Early Development and ParentingEarly Dev. Parent. 7: 145–146 (1998)

Innate Representations?Gary F. Marcus*Department of Psychology, New York University, New York,NY 10003, USA

Has Mark Johnson refuted nativism? The cen-tral claim of Developmental Cognitive Neuroscienceis that representations are not innate. Johnson’sargument that representations are not innateseems to rest on two planks; first, Johnson takesa representation to be innate if every node isfixed in advance, and second, Johnson argues(convincingly) that neurons are not in fact fixedin advance, born predestined to their locations.

This would be a good argument if the nativistposition actually held that representations areinnate only if every node is fixed in advance.But Johnson has strawmanned the nativist posi-tion by taking it to rest on the question ofwhether or not nodes are predestined to theirlocations; he equates the possibility of innaterepresentations with a situation in which ‘boththe basic architecture of the network and thepatterns and strengths of links between nodesare innate’.

Yet no serious nativist would make such aclaim, for the fact that neurons are not bornpredestined to their location is not news. AsDawkins (1987) has pointed out, the geneticcode rarely provides a point-by-point blueprintspecifying the location of every cell; instead,what the genetic code supplies is more like arecipe, a set of building instructions.

Thus, the real question, not taken seriously byJohnson, is whether something can be innate—in the sense of not depending on informationfrom the senses—without having every neuronpredestined to its location. Johnson offers littleagainst this possibility beyond a poverty of theimagination argument (i.e. the fact that nobodyhas yet identified a detailed account for howinnate knowledge would be instantiated in thebrain).

In fact, one can imagine a variety of ways ofbuilding innate representations without de-pending on prespecified point-to-point wiring.All that is required for something to be innate isthat it does not depend on information from thesenses. Just as the structure of the heart can beplausibly construed as being innate, eventhough no individual cell is born ‘knowing’ thatit will be a heart cell, the structure of somekinds of representations could be innate eventhough no individual neuron is born ‘knowing’its fate. Rather, a cell might need only carry‘instructions’ like ‘if I am in the neighbourhoodof a heart cell of type 1, become a heart cell oftype 2’, ‘if I am in the neighbourhood of aneuron of type 1, become a neuron of type 2’, or‘if I receive a chemical signal of type 7, activatethe master gene that controls the operation ofgenes 43–57’, and so forth. As Pinker (1997)puts it

. . . to make precise connections . . . the baby neu-rons must begin to function, and their firing pat-tern carries information downstream about theirpinpoint connections. This is not ‘experience’ as itall can take place in the pitch-black womb, some-times before the rods and cones are functioning,and many mammals can see almost perfectly assoon as they are born. (p. 35)

Much of developmental biology seems to workthis way: pinpoint connections are not prespeci-fied by a blueprint but formed by cellular mech-anisms that are sensitive to internally generatedinformation. One piece of evidence that this sortof developmental mechanism plays an impor-tant role in neural development comes from theresearch of the neuroscientists Katz and Shatz(1996), whose work, barely discussed by John-son, suggests that ‘early in development, inter-nally generated spontaneous activity sculptscircuits on the basis of the brain’s best guess atthe initial configuration of connections neces-

* Correspondence to: Department of Psychology, New YorkUniversity, 6 Washington Place, New York, NY 10003, USA.E-mail: gary@psych.nyu.edu

CCC 1057–3593/98/030145-02$17.50© 1998 John Wiley & Sons, Ltd.

Received 11 December 1997Accepted 11 December 1997

G.F. Marcus146

sary for function and survival’ (p. 1133), leadingthem to conclude that

. . . visual experience alone cannot account formany features of visual system development. Innon-human primates, for example, ocular domi-nance columns in layer 4 begin to form in uteroand are fully formed by birth. Thus, althoughvisual experience can modify existing columns,initial formation of the stripes is independent ofvisual experience. Other features of cortical func-tional architecture, such as orientation tuning andorientation columns, are also present before anyvisual experience . . . (p. 1134)

At a minimum, such mechanisms would appearto underlie the abilities of some mammals towalk at birth and the human infant’s swallow-ing reflex; more generally, cells that carriedinstructions for how to respond to internallygenerated signals could lead to richly differenti-ated structures (and representations) withoutrequiring that the fates of individual cells beprespecified and without requiring sensory in-put from outside the developing embryo. In thisway, representations could be innate withouthaving individual cells be predestined. Al-though current neurophysiological evidence de-scribed in Developmental Cognitive Neurosciencemilitates against a straw person genetic-code-as-blueprint version of nativism, it leaves morerealistic genetic-code-as-recipe views of na-tivism untouched.

Johnson has apparently taken a limitation ofthe simplistic first-generation connectionist

models that he describes to be a limitation ofneuroscience. In his connectionist models (de-scribed in Chapters 9 and 10), the only way forsomething to be innate is if all the connectionsare prewired. But that is a problem with themodels, not a fact about the brain. The braindoes not have to work in the same way; there isevery reason to think that the developmentalmechanisms that are available are far moresophisticated.

Rather than opening our minds to a variety ofnew alternative conceptions of how the brainworks, Johnson’s Developmental Cognitive Neuro-science (read, connectionism) leads us to a nar-row conception for how the brain could work, aproblem that is compounded by the fact thatJohnson barely touches most of the develop-mental neuroscience that conflicts with his posi-tion, such as the review article by Goodmanand Shatz (1993); the latter, by the way, is a fineplace to start for readers interested in a morebalanced view.

REFERENCES

Dawkins, R. (1987). The Blind Watchmaker. New York,NY: Norton.

Goodman, C.S. and Shatz, C.J. (1993). Developmentalmechanisms that generate precise patterns of neu-ronal connectivity. Cell, 72, 77–98.

Katz, L.C. and Shatz, C.J. (1996). Synaptic activityand the construction of cortical circuits. Science,274, 1133–1138.

Pinker, S. (1997). How the Mind Works. New York, NY:Norton.

© 1998 John Wiley & Sons, Ltd. Early Dev. Parent. 7: 145–146 (1998)