Abstract
Distinct brain circuits detect luminance-defined motion (LDM) and contrast-defined motion (CDM). Patterns of transfer of perceptual learning are diagnostic of the amount of overlap between these circuits.
Method: The stimuli were mixtures of dynamic noise and moving texture carriers (speed 10°/sec, duration 400 msec). The textures were patches of static noise filtered with a bandpass filter centered on 1 cycle/deg. CDM stimuli were multiplicative mixtures that contained no net directional Fourier energy (Chubb & Sperling, 1989). LDM stimuli were additive mixtures, titrated to approximate the initial CDM difficulty in a pilot study. Each participant was pre-tested and post-tested on LDM and CDM in a direction discrimination task at two orthogonal directions (-45 vs. -65; 25 vs. 45 deg). Twenty-two participants trained on either LDM or CDM in a constant direction for four sessions, counterbalancing between subjects. To allow for estimation of the psychometric slope, training contained 10 and 6 degree discriminations.
Results: The average d' improved from 1.4 to 2.2 for the LDM group and from 1.2 to 1.8 for the CDM group on the trained direction of motion. There was full transfer to the orthogonal direction in both groups. Critically, there was full transfer from CDM to LDM, and partial transfer (65%) from LDM to CDM.
Conclusion: This substantial amount of transfer suggests substantial overlap between LDM and CDM circuits. Zanker (1999) studied perceptual learning of the global motion of a figure defined by local dot motion. He found very little transfer from primary (local collinear with global) motion to secondary (local orthogonal to global) motion. Our results suggest that the circuits for local LDM and CDM overlap more than the circuits for global primary and secondary motion.