How does sentence context affect word frequency effects? A within-task investigation

Single-word reading studies form the backbone of reading research. In such studies, participants respond to single words and computational models simulate the obtained patterns of results. A more ecologically valid paradigm involves tracking participants' eye movements as they silently read sentences. In the current study, we examine whether the strongest marker effect in the literature on reading, the word frequency effect, differs between single-word reading-aloud studies and eye-movement data, and if so, why. In the first, pre-registered experiment, we collected single-word reading-aloud data from two conditions: (1) reading aloud of sentences, where each word is presented one at a time, and (2) reading aloud of words, presented one at a time, without sentence context and in random order. The materials were taken from the Potsdam Sentence Corpus, which allowed for a comparison with eye-tracking data (single fixation durations). In reading-aloud data, we find stronger effects of frequency, length, and predictability, suggesting that single-word studies may overestimate the importance of the underlying processes. In a second experiment, we further explore whether these differences are due to the lack of a preview of an upcoming word when participants see the whole sentence. In a reading-aloud task, multiple words were presented simultaneously: either sentences, or words in random order. Here, we obtained mixed evidence. Thus, single-word reading-aloud shows weaker effects than eye-movement data, which may be partly driven by the processing of upcoming words.