Evidentials of Inference

[Koontz John E]

In the current issue of IJAL:

James, Deborah and Clarke, Sandra.  2002.  The encoding of information
source in Algonquian:  evidentials in Cree/Montagnais/Naskapi.  IJAL
67.3:229-263.

p. 247

"It has been noted by a number of writers (e.g., Comrie 1976, Anderson
1986, Willet 1988, and Bybee, Perkins and Pagliuca 1994, and see the
articles in section I of Guentche'va 1996) that evidentials which signal
inference frequently derive historically from morphemes with present
perfect meaning.  Such a development has taken place, for example, in
Latvian, Albanian, Bulgarian, Turkish, Georgian, the Tibetan languages, a
number of Uralic languages, Inuktitut, Tucano, and Chinese Pidgin Russian
(Comrie 1976:110, Willett 1988:79, and Bybee, Perkins, and Pagliuca
1994:95).  It seems most likely that perfects develop into inferentials
via a "resultative" stage (Bybee, Perkins and Pagliuca 1994:96).  With
perfects, there can develop over time an increasing emphasis on the result
of the past action, that is, on the way in which that action is relevant
to the present.  This resulting state exists because of the past action.
In this shift of focus, the perceived resulting state can come to be
viewed as evidence on the basis of which the speaker draws an inference
about what the past action must have been which brought about the
resulting state.  Thus a marker of the perfect changes into an evidential
signaling inference.  This is one of the "universal paths of development"
posited by Bybee, Perkins and Pagliuca (1994:95-97)."

I (JEK) might add to this that any given point in this development the
data are going to represent a range of possibilities along
this progression, or at least some somewhat complex compromise between
successive "pure" cases.  This will be especially true if the data are due
to a variety of speakers of different ages and backgrounds, but I think
even a single speaker might be expected to produce a mixture of more or
less progressive or conservative uses, perhaps with special conditioning,
e.g., resultative use with an inferential reading for verbs of motion, to
make up a plausible sounding situation.

I'd have to add that I'm not positive there is any reason to regard the
Dhegiha inferential (evidentials) as old present perfects, in any
classical Indo-European sense.  But it might make sense to regard then as
old resultatives, e.g., 'it stands that/there is a pile of results which
someone X-ed', which becomes 'it stands that/there is a pile of results
such that someone must have X-ed'.  The positional sense of 'stand/be a
pile of' would be a secondary reading, perhaps, of the from earlier *ra-he
with some other reading.

JEK