458

aspiration. Other variations in phonological shape are
attributable to different processes of morphological change
with preservation or loss of different (now submorphemic)
recurrent partials. In two tenses, D-1 and D-2, exten-
sive variation results from the apparent falling-together
of two separate paradigms in each tense. Still other vari-
ations in shape are due to processes of analogy at work
within a dialect, resulting in symmetrical variation

across all its tenses. The most striking instance of

this is the shape of 3+4 suffixes in all tenses elicited
for Juli. All have allomorphs frozen to the NI suffix
-chi and thus homophonous, in that person/tense, with

NI compound verb forms.

Not only is there variation in shape across
dialects; there is considerable variation in allomorphs
within dialects, and a larger sample of speakers would
probably turn up more forms. While most Aymara speakers
have a receptive competence in inflectional variants
they do not use themselves, recognizing them as the way
some people talk, not all variants are equally accept-
able. What belongs in one tense in one dialect may
belong in a different tense in another. For example,
speakers who have only /-fiani/ for 4+3 F reject the use
of /-tana/ for it, saying it means a past action, not
future. There are also some overlaps of person and

tense, a given suffix being used for a certain subject,