Investigating the Effects of Transparency and Ambiguity on Idiom Learning.

The purpose of this thesis was to learn how transparency and ambiguity affect idiom learning. Experiment 1 was a norming study to measure the transparency, ambiguity, and familiarity levels of translated French idioms into English. Experiment 2 was a training study where 25 native English speakers learned 32 of the normed idioms from Experiment 1. The procedure was distributed over three days and included two learning sessions and one testing session. In addition, each participant completed individual difference tasks for working memory, creativity, and figurative language proficiency. We ran Linear Mixed Effects Regression models which rendered a significant effect of transparency on performance. Given the ceiling effects from Experiment 2, we conducted Experiment 3 which differed from Experiment 2 in that it reduced the amount of time subjects spent learning and practicing the idioms and also included a semantic relatedness test. Experiment 3 showed a significant main effect of transparency and an interaction between transparency and the type of test such that performance was better for high transparency idioms on a recall test but worse on a semantic relatedness test. Across the two training experiments, we found neither an effect of ambiguity nor any significant relation between individual difference tasks and performance.