Perception and Attention - Computational Explorations in Cognitive Neuroscience - page 255

Effects of Spatial Pathway Lesions parietal lesions, which are associated with Balint's syn-

drome (Husain & Stein, 1988). Under the disengage
model, one would expect that bilateral parietal dam-
age would produce bilateral deficits in disengaging, and
greater amounts of slowing in the invalidly cued trials
in both directions. However, Balint's syndrome patients
instead exhibit smaller amounts of slowing on invalidly
cued trials, and in general show deficits in deploying
attention. This is consistent with what happens in the
Cohen et al. (1994) model, namely that there is a bilat-
eral reduction in top-down spatial activation, and thus
an overall reduction in the size of attentional effects.
Our model provides a more robust (and somewhat
richer) instantiation of the basic points made by Co-
hen et al. (1994), in part because our model is based
on the architecture shown in figure 8.23, which has lat-
eral interactions between the spatial and object process-
ing pathways as discussed previously. Thus, whereas
Cohen et al. (1994) had to use very carefully chosen
parameter values to achieve reasonable performance,
our model behaves reasonably over a wider parameter
range. Overall, this modeling exercise provides an ex-
cellent example of how the use of biologically based
principles for understanding cognition can give rise to
very different functional descriptions of cognitive phe-
nomena than traditional box-and-arrow models.

The effects of lesions in the parietal cortex on per-
formance in the Posner spatial cuing task provide an
important source of constraint on the model. As dis-
cussed previously, unilateral lesions (i.e., to only one
hemisphere) of the parietal cortex generally lead to
the phenomenon of hemispatial neglect — the inabil-
ity to focus attention to the corresponding half of vi-
sual space (i.e., the left half of visual space for a right
hemisphere lesion). Unilateral parietal damage can re-
sult in varying degrees of behavioral deficit. Patients
with relatively mild impairments exhibit disproportion-
ately slower attention switching on invalidly cued trials
when the target appeared in the damaged half of space.
More severely impaired neglect patients can be com-
pletely unable to switch attention into the damaged half
of space. Although this pattern of data is generally con-
sistent with the idea that parietal damage results in dif-
ficulty focusing attention on the damaged side, Pos-
ner et al. (1984) have argued based on a box-and-arrow
model that the parietal cortex implements a disengage
mechanism (figure 8.24). They conclude that the ob-
served behavior comes from the inability to disengage
attention from the good side of space.
Using a model of spatial attention based on the archi-
tecture shown in figure 8.22, Cohen et al. (1994) were
able to account for the observed parietal lesion data on
the Posner spatial cuing task. This model has no specific
“disengage” mechanism. Instead, the process of disen-
gaging attention from one location occurs as a natural
consequence of engaging attention elsewhere. Thus, in
these terms, the parietal lesion data can be explained as
the inability of the damaged spatial representations to
provide sufficient top-down support to the correspond-
ing half of visual space. Thus, when the target appears
in the damaged half of space, patients are impaired in
engaging attention there. This explanation is also gener-
ally consistent with the hemispatial neglect syndrome in
patients with more severe damage, where patients have
difficulty focusing attention in the damaged half of vi-
sual space.
An interesting source of empirical support for the
Cohen et al. (1994) model over the Posner et al. (1984)
disengage model comes from the effects of bilateral

Object-Based Attention

Our model includes top-down projections from the ob-
ject pathway to V1, and lateral projections to the spatial
system from the object pathway, both of which allow
for the possibility of object-based attentional effects.
There is considerable debate in the literature regard-
ing the exact nature of object-based attention — is it
essentially just attention to the spatial region (and fea-
tures) where the object is located (e.g., Mozer, Zemel,
Behrmann, & Williams, 1992; Vecera & Farah, 1994),
or is it specifically attention in the object pathway (e.g.,
Duncan, 1984)? Our model allows for both, with poten-
tially complex interactions between the two processing
pathways that could give rise to all sorts of interesting
effects, some of which are explored later.