Higher-Level Cognition - Computational Explorations in Cognitive Neuroscience - page 371

Chapter 11

Higher-Level Cognition

intrinsically more difficult to model — one can imagine
combining several of the models from previous chapters
to get a sense of the complexity involved. For these and
other reasons (e.g., the lack of suitable animal models,
differences in learned strategies between individuals),
progress in understanding the neural basis of higher-
level cognition has been slower than for the more basic
mechanisms covered in previous chapters. Thus, much
of our treatment of this topic will be based on ideas
within one particular framework (Cohen et al., 1996;
O'Reilly et al., 1999a; Braver & Cohen, 2000) that has
been developed relatively recently. These ideas should
be regarded as relatively speculative and not necessar-
ily widespread. We present them to give some sense
of modeling at this level, and also in the hope that they
may provide a solid basis for future research in this area.
This overview section is a bit longer than in previous
chapters to allow us to sketch the overall framework.

Contents

11.1Overview ......................379
11.2BiologyoftheFrontalCortex ...........384
11.3ControlledProcessingandtheStroopTask....385
11.3.1 BasicPropertiesoftheModel........387
11.3.2 ExploringtheModel.............388
11.3.3 SummaryandDiscussion ..........391
11.4DynamicCategorization/SortingTasks......392
11.4.1 BasicPropertiesoftheModel........395
11.4.2 ExploringtheModel.............397
11.4.3 SummaryandDiscussion ..........402
11.5 General Role of Frontal Cortex in Higher-Level
Cognition ......................403
11.5.1 Functions Commonly Attributed to Frontal
Cortex....................403
11.5.2 Other Models and Theoretical Frameworks . 407
11.6 Interacting Specialized Systems and Cognitive
Control .......................408
11.7Summary ......................409
11.8FurtherReading ..................4 10

Framing the Challenge of Higher-Level Cognition

Although neural network models of higher-level cog-
nition are in their relative infancy, the more traditional
symbolic approach to this domain has a long history of
successful models (e.g., Newell & Simon, 1972; An-
derson, 1983). One of our objectives in this chapter is
to sketch the relationship between these symbolic mod-
els and the biologically based models discussed here. In
symbolic models, the relative ease of chaining together
sequences of operations and performing arbitrary sym-
bol binding makes it much more straightforward to sim-
ulate higher-level cognition than in a neural network.

11.1 Overview

Higher-level cognition includes important and complex
phenomena such as planning, problem solving, abstract
and formal reasoning, and complex decision-making.
These phenomena are distinguished in part from those
discussed in previous chapters by involving the coordi-
nated action of multiple more basic cognitive mecha-
nisms over longer periods of time. Therefore, they are

379