Early Middle Japanese is one of the oldest attested stages of Japanese.
Its rich legacy consists of several literary works from the Heian era (7
th
to 11
th centuries), some of which are still appreciated and widely read
today. Despite a long tradition of research both within and outside
Japan, quite a few details of the language remain incompletely
understood. The present study addresses a long-standing question in
the verbal domain of Early Middle Japanese, namely the semantics of
the two so-called “past tenses” in
-ki and -ker-. I tested the major
hypotheses regarding their use by means of qualitative, corpus-based
methods. Specifically, I trained a machine learning algorithm to
predict which is likeliest of
-ki and -ker- given a set of grammatical and
semantic variables. Analysis of the results indicates that the suffixes
likely embody a contrast between witnessed and non-witnessed past
tense. It is also possible that mirativity—the grammaticalized
expression of surprise at learning something unexpected—and aspect
influence the choice of past tense suffix.
2018.