If this is to be seen as a successful analysis of representational content in terms of intrinsic properties of the brain, we can conclude that the triadic picture alone—with its analysis of the relational character of representational content in terms of dispositions—already solves the content causation problem. Hence, this problem, by itself, offers criteria that could be turned into an argument for or against any specific theory of content determination.
Such a theory, as I understand it, serves two purposes: it explains how the contentful vehicles of which the theory of interpretation makes use are individuated, and it explains “how relations between mental vehicles and their represented objects can endow subjects with the capacity to respond selectively to those very features of the world” (O’Brien this collection, p. 6). It thus provides the background information necessary for understanding why the theory of interpretation is able to do what is required of it. The second desideratum is only made clear if the triadic account of content causation is adopted. According to O’Brien, it can only be fulfilled if we adopt the resemblance theory of content determination, because “[t]he obtaining of causal relations between external conditions and inner vehicles cannot explain how the latter endow systems with the capacity to respond in a discriminating fashion towards the former” (ibid., p. 12), whereas “resemblance does offer some prospect of a solution to the content causation problem. The key here is that the mere obtaining of the resemblance ensures that the former have properties that can be exploited to shape the behavioural dispositions of cognitive systems towards the latter” (ibid., p. 12).
Let us recapitulate the steps that took us from the content causation problem to the second-order resemblance theory of content determination. The content causation problem motivates the triadic account of representation if we assume that it is a problem about reduction and if we assume that there is a sharp distinction between relational and intrinsic properties that forbids an analysis of the former in terms of the latter, thus preventing the reduction of a theory of causally efficacious mental phenomena to a theory of brain-based causation of behaviour. The triadic account of representation solves this problem by showing us that we need not assume that representational content owes its specific character to relational properties. Dispositions, understood as non-relational second-order properties, do justice to our concept of “relational character”. The triadic account then needs to be enriched by a theory of content determination, and its use of the concept of a “disposition” leads to a new requirement: to explain how inner vehicles can enable a subject to respond selectively towards external objects. Supposedly, only second-order resemblance can fulfill this requirement. Thus understood, a second-order resemblance theory of content-determination is only indirectly relevant to a solution of the content causation problem.
Yet I would like to point out a way in which the second-order resemblance theory itself relates to the content causation problem. Second-order resemblance between vehicles and objects tells us that there is a mapping from objects to vehicles that is pattern-preserving or, in other words, some objects and some vehicles are alike in some of their relational properties. Nevertheless, the kind of pattern involved is to be “sustained by constraints inherent in the vehicles, rather than being imposed extrinsically” (O’Brien this collection, p. 11, footnote 11). The relations constituting a structure or pattern collectively supervene on the distribution of intrinsic properties of objects and vehicles, although individual extrinsic (and specifically relational) properties of objects and vehicles do not. This strategy of explanation is in principle also available to our understanding of content: contents, fixed by a structural organisation of vehicles, are relational in the same, completely unproblematic sense. Representational contents may not individually be determined by intrinsic properties of the brain, but there is a sense in which they are so collectively. But this might count as evidence against the second thesis of the content causation problem: that “[t]he representational contents of mental phenomena are not determined by the intrinsic properties of the brain” (ibid., p. 2). One might then even say that content is not relational at all, and that the puzzle that actually troubles us is the question of how representations can have something to which they are applied, namely a “target” (Cummins 1996, p. 8).[8] This could still be accounted for by the triadic picture of representation, but it would then not amount to solving the content causation problem, but to rejecting it.