In this talk, I discuss a series of experiments and simulations that investigate the role of vowel and consonant categories in first (L1) and second language (L2) perception. The talk is roughly split into three parts. In Part I, I present a computational model used to assess categorical effects in native language speech perception. By fitting the model to behavioral data from past experiments, I show how our model is able to capture a range of categorical effects (e.g., stop consonants, nasal consonants, fricatives, vowels), suggesting that all share a common cognitive mechanism. In Part II, I extend the model to second language perception to examine how categories change over the course of learning. Specifically, I investigate how English speakers learn Russian vowels, and use our model to extract optimal parameters (means and variances) for a newly learned Russian vowel category. In Part III, I discuss how phonetic training may improve L2 vowel learning. Using insights from infant category learning, we developed a novel intervention for adults, in hopes that our investigations will eventually lead to directly applicable pedagogical tools.