Pasted below is from Prof. Bruce Fraser of Boston University.
The terminology of this area is a huge mess. You are correct that I view DMs
as a subclass of PM, with Basic, Commentary, and Parallel markers being the
other three subclasses. But others use different taxonomies. Some take
Pragmatic or Discourse Particles to be the superset, while other take DMs to
be the same as Discourse Connectives (Blakemore). I haven't found anyone who
has dealt successfully with this mess. Below is what I have used as a way of
introducing a recent article. I hope it helps.
Particles are usually thought of as “the small words of language,” words
which are often a members of a syntactic class, but typically in their use
as particles does not reflect the usual function of their nominal syntax. I
start from the observation that there are (at least) two types of particles
in language: Sentence Particles (SP); and Discourse Particles (DP).
Sentences Particles are typically sentence-internal and contribute to the
semantic meaning of the sentence. Examples include Spanish clitics (e.g., te
in Yo quiere amarte), Japanese subject marker (e.g. ga), declension and
conjugation morphology (e.g. Cs, Cing, -ed), verb particles (e.g., up in
look up), Modal Particles (e.g., indeed) and Focus Particles (e.g. just).
Discourse Particles, in contrast, are typically sentence-initial and do not
affect the semantic meaning of the sentence of which they are a part, hence
do not affect the truth conditions of the sentence. They typically serve to
conceptually organize the discourse and in many cases, though certainly not
all, they may be omitted without a change in sequence interepretation. For
example, now, in example (1a),
(1) a) I think we are finished with the first matter. Now, let’s go on to
the second matter.
b) He should be better behaved. That is to say, he shouldn’t complain all
c) A: The movie is over by now. B: So, I suppose we shouldn’t go over
serves to orient the hearer to the change of material which follows, in
(1b), that is to say serves to notify the hearer that a reformulation is
about to occur, while in (1c) the so signals that second segment of the
sequence follows as a conclusion based on the content of the first.
There are several types of DPs:
(2) a) connective particles (e.g., and, but, so, however, furthermore, as
a result), which signal that the speaker posits a specific relationship
between two adjacent discourse segments, S1 and S2;
b) orienting particles (e.g., ok, now, let’s see, I mean, y’know), which
signal that the speaker intends to reorient the conversation;
c) pause particles (e.g., hummm, ah, uhhhh, well), which signal that the
speaker intend to hold the floor.
I am interested here in the first type of Discourse Particle, Connective
Particles, which are known as Discourse Markers (Schiffrin, 1987; Fraser,
1989, etc.) as well as other terms such Discourse Connectives (Blakemore,
2002), Discourse Operators (Redeker, 1992), and Cue Phrases (Knott), as well
as discourse particles, hearsay particles, phatic connectives, pragmatic
connectives, pragmatic particles, sentence connectives, stylistic disjuncts.
However, with some minor variations, they are simply alternative names for
the same functional class of formatives. I will refer to them as Discourse
Discourse Markers are defined as follows (cf. Fraser 2004 for a
(3) For a sequence of discourse segments S1 C S2, each of which encodes a
complete message, a lexical expression LE functions as a discourse marker
if, when it occurs in S2-initial position (S1 C LE + S2), LE signals that a
semantic relationship holds between S2 and S1 which is one of:
a) elaboration, where the DM such as and, furthermore, in addition,
moreover, for example, more precisely,...signals that S2 is an elaboration
of some aspect of S1
(John didn’t want to go and he was going to make everyone life
b) contrast, where the DM such as but, however, nevertheless, on the
contrary, still,... signals a contrast between S2 and some aspect of S1
(John tried to excel in everything, but Harry didn’t even try to be
c) implicative, where the DM such as so, thus, therefore, consequently,
because,… signals that S2 is implied by S1
(The water won’t boil, so we definitely are not going to have tea)
4. Temporal, where the DM signals the time of S2 relative to S1 such
then, when, before, after, now,…(We went to the movies. Then,
we went to bed.
In general on contrastive discourse markers (CDMs),
The quote is in an article called "The DM "but" Across Languages" to appear
in a volume not yet named. Just refer to it as "forthcoming."