Higher-Level Cognition - Computational Explorations in Cognitive Neuroscience - page 398

support a limited number of balls, there are constraints
as to which balls can be moved at any given point to
achieve the goal state. To achieve the goal state in a
minimal number of moves, the sequence of moves must
be planned to take into account the constraints. Frontal
patients are significantly impaired on this task, requir-
ing many more moves than controls to achieve the goal
state, and generally making moves in a rather aimless
fashion (Shallice, 1982). Further, neurologically intact
individuals show activation of the frontal pole as they
solve this task (Baker, Rogers, Owen, Frith, Dolan,
Frackowiak, & Robbins, 1996).
As should be clear from our activation-based process-
ing framework, we view activation-based memory and
controlled processing as two sides of the same coin
(e.g., O'Reilly et al., 1999a; Cohen et al., 1997), but this
view is certainly not universally accepted (e.g., Badde-
ley, 1986; Grafman, 1989; Petrides, 1996). Although an
abstract problem solving task like the Tower of London
may require different sets of goallike control represen-
tations as compared to a more stimulus-based catego-
rization task, the basic mechanisms of rapid updating
and robust maintenance of information are clearly im-
portant for formulating executing a plan. Planning in
particular requires that activations be updated not based
on current stimuli, but rather on internally activated rep-
resentations of future situations. This requires that the
representations involved be protected from interference
from ongoing perceptual processing. Also, once a plan
has been activated, goal states must be maintained and
updated as processing proceeds.

Braver, and Cohen (1998) showed that this area could
alternatively be interpreted as representing the level of
response conflict (i.e., the extent to which two different
responses are strongly competing with each other, as
opposed to there being one dominant response), which
is correlated with the probability of making an error.
Processes like error monitoring are clearly important
for controlling goal-driven kinds of behaviors. Indeed,
we showed in chapter 9 that the adaptive critic mecha-
nism from the temporal-differences learning procedure,
which computes differences between expected and ac-
tual reward, could regulate the updating and mainte-
nance of active memories in a simple working mem-
ory task. We think this kind of monitoring process
is an essential aspect of the activation-based process-
ing in frontal cortex. In addition to the anterior cin-
gulate, neighboring regions of ventromedial prefrontal
cortex are likely to be important for this function (e.g.,
Bechara et al., 1996).

Summary

To summarize, the frontal cortex appears to be impor-
tant for a range of different behavioral manifestations
of what we have characterized as activation-based pro-
cessing. Beyond the basic function of actively main-
taining information, this type of processing can pro-
vide flexibility and fluency, and serves to inhibit task-
irrelevant information via intrinsic competition while
reducing perseveration by being more rapidly updatable
than weight-based systems. Executive-like controlled
processing can emerge out of the effects of this system,
combined with an effective monitoring and evaluation
system and other nonfrontal areas, as elaborated later.
Although we think that the common mechanism of
activation-based processing underlies the full spectrum
of frontal functions, this has not yet been conclusively
demonstrated, and must be regarded as a speculative hy-
pothesis — one that, as we have pointed out, is not nec-
essarily widely held.
Finally, it is important to note that despite the wide
variety of effects of frontal damage, frontal patients of-
ten have completely normal IQ scores. Indeed, as em-
phasized by Shallice (1988), the frontal cortex was
regarded as largely unimportant for several decades

Monitoring/Evaluation

For controlled processing to be effective, the current
state of processing in the cognitive system must be mon-
itored and evaluated. There is growing evidence that
areas of the frontal cortex are important for these mon-
itoring and evaluation tasks. One area that has received
considerable attention is the cingulate, specifically the
anterior portion (see figure 11.2). Gehring, Goss,
Coles, Meyer, and Donchin (1993) observed that the
anterior cingulate appeared to be the source of an error-
related activation signal that arises when people detect
that they have made an erroneous response. Carter,