In studies using Averbach and Coriell's (1961) partial-report bar-probe task, the subject is requested to report the name of the indicated letter in a briefly presented visual multi-letter display. In the error analysis for this task, Townsend (1973) distinguished errors that involved report of an item displayed in another than the indicated position (location errors) and errors that involved report of an item that was not displayed (item errors). Furthermore, a location error was taken as an indication for inaccurate target localization and an item error for inaccurate target identification. Unfortunately, bar-probe studies that applied this error analysis have yielded contradictory results as to the most frequent type of errors that was observed. Consequently, opposed interpretations were formulated as to the limiting factor (either localization or identification) in human information processing capabilities for briefly presented material. These interpretations have in turn led to opposing views on the locus of selection of incoming information (later or early).In the present study, it is shown that (a) the shape of the exposed letter arrays (linear or circular arrangements) and (b) the letter identities and sampling procedure that are used to compose the arrays (the whole alphabet and random sampling or a limited set of letters and sampling while taking interletter confusabilities into account) can explain the empirical incongruency. These two variables appear to affect (a) the accuracy of localization and identification processes and (b) the degree to which localization difficulties are observable as location errors (and not as item errors) and identification difficulties as item errors (and not as location errors). In contrast with earlier claims, it is concluded that performance in partial-report bar-probe tasks is constrained by two factors: localization and identification.