Fig 1.
AM induces decreased contrast detection performance.
(A) Illustration of part of the trial sequence in the AM condition of the 2AFC task. The figure only shows the fourth sequence of AM during which the target grating was briefly presented. Contrast of the target grating has been increased for illustration purposes. (B) Maximum contrast detection performance of a typical observer (AV) is considerably lower in the AM condition (red) than in the Flicker condition (blue) when the orientation of the target grating and inducers is identical. Full lines depict the best-fitting logistic psychometric function. (C) Maximal detection performance (1 − λ) is lower in the AM condition than in the Flicker condition for all observers. Symbols denote different observers. Error bars represent the 95% confidence interval.
Fig 2.
Orientation tuning of AM masking and model fits.
The pooled data of the five observers are shown for each orientation level. Red and blue symbols represent the AM and Flicker conditions, respectively. Dashed lines depict the best-fitting logistic psychometric function, while full lines represent the best-fitting contrast normalization model. Symbols denote different observers.
Fig 3.
Dependence of maximal performance on orientation of the target grating.
Full lines depict regression lines reflecting maximal performance for orientation differences between target and inducers in the 0°–45° range. Error bars denote the 95% confidence interval. When orientation difference increased from 0° to 45°, maximal performance (1 − λ) increased significantly in the AM condition (red), while performance remained constant in the Flicker condition (blue). Maximal performance was not affected when increasing the orientation difference from 45° to 90° in either the AM or Flicker condition.
Fig 4.
Schematic overview of possible AM-induced effects in the contrast normalization model.
Following the standard contrast normalization model, the target grating in our task is encoded by a population of V1-like neurons tuned to orientation, which are subject to response acceleration and divisive inhibition. This standard model is extended by including the effects of AM, which can modify the encoding of gratings in our model in three ways. (1) AM can excite linear receptive fields sensitive to the orientation of the inducers controlled by α, evoking responses as if the inducer was physically present at the target location (as during real motion). This would correspond to a “filling-in” process along the AM path. (2) AM can induce divisive normalization via β by exciting neurons in the gain control pool tuned to the orientation of the inducers. A similar divisive normalization signal would be observed when the inducer would be positioned at the target location. As such, AM-induced inhibition is also in accordance with a “filling-in” account of AM. (3) AM can scale down the contrast response functions due to the suppressive effect exerted by neurons tuned to the inducers’ orientation via γ.
Fig 5.
Comparison of contrast normalization models in capturing AM masking.
The pooled data of five observers are shown for the condition in which the orientation of target and inducers is identical. The predictions of the best-fitting contrast normalization model including AM-induced effects is indicated by the full red line (AM condition) and full blue line (Flicker condition). The dashed line shows the model prediction for the AM condition when excitation and divisive inhibition effects are removed after fitting (α = β = 0). The dash-dotted line represents the prediction when the suppression effect is set to zero after fitting (γ = 0). In contrast to the model without excitation and divisive inhibition, the model without suppression fails to account for the observed masking at high contrast levels. In addition, this model predicts facilitation of contrast detection at low contrast levels, which is not supported by the data. All three models predict the same performance for the Flicker condition, as all parameters controlling AM effects (α, β and γ) are set to zero for this condition.