While the low-lever processes mediating the detection of primary visual att
ributes are well understood, much less is known about the way in which thes
e attributes are assigned to objects in the visual world. For example, when
a region of the retinal image contains multiple motion signals at a range
of spatial scales, how do we know whether these signals come from a single
object or multiple objects? Here, we present data from four neurological pa
tients on a psychophysical task requiring them to report whether the two co
mponents of a plaid pattern appear to move coherently or transparently. The
spatial frequency of one component of the plaid is held constant while tha
t of the other is manipulated. While some of the patients perceive coherent
motion over a much smaller range of spatial frequencies than normal contro
ls, others report coherence over almost the entire range tested. We discuss
the implications of these findings for computational theories of motion pe
rception and higher-level visual processing. (C) 1999 Published by Elsevier
Science B.V. All rights reserved.