Abstract
Primates use both shape and color of objects to evaluate their environment. However, the neural mechanisms governing how shape and color cooperate and compete to determine the utility of objects is not well understood. We previously showed that monkeys learn shapes faster than colors, much like humans during development. After training on a collection of 14 colored shapes, monkeys showed that they could correctly select the color of an achromatic shape, and choose the proper shape from a simple color cue, indicating integration of color and shape. Although color and shape are both used to identify an object, are they used equally? We again tested two monkeys on matching colored shapes. The cue was one of the 14 colored shapes they had learned. After the cue disappeared, the choices were shown. On 80% of the trials, one of the choices matched the cue. On 20% of the trials, one of the choices had the same shape as the cue but a different color, and the other had the same color as the cue but a different shape. The monkey was rewarded randomly at a 50% rate regardless of choice. The monkeys usually chose the matching shape (75.0%, p < 0.001), suggesting that shape is more diagnostic for determining stimulus identity. We duplicated this test on Amazon Mechanical Turk for >100 human subjects. Humans were presented with only six different stimuli to accelerate the learning of shape/color associations. Although the task was slightly different to accommodate differences between subjects and platforms, the probe trials were virtually identical. Humans also typically chose shape over color as the matching feature (78.9%, p<0.001). We conclude that shape is more diagnostic for object recognition than is color in primates, and suggest that the integration of color and shape into object identity is not equal.