Perception and Attention - Computational Explorations in Cognitive Neuroscience - page 253

Valid Trial Invalid Trial

Spatial Object

Cue + +

Target + +

* *

V1
(features x
location)

Figure 8.21: The Posner spatial attention task. The cue is
a brightening or highlighting of one of the boxes that focuses
attention to that region of space. Reaction times to detect the
target are faster when this cue is valid (the target appears in
that same region) than when it is invalid.

Figure 8.22: Representations of space and object could both
interact via a low-level spatially mapped feature array like V1.
Spatial attention would result from top-down activation by the
spatial pathway of corresponding regions in V1, which would
then be the focus of processing by the object pathway.

cific case of the interactions between spatial and object
representations. Specifically, more parallel processing
takes place at lower levels within the system, and higher
levels enforce greater focus on single objects or loca-
tions in space.

trial). Although the target is typically a rather sim-
ple shape, we will assume that its detection occurs via
the object processing pathway, or at least the relatively
early stages thereof.
The Posner task provides a simple way of exploring
the interactions between the spatial and object process-
ing pathways. Specifically, activation in a specific loca-
tion in the spatial processing pathway should facilitate
the processing of objects (targets) that appear within
that part of space, and impede processing of objects in
other parts of space. In addition to this basic attentional
modulation effect, the model must also be capable of
shifting attention to new locations in space as new cues
or objects come into view. For example, in the Posner
task, one must be able to switch attention to the loca-
tion where the target appears, even though doing so may
take longer on invalidly cued trials. In general, there is a
tradeoff between focusing attention in one location and
switching it to a new one — by focusing attention on
one location and impeding processing elsewhere, this
makes it less likely that the system will process and shift
attention to other locations. However, attention must be
dynamic and movable if it is to be useful, so a reason-
able balance must be achieved.
The model could be constructed such that both spa-
tial and object representations interact via top-down ef-
fects on a V1-like spatial feature map, that provides
the inputs for both subsequent levels of processing (fig-
ure 8.22). In such a model, the activation of a region in

8.5.1 Basic Properties of the Model

In the first attentional model, we begin to explore the
ways that spatial representations can interact with the
kinds of object-based processing developed in the pre-
vious model. Thus, we are simulating some of the ways
in which the dorsal “where” processing stream can in-
teract with the ventral “what” processing stream. As
discussed previously, parietal cortex contains many dif-
ferent types of spatial representations, and lesions in
this lobe cause deficits in spatial processing. Although
there are probably many different types of spatial rep-
resentations in the parietal cortex (e.g., having different
reference frames), we will not make many assumptions
about the nature and other functions of these parietal
representations — we will just use a simple maplike
spatial representation.
One of the simplest and most influential paradigms
for studying spatial attention is the Posner task (Pos-
ner et al., 1984), illustrated in figure 8.21. When atten-
tion is drawn or cued to one region of space (e.g., by
highlighting a box on one side of the display), this af-
fects the speed of target detection in different locations.
When attention is drawn to the same region where the
target subsequently appears (i.e., a validly cued trial),
subjects are faster to detect the target than when atten-
tion is drawn to the opposite region (an invalidly cued