Abstract
Reading is a complex task that is integral to the daily lives of humans in the modern era. When reading, the visual system is confronted with many words simultaneously. How much of that visual information can a reader process at once? Previous fMRI and behavioral data have demonstrated that lower-level visual features, such as color, can be processed in parallel for multiple stimuli simultaneously. However, higher-order lexical and semantic processing has been shown to occur serially, one word at a time. This implies that an internal bottleneck lies somewhere between early visual and lexical analysis. We used a behavioral paradigm to investigate whether the bottleneck lies before or after the stage of letter recognition. On each trial, participants viewed two strings of 5 letters that were flashed briefly, one above and one below fixation, and then masked. Each letter string was composed of either all consonants or four consonants and one vowel. The task was to indicate whether a vowel was present in a particular letter string. We compared accuracy in a divided attention condition, in which participants had to judge both strings independently, with a focused attention condition, in which participants judged just one string. Accuracy in the divided attention condition was so much worse than in the focused attention condition that it supported an all-or-none serial-processing model. That means that participants were only able to judge the presence or absence of a vowel in one letter string per trial. Furthermore, in the divided attention condition, we found a stimulus processing tradeoff that is characteristic of serial processing: if participants judged one letter string correctly, they were less likely to judge the other string correctly. Future studies investigating other aspects of word recognition, such as phonological decoding, are needed to fully characterize capacity limits in word recognition.