Fig. 2: Stimuli and experimental design.

a Pre/post training orientation and color discrimination tasks (top and bottom rows, respectively). In separate runs, a 500 ms fixation period was followed by a 200 ms presentation of a grating or colored disc, respectively. Response targets were presented on the left/right or top/bottom of the fixation cross for orientation and color task, respectively. Subjects responded by directing their gaze to the target. In all experimental groups, these tasks were conducted before training to determine individual difficulty levels, and repeated after training course to establish learning effects on individual features. b Colors were uniformly sampled from the red-purple spectrum of the LAB color space (black dots signify color levels). c Subjects had to determine whether a grating was tilted clockwise or anti-clockwise with reference to the diagonal (45°) in the orientation discrimination tasks. d Six individual difficulty levels for color and orientation discrimination, respectively, were determined from the psychometric curves of the two tasks. e In the main experiments, a 500 ms fixation period was followed by a 200 ms presentation of the stimuli (as in the pre/post measurements). The stimuli were chromatic gratings with individually determined orientation/color difficulty levels. In Experiments 1 and 2, subjects were trained on orientation-color conjunction discrimination. In Experiment 1, we provided global feedback on each trial, i.e., feedback informed about accuracy on the conjunction level (feedback was provided by changing the color of the chosen target; green signifies correct and red incorrect responses). In Experiment 2, we provided feature-specific feedback (e.g., orientation incorrect, color correct). In Experiment 3, subjects were exposed to the same stimuli as in Experiments 1 and 2, but were post-cued to respond only to one feature in a given trial to encourage separate feature learning. The location of the targets (top/bottom or left/right) indicated which feature to respond to. The order of features was pseudo-randomized. Feedback was provided for the given response. In Experiment 4, subjects were trained on a discriminating a single feature, orientation. Color varied from trial to trial but had to be ignored. Color difficulty levels were not individually determined but chosen to maximize between-level distances in the sampled LAB space to ensure high saliency in the color dimension.