This thesis is about linguistic variation in swearing and its consequences for how speakers are socially evaluated. Abundant research has established that, beyond its perception as rude or impolite, swearing is hugely socially meaningful in a variety of ways (Stapleton, 2010; Beers Fägersten, 2012). Swearing has been shown to index solidarity (Daly et al., 2004), intimacy (Stapleton, 2003), differing forms of masculinity (De Klerk, 1997) and femininity (S. E. Hughes, 1992), honesty (Feldman et al., 2017), believability (Rassin & Heijden, 2005) and lack of intelligence (DeFrank & Kahlbaugh, 2019), among other traits. The activation of these social meanings also depends on language-external factors such as speaker gender (Howell & Giuliano, 2011), ethnicity (Jacobi, 2014) and social status (T. Jay & Janschewitz, 2008). What has not been established is whether this also depends on language-internal factors such as pronunciation, word formation or sentence structure. This thesis investigates the effect of variation from three different domains of language - phonetics, morphology and semantics/pragmatics - on social evaluation of a speaker. To do so, the thesis takes an experimental approach using the variationist sociolinguistic framework. For variation in each domain, two experiments were used to test for different levels of awareness, following Squires’s (2016) approach for grammatical variation (see also Schmidt, 1990). One experiment tested whether people perceived the variation, while a second tested whether people noticed the variation in the process of social evaluation; the concepts of perceiving and noticing roughly map to the Labovian concepts of the sociolinguistic indicator and marker respectively (Labov, 1972). At the level of phonetics, variation in the realisation of variable (ING) in swearwords (e.g., fucking vs fuckin) was first tested using a variant categorization task, revealing that listeners have an implicit bias towards the velar [IN] variant when hearing swearwords, compared to neutral words and non-words. An auditory matched-guise task then revealed that this same bias affects how listeners extract social information from (ING) tokens attached to swearwords in relation to social meanings typically associated with the variable (Schleef et al., 2017). This result suggests that, rather than pronunciation affecting how swearwords are socially evaluated, swearwords can affect how other phonetic sources of social meaning are evaluated.