Abstract
Image features (e.g. line terminators) play an important role in the computation of 2D motion. It has been suggested that such features are processed by specialised detectors (end-stopped cells). However, the motion of line terminators may also be processed by a combination of first- and second-order motion information (Loffler & Orbach, 1999). According to this model, first-order signals are biased in the direction orthogonal to the line orientation, whereas second-order signals are biased in the direction of the line orientation. Are these biases evident in the perception of moving lines?
Observers were presented with an array of tilted line segments (each measuring 0.5 by 2.2 degrees, with orientation ±20 degrees to the vertical and 30% – 90% contrast) moving along one of several linear trajectories (±60 degrees to the horizontal). Lines were defined by either first- or second-order information (i.e. luminance or contrast modulated dynamic 2D noise). Observers' task was to state whether the trajectory of the lines was above or below horizontal. Perceived horizontal motion, the trajectory where ‘above’ and ‘below’ responses are equally probable, was inferred from psychometric functions.
For first-order lines, perceived horizontal motion was biased in the direction orthogonal to the line orientation (average bias = −31.6 degrees). For second-order lines the bias was in the opposite direction, towards the line orientation (average bias = 27.2 degrees), as per model predictions. Our results suggest a novel role for second-order processing: it may be crucial for the coding of veridical two-dimensional feature motion.
This research is supported by EPSRC Grant No. GR/S59239/01