Research
- M. T. Banich
II.
Brain regions involved in attentional selection as assessed by fMRI
My work has systematically used variants of the Stroop task as
a means for identifying the role played by different components of the
network controlling attention in the human brain.
In the Stroop task, individuals are asked to name the ink color in
which a word is printed while ignoring the meaning of the word.
Compared to a neutral baseline word which has no intrinsic meaning
with regard to color (e.g., “blank”), responses are facilitated if the
word is the same as the ink color (e.g., the word “red” in red ink) and
responses are slowed if the word names a different ink color (e.g., the word
“blue” in red ink). This
phenomenon is thought to result because word reading is relatively automatic
compared to naming the ink color, and hence it is difficult to direct
attention to the ink color over the word (see Cohen, et al., 1990 for a
connectionist model of this phenomenon).
Previous neuroimaging research using this task has implicated a
variety of structures including frontal regions, posterior regions, and the
anterior cingulate (e.g., Bench et al., 1992; George et al, 1994, Pardo et
al., 1990).
What has not been clear from all these studies, however, is the exact
aspect of attentional control that is being performed by each of these
structures. To explicate this
issue, our approach has been to employ variants of the Stroop task that allow
for a systematic examination of different aspects of attentional control and
their neutral underpinnings. To examine the nature of the network that allows for
attentional selection of an attribute, we have varied the nature of the
attended dimension. To
understand the degree to which to-be-ignored attributes are processed, we
have varied the nature of the unattended dimension.
And finally, to understand the selection of a response set, we have
manipulated the nature of response requirements.
For example, we have compared the pattern of brain activation for a
color Stroop task, in which color is the attended dimension and the word’s
meaning is the unattended dimension, to a spatial Stroop task in which the
spatial relation between a word and a box is the attended dimension and the
word’s meaning is the unattended dimension (Banich et al., 2000 JCN).
Thus, these two tasks have different attended dimensions, but share
the same unattended dimension, namely the meaning of the word.
In each of the two tasks, we manipulated the degree to which
attentional selection was required for task completion by varying whether
information in the unattended dimension directly conflicted with that in the
attended dimension (i.e., the word “blue” colored in yellow) (known as
incongruent trials) or was unrelated (e.g., the word “blank”) (known as
neutral trials). The comparison
between incongruent and neutral trials indicated activation of similar brain
regions for both tasks: anterior cingulate, prefrontal, extrastriate and
parietal cortices. Although the areas activated by each task overlapped,
there were also some areas that were uniquely activated by each of the two
tasks. Thus, although the same
general network is activated when a dimension must be attended, there are
small variations in exactly how this network gets activated, which are
important as they represent specific tuning depending on the nature of the
attended attribute.
In another study we compared two Stroop tasks in which the
attended dimension was the same but the unattended dimension was different.
Here we found clear evidence of a dissociation in processing between
the frontal and posterior attentional systems (Posner and Dehaene, 1994).
Both tasks activated very similar frontal regions, which we
interpreted as reflecting the role of frontal regions in setting an
attentional set for selecting or attending to a particular item attribute.
In contrast, we found that activation in posterior regions depended on
task demands, which coupled with our prior results suggests that both the
attended and unattended dimensions are processed in posterior areas, but
within different brain regions depending on the nature of the attribute.
To further examine the hypothesis that frontal regions create
an attentional set or expectancy that is independent of response
requirements, we performed another study in which participants needed only to
attend to a single dimension (e.g., the color of a word) by monitoring for an
oddball (e.g., a purple word) (Banich et al., 2000, CBR).
Unlike our prior studies in which selection between a variety of
responses was required (e.g., press one button for words printed in red,
another for words printed in blue, and yet another for words printed in
green), no such response selection was required here.
As predicted, this task also activated the same frontal regions as
seen when selection between a variety of responses was required, suggesting
that frontal activation is not response related but reflects an attentional
set or expectancy.
We have also examined the role played by the anterior cingulate in
attentional processes. Some
researchers argue that this brain structure mediates between “conflict”
in general (Carter et al., 1998) whereas others suggest that it is mainly
involved in response processes (Paus et al., 1993).
Our work suggests that the latter interpretation is more correct for
three reasons. First, in the
study described above in which individuals did not need to select between a
variety of responses, we did not obtain cingulate activity.
Furthermore, when we compare activation of congruent trials in the
color-word Stroop task (e.g., “red” printed in red ink) to neutral trials
(e.g. “hope” printed in red ink), the cingulate was not active.
Finally, we also examined the degree to which the cingulate is
involved in response selection by comparing activation for neutral trials to
two types of incongruent trials: those in which the word’s meaning is
incongruent with the correct response and which also leads to activation of
an incorrect motor response (e.g., “red” printed in green ink, when red,
green and blue are valid responses), and those in which the word’s meaning
is incongruent with the correct response but which is neutral with regard to
possible responses (e.g., “purple” printed in green ink).
These latter trials, which are response neutral, did not produce
cingulate activity (Milham et al, submitted).
This research has furthered our understanding of the neutral
network controlling attention in the following ways:
1)
We have provided evidence that the frontal attentional system acts to set
expectancies about the salient or to-be-attended dimension of an attribute
when discrimination between attributes is required.
2)
We have shown that “unattended” attributes, like attended ones, are
processed, and that such processing occurs in posterior regions of the brain.
The specific posterior regions activated depend on the nature of the
attribute (e.g., color vs. form).
3)
We have shown that activity in the anterior cingulate cortex is mainly
involved in attentional processes that serve to mediate response conflict as
compared to conflict at other levels of processing (e.g., semantic,
perceptual).
III.
The modulation of attentional functioning by interaction between brain
regions