Connectionism, Analogicity and Mental Content
12
rules (independent, that is, of the representational substrate of cognition). To put it bluntly, the
“shape” of these vehicles matters; it’s their “shape” that drives cognition.11 And this completely
changes our view of one of the most important topics in the contemporary philosophy of mind:
the determination of mental content.
Philosophers, as H&T observe (1996, p.13), have long thought that the issue of
intentionality is orthogonal to the question of cognitive architecture. This view is the legacy of
classical cognitive science. Given that digital computations inherit their semantic coherence from
rules that are quite distinct from the structural properties of the symbols they apply to,
classicism appears to place few constraints on a theory of mental content. As long as these
symbols are transformed according to these rules, it matters not where their representational
content derives from. Thus the causal, functional and teleofunctional theories of content that
dominate the current literature are all, prima facie, compatible with the idea that mental
representations are symbols. In fact, of all the psychosemantic theories in the philosophical
marketplace, there is only one that would seem incompatible with classicism. This is the
resemblance theory of mental content, according to which representational vehicles are
contentful in virtue of resembling, in some way, their representational objects. As Robert
Cummins notes, “[classical] computationalists must dismiss similarity theories of representation
out of hand; nothing is more obvious than that data structures don’t resemble what they
represent” (1989, pp.30-1).
But all of this changes when we replace classicism with connectionism. Connectionism,
because it is grounded in analog computation, and because analog computation requires the
presence of a structural isomorphism between representational substrate and target domain,
completely closes the gap between the issues of intentionality and cognitive architecture. In
connectionism, the content a mental representational vehicle carries through the mind cannot be
independent of its “shape”. As a consequence, connectionists just don’t have the same luxury as
classicists when it comes to mental content determination. They are forced to explain mental
content, at least in part, in terms of resemblance relations between representational vehicles and
their representational objects. They are forced, in short, to embrace a resemblance theory of
mental content.
A few years ago this would have seemed a serious objection to connectionism. The
resemblance theory was thought to suffer from a number of fatal flaws, and hence, in most
philosophers’ minds, was not much more than an historical curiosity (see Cummins, 1989,
chp.3). But the last few years have seen the beginnings of a seachange in the philosophy of mind.
A number of theorists are taking resemblance very seriously again, especially in the form of
second order structural isomorphism.12 And if the line of reasoning that has been developed in
this paper is at all on the right track, this group of theorists will soon have its number increased
dramatically.
11 This is not to say that the “shape” of mental symbols is wholly unimportant in the classical theory of mind. But the
importance of shape in this latter case is relative - relative to the rules according to which these symbols are
transformed. The point is that so long as these symbols are transformed according to these rules, it doesn’t matter
what shape they are.
12 Two theorists who have kept the torch of resemblance, in the form of second order structural isomorphism, burning
over the years are Stephen Palmer (1978) and Roger Shepard (Shepard and Chipman, 1970; Shepard and Metzler,
1971). But more recently, James Blachowicz (1997), Robert Cummins (1996), Shimon Edelman (forthcoming), Craig
Files (1996), Peter Gardenfors (1996), Daniel Gilman (1994), and Chris Swoyer (1991), have all explored, though in
different ways, resemblance relations between representational vehicles and their representational objects.