diverted our subjects’ attention by asking them to report first a digit shown at
each side of the display and only then the colored letters. In about one trial in
three, the subjects reported the wrong combinations—perhaps a redX,agreen
O,orablueT.
The subjects made these conjunction errors much more often than they
reported a color or shape that was not present in the display, which suggests
that the errors reflect genuine exchanges of properties rather than simply mis-
perceptions of a single object. Many of these errors appear to be real illusions,
so convincing that subjects demand to see the display again to convince them-
selves that the errors were indeed mistakes.
We have looked for constraints on the occurrence of such illusory conjunc-
tions. For example, we have asked whether objects must be similar for their
Figure 16.1
Boundaries that ‘‘pop out’’ of a scene are likely to reveal the simple properties, or features, of the
visual world that are seized on by the initial stage of visual processing. For example, a boundary
betweenTsand tiltedTspops out, whereas a boundary betweenTsandLsdoes not (a). The impli-
cation is that line orientations are important features in early visual processing but that particular
arrangements of conjunctions of lines are not. A boundary betweenOsandVspops out (b). The
implication is that simple shape properties (such as line curvature) are important.
Features and Objects in Visual Processing 401