Asymmetric detection of relative speed during motion-based figure-ground segregation

Abstract

Visual motion can support figure-ground segregation when a local region moves at a different speed from its background, but whether faster- and slower-than-background targets are detected equally well, and how any asymmetry depends on background speed and motion reliability remains incompletely understood. We measured relative-speed detection thresholds in four human observers using an eight-alternative forced-choice localization task. On each trial, one of eight random-dot patches at 10-degree eccentricity moved either faster or slower than a surrounding moving background. Relative-speed detection thresholds were measured with adaptive staircases at five background speeds (2.5-30 deg/s) under 100% and 60% motion coherence. In every observer, slower targets required larger relative-speed differences for detection than faster targets. This increment-decrement asymmetry was strongest at low background speeds and diminished as speed increased. Reducing motion coherence elevated thresholds overall but disproportionately impaired detection of slower targets, amplifying the asymmetry, most strongly at the lowest measurable speeds. Under 100% coherence, increment detection showed approximately Weber-like scaling from 10 to 30 deg/s, whereas under reduced coherence both increment and decrement Weber fractions decreased with increasing background speed. Relative-speed detection therefore depends on the sign of the speed difference, background speed, and motion reliability. The faster-target advantage is consistent with faster-speed weighting reported in cortical area MT and with natural-scene statistics showing that figural regions tend to move faster than their backgrounds, suggesting that relative-speed sensitivity may contribute to motion-based figure-ground segregation. More broadly, such increment-decrement asymmetries can reveal biases in how sensory signals are represented and read out.