Abstract
A long-standing, though controversial, hypothesis is that colour and form information are segregated early in visual processing. Many colour-preferring cells in primate V1 are insensitive to orientation (Lennie et al., 1990), and there can be a loss of tilt induction at isoluminance (Livingstone and Hubel, 1984). However, others have found maximum tilt aftereffects for gratings modulated in colour along the principal axes of colour space (Flanagan et al., 1990) and a recent report showed that many orientation-selective cells in primate V1 are responsive to both colour and luminance modulation (Johnson et al., 2001). This prompted us to re-examine interactions between colour and luminance in human orientation perception using the tilt illusion.
We measured the effect of a surround grating oriented at 15° to the vertical upon the perceived orientation of a central test grating. We varied the axis of colour space along which the two (1cyc/°) gratings were modulated; subjects were tested on the three cardinal axes and six non-cardinal axes. At low contrasts (5–20 × detection threshold) we found that some subjects experienced large tilt illusions for subjectively isoluminant stimuli while others experienced none. At higher contrasts (30–40 × detection threshold) all subjects showed tilt illusions of approximately equal magnitudes when test and surround were modulated along a single chromatic axis, regardless of that axis. When the modulation axis of the surround was varied but that of the test was fixed (with contrast equated at 30× or 40× detection threshold), significant tilt illusions were also consistently observed, though the maximum tilt illusion always occurred when test and surround were modulated along the same axis.
These results demonstrate the existence of orientation-selective mechanisms in the human visual system tuned to non-cardinal directions of colour space, and suggest that the processing of colour and orientation are intimately coupled (Lennie, 1998).