To the editors:

I found Robert Berwick and Noam Chomsky’s essay to be concise and insightful. I was, however, bothered by what seems to be the core of the argument: the ability of an evolutionary process to generate human language.

So far as I can tell, the strongest case made for this in their essay is as follows:

A slight rewiring can sometimes result in a large transition. … The ability to process sequential information is shared across many vertebrate species—perhaps all. A slight alteration in the wiring of a simple sequential processor is sufficient to endow it with a push-down stack. This makes for a significant improvement in its computational power. It is a point of some significance: a push-down stack is needed to process hierarchical structures.

First of all, this is just an analogy. The processes undertaken by computers are sort of like human language, the argument runs, therefore the model of the computer explains human language—QED. This is the weakest form of reasoning and has never worked well in the sciences; it has generally given rise only to fads and fashions. Today humans are supposed to be advanced computers; in the eighteenth century, they were like crude mechanical automatons. We become whatever is lying around the conceptual workshop.

But even if we think this weak analogical reasoning is valid, I find the claims made by the authors to be rather extreme. The alteration of the wiring in their simple computer circuit is described as slight. I count 10 alterations to each of their four boxes. That is 40 alterations. What is more, the alterations have to be specific; if one of the circuits is connected in the wrong way, the system breaks down. This means that the number of potential ways to wire the boxes would be exponential, and so too the chances that a functioning push-down stack emerges to be a negative exponent.

That seems pretty complex to me. History appears to be on my side, too. Some amateur archival work on my part suggests that a simple push-down stack was first described by Allen Newell et al. at the RAND Corporation in 1959.¹ If the alterations to the circuit were as slight as Chomsky and Berwick claim, one wonders why some workshop tinkerer did not stumble upon it earlier. Volta had an operating electrical circuit in 1800, and by the late nineteenth century, Edison and company were producing all sorts of technological marvels.²

Philip Pilkington

Robert Berwick and Noam Chomsky reply:

Philip Pilkington is “bothered by what seems to be the core of the argument: the ability of an evolutionary process to generate human language.” Since it is a truism that language emerged through an evolutionary process, we assume he meant something different. Perhaps he is bothered by our particular way of approaching the evolutionary process, which he misunderstands. What he describes is neither an analogy nor a proposal as to how language evolved. Rather, it is a notional illustration of how a slight alteration of a sequential processor could endow it with enough computational power to process hierarchical structures, as required for language processing. It is a kind of existence proof. Its units are silicon, but neural connections can be grown in more directed and non-random ways than silicon. To be sure, Pilkington is right to inquire into the step-by-step details about any such example, for it remains to be determined how the capacity to process hierarchical structures arose through an evolutionary process.