One of the first attempts at a naturalistic account of content (or intentionality) in the philosophy of mind was Dretske’s (1981) informational semantics, according to which a sign or signal s carries information about property F iff there is a nomic (or lawful) covariation between instances of F and tokenings of s. As Millikan (1984, 2004) emphasized shortly after, informational semantics faces the puzzle of accounting for the possibility of misrepresentation. If the conditional probability that F is instantiated given s is 1, then how could s ever misrepresent instances of F? This puzzle is neatly solved by teleological approaches: if a representation has a function, then it can fail to fulfil its function and thereby misrepresent what it is designed to represent (Millikan 2004, Ch. 5). According to Dretske’s (1988, 1995) own later attempt at preserving informational semantics as part of teleosemantics, a sign or signal s could not represent some property F unless s had the function of carrying information about (or indicating) instances of F.
Millikan’s (1984, 2004) teleosemantic approach sharply departs from Dretske’s (1988, 1995) information-based framework in at least two fundamental respects. First of all, in Millikan’s earliest (1984) teleosemantic framework, there was no room for information-theoretic notions at all. In her later (1989a, 2004) work, she argued that carrying information could not be a teleological function of a sign on the following grounds. Whether a sign carries information about some property depends on how the sign was caused or produced. But according to the etiological theory of functions, the function of a sign is one of its own effects, i.e., the selected effect that explains the continued reproduction of tokens of signs of this kind. How a sign was caused cannot be one of its effects, let alone its selected effect. If and when a sign happens to carry information about something, carrying information cannot be its selected effect, i.e., its etiological function.[4]
Second, Dretske’s (1981) informational semantics could only be suitably naturalistic in the required sense if information is construed as the converse of nomological covariation (or necessity), i.e., as an entirely non-intentional and/or non-epistemic commodity. But as Millikan (2004, pp. 32–34) argues, if signal s could not carry information about F unless the probability that F is instantiated when s is tokened were 1 (in accordance with some natural law), then no animal could ever learn about F from perceiving tokens of s.
In her 2004 book, Millikan elaborates a notion of natural sign that is more “user-friendly” precisely because “it is at root an epistemic notion” (Millikan 2004, p. 37). On Millikan’s (2004) account, natural signs (e.g., tracks made by quail) are locally recurrent signs within highly restricted spatial and temporal domains: relative to one local domain, such tracks are natural signs of quail. Relative to a neighboring domain, the very same tracks are made by pheasants and are therefore natural signs of pheasants, not quail. Locally recurrent signs afford knowledge of the world for animals who can learn how to track the circumscribed domains relative to which they carry reliable information. Furthermore, locally recurrent natural signs can form transitive chains (or be productively embedded) within circumscribed domains. For example, retinal patterns can be a natural sign of tracks in the ground, which in turn may be a natural sign of quail within a circumscribed local domain. Perception is what enables non-human animals and humans alike to track the meanings of locally recurrent natural signs in their circumscribed domain of validity and thereby to acquire knowledge of the world (Millikan 2004, Ch. 4).
The application of Millikan’s (1984, 2004) teleosemantic framework to the meanings of intentional conventional signs results from the combination of three related ingredients: (i) the etiological view of functions; (ii) acceptance of the sender-receiver structure as a necessary condition on the contents of intentional representations; and (iii) a naturalistic account of the reproduction of conventions.
As I said above, on the etiological view, the function of some trait is its selected effect that explains the continued reproduction of past tokens of this trait. This is what Millikan (1984) calls a device’s direct proper function. But a device may also have what she calls a derived proper function. For example, it is the direct proper function of the mechanism of color change in the skin of chameleons to make them undetectable from the local background by predators. It is a derived proper function of this mechanism in a particular chameleon, Sam, at a particular place and time, to make the color of its skin match the color of its particular local background at that time so as to make it undetectable by predators there and then.
While Millikan’s teleofunctional framework based on the etiological approach to functions primarily fits biological traits, it applies equally to non-biological items such as non-bodily tools—including public-language forms. For example, screwdrivers have the direct proper function of turning (driving or removing) screws. This is the effect of screwdrivers that explains their continued reproduction. Clearly, a screwdriver may also be intentionally used for the purpose of driving a screw with a particular metallic structure, length, and diameter into a particular wooden material at a particular time and place. If so, then driving this particular screw into this particular wooden material at a particular time and place will be the derived proper function that this particular screwdriver inherits from the agent’s intention.
According to the sender-receiver framework, a sign or signal R can be an intentional representation (as opposed to a natural sign) only if it is a relatum in a three-place relation involving two systems (or mechanisms), one of which is the sender (who produces R), the other of which is the receiver (who uses R). By application of the etiological view of functions, the sender (or producer) and the receiver (or consumer) have co-evolved so that what Millikan (1984, 2004) calls the Normal conditions for the performance of the function of one depends on the performance of the other’s function and vice versa. In a nutshell, the producer and the consumer are cooperative devices, whose interests overlap and whose activities are beneficial to both. Thus, the cooperative ternary sender–receiver structure naturally applies to the contents of intentional mental representations that mediate between cognitive mechanisms located within a single organism.[5]
In virtue of the fact that intentional mental representations can have two basic directions of fit, the evolved cooperation between the producer and the consumer can take two basic forms. If and when the representation is descriptive or has a mind-to-world direction of fit, the producer’s function is to make a sign R, whose content matches some state of affairs S, for the purpose of enabling (or helping) the consumer to perform its own task when and only when S obtains. If and when the representation is directive (or prescriptive) or has a world-to-mind direction of fit, the producer’s function is to produce a representation whose content will guide the consumer’s action, and it is the consumer’s function to make the world match the content of the sign by its own activities. Furthermore, Millikan (1995, 2004) has argued that the most primitive kinds of intentional mental representations (shared by humans and non-human animals) are what she calls pushmi-pullyu representations, which are at once descriptive and prescriptive, with both a mind-to-world and a world-to-mind direction of fit.
The third component of Millikan’s teleosemantic approach to the meanings of intentional conventional signs involves her (1998) naturalistic account of conventions. On her account, so-called natural conventionality rests on two elementary characteristics: first, natural conventions are patterns that are reproduced (or that proliferate). Second, they are reproduced (or “handed down”) “owing to precedent determined by historical accident, rather than owing to properties that make them more intrinsically serviceable than other forms would have been” (Millikan 2005, p. 188). The fact that conventions rest on historical precedent to a large extent accounts for their arbitrariness.[6] On the basis of her naturalistic account of the continued reproduction of natural conventions, Millikan further offers a purportedly naturalistic account of the continued reproduction of conventional public-language signs, whose function is to coordinate the transfer of information between speakers and hearers. She thereby extends the cooperative ternary sender–receiver structure to the meanings of intentional conventional signs (or public-language forms) that mediate cognitive mechanisms located within pairs of distinct organisms.
Conventional public-language forms are tools or memes in Dawkins’s (1976) sense: they have been selected and have accordingly been reproduced because they serve coordinating functions between a sender (the speaker) and a receiver (the addressee), whose interests overlap. But like any other tool, in addition to its direct memetic (or “stabilizing”) function (which explains its continued reproduction), a particular token of some public-language form may also have a derived function or purpose, derived from the purpose of the speaker who produced it at a particular place and time. Thus, a token of a public-language form has two kinds of purposes: a memetic purpose and the speaker’s purpose, which may or not coincide (cf. Millikan 1984, 2004, 2005).