When attention is divided between spatially distinct objects, the time
to detect a target decreases when two or more targets are present. Th
is redundancy gain can be accounted for by an interactive race model (
Mordkoff & Yantis, 1991) in which separate decisions are made about ea
ch object, but environmental contingencies among the objects can influ
ence decision times. In the present study, we examined whether the mod
el also accounts for performance when attention must be divided betwee
n stimulus attributes other than spatial location. Subjects made targe
t-present responses when displays included a prespecified color, a pre
specified letter, or both target features. The data violated the predi
ctions of all separate-activations models, including the interactive r
ace model. Two control experiments ruled out an alternative account ba
sed on task complexity. We conclude that coactivation occurs when targ
et attributes from two separable dimensions are simultaneously present
, but not when target attributes come from the same dimension. A modul
ar hybrid of race and coactivation models is tentatively proposed.