3.4 Accounting for the Change 89persistence is evident in the continued semantic meaning of ‘formerly’ in the
new adjectival form. Layering exists as well, though it is whilom that consti-
tutes the older layer, with newer forms such as then or former replacing whilom
or quondam. Divergence holds only weakly because adverbial whilom is now
shown as “obsolete” or “archaic” in its adverbial uses in the OED (s.v. whilom ,
def. 1 and 2a; also see the examples in (6) above). Finally, the fact that the
change occurs in a highly constrained discourse context – namely, in the con-
text of nouns of position, role, or rank – agrees with current thinking about
grammaticalization (see, e.g., in Traugott 2003a ).
Semantically, the development of adjectival whilom is inconsistent with gram-
maticalization, which involves semantic expansion (Himmelmann 2004 : 33) and
the development of polysemies. In this case, however, we see semantic contrac-
tion, from a more general meaning of ‘formerly, once’ (in respect to an action or
state of being) to ‘former,’ narrowly applied to a person, and more specifi cally to
the person’s role. Semantic narrowing is inconsistent with grammaticalization,
as Hopper and Traugott conclude (2003:103): “In sum, as grammaticalization
progresses, meanings expand their range through the development of various
polysemies ... The important claim should not be that bleaching follows from
generalization, but rather that meaning changes leading to narrowing of meaning
will typically not occur in grammaticalization.”
Hopper ’s principle of decategorialization is also problematic. Not only does
adjectival whilom fail to acquire all of the morphosyntactic characteristics of
the new category – i.e., the ability to occur in predicate position, be infl ected for
degree, be modifi ed by intensifi ers – but the change from adverb to adjective con-
tradicts the (much debated, but widely believed) principle of unidirectionality in
grammaticalization. I will briefl y discuss the concept of unidirectionality before
moving on to alternative accounts for changes affecting whilom.
3.4.1 Aspects of Unidirectionality
The view of grammaticalization as unidirectional is propounded by most
scholars.^18 Lessau ( 1994 : 885) sees unidirectionality as “one of the strong-
est claims about grammaticalization,” while according to Hopper and Traugott
( 2003 : 99), grammaticalization is, from a diachronic perspective, “prototypi-
cally a unidirectional phenomenon.”^19
18 Discussions of unidirectionality form part of almost every treatment of grammaticaliza-
tion: see, e.g., Heine , Claudi , and Hünnemeyer (1991: 4– 5, 212, passim), Bybee , Perkins , and
Pagliuca (1994: 12– 14), Lessau ( 1994 : 885– 892), Lehmann ( 2002a : 17), Hopper and Traugott
( 2003 : Ch. 5), Heine ( 2003a ), Himmelmann ( 2004 ), and Börjars and Vincent ( 2011 ).
19 The principle of unidirectionality is inherent in more recent approaches to grammaticaliza-
tion. For example, Himmelmann ( 2004 ) conceptualizes grammaticalization as involving “host-
class expansion,” syntactic context expansion, and semantic- pragmatic context expandion. In