Are letter identity and letter order encoding functionally separable? Evidence from masked priming with visual similarity and letter transpositions

Skilled word reading requires the rapid encoding of both letter identities and letter order. Whereas many models of visual word recognition compute letter order over abstract, case-invariant letter identities, other accounts allow uncertainty about letter identity and letter position to remain jointly constrained during early orthographic processing. We pit these competing claims against each other in a series of masked priming experiments. As a manipulation check for the use of uppercase primes and lowercase targets, Experiment 1 (substitution priming) confirmed a visual-similarity priming effect: Priming was larger for visually similar than visually dissimilar substitution primes. The key tests were Experiments 2-3, which contrasted transposed-letter with replacement-letter control primes in lexical decision and same-different tasks, while varying the visual similarity of the transposed pair. Transposed-letter priming was sizable across tasks and, critically, its magnitude was comparable for visually similar and visually dissimilar transpositions. Taken together, our findings favor a functional dissociation between letter-identity and letter-order encoding. These data place quantitative constraints on accounts in which early visual-form uncertainty is expected to propagate directly into uncertainty about relative letter position in masked priming, and they are more naturally accommodated by models in which letter order is encoded over abstract orthographic representations.