In these experiments we tested whether physical differences between letters and digits could account for the category effect in visual detection and partial report. In our task, observers decided whether a target item matched any item in a briefly presented array. This yielded a visual detection task when the target preceded the array and a type of partial-report task when the target followed the array. In Experiment 1, the stimulus set consisted of nine digits and nine letters modified to match the digits on the basis of visual similarity. Partial-report performance was better in a mixed-category condition than in a single-category condition, but no such effect occurred in visual detection. However, the similarity of items between categories may have biased observers against using category information to perform the task. When the similarity was controlled both within and between categories in Experiment 2, a category effect emerged in visual detection as well. A third experiment, using a same-different reaction time task, verified that the stimuli were equally similar within and between categories. The results indicate that physical differences alone cannot explain the category effect.It has been found in a number of tasks that the visual processing of an alphanumeric display is more efficient when some of the items come from an irrelevant category. For instance, it takes less time to find a digit in an array of letters than in an array of other digits (e.g., Egeth, Jonides, & Wall, 1972), and detection of a digit in a brief display is more accurate when the distractors are letters rather than digits (e.g., Schneider & Shiffrin, 1977). This finding can be referred to as a category effect. The category effect is important in theories of visual information processing because it has suggested to some that, at some level, characters are processed in parallel with unlimited capacity (e.g., Duncan, 1980;Gardner, 1973). However, this conclusion has not been universally accepted, and the mechanism underlying the category effect remains controversial (Deutsch, 1977;Kahneman & Treisman, 1984;Krueger, 1984). In this study, we tested an often proposed counterexplanation of the category effect that can be termed the partial analysis account.The unlimited parallel processing view of the category effect assumes that the items in a visual display are identified and categorized in parallel without requiring effort or attention on the part of the observer. However, such subsequent operations as transferring items to short-term memory or comparing items may limit performance. In this view, the category effect occurs because information about the category of items in the visual display can be used to separate relevant from irrelevant items. For instance, if the observers in a visual detection task know This research was supported by Natural Sciences and Engineering Research Council of Canada Grant A8263 to the first author. We would like to thank Vince Di Lollo for comments on an earlier draft. Requests for reprints should be sen...