120 Agnes Petocz
KR is widely agreed to be: What is the nature of knowledge and how do we represent it? This
sounds fine. But closer scrutiny reveals that the question applies to how knowledge is
represented in us (as in computers), and how we use internal representations (as do
computers) in coming to know anything. Thus KR is typically concerned with representing
knowledge in computers (i.e., via computer software programs) to enable such systems to
model human reasoning (van Harmelan, Lifschitz and Porter, 2007). In their classic paper
"What is a knowledge representation?" Davis, Shrobe and Szolovits (1993) claim that "the
fundamental task of representation is describing the natural world" (p. 32), since
representations are "the means by which we express things about the world" (p. 27). So far so
good. But they also argue that the primary role of a KR is that it "functions as a surrogate
inside the reasoner, a stand-in for things that exist in the world" (p. 18). Another role is that it
is "a medium for pragmatically efficient communication, that is, the computational
environment in which thinking is accomplished" (p. 17), since "reasoning in machines (and
somewhat more debatably, in people) is a computational process" (p. 26). Clearly, this is the
language-like medium (cf. Fodor's (1975) famous LOT - "mentalese") constituting the
internal process of semiosis without which we would be unable to think.
But the irreducible triadic relation of signification precludes this on three counts. First,
since signification is essentially a psychological process, the three-term relation (person,
signifier, signified) presupposes the two-term cognitive relation (knower, known). Since the
person must be able to cognise separately both the signifier and the signified, cognitive
relations are not only necessarily part of the three-term meaning relation, but also
presupposed by it; they are logically prior. Since there cannot be meaning without prior
cognition, semiosis cannot be the fundamental basis of cognition. The second point is that the
idea of representation as internal cognitive mediator violates the requirement that one of the
terms in the meaning relation, the cognising organism, have direct and independent
perceptual/cognitive access not only to the signified but also to the signifier or representation.
In this, semiotics converges with the traditional arguments against representationism. Since
the cogniser has no internal perceptual access to the supposed mental representation, the
theory requires the postulation of some other internal perceiver, leading to the famous
homunculus problem - the problem of a vicious infinite regress in which the explanandum -
perception - is begged at every step. But, thirdly, even if there were some kind of internal
perceptual access, the representation could not be intrinsically meaningful, as it is typically
held to be for mental representations, for meaning is a relation extrinsic to the signifier. It is
implicitly recognised even in computer programming that the representational function of any
representation must be external to it; hence the focus on the independence of syntax from the
semantics that is allocated via the software program. Thus, as has been pointed out, the
computer, far from being an existence proof of internal mental representations, is an existence
proof that only extrinsic representation is coherent (cf. Michell, 1988). And the whole
trajectory of the theory of mental representations, now moving away from them as
linguistically conceived and coming full circle back to Lockean pictorial "non-propositional"
forms (Holenstein, 2008), has been a misconceived detour resting on a mistake (Heil, 1981).
In sum, then, recognition of the irreducible triadic semiotic relation at the heart of
meaning reveals the explanatory bankruptcy of the pervasive representationism in psychology
and cognitive science, showing that the standard calls for unification in the information
sciences are misguided, because the type of unification envisaged is simply impossible. To
repeat, any attempt to unify psychology and semiotics via the background assumption that