Buahahah, another reason to avoid preparing for my anthropology exam (and I feared I had exhausted them all...)
If we're talking about evidentials, I feel like I have to point out that evidentials are not restricted to assertions - some languages (although this is typologically rare) allow evidentials to tack onto questions and imperatives, and for this reason I prefer to use Faller 2002's broad definition of an evidential. She defines evidentials are as elements that "encode the grounds for making a speech act," which for the speech acts of the type 'assertion', are typically one's information source (Faller 2002:81). Just to frame things, Faller defines epistemic modality as referring to a judgment of possibility or necessity regarding the truth of a proposition (Faller 2002:81). The main difference she makes between evidentials and epistemic modals, as I read it, is whether or not an element encodes as opposed to implicates the relevant semantic material. (Whether or not this can be conflated with Matthewson et. al 2005's notion of 'fixed' versus 'context-varying' values is a headache I feel I should have figured out but still haven't). Basically, an evidential encodes one's grounds for making a speech act, and may or may not implicate judgments/evaluations regarding the the validity of the speech act. An epistemic modal encodes a judgment/evaluation, and may or may not implicate one's grounds for making the speech act. And, as Aislin mentioned, language can be sticky in that an element might encode both. So, if Aislin's in De Haan's camp, you can say that I'm camping out with Faller. I like her definitions.
Now, my issue is not so much how these semantic definitions hold up to other semantic definitions, but how these semantic definitions correspond with the syntactic tests used to categorize an element as either an evidential or an epistemic modal. Tests for evidentials often involve trying to embed them under some sort of propositional operator (such as my favourite propositional operator, negation), where dedicated evidentials are supposed to scope outside of the propositional content, and epistemic modals are ...well...tricky. Some of them can scope under negation (eg. English 'can') but some of them scope outside of negation (eg. English 'must')...but generally, if something scopes under a propositional operator, it's probably a modal (sufficient, but not necessary). At least this is the impression I get. My issue with this is that it seems to assume i) that a notion like syntactic c-command at LF correlates with semantic scope, and ii) that an element is either under the scope of a propositional operator, or outside of the scope of a propositional operator.
ii) is the one that makes me think the most. As noted above, things can encode more than one kind of semantic material (like English tense/aspect being conflated in the Perfect, etc.). So say you have a morpheme /MORPHEME/ that encodes both a semantic feature A and also another semantic feature B. What stops A from escaping the scope of a propositional operator while B stays within the aforementioned propositional operator's scope? Am I just a crazy person, or does this make sense to other people too?
On a side note...I'd also like to welcome Meaghan Fowlie to the blog!
And another side note: I am a dork, so when I saw this footnote:
**ie, by me, in my arrogant undergrad seminar papers. But by actual, non-proto linguists, as well
It made me think that Aislin was reconstructed from comparative research on extant linguists, and should really be denoted as *Aislin. (Man, linguists get a lot of mileage out of that asterix...remember phrase structure rules where a star AFTERWARDS means recursive???)
And since three is a good number, another side note: I totally want a linguistics t-shirt, but am torn between the several options of what to put on the t-shirt...whether to do an "I *heart* Saussurre", or something that could be taken entirely the wrong way by non-linguists like "I raise my diphthongs before voiceless obstruents"...