An adaptationist criterion for signal meaning

To think that use constitutes meaning is an appealing idea, but also notoriously di cult to make precise. After reflecting on the notion of use-constituted meaning that signaling theory, in the vein of Lewis (1969) and Skyrms (2010), gives us, I suggest a notion of signal meaning that draws on adaptationist theories of biological communication. Such a criterion has interesting consequences, such as that just because language use is vague does not necessarily entail that language is vague. “Wenn sich alles so verhält als hätte ein Zeichen Bedeutung, dann hat es auch Bedeutung.” (Wittgenstein, TLP 3.328) Communication takes place between two separable entities. These need not necessarily be two di↵erent individuals, for we want to allow for communication to be possible between, say, neurons in the brain. Communicating entities also need not be of the same species, for we want to allow for communication between, say, a flower and a bee. Stereotypically, we might think of communication as taking place between human speakers and listeners, but that is just a special case. Opinions diverge on how special it is, and why it should be so. In any case, communication often has a certain directedness from one entity to the other. The former will be called sender, the latter receiver. We expect communication to arise when the utilities (think: happiness, longevity, replication or procreation probability, biological fitness, . . . ) of sender and receiver are dependent in appropriate ways. A comprehensive, versatile and highly expandable model of such a situation is found in signaling games as introduced by David Lewis (1969). The sender shows or performs some observable pattern, henceforth called a message m. This may possibly be conditional on some state of nature t. The receiver shows some kind of reaction a to the observable messages m. (The receiver may have his own informational resources, but in the simplest case that we assume here the receiver is ignorant of which state of nature obtains.) The triple t, m and a is an outcome of the game. How valuable an outcome is to sender and receiver is given by a utility function that maps outcomes onto reals (the higher the number, the better). In the simplest, non-trivial case there are two states, two messages and two acts. There is a one-toone mapping between states and acts in the sense that sender and receiver both get a high payo↵ when the receiver chooses the act which is right given the state and a low one otherwise. We assume that a1 is appropriate for t1, and a2 for t2. There are four non-probabilistic patterns of sender behavior: (1) a. t1 7! ma t2 7! ma b. t1 7! ma t2 7! mb c. t1 7! mb t2 7! ma d. t1 7! mb t2 7! mb Similarly, there are four principled types of non-probabilistic receiver behavior: (2) a. m1 7! a1 m2 7! a1 b. m1 7! a1 m2 7! a2 c. m1 7! a2 m2 7! a1 d. m1 7! a2 m2 7! a2 Two combinations of sender and receiver behavior are special; Lewis called them signaling systems: Signal meaning Michael Franke