Abstract
Is it possible to recognize two words at once? This question relates to the more general investigation of divided visual attention, and to a fierce debate about whether multiple words are processed serially or in parallel during natural reading. We recently found evidence of a serial bottleneck in a semantic categorization task (White, Palmer & Boynton, Psychological Science, in press). In that study, observers viewed masked pairs of words and either focused their attention to judge one word (single-task condition) or attempted to divide their attention to judge both words independently (dual-task condition). Accuracy was so much worse in the dual-task condition that it supported an all-or-none serial model: observer could distinguish the semantic category of only one of the words and had to guess about the other. In this presentation, we delve further into the data and present new experiments to investigate whether parallel processing is possible in some circumstances. Specifically, we first examine the effects of word length and lexical frequency on semantic categorization. Is parallel processing possible for two short (3-4 letter) words, or for two high-frequency words? The answer is no: accuracy in the single-task condition was affected by length and frequency, but dual-task accuracy was always so low that it supported the serial model. We also present a new lexical decision experiment that was similar except the task was to distinguish real words from pseudowords. If the serial bottleneck lies in high-level semantic processing, then this task should show less of a dual-task deficit and evidence of parallel processing. Alternatively, if the bottleneck applies to any kind of lexical processing, then this task should also show a severe dual-task deficit. Preliminary evidence indicates that even in the lexical decision task, a serial bottleneck prevents more than one word from being processed at a time.
Meeting abstract presented at VSS 2018