Abstract
Meshing ideas from the reading literature with our work on crowding, we present a three-process model of reading to account for reading rate in word/min. We use crowding to distinguish holistic vs by-parts recognition. In crowding there is a critical spacing beyond which neighbors no longer interfere. If the object is recognized holistically, then it can be identified even when the whole object lies within a critical spacing, without isolating any part (Martelli, Majaj, and Pelli, 2004, Journal of Vision, in press). If recognition is by parts then object identification will be possible only when each part is isolated from the rest of the object by the critical spacing. Words are recognized in three ways: by parts, holistically, and through context. Stimulus degradations — shuffling, capitalizing, and scrambling — each knock out one of the three processes. We call the processes L, W, and S, which stand for Letter, Word, and Sentence. We characterize reading as identifying words, one after another. All three processes — L, W, and S — identify words. They differ in what part of the stimulus they use to identify the word. L does it on the basis of word content (i.e. the letters), W does it on the basis of word shape (i.e. holistic), and S does it on the basis of word context (i.e. the rest of the sentence). We have measured the contributions of these three recognition processes to RSVP reading as a function of location in the visual field. Preliminary results find that the W process contributes 50 word/min everywhere, the S process contributes 100 word/min everywhere, and the L process contributes 210 word/min centrally, but only 5 word/min in the periphery. These numbers make sense. The letter-decoding process (L) makes the biggest contribution, as expected from the learning-to-read literature, and is available only in central vision, as expected from crowding. Peripheral reading is much slower because only W and S contribute.
Supported by grant EY04432 to Denis Pelli