General Discussion

Trying to express anger using a swear word full of gentle, soft sounds... would be the verbal equivalent of angrily trying to slam a door fitted with a compressed air hinge.

~ Rebecca Roache (2016)

Swear words have a unique linguistic power. Swearing in public is illegal in many countries and profanity is a major target for censorship in the arts and entertainment industries (Bergen, 2016). Swearing elicits physiological responses such as elevated heart rate and increased galvanic skin response (Bowers & Pleydell-Pearce, 2011; Buchanan et al., 2006; Harris et al., 2003). Moreover, swearing aloud increases tolerance to pain (Stephens et al., 2009; Stephens & Robertson, 2020; Stephens & Umland, 2011) and boosts physical performance (Stephens et al., 2018).

What gives swear words their potency? Part of the answer, of course, may lie in what these words literally refer to: after all, the usual suspects include taboo topics such as excretion and sex. However, the sounds in swear words may also play an important role.

Our findings indicate that not all sounds are equally suitable for profanity. In an initial, pilot study we explored statistical regularities in the sounds of swear words across a set of typologically distant languages. The most promising candidate for a universal phonemic pattern in profanity to emerge from this analysis was the absence of approximants (sonorous sounds like lrw and y). Study 1 confirmed that native speakers of various typologically distant languages were relatively unlikely to identify words containing an approximant as swear words. It may be that approximants are sound-symbolically associated with calm and contentment, and so are unsuitable for giving offence (Nielsen & Rendall, 20112013; Yardy, 2010). In Study 2 we found that sanitized versions of swear words – minced oaths – contain significantly more of these sounds than the swear words they were derived from. According to Hazen (2020), minced oaths allow “for restrained fist-shaking at the universe”. Our findings suggest that approximants are a relevant “restraint” – the verbal equivalent of fitting a compressed air hinge to a door (Roache, 2016).

Though we focused primarily on approximants, future studies may identify other phonemic groups particularly suitable or unsuitable for profanity. For example, studies with larger samples from more languages might allow other, weaker, effects to emerge. Alternatively, it may be that some phonemic effects occur only in certain word positions (e.g., stressed syllable, word ending, vocalized position) or only in combination with other phonemes. For example, Bergen (2016) reported that monosyllabic nonwords sound more profane to native English speakers when they end in a consonant than a vowel.

Our findings demonstrate that sound symbolism is more pervasive, with a broader functional role, than has previously been appreciated, extending beyond single concepts (such as object size) to broad pragmatic functions. This has both practical and theoretical implications. At the practical level, using words rich in approximants may help defuse tense social situations and so may be important in a range of real-world contexts (e.g., relationship conflict, diplomacy, hostage negotiation). At the theoretical level, our results suggest a functional role for sound symbolism that extends beyond supporting language acquisition in childhood (Imai et al., 2008; Kantartzis et al., 2011; Monaghan et al., 2014; Perry et al., 2018; Thompson et al., 2012). Specifically, our findings suggest that sound symbolism can facilitate the pragmatic expression of emotion, attitude or arousal (Nielsen & Rendall, 20112013). Other fields of linguistics, such as semantics and historical linguistics, may benefit from considering how sounds can be modified to better exploit other pragmatic functions. While Study 2 focussed on minced oaths and offensiveness, word alterations in other vocabulary domains may reflect other sound symbolic patterns and serve other pragmatic functions (e.g., cajoling, appeasing, expressing authority).

We acknowledge three caveats. First, while our results demonstrate clearly that speakers of a range of languages tend to judge that words with approximants are not good candidates to be swear words, this finding is about perceptions of swearing, rather than swearing itself. Nevertheless, our experimental approach is consistent with a long tradition in sound symbolism research. For instance, it has long been accepted that there is a sound symbolic association between high front vowels such as i and small size, yet for decades this consensus relied solely on experiments (e.g., Newman, 1933; Peña et al., 2011; Sapir, 1929; Tarte & Barritt, 1971; Thompson & Estes, 2011). It was only very recently that an association between the vowel i and the concept “small” was demonstrated to be a statistical regularity across actual human languages (Blasi et al., 2016).

Second, we do not mean to suggest that the presence of approximants is sufficient to render words inoffensive: again, our findings are probabilistic rather than deterministic. What our results point to is an underlying cognitive bias, a predisposition that will have acted in concert with historical accident to shape the evolution of swear words. Just as the association between nasal sounds and words for “nose” does not manifest in every language – or even in most languages (Blasi et al., 2016) – we should not expect that the pattern we have identified will manifest in every language, and even languages that reflect the pattern are likely to have swear words with approximants, though fewer than would be predicted by their sound system.

Third, although we recruited speakers of different languages for Study 1, they were all familiar with English. We cannot exclude, therefore, the possibility that their performance in the “sweardar” experiment reflected their familiarity with phonetic patterns in English. We think this unlikely, as native French speakers, whose language does not exhibit this pattern, demonstrated numerically the strongest effect in the experiment. This suggests that performance in the experiment does not simply mirror linguistic knowledge. Moreover, our pilot study results indicated that approximants were at least as under-represented in the swear words of some of the other languages we investigated (e.g., Hungarian, Russian) as they were in English.

So, are swear words “universally patterned on the basis of sound” (Wajnryb, 2005, p. 205)? Our results point to a robust cross-linguistic sound symbolic association in the minds of human speakers. As to the wider universe, the jury is out: surprisingly, according to The Hitchhiker’s Guide to the Galaxy, the rudest word in the universe is “Belgium”, which contains an approximant.Footnote4