Abstract
Background: How much information can the human visual system process at once? This critical question is often studied by having observers make independent judgments of two simultaneous stimuli. Capacity limits impair accuracy in such a "dual-task" condition compared to a "single-task" condition, in which only one stimulus must be judged. The most extreme capacity limit manifests as serial processing: the observer can process only one stimulus and therefore guesses about the other. This has been rarely observed. Here we investigate such capacity limits for two measures of word recognition. Methods: In the first experiment, words appeared in rapid serial visual presentation (RSVP) simultaneously to the left and right of fixation. In separate blocks, observers either detected semantic targets (nouns belonging to a category such as "animals") or color targets (words colored slightly red). For both types of judgments we compared a dual-task condition (observers judge words on both sides) with a single-task condition (observers judge only one side and ignore the other). The second experiment was similar but used a masking paradigm with only one word presented at each location per trial. Results: In both experiments, the difference between dual- and single-task accuracy for semantic categorization was so large that it supported a serial processing model. This serial model was also supported by a negative correlation between accuracy on the two sides: judgments of one side were less accurate when the other side was judged correctly. In contrast, performance in the color detection task with identical stimuli was consistent with an unlimited-capacity, parallel processing model. Thus, the severe capacity limits are due to the requirements of the semantic task and not the stimulus properties alone. In discussion, we will contrast our new results with conflicting prior results on serial processing in word recognition.
Meeting abstract presented at VSS 2017