Another criticism of the computational approach is that its formality renders it universally realisable - Putnam [11] and Searle [13] argue that any physical system can be interpreted as realising any formal automaton. This has the consequence that an account of cognition cannot be in terms of formal computation, since any particular formal structure, the realisation of which is claimed to be sufficient for cognition, can be realised by any physical system, including those that are obviously non-cognitive.
The previous two subsections were examples of two different situations: cases in which the notion of computation could be extended to cover the phenomenon in question, and cases in which it could not, respectively. This subsection is an example of a third situation: cases in which the notion of computation need not be extended to new phenomena, but rather re-conceived concerning its application to central, undeniably computational phenomena. Specifically, the solution that has been proposed (by [1], [2], et al) has been to acknowledge an aspect of traditional automata theory that has lain dormant until these objections were raised: the essentially causal nature of state transitions. It is only this causal requirement that explains our pre-theoretic view that although a PC may be a realisation of a Turing machine T, the same is not true of the set of display screen states graphically depicting T, even though both PC and screen go through an isomorphic pattern of states. Once this is acknowledged, then it is seen that computation is not universally realisable, since the requisite counterfactual-supporting causal transitions necessary for the realisation of a particular automaton will not be found in an arbitrary physical system (pace Putnam).
With respect to transparent computationalism, this kind of reply seems to be a borderline case. Does making an implicit aspect of the traditional notion of computation explicit count as changing the concept of computation? Inasmuch as the answer is yes, this line of response to the universal realisability objection to computationalism adopts the transparent reading of computationalism. Inasmuch as such explicitation does not count as conceptual change, then it is a refutation of a more direct sort - the transparent approach is not required.