A developmental chronology
Memory research
Adaptationism and Modularity
Steen's stab at a clarification
(5 September 1997, comment added 2 October 1999)

Question: When you speak about an adaptationist approach, does that imply a modular model of the mind?

Answer:  The way I see it, if you think certain kinds of faculties are adaptations, this implies you don't think the mind is only a domain-general problem-solver. However, it doesn't mean you are committed to a strictly modular mind.1

Many of the strongest claims for specific adaptations lie fairly close to the sensory systems--Spelke's stuff on object perception; Baron-Cohen on intentionality detection, gaze-direction detection, and shared-attention detection; Buss on mate-selection based on symmetry and waist-hip ratios; face detection; Shawitz on phonological parsing; preferences for sweet and fat foods and odours, and a disgust-reaction to putrid smells; snake-avoidance; a large number of generalized algorithms in the visual system (body-oriented locational maps, a stable world despite saccades and body movements, looming inferences, parallax inferences, etc.). These sensory response systems are typically fast and relatively resistant to conceptual interference, take place below the horizon of conscious awareness, are typically present at birth or soon after, and generally fit fairly well with Fodor's idea of modularity. This list does not include innate motor skills, or the motor component of such sensory-based inference engines--people are doing interesting work in embodied cognition to study this (e.g. Clark). In most cases, the mind is not compelled to act on the information these modules make available. (The sensory and motor areas have been mapped quite precisely onto the neocortex; for the classic homunculi, see diagram, new window.)

Then there's a second level of quite convincing functional specializations (Tooby's preferred term for "modules"), such as emotions and emotion detection (Ekman), the capacity for metacognition (Leslie, Sperber), a theory of mind (Baron-Cohen), language acquisition (Pinker), certain biases in classification (Markman), and social reasoning of various kinds (Cosmides and Tooby). Here the systems are clearly no longer "modular" in Fodor's sense--they are, for instance, open to and often dependent on input from memory as well as from the senses. Characterizing how these systems work and how they differ from the more automated sensory modules is an on-going task. Like the sensory modules, they operate through certain conceptual primitives and inference algorithms; however, at least elements of the information processing are accessible to conscious awareness, and intentions, knowledge, and desire can moderate the results. They are involved in refining information and setting priorities. They are not informationally encapsulated, bottom-up informational-processing structures.

It is this second level of functional specializations that are contentious. The debate is computational, neurological, developmental, and evolutionary.

What kind of a computational system do we need to have to solve those problems without help from a biologically evolved domain-specific specialized circuitry? Can we account for the skills through generalized algorithms, or do we need to invoke evolved preferences for framing the situation in certain ways? Is it possible for a generalized cognitive system to deduce the existence of minds, biological kinds, and cost/benefit ratios? How much structure do we need in the mind to start with, and how does the mind pick up new structuring information?

Do we in fact have subsystems in the brain that perform these specialized tasks? Do we find that people process the same kind of information in the same areas of the brain? Do we get dissociative deficits in neurological pathologies? (But see also Bickerton's comment, below.)

Do children develop these skills even in the absence of explicit instruction? Does the development tend to happen at the same time in all children? Are these skills associated with specific developmental deficits?

Do these faculties represent solutions to problems that were likely to have had significant fitness consequences in the EEA? Are they present in all human populations?

In conclusion, the adaptationist claim includes Fodor-type modules, but also makes additional claims about functionally specialized subsystems that take a variety of input and whose processing is in part open to introspection. One need not be a priori committed to a panhuman neurological implementation of such functional specializations, though within certain limits this is likely to be the case. As for the differences between the cognitive constructivist and the adaptationist approach, it is hard to see a fundamental conflict of interest; adaptationism is itself computational, and even the most ardent cognitive constructivist admits evolution has played a role in developing the human cognitive system.

Finally, since human beings exist and emerge from a complex web of physical and representational relations, our own descriptions of ourselves should not aim only for Platonic abstractions, but also explicitly discuss what kind of understanding would actually be beneficial.

Francis Steen
September 1997

Comment (1 October 1999): In an informal response on the Evolutionary Psychology Mailing List to the published critique of domain specificity by Elman, Bates, and others, Derek Bickerton writes,

A large part of the problem lies in confusion between domain specificty and localization.  The two are often taken as going together, so once you've downed localization (as brain imaging has fairly well done for language functions) you've downed domain specificity, or as I'd rather call it, dedicated circuitry.  No way.  Once you grasp that the same cell or cell assembly may belong to several different circuits, the problem diminishes. Assume a brain of Dennettian dumb homunculi, with no executive suite, engaged in Darwinian competition for "spare" neurons à la Calvin, and it disappears altogether.

Quoted with permission. See also Bickerton and Calvin's project Lingua ex machina (forthcoming and on-line).

A developmental chronology
Next: Memory research
Memory research
Back to CogSci index
© 1997 Francis F. Steen, Communication Studies, University of California, Los Angeles


1. Marr (1992) uses the term "module" when speaking about aspects of the visual system, and Fodor (1983) generalizes and refines the idea. In Reflections on Language (1975), Chomsky refers to "mental organs" or "cognitive competences." Hirschfeld and Gelman (1994) and other experimental psychologists generally use "domain specificity," while Barkow, Tooby and Cosmides (1992) and the evolutionary psychologists prefer "psychological adaptations" or "functional specializations." (Thanks to Brian Boyd (1997) for these clarifications.) Back