Two or three decades ago, the status of explanations by reasons was uncertain. Then, with the assimilation of the view that reasons are causes, philosophers stopped worrying about the fate of reasonsexplanations: To explain a person's behavior in terms of her reasons was just another way to explain behavior in terms of its causes. Recently, however, a new threat to reasons has come to light. Even if reasons are taken to be internal events that cause behavior, does the fact that the cause is a reason make any difference to the production of behavior?
The problem is this: Not every aspect of a cause is relevant to its producing an effect. To borrow a telling example from Fred Dretske, a soprano's high-C may shatter glass and it may mean, e.g., "Help!" But its meaning "Help!" had nothing to do with its effect on the glass, which is explained wholly in terms of the acoustic properties of the sound. If the high-C had meant something else or nothing at all, the glass would still have shattered. The worry is that the meaning of reasons may be as explanatorily irrelevant to behavior as the meaning of the soprano's high-C is to the shattering of the glass.
Dretske has developed an influential account of "how ordinary explanations, explanations couched in terms of an agent's reasons, explain. ''1 In order to show how reasons explain, Dretske tightens the issue in the following way: Supposing that reasons are attitudes like beliefs, desires and intentions, and that beliefs, etc., are identified by their meaning, the task is to show how meaning (e.g., what one believes) can help explain behavior. The goal is to find an explanatory role for meaning in a world of causes.
Although Dretske's account is rightly regarded as ingenious, I believe that it falls to circularity. For, as I shall try to show, Dretske is