Abstract
Goal. While feature-based attention (FBA) enhances perception to an attended spatial frequency (SF), the precise profile of such attentional modulation remains unclear. The feature-similarity gain model predicts a monotonic modulation profile. However, the surround suppression model suggests a non-monotonic (Mexican hat) profile. Here, we investigated how the attentional modulation systematically varies as a function of the difference between the stimulus SF and the attended SF. Methods. We employed filtered Gaussian noise stimuli as target and unfiltered noise as mask in a two-interval forced choice paradigm. One interval contained the target of a particular SF and the other interval contained a scrambled version of the target as noise (masks were presented in both intervals to reduce visibility). Participants reported the interval that contained the target in two conditions. In a cueing condition, a fixed SF precue was presented before the stimuli, indicating the target SF in 50% of trials (valid). In the remaining trials (invalid), the target could be ±0.5, ±1, ±1.5, ±2 octaves away from the cued SF. In the neutral condition, no precue was presented which served as the baseline to assess the attentional modulation. Results. There was an enhancement for valid condition relative to the neutral condition. Importantly, for the invalid conditions, we found a suppression at +1 octave followed by a rebound at +1.5 octaves. However, the lower frequencies (i.e., −2 to 0 octaves) showed a monotonic profile. Thus, FBA elicited a surround suppression modulation for higher SF and a feature-similarity modulation for lower SF. This asymmetry in attentional modulation might be related to a wider neuronal tuning bandwidth toward higher frequencies on a linear scale. We performed model simulations to explore the potential interactions between neuronal tuning and FBA’s modulation profile.