Ford, Laura2026-07-022026-07-022026-06-302026-06-30http://hdl.handle.net/1993/39834This thesis uses mixed-methods to investigate variation in the lexical variable ‘DIE,’ focusing on the variants DIE, PASS away, PASS on and PASS. First, taking a qualitative approach, it analyzes attitudes towards variation expressed in a corpus of Reddit commentary made up of four threads and 254 comments. Using Thematic Analysis and a language attitudes framework, I identify and explore four themes centred on variation of ‘DIE’: nuances of variant meaning, the relevance of emotions, communitive context, and associations with macrosocial categories. Next, guided by the results of the qualitative analysis, I follow quantitative variationist methods to analyze ‘DIE’ as a sociolinguistic variable. To capture the variable in use, data was collected from the spoken portions of the Canadian Business and Current Affairs database and the Corpus of Contemporary American English, coding 1000 tokens for linguistic and social factors that may influence variant choice. Multiple statistical methods including random forest analysis, conditional inference tree analysis, and logistic regression modeling allow exploration of the influence and interrelationships of many potential predictors. Of these, DEATH TYPE, PROGRESSIVE, DEATH HAS OCCURRED and PERSONAL RELATIONSHIP stand out. Violent or sudden death, deaths that have not occurred and possibly use of the progressive favoured the direct language variant DIE, while contexts where the speaker had a personal relationship with the deceased favour use of the euphemistic variants PASS away, PASS on and PASS. Both the qualitative and quantitative studies highlight the importance of the metaphorical nature of PASS-type variants to factors that condition variation. Together, the two studies characterize DIE as informative, bodily focused, and impersonal and PASS away, PASS on and PASS as specialized variants recruited principally for relationship-focused communication.englexical variation‘DIE’Canadian Englishmixed methodscorpora