This was originally posted to Facebook on October 24, 2012.
WYSIWYG (What You See Is What You Get): This is popular among semanticists. I have heard similar ideas expressed explicitly on at least three occasions, and it is implicit in many papers written by semanticists. The basic idea is that if one is analyzing a sentence, then one has the right to postulate syntactically the morphemes that one actually hears (or sees in writing and sign language), but one has no right to postulate other morphemes. Therefore, when figuring out the compositional semantics of the sentence, one assigns semantic values to the overt morphemes only (since these are the only ones there). Problem number 1: after more than 50 years of generative syntax it is 100% clear that WYSINWYG ("What you see is not what you get"). Simple examples abound. Ross' discussion of sluicing, later picked up by Merchant, show without a doubt that the sluiced site has to be deleted syntactic copy. A purely semantic approach to the phenomenon simply cannot work. Or take a simpler example, "John's house" vs. "our house". One could say that since there is no overt possessive 's in "our house" that it is not present. But a look at NP deletion falsifies that claim ("I didn't see ours"). It looks like in "our house" hat the possessive -'s has simply been deleted. Problem number 2: WYSIWYG presupposes that the semantic values of morphemes do not count for WYSIWYG. So even if one has huge ungainly semantic values for particular morphemes, it doesn't matter. So long as they are huge ungainly semantic values of overt morphemes. Of course, the other route would be to have some covert morphemes, and to distribute the components of the semantic values to these different morphemes, yielding a simpler more transparent compositional semantics. Problem number 3: WYSIWYG has the effect of inhibiting actual syntactic analysis. What tests would allow you to see if some morpheme is actually there? What is the full range of alternatives, permutations, uses of a particular construction that might allow one to probe the syntactic structure. If one says, WYSIWYG, then one limits oneself to a certain set of syntactic hypotheses, and hence limits oneself in formulating relevant data questions. All of this is relevant to a particular agenda for me. I am fully convinced by the early Katz-Postal idea that there are null morphemes all over the place, recently picked up by Kayne.