Abstract
Previous work (Wang, Cavanagh, & Green, 1994; Malinowski & Hübner, 2001; Shen & Reingold, 2001) suggests that search is easier when the distractors are familiar. Letters of the alphabet are all highly familiar, but in any given language certain letters occur more frequently than others. From the point of view of visual search, are all letters equally familiar, or does performance vary with the frequency of occurrence of letters in standard text? Observers (first language English, no known reading deficits) searched for any mirrored capital letter (possible letters were E, N, R, S, D, B, K, Q, J, and Z). In the two conditions, the heterogeneous distractors were ordinary capital letters that either commonly occur in English (E, T, N, R, S, and D) or occur less frequently (B, X, K, Q, J, and Z). Interleaved staircases measured the threshold display time required to perform a 2IFC search task at 90% correct. Observers required significantly less time to perform the search task when the distractors were common letters than when they were less common. Less common distractors required as much as twice the search time as more common distractors. This was true in spite of the fact that a low-level pattern-discrimination model applied to these sets of letters predicted equal or worse performance with the common distractors. It makes sense for the visual system to have access to letter frequency information, as it could be useful for reading. On the other hand, these results suggest that the “familiarity” effect is a bit of a misnomer, as by a lay definition there is little difference between the familiarity of more and less frequent letters. Possible explanations for this phenomenon abound and will be discussed, including novelty as a feature, improved grouping for more familiar items, reduced internal noise for more common items, and an information theoretic explanation in which more common letters are encoded with fewer bits, as in Huffman coding.