2 Expectations
Koniecz ny and Döring (2003) investigated whether having an additional con-
stituent dependent on the verb leads to a faster reading time on the verb. In
German, a tensed verb is placed clause-fi nally in the embedded clause, and thus
all constituents, both arguments and adjuncts, appear before the verb. In example
sentences in (2), a clause was embedded in the noun complement structure
with an NP die Einsicht ‘the insight’. In (2a), the embedded verb verkaufte
‘sold’ was preceded by three constituents (a subject, a dative object and a direct
object with a modifying PP), while in (2b), there are two constituents (a subject
with a genitive NP and a direct object with a modifying PP) before the verb.
The only difference between the two examples was the case-marker of the
determiner dem/des ‘the-dative/the-genitive’ of the second NP.
(2) a. NP-dative, NP-modifying PP
Die Einsicht, dass der Freund dem Kunden das Auto aus Plastic
the insight that the friend to the client the car out.of plastic
verkaufte,...
sold,...
“the insight that the friend sold the client the car made from plastic.. .”
b. NP-genitive, NP-modifying PP
Die Einsicht, dass der Freund des Kunden das Auto aus Plastic
the insight that the friend of the client the car out.of plastic
verkaufte,...
sold,...
“the insight that the friend of the client sold the car made from plastic.. .”
In an eye-tracking study, Konieczny and Döring found that the mean reading
time (regression path durations^1 ) for the embedded verb in (2a) was about
200ms shorter than in (2b).
These observations indicate that having more information helps the reader
make more precise predictions about the upcoming verb. In the NP-dative
condition, the reader may predict the verb to be a ditransitive verb compatible
with the dative NP. When the reader saw the right kind of verb that the reader
predicted, the processing cost for the verb was low. In the NP-genitive condi-
tion, on the other hand, the reader was able to predict that, given the presence
of the accusative NP, the verb must be one of the transitive verbs, but the
number of possible verbs was still large, compared with the NP-dative condition.
The verb in the example was compatible with the prediction by the reader, but
there were other types of verbs that were also compatible with the prediction.
It is plausible that the processing cost of the verb was higher when the reader
did not have a specifi c verb in mind and was open to other possibilities.
In the same experiment, there were in fact two other conditions in which the
V-modifying PP aus Fraude ‘just for fun’ was used instead of the NP-modifying
PP. No statistically reliable effect of the PP-type was found. It is then clear that,
although adding a V-modifying PP makes a more complex VP structure, this
Make a good prediction 85