Bayesian vocabulary tests

Abstract

This paper proposes a new way of looking at productive vocabulary in L1 and L2 speakers. An experiment was conducted where 160 participants provided six words for five different picture prompts they were presented with. Data from this minimal vocabulary test was analysed using Bayesian statistics in order to decide whether a set of responses were generated by an L1 speaker or by an L2 advanced learner. Results obtained provide some interesting insights into the viability of minimal vocabulary tests (small sets of words can carry large amounts of information on vocabulary use), as well as some indications of how Bayesian methods could help us explore productive vocabularies of L2 speakers at different proficiency levels.