I've seen the term, and it seems to influence how claims are interpreted, but most of what I've read on the subject (IANAL!) rapidly gets confusing. Can anyone provide a good definition of what it is and what it means for patent interpretation?
In patent law, it is the "claims" that define the scope of the exclusive patent right. Claims that are more broadly written cover more potentially infringing activity and are thus seen as valuable (though more likely to be attacked on validity grounds).
One potential way to greatly increase claim scope is to draft a claim with functional limits rather than structural limits. For a pencil, you might claim "a marking element" that in many cases will be much broader than "a cylindrical piece of lead surrounded by an encasing material." The functional version of a claim is also much easier to write because you don't have to actually explain how the invention works, only what it accomplishes.
For various reasons, pure functional claim limitations are disfavored in the patent system. If written improperly a purely functional claim will be found invalid as unduly indefinite or abstract. If written properly, the claim will be quite narrowly interpreted. Let me explain. A proper functional limitation will be written in what is known as "means-plus-function" form, such as "a means for marking." See 35 U.S.C. 112(6). Although the language "means for marking" seems quite broad, over the past 25 years, courts have given these claims increasingly narrow scope. As such, patent attorneys have largely moved away from using MPF language. On this point, see my post on Patently-O.
The more common approach is to use a hybrid claim form that includes some amount of structure and also some amount of function. Thus, you might claim "a circuit configured to receive data" or "software configured to display an inverted image." In these cases, some amount of generalized structure is provided (a circuit; and software). However, the claims do not describe the exact design of the structure but instead provide a functional limitation. The patentee hopes (and many courts have agreed) that this hybrid claim structure allows the patent to cover any and all forms of the generalized structure that is designed in a way that accomplishes the claimed function. This obviously gives the patent quite broad coverage.
Professor Mark Lemley recently wrote an article where he argues that claim language in the form of "software for accomplishing x" should be seen as a pure functional claim (rather than a hybrid) and thus should be either rejected as invalid or narrowly interpreted. Lemley Article. Lemley's suggestion has not (yet) been adopted by the courts or the USPTO.
If you make a functional claim (a "means-plus-function" claim) then what it means is that you get the function, but only if you disclose sufficient structure for those means in the specification. For literal infringement, you only get that function when tied to the relevant structure - you can also get the equivalent structure under the doctrine of equivalents, so long as you haven't surrendered it during the prosecution history.
Here's a pretty good explanation from a Federal Circuit case: http://www.patenthawk.com/blog/2006/05/meansplusfunction_claims.html