The use of local landmarks by foraging goldfish

Abstract

Goldfish, Carassius auratus, were introduced into an arena tank containing two symmetrically placed gravel-filled trays, one of which contained hidden food. In the absence of additional structures in the tank, the accuracy of food-patch choice (i.e. the proportion of individuals contacting the full tray before the empty tray) was close to that expected assuming random binary choice. Accuracy improved significantly, however, when landmarks (a rock/plant combination or multicoloured plastic columns) were positioned close to the food patch. Choice performance also improved when the plastic columns were placed in the opposite (non-food) half of the tank, indicating a capacity for indirect landmark usage as distinct from direct visual cueing. Latencies to feeding decreased with experience in all tratments, but were significantly lower in the presence of landmarks. Exploratory sampling behaviour, as indicated by the number of visits to the empty patch once food had been found during a trial, was significantly less frequent in the presence of strong landmarks. Fish in all treatments increased the frequency of sampling visits following a reversal of full and empty patches. Reliance on memory-based spatial behaviour may therefore vary according to the nature and complexity of the visual environment.