According to the parafoveal-processing hypothesis, letters are more often m
issed in function words than in content words because the former are more l
ikely to be identified in the parafovea, where letter processing is not ava
ilable. Contrary to previous demonstrations, more omissions occurred in fun
ction words than in content words when parafoveal processing was not availa
ble because words were displayed in column format, text was read through a
5-letter window, or words were presented I at a time on a computer screen.
In all experiments, impeding parafoveal processing decreased omission rates
for function but not for content words. In the last experiment, direct mon
itoring of eye movements revealed that, for both fixated and skipped words,
letters in function words are missed more often than content words. These
results are best interpreted within a model including the structural preced
ence hypothesis and stressing the importance of visual factors.