Abstract
The ability to read depends on a region in the ventral temporal cortex known as the “visual word form area” (VWFA). Although it responds most strongly to written words, its selectivity is not absolute, and it exhibits top-down modulations related to task demands. Here, we used fMRI to investigate how stimulus features and task demands interact to drive activity in the VWFA. Participants viewed three types of character strings: familiar English words, unfamiliar pseudowords, and unfamiliar false fonts with visual properties matched to the words. We presented these stimuli in randomized sequences while participants performed three different tasks: discriminating the lexicality of the character string (word vs. nonword), discriminating the color of the character string, and discriminating the color of the fixation mark. If attending to any character string is sufficient to activate the VWFA, its responses should be elevated in the color task compared to the fixation task, for all stimulus types. If voluntary linguistic processing is necessary and sufficient to engage the VWFA, then for all stimulus types its activity should be highest in the lexical task. Contrary to both of these predictions, we found a strong interaction between stimulus type and task: compared to responses in the fixation color task, the VWFA’s response to text was elevated in the lexical task, but not in the character string color task. Moreover, its response to false fonts was not higher in the lexical task than in the fixation task. Therefore, activation of the VWFA is not merely dependent on the presence of words nor on engagement in a lexical task, but on the conjunction of both. These results highlight how the exquisite specialization of ventral temporal cortex involves top-down feedback that is distinct from general forms of attention that operate in early visual cortex.