Transfer between Tasks and Modalities

An abstract concept of number should not be tied to a sensory modality. Do animals appreciate that two sounds and two visual objects share "twoness"? Fernandes and Church (1982) found that rats trained to discriminate two vs. four sound bursts immediately transferred the discrimination to two vs. four light flashes. Even more impressively, rats trained to make one response after two light flashes or tones and a second response after four light flashes or tones made the four-response to a compound stimulus of two light flashes and two tones (Church and Meck, 1984). These results suggest that the rats summed over the two modalities. However, rats trained by Davis and Albert (1987) were unable to transfer an auditory numerical discrimination to the visual modality when a more complex task was used that required rats to identify the intermediate value 3 from the adjacent values, 2 and 4 (see also Pastore, 1961; Salman, 1943).

Another indication of an abstract numerical concept would be if animals could transfer a numerical discrimination learned in one task to another task (Seibt, 1982). For example, Meck (1997) trained one group of rats to behaviorally discriminate two vs. six light flashes and a second group of rats to discriminate two distinct wavelengths. Both groups were then tested in a 12-arm radial maze task where 4, 6, or 8 of the 12 arms were baited with food. Rats were required to obtain the N food items and then signal that they had completed foraging by entering a stop box. Amazingly, the rats that had been trained in the numerical discrimination task to discriminate 2 vs. 6 learned to retrieve six food items more readily than the rats that had been trained to discriminate two wavelengths. In addition, rats trained on the 2 vs. 6 discrimination performed better when they were required to retrieve 6 food items compared to 4 or 12 food items. In other words, rats trained to discriminate two vs. six light flashes exhibited savings when trained to retrieve six food items from a 12-arm maze. In contrast, when the chimpanzee Ai's ability to numerically label random dot arrays was tested, she showed no savings from her original training on numerically labeling arrays of household objects (Matsuzawa, 1985; Murofushi,

1987). More studies are needed to determine whether animals can transfer numerical discriminations learned in one task or modality to another.

0 0

Post a comment