Matthewson 2004's paper on semantic fieldwork distinguishes three kinds of judgements:
- grammaticality judgements (i.e., syntactic/morphological well-formedness),
- truth-value judgements (i.e., true or false in a given context), and
- felicity judgements (i.e., acceptable or not acceptable in a given context).
The discussion of (in)felicity judgements suggests that they deal with a speaker's failure to make sure the context satisfies either
(i) (lexically-encoded) presuppositions, or
(ii) Gricean Maxim-induced implicatures.
For instance, she uses the following example to contrast truth-value judgements from felicity judgements.
(1) Context: Two cats are in the room, and they are both asleep.
a. The cats are awake.
b. The cat is asleep.
(1a) is FALSE, because the cats are both asleep. (1b) is INFELICITOUS because 'the' presupposes something like: the entity referred to is the maximal entity that satisfies the NP restriction in the context, and the maximal entity in the context is both cats, not a single cat
An example of infelicity via failure to satisfy Gricean Maxims would be something like (c):
c. One of the cats is asleep.
(1c) is TRUE, but INFELICITOUS because it implicates that the other cat is awake, and this violates the Maxim of Quantity.
If you consider presuppositions to be grammatically encoded (which I think is the standard view, eg., Heim 1983), then instances of infelicity via presupposition failure would be considered a subset of the 'grammatical errors.'
Similarly, if you consider implicatures to be grammatically encoded (not the standard view, but cf. Chierchia, Fox and Spector 2008 on 'scalar implicatures'), then instances of infelicity via Gricean-violations would also be a subset of 'grammatical errors.'