Quantifying the Relationship Between Search Efficiency and Perceptual Similarity in Color Space Across Different Efficient Search Tasks

When looking for a specific object in the environment, the visual system guides attention toward objects in the scene that contain features that are similar to those of the object in mind, also known as the target template. However, the precise relation between search performance and perceptual similarity (between objects in the scene and the target template) has not been properly characterized. Recently, target-contrast signal theory proposed an explicit relationship linking search performance to the concept of top-down “target–distractor contrast,” with contrast being a measure of the amount of perceptual evidence that allows peripheral processing to differentiate target from distractors. We used a well-characterized color space to investigate the relationship between target–distractor similarity and search efficiency. We compared three different models relating color distance to search performance: the universal law of generalization, and two implementations of target-contrast signal theory. In the first, target–distractor distance indexes the target–distractor contrast, while the second uses the signal-to-noise ratio (SNR) between the neuronal responses to the attended color and the distractor color to index target–distractor contrast. When the target color is known but the distractor color cannot be anticipated, perceptual distance best predicts performance (Experiments 1, 2A, and 2B). When target and distractor colors repeat from trial to trial, the SNR measure best captures performance (Experiments 3A and 3B). Finally, when neither the target nor the distractor color is known to observers, performance deteriorates significantly and is no longer indexed by either of these two measures of target–distractor contrast (Experiment 4).