5.6 Development 163development” and suggests that it would be fruitful “to investigate the extent to
which [epistemics] intersect with evidentials” in their development.
In the case of fi rst- person epistemic parentheticals, the route of semantic
development that I propose is the following:
act of cognition > mode of knowing (evidential) > (un)certainty (epis temic)^32In the fi rst stage, verbs meaning ‘think,’ ‘believe,’ and ‘guess,’ used to denote
a mental act, come to be used to denote a state of mind. This development of
mental- construct evidential markers from verbs of cognition is analogous to
the development of hearsay, sensory, and inferential evidentials from verbs of
perception meaning ‘hear,’ ‘see,’ and ‘feel’ (see Anderson 1986 : 278– 286; cf.
Hanson 1987 ). While Anderson ( 1986 : 286) sees “weakening and generaliza-
tion” of meaning in this shift, Matlock ( 1989 : 219– 221; see also Willett 1988 :
80) considers it metaphorical, based on the underlying metaphor “knowing
is seeing.” The development of mental- construct evidentials does not involve
a comparable metaphorical transfer from concrete sensory to abstract men-
tal domain, but there is a clear increase in subjectivity , in which the expres-
sion comes to express a subjective belief state or attitude toward the situation
(Traugott and König 1991 : 209). Traugott ( 1995b : 39) notes further that in
expressions such as I think , “the subject is losing referential (objective) prop-
erties, and becoming simply the starting point of a perspective,” a shift from
subject of proposition to subject of utterance.
The second stage in the process, from evidential to epistemic, seems to be
an extremely clear example of the conventionalization of invited inferences.
Evidentiality, defi ned in a narrow sense as encompassing modes of knowing
and sources of evidence, invites inferences concerning reliability and confi -
dence. That is, evidentials imply uncertainty but do not necessitate it. Like
all inferences, these expressions of uncertainty are defeasible (Levinson
1983 : 114). For example, although a hearsay evidential would normally imply
some degree of uncertainty ( I hear he’s coming, but I’m not certain ), the impli-
cation of uncertainty can be canceled ( I hear he’s coming, and I’m certain he
is ); such cancelations are also possible for the inferences arising from mental-
construct evidentials, e.g., I {guess , am guessing} he’s coming to the party,
and I’m quite certain he will. If, as I have argued above, the primary – and
conventional – meaning of the fi rst- person epistemic parentheticals in Middle
English is epistemic uncertainty, not evidential nonactuality, then there has
32 As and Boye and Harder (2007: 593) point out, the “act of cognition” meaning corresponds to
the “lexical” and discourse primary meaning they attribute to these forms, and the “evidential/
epistemic” meaning corresponds to their “grammatical” and “discourse secondary” meaning.
See further Section 5.1.