Connectionism, Analogicity and Mental Content
physically inscribed in the form of concatenations of high and low voltages. One must next find
a means of “reading” and “writing” the symbols in this representational medium. This requires
a device capable of detecting and transforming these symbols on the basis of their syntactic
properties, rather than their microphysical details. This is achieved in conventional digital
computers by grouping transistors together to form “gates”, which are differentially responsive
to high and low voltage states, rather than precise voltage values.
Together these two steps lead to the construction of a syntactic engine: a physical device
whose casual behaviour is determined by the syntactic properties of the symbols recorded in its
representational medium. But left to its own devices, a syntactic engine is not terribly useful; it
doesn’t “naturally” transform symbols in a manner that is sensitive to their semantic content,
and hence doesn’t “naturally” perform any computations. In general, there is nothing intrinsic to
the syntactic structure of symbols that ordains that they will be manipulated in a semantically
coherent fashion (eg, there is nothing in the particular sequence of high and low voltages
comprising one symbol in a conventional digital computer that dictates that it must bear some
sensible semantic relation to the sequence of high and low voltages comprising another).
Consequently, it takes work to harness the power of a syntactic engine; the engine must be
forced to perform certain kinds of symbol manipulations (ie, those that are semantically
coherent) and avoid others (those that aren’t). The engine’s behaviour must, in short, be rule-
governed. These rules, while they are syntactically applied, are shaped in accordance with the
semantic relations that obtain between the symbols on which the syntactic engine operates, and
hence contain the semantic coherence of the device. They dictate that this engine will transform
one symbol into another only if their is a sensible semantic relation between them. It is only
when syntactic engines conform to such rules that they become digital computers.
In practice, of course, these rules, like every other feature of the device, must be
physically implemented. It turns out that many of these rules can be explicitly coded in the form
of symbols inscribed in a digital computer’s representational medium (ie, the notion of a stored
program). Ultimately, though, a set of primitive rules (analogous to the primitive computational
instructions resident in the read/write head of Turing’s abstract machine) must be “hardwired”.
This is done, in standard digital computers, by subtly wiring together groups of transistors - the
read/write “gates” mentioned earlier - so that the voltage transformations they execute are
always in accordance with the primitive rules. This microcircuitry thus literally embodies a set of
primitive instructions and with it a basic computational competence
This is Turing’s way of mechanising computation. And when Turing’s way is applied to
the mind, the result is classicism: the theory that cognitive processes are digital computations.
Classicism takes the generic computational theory of mind (the doctrine that cognitive processes
are semantically coherent operations over neurally implemented representational vehicles) and
adds a more precise account of both the representational vehicles (they are complex symbol
structures possessing a combinatorial syntax and semantics) and the computational processes
(they are rule-governed, syntactical transformations of these symbol structures). The rich
diversity of human thought, according to classicism, is the result of a colossal number of
syntactically-driven operations defined over complex neural symbols.
All of this is standard fare, as classicism is normally understood in the literature as the
digital computational conception of mind. According to H&T, for example, the most
fundamental assumption of classical cognitive science is a commitment to what they term
programmable representation-level rules: “cognitive processing conforms to precise, exceptionless
rules, statable over the representations themselves and articulable in the format of a computer
program” (1996, p.24). Their focus on such rules is quite justified, given the foregoing
description of digital computation, as it is these rules, and these rules only, that embody the
computational competence of a digital device and ensure that it behaves in a semantically