that a facilitation based on the semantic information of this kind did not occur,
or at least was not strong enough to make an impact on the reading time of
the verb.
In this experiment, they observed the locality effects in (5d), compared with
(5c). When there were both adverbial and dative NPs in the relative clause
structure as in (5d), the reading times for the verb in the relative clause were
longer. Levy and Keller argue that having a dative NP leads to the facilitation
while also adding an adverbial cancels out such a facilitation effect. They also
suggest that using a structure that requires a sufficient amount of working
memory seems to be an important factor for obtaining both effects at the same
time. In their paper, they reported another experiment in which they used
simpler structures, and they did not observe a slowdown such as the one in
(5d). Another consequence of their results is that a distance-based theory such
as DLT is not sufficient to account for the whole pattern of results.
Related to Levy and Keller (201 3), there is another study, Husain, Vasishth
and Srinivasan (2014), showing the emergence of expectations and locality
effects. In their experiment, they manipulated the strength of the expectation
and investigated how it influences the (anti-)locality effects in Hindi. Just
like in German, it has been shown previously that the additional constituent
before a verb reduces the processing cost of the verb in Hindi (an anti-locality
effect, e.g., Vasishth and Lewis, 2006). Husain et al. (2014) manipulated the
strength of expectations not by the presence of constituents in the structure,
but by the use of different object NPs, which varied in the degree of the
expectations they would invoke for a particular verb. They prepared high and
low expectation conditions, crossed with a distance factor (long vs. short).
The short version of the high expectation condition is illustrated below (only
the crucial part is shown). According to Husain et al. (2014) , in the example
of the high expectation condition, when placed in an appropriate context,
the object khayaal ‘care’ provides a strong clue for the identity of the verb
(in this case, rakhe ‘keep’), due to the high collocations of the noun-verb
sequence in Hindi. In the low expectation conditions, they used another
noun, gitaar ‘guitar’, as the object that does not have any strong relation
with the verb rakhe ‘keep’ although they are semantically fully compatible
(the meaning of the sentence will be “she properly keeps the guitar”). Fur-
thermore, they prepared the long version of the target sentences where an
extra adverbial, binaa kisi laaparvaahi ke ‘without any carelessness’, was
inserted after the object, so that the object was separated from the verb by
several words. They were interested in the reading-time patterns of the verb,
rakhe ‘keep’.
(7) High expectation, short
.. .vah apnaa khayaal achC se rakhe,...
she her care properly keep
“... she keeps good care of herself,.. .”
Make a good prediction 89