158 Michael Fortescue
Note that the independent pronoun in (14) is in its subject form, not its
possessor form: the first-person possessor affixed to the first word has
subject properties, as Nakayama puts it, and the possession marker itself
produces a verbal expression. In such constructions the possessor is topic,
i.e. salient in the discourse, and treated as a grammatical argument of the
predicate expression. Nakayama argues elsewhere that the category of
syntactic ‘subject’ (and ‘object’), as opposed to semantic ‘agent’ (and
‘undergoer’), is in general difficult to maintain for this language (Naka-
yama 1997: 99ff.).
- Nootka phrasal affixes and zero-verbalization
Now it could be argued that the ‘suffixes’ attached obligatorily to the initial
constituent in the serialization type of construction are enclitic and that that
constituent is not ‘really’ the predicate of the sentence at all. But Naka-
yama argues that this is not the case, for whether the suffixes concerned are
inflectional or derivational they have properties that are un-clitic-like. Thus
in Nakayama (1994: 266) he compares the following sentences, all contain-
ing ‘phrasal affix’ –i: ‘make’:
(15) ča:pac-i:
canoe-make
‘He made a canoe’
(16) u-i: čapac
nice-make canoe
‘He made a nice canoe’
(17) mu:kw-i: u čapac
four-make nice canoe
‘He made four nice canoes’
As he argues, the ‘verbalizing’ suffix here takes a whole phrase as its base
(though it must morphosyntactically stand on the first word of the phrase,
whatever that is). Its scope properties are directly comparable to those of
such West Greenlandic affixes as -liur- ‘make’, which can also be added to
an NP leaving the modifying element stranded (in the instrumental case).^11
Whether one can jump from this to the claim that these suffixes are not en-
clitics at all is perhaps a matter of definition: they are at least not
prototypical enclitics, being lexically much weightier than clitic elements
usually are.