What is the difference between evidential, epistemic, and evaluative fragments?

Question

Following a discussion about introductory expressions like “It is not the case that…” [1], I was found out that these expressions fall into the category of evidential/epistemic/evaluative fragments proposed by Thompson (2002).

Intuitively, I can get that 1 is an evaluative fragment, whereas 2 is either an evidential or epistemic fragment.

1. It is interesting that...
2. It is possible that...

However, I'm not sure about how linguists understand the evidential/epistemic distinction. (Nor am I 100% sure that I understand their evidential/epistemic vs evaluative distinction.)

I would like to have definitions and examples for each of these three kinds of fragments.

• Thompson, Sandra A. 2002. “Object complements” and conversation: towards a realistic account. Studies in Language 26(1). 125–63.

Answer 1

Basically, the definitions usually used in the syntax and semantics literatures are:

• If a linguistic form expresses evidential meaning, you are talking about the source that you got the information from, regardless of how sure you are about it
• If a linguistic form expresses epistemic meaning, you are talking about how sure you are about the information, regardless of where you got the info from
• If a linguistic form expresses evaluative meaning, you are talking about what you feel about that information (whether it's good or bad, etc.)

These are the definitions that Thompson assumes in the paper, and in the case of the English evidential/epistemic/evaluative fragments, I think the distinctions are quite clear-cut. (There are, however, a lot of edge cases in other places, and if you dig into the semantics literature further, you'll find a lot of debate about whether something is best treated as evidentiality or epistemic modality, whether evidentiality is a part of modality, etc.)

Note that the pragmatic/discourse literature on stance is a bit less strict about these definitions, and you will frequently find people subsume evidential into epistemic or vice versa.