Two years ago, I ran a survey of semantic intuitions concerning empty names, that is, names that appear to be without a referent (e.g., 'Santa Claus', 'Superman').  Some of those who kindly participated in the survey asked me to post the results of the survey.  (Thanks again for your help, folks.)

The question was whether sentences and pseudo-sentences such as 'Santa Claus is fat', 'Santa Claus doesn't exist', and '#^*g~)# doesn't exist' are true, false, or have no truth value.  I found the results quite interesting.  Very briefly, (1) most people have no difficulty assigning truth values to most sentences containing empty names or words that look like empty names, but (2) they balk at assigning truth values to sentences that contain non-words in the place of an empty name.  However, (3) different people disagree on which sentences have truth values and in some cases, on which truth values certain sentences have.

I think these results (in combination with some other considerations) cause trouble for Millian theories of proper names.  A paper that presents my case against Millian theories (co-authored with Sam Scott) and in favor of Larson and Segal's theory of names is forthcoming in Canadian Journal of Philosophy.  Among other things,  the paper contains a discussion of ways in which we can recover what is said by uttering a sentence.  To this effect, we propose a "generalized cancellation test" that is a bit more general than Grice's cancellation test for implicatures.  Among the data presented in the paper is a table with the results of the above-mentioned survey.  

Some of the ideas and methods presented in this paper might be worth pursuing further, although I surely don't have time for that in the near future.

Incidentally, as Dan Sperber points out here, Mind and Language has recently published an entire issue on proper names.