Choice and complexity: In naturally occurring data, absolute complexity does not necessarily trigger relative complexity

This article interrogates two related assumptions widespread in many approaches to language: (1) languages do not like synonymy; (2) absolute complexity (i.e. the length of the grammatical description of a language) tends to be proportional to relative complexity (i.e. difficulty). Against this backdrop, we explore the link between syntactic synonymy (i.e., grammatical variation and optionality) and relative complexity (i.e., cognitive load) using methods from both corpus and psycholinguistics. We test two predictions: First, if synonymy avoidance is a design feature of human language, then grammatical variation should be sub-optimal and cause a measurable increase in production difficulty. Second, optionality will necessarily increase the absolute complexity of a language system. This increased absolute complexity will, in turn, increase relative complexity, i.e., cognitive load, also measured by increased production difficulty. Contrary to these predictions, analyses based on the SWITCHBOARD corpus of American English shows that the presence of choice contexts does not positively correlate with two metrics of production difficulty, namely filled pauses (um and uh) and unfilled pauses (speech planning time), not even when a typology of grammatical alternation type (insertion/deletion, substitution, permutation) is taken into account. These results challenge the view that grammatical optionality is sub-optimal and difficult for speakers, and that absolute complexity is necessarily proportional to relative complexity.