It is not always easy to distinguish empirically between evidential markers and epistemic modals. Tense and aspect markers can also be a problem, because they too can have secondary evidential functions or associations. Perfect aspect in particular often carries an indirect evidential connotation, and indirect evidence markers frequently develop out of perfect aspect markers.8 For example, in Iranian Azerbaijani (closely related to Turkish) the suffix -miş is polysemous between an older perfect sense and a more recent evidential sense.9 We can see that the two senses are distinct in the modern language, because they can cooccur in the same word as seen in (5).
(5) zefer qazan-miş-miş-am
victory win-prf-indirect-1sg
‘reportedly I have won’ [Noah Lee, p.c.]
So then, when we encounter a grammatical marker which seems to indicate source of information in at least some contexts, but has other functions as well, how can we decide what to call it? In other words, how do we determine its “primary function”? The key is to search for contexts where the expected correlation does not hold, so that the two possible analyses would make different predictions.
David Weber (1989: 421ff.) compares his analysis of the Huallaga Quechua evidential clitics with an alternative analysis which treats them as validational markers, that is, indicators of the speaker’s degree of commitment to the truth of the proposition being expressed. The choice between these two analyses is not immediately obvious, because there is a correlation between source of information and speaker’s degree of commitment. As we have noted, a speaker is likely to be more certain of knowledge gained through direct experience than of knowledge gained through hearsay or inference. In many contexts the direct evidential =mi (which is optional) can be used to indicate certainty; and hearers may sometimes interpret the hearsay evidential =shi as indicating uncertainty on the part of the speaker.
However, when there is a conflict between source of information and degree of commitment, it is source of information that determines the choice of clitic. For example, if someone were to say ‘My mother’s grandfather’s name was John,’ the direct evidential =mi would be extremely unnatural, no matter how firmly the speaker believes what he is saying. The hearsay evidential =shi must be used instead, because it is very unlikely in that culture for the speaker to have actually met his great-grandfather. Similarly, in describing cultural practices which the speaker firmly believes but has not personally experienced (e.g., ‘Having chewed coca, their strength comes to them’), the hearsay evidential is strongly preferred.
The general principle is that when we are trying to identify the meaning of a certain form, and there are two or more semantic factors that seem to correlate with the presence of that form, we need to find or create situations in which only one of those factors is possible and test whether the form would appear in such situations.
8 Izvorski (1997); Bybee et al. (1994).
9 Lee (2008).
