Honeybees were tested in delayed conditional discrimination procedures (matching-to-sample and nonmatching-to-sample), using color stimuli presented on a video monitor. A small but reliable tendency to choose the color presented as the conditional cue was found, regardless of whether the contingencies reinforced or discouraged this tendency, The perseverative tendency occurred even with a delay of up to 1-2 min between the conditional cue and the choice. The tendency cannot be explained by changes in the associative value of the colors. Explanation of the results requires some form of working memory for color.Many vertebrate species have been shown to exhibit the flexible, dynamic form ofshort-term memory that is often termed working memory. Working memory is typically thought of as holding discreet items of information, usually for relatively short periods of time (see, e.g., Honig, 1978). Two particularly well-studied examples are the memory for form or color stimuli exhibited by pigeons in the matching-to-sample paradigm (e.g., Roberts & Grant, 1974) and the memory for spatial locations exhibited by rats in the radial-arm maze (Olton & Samuelson, 1976). Brown and Demas (1994;Demas & Brown, 1995) recently presented evidence for spatial working memory in honeybees. Their procedure was designed to be an analogue ofthe rat radial-arm maze procedure. They reported a small but reliable tendency for bees to avoid revisits to locations recently depleted of food, just as rats display a very robust tendency to avoid revisits to locations recently depleted of food. Brown, Moore, Brown, and Langheld (1997) replicated and extended these results under a variety of experimental conditions. In contrast, Burmeister, Couvillon, and Bitterman (1995) failed to find evidence that spatial choices were controlled by the identity of previous spatial choices. More recently, Isnec, Couvillon, and Bitterman (1997) reported four experiments in which bees' spatial choices were controlled by the identity of previous choices. However, rather than the "winshift" (alternation) tendency found in our laboratory, this control was expressed as the opposite "win-stay" (perseverative) tendency in Isnec et ai.'s subjects. Demas and Brown (1995) reported that bees are predisposed to alternate among reinforced spatial locations but could learn to perseverate under the appropriate reinforcement contingencies. In contrast, Isnec et al.s subjects exhibited perseveration regardless of the contingencies. Although the explanation for the discrepancies in the manner in which previous visits affect subsequent choices is unclear, there is agreement that the results of these experiments require a memory system that allows discrimination oflocations previously visited within a trial from those not yet visited (Bitterman, 1996; Brown et aI., 1997; Isnec et aI., 1997).The existence of working memory in honeybees is of interest because of the vast differences between the structure of the honeybee's nervous system and those of the vertebrates in which working mem...