5dez 2021
00:00 UTC
#linguistweets
#abralin

Redundancy: do dutch and english speakers like it?

We investigate redundancy –specifically, redundant use of argument disambiguation strategies – as a potential case of competition between production efficiency and robustness: While redundancy should be dispreferred in a system geared towards maximal efficiency, redundant marking may have advantages from a perception perspective due to increasing robustness of transmission in noisy channels.

We test these two opposing hypotheses in a specific case of argument disambiguation, viz. strategies to distinguish between agents and recipients in Dutch and English ditransitive clauses. These include (i) prepositional marking of the recipient (as in e.g. They gave a book to the librarian), (ii) formal differentiation (by means of distinctive subject vs object forms of personal pronouns, as in They gave me a book vs I gave them a book), (iii) subject-verb agreement (e.g. The student gives the librarians a book vs The librarians give the student a book), and (iv) constituent order (e.g. agents in pre-verbal vs recipients in post-verbal position in SVO languages). A further strategy, namely exploiting biases in semantics and pragmatics of the arguments, such as prototypical expectations about animacy, does not hold in this particular case – agents and recipients are both typically animate, given, and definite, among other things.

Analysing ditransitive tokens from corpora of Present Day Dutch and English, we show that redundant marking, i.e. using more than one disambiguation strategy in a single utterance, is the default. At the same time, however, we find that redundancy is constrained: a large majority of cases exhibit double marking, while triple or even quadruple marking is rare. Additional support for a preference for redundancy within limits comes from diachronic data from the history of English – periods of greater redundancy are temporary, and redundant marking is soon reduced. We take these results to confirm both of the initial hypotheses, with the extent of redundant marking indicating a balance between efficiency and robustness.