Date of Degree


Document Type


Degree Name





Michael Devitt

Committee Members

Michael Levin

David Rosenthal

Subject Categories

Metaphysics | Philosophy of Language | Philosophy of Mind


Brandom, Dretske, Moltmann, Pietroski, Godfrey-Smith


I propose an approach to naturalize semantics that combines the use-theory of meaning with teleosemantics. More specifically, I combine Horwich’s claim that the meanings of words are engendered by the acceptance of basic sentences that govern their deployment with the teleosemantic model, developed by Millikan, Papineau and Neander, according to which the meanings of symbols are related to functions determined by the history of their use and of the underlying biological mechanisms responsible for it.

Horwich’s account is general enough to offer plausible explanations of the meanings of all kinds of words and provides a plausible explanation of how meanings govern the uses of words in inference. But, as Devitt shows, his claim that meanings are law-like regularities in the use of words does not make enough room for misuses due to ignorance or error, which may be regular. I argue that this problem can be overcome by adopting a teleonomic account of the functions of symbols, which allows for failures in performance. The teleonomic account characterizes functions as what items are supposed (but may fail) do to, based on their selective history. While this is a biological notion, Millikan and Papineau have proposed plausible ways to extend it to acquired representations. Available teleosemantic theories are truth-referential and are usually regarded as competing with use-theories that are motivated by deflationary views of truth and reference. I argue that we need the basic-acceptance account independently of the fate of deflationism and that it can be articulated in truth-referentialist terms. Additionally, I argue that we need to combine it with teleosemantics. The resulting basic-acceptance teleosemantics claims that some basic sentences containing a word are supposed (but may fail) to govern its overall use. This account, unlike Horwich’s, makes plenty of room for words being misused due to ignorance or error.

Basic-acceptance semantics applies to symbols that play roles in inferential processes. For more basic animal representations that have direct perceptual causes and behavioral effects, I propose instead an account that combines Millikan’s effect-based teleosemantics with Neander’s cause-based teleosemantics. Millikan’s theory explains meanings in terms of the conditions in the world that representations are supposed to covary with in order to have the effects they have the function of producing. I argue that this theory has the advantage of making enough room for misrepresentation, due to its output-based character, but can ascribe meanings that are implausible because it ignores the causes of representations. Neander’s theory explains meanings in terms of the causes that the perceptual mechanisms are supposed to respond to. I argue that this theory has the advantage of ascribing meanings that are plausible given the perceptual capacities of organisms, but it does not make enough room for misrepresentation because it ignores the effects of representations. According to the hybrid account I propose, the meanings of basic representations are determined by what is supposed to cause their tokens in order for them to bring about the effects they have the function of producing. I argue that the hybrid account makes enough room for misrepresentation while ascribing meanings that are plausible given the capacities of the perceptual mechanisms that produce them.