Algorithmic behaviour and spatial memory are used by two intertidal fish species to solve the radial maze.

We used an eight-arm radial maze to assess the relative contributions of learned patterns of movement (algorithmic behaviour) and spatial memory to the foraging efficiency of two sympatric rocky-shore fish, fifteen-spined stickleback, Spinachia spinachia, and corkwing wrasse, Crenilabrus melops, exploiting nonrenewable food sources. To forage efficiently, subjects had to avoid arms already depleted within a trial. In the absence of spatial cues, sticklebacks and wrasse improved their foraging efficiency by developing the algorithm of visiting every third arm. In the presence of spatial cues (coloured tiles) algorithmic behaviour was largely subsumed by the use of spatial memory. Imposition of a delay within trials reset the behavioural algorithm, so depressing foraging efficiency in the absence of cues, but not in their presence when memory could be used to guide behaviour. Memory retention for previous choices (working memory) lay within the range 0.5-5.0 min, consistent with the characteristic timescale expected for habitats where prey distribution changes rapidly during the tidal cycle. We considered two hypotheses on the type of information memorized: the cue list hypothesis and the spatial configuration hypothesis. The cue list hypothesis predicts that neither random repositioning nor fixed rotation of spatial cues should impair foraging efficiency, whereas the spatial configuration hypothesis predicts that efficiency should be impaired by random repositioning of cues but not by rotation. Data supported the spatial configuration hypothesis. Copyright 1999 The Association for the Study of Animal Behaviour.