Abstract
Aim: To test whether binocular rivalry suppression entails low-level, non-selective eye suppression, or higher-level feature-selective suppression. Methods: Subjects (n=4) viewed 8°×5° rival stimuli, either faces (one tinted red, the other green, equal luminance & RMS contrast) or global motions (expansion vs. contraction, 150 dots, 100% coherence, 6.5°/s). During dominance (and suppression), thresholds were measured for two kinds of brief (100 ms) probe, cross-faded faces (identity discrimination) and rotary motions (direction discrimination). The ratio of dominance to suppression thresholds quantifies rivalry suppression depth (1=no suppression, 0=max suppression). Rivalry coherence was also measured (proportion of 5 min period in which one or other image was exclusively visible). Results: For rivalry between two faces, suppression was deep for face probes (.27) while global motion probes were not suppressed at all (∼1). The converse was true: rivalling global motions deeply suppressed global motion probes (.30), but not face probes (∼1). For mixed rivalry pairs (global motion/face), suppression was shallow for both global motion probes (.64) and face probes (.74). Rivalry coherence measures showed face pairs (.85) of global motion pairs (.83) rivalled coherently, with little piecemeal rivalry. Coherence for mixed rivalry pairs was lower (.53). Conclusions: (i) Rivalry between stimuli represented in extrastriate areas (e.g., global motions; faces) is deep. (ii) This deep suppression is limited to the processing stream or area representing the rival stimuli. (iii) Sensitivity to probes not represented in the rivalling area is unaltered, during dominance and suppression. (iv) Shallow suppression observed for mixed global motion/face stimuli is consistent with the Highest Common Denominator hypothesis (Alais & Melcher 2005). (v) Rivalry suppression does not affect an entire eye- it is selective for stimulus attributes and leaves others unaffected.