Figure 2. Integration and non-integration models produce similar psychometric functions and psychophysical kernels in a fixed stimulus duration task (stimulus duration = 1 second).

(A) Proportion of positive choices as a function of stimulus strength for each model simulation (N = 30,000 trials per simulation). The inset displays the SNR parameter ($\kappa$) for each of the three models. All three models are capable of producing sigmoidal psychometric functions of similar slope but require different SNR parameters. (B) Psychophysical kernels produced from the choice-data in A. Each simulated trial contained a small, 100 ms long stimulus pulse that occurred at a pseudorandom time during the trial. Kernels are calculated by computing the pulse’s effect on choices (coefficients of a logistic regression, Equation 16) as a function of pulse time. Shaded region represents the standard error. Like integration, non-integration models are capable of producing equal temporal weighting throughout the stimulus presentation epoch.

Figure 2—figure supplement 1. Extension of the exercise in Figure 2 to a low sensitivity regime (A) and a high sensitivity regime (B).

N = 30,000 trials per simulation.