Event-related potentials were measured in a task that combined the classic selective attention paradigm with memory search and mental rotation paradigms. Subjects were required to attend to stimulus letters in one color and to ignore stimuli in a different color. Within the attended category subjects searched for target letters from a prememorized set (the memory set) and indicated whether they were presented normally or in mirror-image. Letters in all stimulus categories were presented randomly in either their upright position or rotated over 60 degrees, 120 degrees, or 180 degrees. The event-related potentials showed that the earliest effect of attending to color was a positivity at the anterior electrodes and a negativity at Oz (onset about 150 ms), followed by the enhancement of a central N2 component (N2b, onset about 220 ms). The early effect is thought to reflect selective processing of elementary stimulus features, the later N2b the covert orienting of attention. A later, prolonged central negativity elicited by attended stimuli covaried with the duration of the memory search process. Target detection resulted in a parietal P3b component, and also in an earlier negativity (in the range approximately 200-300 ms) to both attended and unattended targets, suggesting an early preattentive target classification. There were two effects of the rotation of stimuli. First there was an early occipital effect (about 200-300 ms), irrespective of attention and stimulus categories. It was argued that this effect reflected the preattentive identification of the orientation of the stimulus letters. A later (onset about 350-400 ms) parietal effect consisted of an increase in negativity as a function of the angle over which letters were rotated. This prolonged negativity had a later onset latency than the search-related negativity, and was restricted to the event-related potentials to target letters in the attended input channel. It was suggested that this component is the manifestation of a mental rotation process. It was argued that rotation-related negativity and memory search-related activity reflect operations in distinct working memory subsystems.