ABSTRACT
Several theories assume that words are parsed into sublexical structures such as syllables, morphemes, or BOSSes as part of the recognition process. Empirical evidence for each of these units has been inconsistent; moreover, the notion that such units function as “access” codes is problematic in light of the properties of English orthography. An alternative view is that the effects of such units derive from orthographic redundancy. The present studies used feature integration errors to examine the perceptual groupings of letters in visual word recognition. Experiment 1 showed that syllables rather than BOSSes influenced feature integration errors. Experiment 2 showed that such errors occur when syllables are marked by low-frequency bigrams. Experiment 3 showed that orthographically similar pairs such as NAIVE and WAIVE act alike with respect to feature integration errors. The results suggest that recovering structures such as syllables or BOSSes is not a necessary stage in processing. To the extent that such units emerge, it is because they consist of spelling patterns that are salient in terms of orthographic redundancy. The results are discussed in terms of connectionist models in which there are no parsing mechanisms or access units.
