Internal noise and sampling efficiency for motion detection, discrimination and summation
Helle Kristine Falkenberg, W. Simpson, Velitchko Manahilov · Journal of Vision · 2010
In an ideal observer framework, human performance is limited by suboptimal sampling efficiency and by additive internal noise. We developed ideal observer models for the detection of drifting gratings, for direction discrimination of oppositely drifting gratings, and for detection of the sum of two oppositely drifting gratings. Human contrast energy thresholds for performing these tasks were measured as a function of the external Gaussian dynamic noise level. We find that the three tasks share a common source of internal noise. The sampling efficiency for discriminating motion direction is much lower than that for the other tasks. One possible explanation of the low efficiency for direction discrimination is that the internal representations of the two drifting gratings are overly similar. An ideal observer cross-correlates the received stimulus with representations of each of the two signals (drifting in opposite directions). If the human observer misrepresents the signals as being highly similar, discrimination performance will be quite inefficient (as we find).