I must say I am bad at semantics, but somehow I can give you some clue. You have to be aware that meaning is not always compositional, the adjective fake in a fake gun and fake teeth will profile different characteristics of the things that it can modify, e.g., a fake gun retains some characteristics of a real gun, such as shape, even material, but fails to exhibit its function, while for fake teeth, they exhibit the shape and functions of real teeth, but fail in terms of material. The phenomenon is extensively discussed in cognitive semantics and conceptual blending theory, to name just a few. I recommand the book Cognitive Linguistics: Foundations, Scope, and Methodology (edited by Theo Janssen, Gisela Redeker, 1999), especially P147-148, you can get a  preview online of these two pages in Googlebooks, and it can pretty much solve your problem (by using conceputal blending theory). Hope this helps.  

I must say I am bad at semantics, but somehow I can give you some clue. You have to be aware that meaning is not always compositional, the adjective fake in a fake gun and fake teeth will profile different characteristics of the things that it can modify, e.g., a fake gun retains some characteristics of a real gun, such as shape, even material, but fails to exhibit its function, while for fake teeth, they exhibit the shape and functions of real teeth, but fail in terms of material. The phenomenon is extensively discussed in cognitive semantics and conceptual blending theory, to name just a few. I recommand the book Cognitive Linguistics: Foundations, Scope, and Methodology (edited by Theo Janssen, Gisela Redeker, 1999), especially P147-148, you can get a  preview online of these two pages in Googlebooks, and it can pretty much solve your problem (by using conceputal blending theory). Hope this helps.  

Thanks a lot, Matthew. Yes, seeing nouns as evoking prototypical categories with several properties that need't always be present and also especially allowing adjectives to profile some but not all of these properties might work as a solution.

Hi Bert,

As Matthew said, 'fake X' has been much studied under the heading of 'conceptual blending'. See for example Coulson and Fauconnier at http://www.cogsci.ucsd.edu/~coulson/Fake/fakeguns.htm. The first mention (as far as I know)  of the non-compositionality of NPs with 'fake' is in Lakoff & Johnson's *Metaphors we live by*.

As John R. Taylor points out in his book *Linguistic categorization*, 'fake' involves more than just non-compositionality but also the notion of an intention to deceive someone else about something relatively important. He shows that 'false' does not have this notion of high-stakes deception: if you discover that someone has false teeth, that is no big deal, but if they pass off fake teeth as proof of some theory of evolution, they stand to be ridiculed as palaeontologists. What does this mean for fake orgasms? Perhaps that you stand to lose your partner('s respect) if you're caught.

Hi John,

Thank you so much for these further references, including Taylor's views on "fake" vs. "false". There's so much broad, encyclopedic meaning that goes into words. The whole discussion of "fake x" came back to my mind when I recently saw Chomsky use "non-" in this sentence:

"Let's review this curious non-interchange"

(http://www.samharris.org/blog/item/the-limits-of-discourse)

Again, if you take this at face value, this might seem to make little sense when it's used what actually does look like an interchange. So, here too, what is needed is a semantic theory that allows the speaker to take into account various perspectives attributed to different speakers (yielding something here like 'what you may consider to be an interchange from your delusional point of view certainly isn't one in my book'). A colleague of mine (Edwige Dugas) is writing a PhD thesis on various uses of non-, which I'm sure will provide further interesting insights.

Again, many thanks, Matthew and John, for reminding me of where to read up on these phenomena.

B

I cannot add to what has been said above. let me sum up the arguments in my words however. The theory of meaning allows the adjective fake to modify nouns in order to deceive. All lies fall in that rough category of creating a world which does not correspond to the real world. This, indeed, is a quintessentially human capability. Lying, fantasy, metaphor and fiction are all based on deviating from phenomenal reality (as does insanity). It is a wonderful subject to explore at length.

I recommend Lakoff and Johnson's "Metaphors We Live By" who first develop their theory of "fake guns" and prototypicality.

A more personal thought, according to a strict literalist interpretation of the word "fake" as the one which proposed most "fake X", if not all, as contradiction in terms, this would amount to saying that the word "fake" itself could never be used and would thus be meaningless. It also reflects a persistent literalist view of language and the world, shunning metaphor and almost denying the fact that paradoxes and contradictions are part of the world we live in even beyond language, and certainly beyond semantics.  I also agree with Tariq Rahman's view.

A linguist should not become a Dickensian character who would not have flower patterens on a carpet using the argument that you would not tread on real flowers.

Most normal people can easily interpret "fake orgasm" with "orgasm" meaning "the perception third parties have of what one is feeling at a given moment", not the actual feeling. In the animal kingdom this is a frequent pattern of behaviour, often for survival purposes. Would you say that an animal's camougflage colours are "fake vegetation colours"? From that point of view, you just need to be more imaginative in your definitions, to adjust them to how people understand language, which is probably the point of linguistics, anyway.

Hi,

I fail to see why 'fake x' should be fundamentally different from other non-intersective adjectives like 'former x' or 'alleged x', all of which are clearly compositional.

Everybody has to live with the established fact that not all adjectives show intersective semantics, and 'fake' clearly belongs to this class.

If 'fake x' has roughly the same semantics as 'pretending to be an x' there is no contradiction, and no non-compositionality either. Basically, a fake of x is not an x.

Hi Tibor,

Thank you for your contribution. So, 'fake x', 'former x', alleged x', 'likely x', 'potential x', etc. are compostional in that the adjective contributes a specific kind of meaning which differs from Adj to Adj, but their semantics is such that they don't just portion off a subset of all elements that objectively (whatever that could mean) belong to the class of x?

Hi Bert,

yes, the relevant adjectives are not intersective. 'Former N' picks out individuals that have been restricted by N prior to the relevant point in time, 'alleged N' picks out individuals that have been claimed to be restricted by N, and 'fake N' picks out individuals (or better objects) that are claimed but fail to be restricted by N for some characteristic property that should be present of they actually were N. This can be expressed by various functions and always yields a compositional meaning. 

While I am delighted with both of Tibor Kiss's contributions I still wonder whether his explanations, which I do not question, can also be complemented with other accounts from a different angle, without cancelling each other out. I think the root of the original question had something to do with paradox, and how it works communicatively. How paradoxes work before they are solved, when they are solved, or when they are never solved, but communication does not break down.  Could we anaylse the issue from the point of view of paradox analysis, and rhetorical devices as complement to semantics?

Thank you, Patrick. Yes: the root of my question is indeed that we use these combinations effortlessly in normal communication but that they can strike us as paradoxical when we start analysing them the way we usually analyse Adj+N combinations. I agree with Tibor that they're not intersective (i.e. the combination fake guns doesn't refer to things that are both fake and guns, since we're not dealing with guns here to begin with). I also agree with him that fake N instead refers to something that is claimed or perceived (by someone else than the speaker) to be an N but that (from the perspective of the speaker) is not an actual N since it lacks an important defining quality (e.g. the capacity to inflict a shot wound). This seems to be also the essence of the Lakovian-slash-mental space mapping account you, Matthew and John referred to.

Meanwhile, I was wondering whether the mechanism of coercion is relevant here at all. Maybe by calling an object a fake N, that object is coerced as a member of the class of N so that it can receive modification. Or is that wholly unnecessary? What I'm referring to here is some kind of cognitive coercing (though I don't know whether this is to be distinguished from purely 'linguistic' coercion as in There's cat all over the driveway). 

I'd be most interested in pointers to literature on this 'paradox analysis' you mention. Rhetorics certainly plays a role in the resolution of certain apparent paradoxes, such as praeteritio (e.g. I won't even mention that he's underage). 

I will think about literature (deconstructionism, polyphony, discourse analysis, critical discourse analysis...), but in the meanwhile, I certainly agree with you, Bert,  that coercion is a most worthwhile hypothesis to explore, at least as a factor. In this respect I find fake orgasm a good example. One I was thinking of is "fake laughter" (in a way, how can you fake laughter without laughing? and if you fake it long enough maybe you end up laughing for real, so where is the borderline here?). Thanks so much for an interesitng exchange and food for thought. I work in translation studies, and much is made of the virtues and vices of literal translation. But I think that not enough attention is paid to "problematic" literal interpretations and single-meaning readings of the source text; so, rather than literal translation as a problematic translational technique the focus shuold maybe turn more towards literal reading as a problematic reading technique for the purpose of translation.  What is in the balance is whether meaning and its construction is "out there" and objectively definable, as perceived in Saussurian semantic approaches, or whether meaning and its construction can be negotiated contingently, more relativistically, closer to hermeneutics in that respect. Translators and translation scholars are often confronted with this dilemma.

Hi Bert,

Pollard and Sag (1994) discuss the "alledged murderer", which is not a murderer either. What is needed is some sort of encapuslation: x, alledge(murderer(X))

For the "fake diamond" you will have an object that is fake and is supposed to look like a diamond. I guess this is similar.

Best

Stefan

Research gate has a bizarre way to display the answers, so I did not see Tibor's answer. My answer is similar, but at least there is a reference in it ...

As a belated answer to my own question: a recent article in Semantics and Pragmatics by Guillermo Del Pinal deals with these adjectives. The paper is very rich in content, references previous accounts and can be considered the state of the art.