MANDARIN CLAUSE LINKAGE 217cations to be attatched to a small number of different complement construc
tion types. This approach will not make the necessary distinctions, how
ever, as shown by the difference between (36a-c) above and the sentences
in (37) below:
(37) a. Wo ting fán le nà shöu gè.
I listen irritated ASP that CLF song
"I'm sick of hearing that song."
b. Wo ting fán le.
I listen irritated ASP
"I'm sick of hearing (it)."
e. Gè ting fán le.
song listen irritated ASP
"The song got listened to to the point of irritation." (i.e.
someone being irritated) but not *"The song got listened to
to the point of being irritated." (i.e. the song being irritated)
Since (37a) and (36a) are identical in terms of syntactic structure, no
specification of argument structure based on syntactic structure can explain
why (36c) is grammatical but (37c) (second reading) is not. It can only be
explained in terms of the semantic structure of the V 2 in each construction:
the V 2 in (36a) pò "to break" takes an inanimate patient, while the V 2
in (37a) fan "to be irritated" takes an experiencer. This difference has
nothing to do with the syntactic structure of the CR construction, nor with
the semantics of the Vx these verbs are connected to, they are independent
properties of the verbs pò and fan that are unaffected by their entering into
the CR construction. This constitutes strong evidence that verbs in a CC are
subject to the same semantic processes and constraints as they are when
occuring separately as main verbs, or when occuring in a SVC. In other
words, verbs in a CC (except in the cases of those V 2 s that have been par
tially grammaticalized and bleached of some of their semantic content) lose
none of their semantic content by virtue of occuring in this construction,
and it will be argued below that the syntactic differences that do exist are
regular effects of the tightness of clause juncture that they occur in, not
because of any qualitative difference between CCs and other kinds of verb
serializations.