who gives what to whom, and how do we know
Eva Zehentner (University of Zurich - Switzerland)
Specifically, I suggest that disambiguation between recipients and themes likely did not present much trouble, based on the prototypical distributional properties of these arguments: while recipients tend to be animate, given, accessible, pronominal, definite, etc., themes usually have very different features, e.g. being low on the animacy scale. However, there is great overlap between recipients and agents regarding these aspects; at a time when constituent ordering was still relatively free, and SVO order was not predominant in the language, ambiguities may have given prepositional patterns formally distinguishing one of them an advantage. I test this hypothesis by means of a quantitative analysis of ditransitive tokens in a corpus of Middle English (PPCME2), coding the tokens on a variety of variables including features of the arguments, as well as case marking salience and aspects of constituent ordering. A behavioural profile/ multiple correspondence analysis demonstrates that while subjects and recipients clearly cluster together and contrast with themes; at the same time, however, a clear effect of subject-related factors on the choice between constructions is only partly confirmed by mixed-effects logistic regression modelling of the data. The results lend support to the idea that loss of specific formal disambiguation strategies is ‘compensated’ by an increase in other strategies if semantic cues are not sufficient; at the same time, the study showcases the complex interplay of different strategies (such as prepositional marking and constituent order), making it difficult to disentangle their individual impact.