Higher-Level Cognition - Computational Explorations in Cognitive Neuroscience - page 396

ventromedial prefrontal cortex in a railroad construc-
tion accident. After the accident, the previously mild-
mannered man began to engage in inappropriate behav-
iors and say rude and inappropriate things, often in out-
bursts of rage. One interpretation is that he could no
longer inhibit these inappropriate urges.
In the laboratory, inhibition has been studied pri-
marily using simple motor response tasks, for example
the anti-saccade task (Roberts, Hager, & Heron, 1994)
where subjects have to saccade (move their eyes) in the
opposite direction of a visual stimulus. Here, the pre-
potent response is to saccade toward the stimulus, so
this must be inhibited. Frontal lesions produce deficits
in this task (e.g., Guitton, Buchtel, & Douglas, 1985).
Inhibition also can be invoked to explain aspects of sev-
eral of the other attributions to frontal cortex, as de-
scribed in subsequent sections.
Although some researchers have proposed that
frontal areas, specifically ventromedial, are specialized
for inhibitory processing per se (e.g., Fuster, 1989;
Diamond & Goldman-Rakic, 1989), other researchers
have characterized the inhibitory functions of frontal
cortex as a consequence of activation-based process-
ing (Cohen et al., 1990; Kimberg & Farah, 1993; Mu-
nakata, 1998; O'Reilly et al., 1999a). For example,
as we saw in the Stroop model, top-down support for
color naming can inhibit the word-reading process as
a consequence of direct competition among these pro-
cesses. Our framework is more consistent with this
activation-based characterization. With widespread in-
hibitory competition throughout our networks (and pre-
sumably the cortex), any differential support for one set
of representations or processes will automatically result
in inhibition of other competing ones.
In many ways, the inhibition versus competition dis-
tinction is similar to the case of the disengage mecha-
nism for attention discussed in chapter 8. Recall that
Posner and colleagues had proposed that to switch at-
tention to a new location, one first needed to disen-
gage (i.e., inhibit) the previously attended location. We
saw that instead of requiring a specialized disengaging
mechanism, competition from the engaging of attention
elsewhere could produce the necessary disengagement.
This competition model seems to be more consistent
with the attentional data from patients and normals, and

arises quite naturally within the modeling framework
developed here.
Further evidence against the notion of a specialized
inhibitory system comes from the nature of the connec-
tions from prefrontal cortex to other areas. Long-range
intracortical connections (e.g., from the prefrontal cor-
tex to other posterior cortical areas) are all excitatory
rather than inhibitory. These excitatory connections
also synapse on inhibitory interneurons, but they do not
do so exclusively. Further, the inhibitory interneurons
have very diffuse patterns of connectivity that would not
facilitate the precise inhibition of only selected types of
information.

Flexibility

The contribution of frontal cortex to flexibility is re-
vealed by the perseverations of frontal patients. Recall
that perseveration is the tendency to persist in making
a previously valid response even when task conditions
have changed. Probably the most well-known paradigm
where perseveration is observed is the Wisconsin card
sorting task as we explored previously, but it is also
found across a range of other paradigms, including the
A-not-B task as discussed in chapter 9.
One standard interpretation of these perseverative ef-
fects is that they reflect a failure of inhibition, where
the frontal patient fails to properly inhibit the previ-
ous response or categorization rule (e.g., Diamond &
Goldman-Rakic, 1989; Dias et al., 1997). However, as
we saw in the dynamic categorization exploration and
the A-not-B model in chapter 9, perseveration can be
explained in terms of a weak active memory system that
cannot overcome the effects of prior learning.
In short, perseveration in frontal patients may reveal
the flexibility benefits in normals of an activation-based
solution over a weight-based one. When the activation-
based solution is eliminated or impaired (via a frontal
lesion), the system resorts to a weight-based solution,
which is slower to react to changes (i.e., exhibits per-
severation) and is generally less flexible. The apparent
need to inhibit prior responses to avoid perseveration
falls naturally out of the competitive model described
earlier — one does not need to hypothesize a specific
inhibitory system.