Fig. 3. Predictions of the two models for choices in the two-alternative condition.

The population model (left panels) assumes that decision-making circuits have access to the activity levels associated with each of the four colors (four gray bars), whereas the summary model (right panels) assumes that decision-making circuits only have access to the highest activity level (single gray bar). In all examples, the dominant circle is white, and subjects are given a choice between white and green. a When the highest activity happens to be at the dominant color, both models predict that the subject would correctly choose the dominant color. b When the highest activity happens to be at the alternative color, both models predict that the subject would incorrectly choose the alternative color. c The predictions of the two models diverge when the highest activity is associated with a color other than the two presented alternatives. In such cases, the activation for the dominant color is likely to be higher than for the alternative color, so according to the population model, subjects would ignore the color with the highest activity (blue color in the example here) and correctly pick the dominant color in the majority of the trials. However, according to the summary model, subjects have no information about the activation levels for the dominant and the alternative colors and would thus correctly pick the dominant color on only 50% of such trials.