floating clitics

[Matthew S Dryer]

I am interested in finding examples of what I call "floating clitics",
clitics which must attach phonologically to some other word, but where the
host phrase or word cannot be defined in terms of some specific position
in the clause (like second position or final position) or in terms of some
specific host (like subject or verb).  My impression is that this sort of
clitic is not uncommon, but does not seemed to be discussed much in the
literature on clitics that I have looked at.

One example that I am aware of is discussed by Spencer (1991: 369 - 374),
what he calls "auxiliary formatives" in Polish which code person, number,
and gender of the subject and which can attach to various clausal
constituents, but not to the negative particle or to prepositions, and not
to constituents that follow the verb.  A second example is an emphatic
clitic in Kannada (Sridhar 1990: 257), which encliticizes to any of
various constituents.

I would argue that these two examples represent two types.  In the Kannada
case, the emphatic clitic attaches to a host with which its meaning is
being associated, and the nature of emphasis (or focus) is such that
various different sorts of constituents could be emphasized.  In the
Polish case, however, it is not clear what determines what the clitic
attaches to and there does not seem to be any sense in which the meaning
of the clitic is specifically associated with the constituent to which it
attaches.

I believe that I have seen various other instances of clitics like the
Kannada one, where the meaning is some sort of focus, such as an
interrogative morpheme, where the position of the interrogative morpheme
is a function of the focus of the question.  But I am primarily interested
in examples like the Polish case, where there is no apparent natural
connection between the meaning of the clitic and the particular host.

My question overlaps with Rachel Nordlinger's recent query about
tense/aspect morphemes on nouns, since clitics of the sort that I am
looking for that code tense/aspect would count as one type of tense/aspect
marking on nouns.

Thanks,

Matthew Dryer
dryer at acsu.buffalo.edu