question on Preferred Arg. Str

Östen Dahl oesten at
Tue Oct 12 14:59:38 UTC 2004

Martin Haspelmath wonders if the PAS constraints proposed by Du Bois are not
predictable from well-known statistical tendencies in discourse. 
In Dahl (2000), I argue that the "one lexical NP per clause" is indeed

"Du Bois (1987) argues for assuming a “one lexical NP per clause” constraint
in spoken language. However, in the G corpus, the number of clauses that
contain lexical NPs in both subject and DO position conforms exactly to the
prediction obtained by multiplying the frequency of lexical subjects with
that of lexical Dos — 75 or 2.4%. In other words, there is no basis here for
postulating an independent constraint that would work against combinations
of two lexical NPs in one clause, even if it may still be said that the
whole system is constructed in such a way that they will not be very
frequent. Note in particular that in the five hours of conversation
represented in the G corpus, there was not a single instance of the
linguist’s favourite sentence type — transitive sentences with two proper
names as in John loves Mary." (Dahl 2000, 50)

("the G corpus" refers to the 65000 word corpus of spoken Swedish that I was

- Östen Dahl


Dahl, Östen. 2000. Egophoricity in Discourse and Syntax. Functions of
Language 7(1): 33-77.

More information about the Funknet mailing list