables (e.g., task instructions), as it is the case for lexical processing. We showed correct and semantically anomalous sentences to university students while recording their eye movements. In a within-subject design, participants were either given no instructions or were required to read for understanding. Target words could either be low- or high-frequency words. The eye tracking metrics relative to an early processing stage showed an increase in fixation duration for low frequency words and for semantically anomalous words, regardless of task instructions. Conversely, the eye tracking metrics relative to successive processing stages revealed a larger number of fixations and longer fixation durations on semantically anomalous words but only in the with-instruction condition. The results indicate that semantic processing, differently from lexical processing, is an automatic process in the very early stages and it is modulated by top-down factors only in later processing stages.