I know, eh?
So there's been all kinds of talk at Language Log regarding the Canadian tag particle eh?. It begins with Mark Liberman discussing a query he received from a reader regarding the particle in relation to modal and affective tags. In trying to find an answer he found a paper by Elaine Gold documenting sociolinguistic attitudes and native-speaker judgements about different functions of the particle.
For the record, tag eh can have both modal and affective readings. That's all I can offer for this issue -- whether they have slightly different intonational contours would require lots more data than I have access to.
Mark suggests that it may not be appropriate to rely on native speaker judgements for this kind of discourse element, at least for anything but the documentation of sociolinguistic attitudes about it. Surveys, though very useful, are not by themselves an adequate way to study such patterns of usage. Instead, actual discourse is a better place to look for data on these patterns.
I have to agree with this point, because of the risk that speakers would under- or over-report how much they've heard it and used it in various functions. (Gold acknowledges this issue in her paper). One issue I would have brought up had I attended that CLA was the interrogative usage, in which the particle is added to a sentence which is already a question with auxiliaries inverted. Gold's example is What are they trying to do, eh? Ignoring for a moment Mark's advice about native speaker judgements, I say this is an impossible usage of the word, unless the question is rhetorical.
Mark follows up, posting about electronically available transcripts of discussions from the Ontario provincial parliament, including instantiations of eh. Then, in Part 3, he comments upon an apparent "filled pause" from one of these transcripts, in which the particle seems to function like "um":
Mr Murdoch: [...] A couple of other ones: the stockyards, the money you talk about, is that the province's money? It is, eh, the money that you're -- who owns them?
I just find it hard to believe that the eh in this utterance is [both a pause-filler and] the same discourse item as the tag and narrative eh. There are at least two alternatives: it could be that the addressee nodded in response to the first question. Murdoch acknowledges the nod by saying "It is, eh?" and continues to the next sentence (starting at The money). In that scenario, it's a tag [and not a filler]. Or, it could be that the filled pause is actually a lax [ɛ], which is difficult to spell any other way, but which is distinct from the tense vowel of the tag (it would also have different intonation). Really, the only way to be sure is to have an audio recording (or very narrow transcription) of the exchange.
Part 4 discusses a parallel with Japanese ne, which has many similar functions, including a stigmatized narrative use. The narrative usage for eh is an interesting one, and I have read in other sources besides Gold's paper that it is one of the innovative Canadian functions of the marker. I haven't thought much about its relationship with discourse structure, but I'm going to hypothesize that it acts as a focuser, but with post-focus position. That is, while focuser-like sets off and highlights the following phrase, focuser-eh sets off the preceding phrase. E.g.: "Not an easy thing to talk about, eh, but you might get the drift". ( = "Not an easy thing to like, talk about, but you might get the drift").
I'm interested now in the intonational contours of eh, especially regarding (probably minute) differences across functions. There is always a drop in tone on the word before eh, and eh then has a rising contour - but the degree of the drop or the end point of the rising contour might differ slightly across modal, affective, imperative, and narrative functions.
Unlike ne (or huh), eh cannot ever have falling intonation, even in its narrative use. Again, I say this with the caveat that I'm weighing in with native speaker judgements rather than with data from recorded discourse. Which I can't get right now, as my usage drops to near zero when I'm not in the company of Canadians or close friends. (Simply because using it triggers an amused response that distracts from the intended conversation).
[Update (15 minutes later): Heidi Harley has already posted some similar thoughts on the eh tokens from the Hansard transcripts.]