Proceedings of the Annual Meeting of the Cognitive Science Society

The debate about how attention is allocated during readinghas been framed in as: Either attention is allocated in a strictlyserial manner, to support the identification of one word at atime, or it is allocated as a gradient, to support the concurrentprocessing of multiple words. The first part of this article re-views reading models to examine the feasibility of both posi-tions. Although word-identification and sentence-processingmodels assume that words are identified serially to incremen-tally build larger units of representation, discourse-processingmodel allow several propositions to be co-active in workingmemory. The remainder of this article then describes an in-stance-based model of word identification, Über-Reader, andsimulations comparing the identification of single words andword pairs. These simulations indicate that, although wordpairs can be identified, accurate identification is restricted toshort high-frequency words due to the computational de-mands of both memory retrieval and limited visual acuity.