A bottlenosed dolphin (Tursiops truncatus) with good underwater and aerial visual acuity was tested in visual matching-to-sample (MTS) paradigms. Attempts to train visual identity MTS directly, using two simple two-dimensional patterns as sample stimuli and as alternatives (comparison stimuli), met with little success, in keeping with previously observed difficulties of this auditory-specialized species for learning complex tasks utilizing simple visual materials. Pairing of each visual sample with a unique sound, to produce a compound auditory-visual sample, while retaining the two visual alternatives, resulted in the dolphin's learning both auditory-visual symbolic matching and visual-visual identity matching. At O-sec delay, performance with the auditory element of the sample alone was equivalent (76%) to performance with the visual element alone; performance with the compound was distinctly better (95%-98% correct). Testing with longer delays using the visual element alone resulted in successful matching through to a maximum delay of 34 sec. These results provided the first demonstration of delayed MTS in a dolphin using visual materials, and complemented other data showing the ready capability of this species for delayed auditory MTS. It appeared that the dolphin solved the visual MTS task by forming auditory codes to represent the visual materials, and that these auditory codes were eventually replaced with purely visual codes.Delayed-matching-to-sample (DMTS) tasks have been used extensively in the analysis of animal memory (Roitblat, 1982). In the typical form of these tasks, a to-beremembered stimulus (the sample) is presented briefly. The animal then indicates its memory for the sample by choosing a matching stimulus from among subsequently offered alternatives. In identity matching, the correct alternative physically matches the sample. In symbolic matching, the alternatives are learned associates of the samples (e.g., D 'Amato & Worsham, 1974;Herman & Thompson, 1982;Roitblat, 1980). For example, a red disk, shownas a sample, may be symbolically represented among the alternatives by an inverted triangle.With rare exceptions, the stimuliused in matchingtests have been limitedto the single,dominantsensorymodality for the species studied. The reason is straightforward: it has proven extremely difficult to teach any animal a DMTS task when task information is given other than through its presumed major sensory interface with its environment. For the pigeon and nonhuman primate, this has been vision. If both sample and alternativesare visual stimuli (visual-visual, or V-V, matching), DMTS is relatively easy to teach these animals (e.g.,
