In his paper in this collection, Gerard O’Brien confronts the task of “explaining how mind matters” (p. 12). He does so, because he— rightly, I believe—identifies the fact “that our minds matter—that our beliefs and desires, and our perceptions and thoughts ultimately have a causal impact on our behaviour” as a ubiquitous and well-accepted, but unexplained phenomenon (p. 1).
O’Brien’s investigation is motivated by the following question: “[h]ow can mental phenomena be causally efficacious of behaviour in virtue of their representational contents if those contents are not determined by intrinsic properties of the brain?” (O’Brien this collection, p. 2) He calls this question the “content causation problem” (ibid., p. 2). This specific way of approaching the matter of mental causation is set in the context of “three widely accepted theses about mental phenomena and their physical realization in the brain” (ibid., p. 2): (i) the supposed causal efficacy of mental phenomena is grounded in their representational contents, which, (ii) are taken to be relational properties of those phenomena (ibid., p. 2–3); and (iii) the results of neuroscience, which already provide us with an explanation of how behaviour is caused, only make use of the brain and its intrinsic properties in their explanation (ibid., p. 2). Hence, there is a need for an explanation of behaviour being caused by mental phenomena in virtue of their relational properties, and this explanation cannot easily make use of the explanation of the causation of behaviour that has already been provided by contemporary neuroscience. At first, it looks as if this shortfall is exactly what O’Brien is addressing.
Philosophical mainstream accounts of representation, O’Brien reminds us, are built on an understanding of representation as a two-place relation. Representational content is thus described in terms of aboutness and/or reference. O’Brien, however, advises us to abandon the traditional understanding of the notion of representation, i.e., the idea that representation is a two-place relation and adequately phrased in terms of one thing being about another (O’Brien this collection, p. 3–4). Instead, he proposes a triadic conception of representation, making representation a three-place relation between a represented object, a representing vehicle, and an interpretation (ibid., p. 5).
In the triadic picture, interpretation is “a cognitive effect [of the object] in the subject”, thereby establishing a relationship between this subject and the represented object (O’Brien this collection., p. 5). The ingenious move here, of course, is that interpretation is explained in causal vocabulary. We should think of interpretation as “presumably implicating the production of mental representing vehicles” possessing new properties, which should in turn be thought of as “bring[ing] the subject into some appropriate relationship to the original vehicle’s represented object” (ibid., p. 5), i.e., “modifying [the subject’s] behavioural dispositions” (ibid., p. 6). At first, when O’Brien further describes those vehicles as “hav[ing] […] cognitive and ultimately behavioural effects” (ibid., p. 6), it isn’t clear exactly which category we are dealing with. I take the relata of the causation relation to be events, but understand talk of vehicles to be talk about objects.[1] I suggest that we understand the vehicles as modifying the system’s behavioural dispositions insofar as, once produced, they have certain properties that are directly and specifically relevant for a causal process to take place (given that some sort of stimulus initiates the causal process). If we adopted a view of dispositions as second-order properties, i.e., the property of having certain properties that can be causally relevant (cf. Choi & Fara 2014), this would allow us to think of the vehicles as modifying global dispositions of a system as a whole, in the sense of providing new ones—that is, by themselves having novel dispositional properties. This way, we can analyze the obtaining of the representation relation as a specific causal process having taken place: the first step of that process is the triggering of the cognitive effect by the representandum (the first relatum); the second relatum is the event of interpretation itself; and the third relatum is the new vehicle produced during the event of interpretation that provides the subject with a new behavioural disposition towards the representandum. The representational character of mental content, in this picture, just rests on what we call “content” resulting from the multi-layer causal process described above.[2]
O’Brien’s triadic account of representation describes the obtaining of the representation relation not in terms of our everyday intuitions about representation, but in terms of the job it is supposedly doing for us: bridging the gap between whatever is going on in the sphere of “the mental” and the external world by alluding to the causal chain that unites the two. It is thus understandable why the dyadic conception might be accused of hiding behind terms like “aboutness” or “reference”: saying that something mental is about something external is just saying that there is a gap being bridged. Saying that something external sets a three-step causal chain in motion with the result that a subject has undergone a specific change in her behavioural dispositions seems much closer to saying what the bridge is made of.
Yet we should still dispose of the vague language of “specific change in her behavioural dispositions”. What exactly makes this change specific? It was called “specific” because it selectively relates back to the object that set the causal chain in motion in the first place and which we would like to keep calling “the represented object.”[3] But how can the change in a subject’s behavioural dispositions make them pick out the exact same object from which this change originates?[4] The answer to this lies in the theory of content determination.
When holding that the representing vehicle brings about a change in the subject’s behavioural dispositions, causal theories of content determination are supposedly to be abandoned because of a “disconnect between world–mind causal relations and a system’s behavioural dispositions” (O’Brien this collection, p. 12). An appropriate theory of content determination—so says a desideratum that we gain from the results of the triadic analysis of representation—must “explain how [inner vehicles] endow systems with the capacity to respond in a discriminating fashion towards [external conditions]” (ibid., p. 12). For fulfillment of this criterion, O’Brien turns to resemblance theories of content determination, which “hold that representing vehicles are contentful in virtue of resembling their represented objects” (this collection, p. 9).
Within the triadic conception of representation, O’Brien identifies two hurdles for a resemblance theory that still need to be overcome: it must be shown how the theory can be compatible with physicalism, and it must be secured that the theory does not leave content indeterminate (ibid., pp. 9–11).
In order to secure the compatibility of a resemblance theory of content determination with physicalism, O’Brien turns away from the notion of first-order resemblance and instead makes use of structural or second-order resemblance (ibid., pp. 10–11). He thus avoids the seemingly naive and implausible thesis that mental representations must actually share properties with what they represent (ibid., p. 10). Resemblance is taken to a more abstract level where, for example, something red can be mentally represented with the representation resembling the representandum, but without them both sharing the property of being red (ibid., pp. 10–11).
The second hurdle might seem redundant at first glance. Explaining how content is determined is basically the job description of a theory of content determination. It is still worth mentioning this as an obstacle, however, because the reliance on second-order resemblance makes this job look particularly difficult: second-order resemblance is too easily established. If a set of mental representations second-order resembles a pattern of colour shades, it might in virtue of the same relational organization also second-order resemble a pattern of locations in a two-dimensional space. Nevertheless, O’Brien trusts that within the triadic conception of representation, second-order resemblance will do the job. The idea is that some of the possibilities for the content of a vehicle that are left open by second-order resemblance are ruled out in the process of interpretation—interpretation is “content-limiting” by “anchoring vehicles” in “domains” (O’Brien this collection, p. 11). The preexisting behavioural dispositions influence the newly developing ones, so that they are not directed towards all domains with a specific relational organisation, but towards a selection of these.
So far, O’Brien has provided us with an interesting account of how mental phenomena are causally efficacious in virtue of their representational contents: the property
x has representational content
is analyzed as the property
x results from a causal process that brings about behavioural dispositions towards the object that triggered the causal process.
These behavioural dispositions, given their respective stimuli, can now yield causal effects.
But if we took this as O’Brien’s only accomplishment, we would miss the most interesting part of his argument. Furthermore, we would take the second step before the first.