Vowel mergers are some of the most well-studied sound change phenomena, particularly in varieties of English. But although sociolinguists, dialectologists, and phoneticians are all interested in providing accurate and precise descriptions of an individual speaker's participation in a near-merger (or near-split), the methods for doing so vary widely, especially for researchers analyzing naturalistic corpora. In this paper, we consider four current methodological approaches to representing and assessing vowel distance and overlap: Euclidean distances between averages, Pillai-Bartlett trace (Hay et al., 2006), mixed effects regression modeling (Nycz 2013), and the spectral overlap assessment metric (Wassink 2006). We compare the advantages and disadvantages of each by applying all four methods to three separate data sets. These represent low vowel realizations by speakers from three different studies of English variation: one undergoing merger (COT and CAUGHT in San Francisco, California), one undergoing split, for the same contrast (COT and CAUGHT among Canadians in New York City), and one undergoing split, but for a different contrast (TRAP and BATH among Scots in England). By comparing the similarities and differences between the data sets themselves, as well as the differing analytic motivations for quantifying speaker-specific vowel overlap, we conclude with practical recommendations.