Fig. 4
From: A flexible readout mechanism of human sensory representations
Readout linking model. The linking model simulates the cortical response evoked by each dot patch according to an existing framework48, which parameterized the contrast response function (orange curves) as a Naka–Rushton and the coherence response function (blue curves) as linear or a saturating exponential function. The model weights the cortical responses from each visual area (β values) evoked by the stimulus (Right or Left) according to the current task. The model then takes the difference between the signals evoked by each stimulus, plus a bias term (βbias) to account for any individual observer’s bias to choose one response over the other. To convert from this weighted signal to probability of choosing the patch on the right, the signal is passed through a cumulative normal distribution (curve on right). The linking model is analogous to probit regression with nonlinear input signals
