Despite the presence of humour in virtually all human cultures, its evolutionary basis is not well understood (Fry, 1994; Provine, 2000). While various evolutionary theories of humour have been proposed (see Polimeni & Reiss, 2006, and Kozbelt, 2019, for overviews), one possibility is that humour is conducive to mating success and thereby subject to sexual selection. Indeed, individuals consistently report a preference for humour in a romantic partner (Bressler, Martin, & Balshine, 2006; Buss, 1988; Daniel, O’Brien, McCabe, & Quinter, 1985; McGee & Shevlin, 2009). However, it remains unclear why, in an ultimate sense, humour is considered an attractive quality. Humour is enjoyable, of course – but there is nothing inherently enjoyable or attractive about the kinds of stimuli that we describe as funny. The question becomes, why did we evolve to find certain kinds of stimuli funny and to be attracted to funny individuals?

Miller (2000) suggests that humour production (attempts to amuse, e.g. making a joke), when successful, is a signal of underlying genetic quality. The idea is that successful humour production requires mental performance (e.g. speed, intelligence, creativity), which in turn requires a high-functioning brain, which in turn requires a low load of genetic mutations. According to the ‘fitness indicator hypothesis’, sexual or romantic preference for funny individuals is therefore beneficial because offspring will inherit lower mutation loads and pass on their parents' genes more effectively. According to this sexual selection hypothesis, both being funny and being attracted to funny people are evolutionarily favoured.

According to parental investment theory (Trivers, 1972), women should be more interested in fitness indicators in a partner than men. This is because women, compared with men, have had considerably greater minimum time and energy costs associated with rearing offspring (e.g. nine months labour, then a long period of breastfeeding), thus making the consequences of a poor choice of partner more costly in terms of fitness. Conversely, given their much lower minimum investment in reproduction (i.e. a single act of sexual intercourse), men should on average be less choosy and more interested in accessing receptive women, to maximise their mating success (Darwin, 1871; Trivers, 1972). Therefore, if Miller's fitness indicator hypothesis is correct, one would expect a difference, on average, in how men and women produce and respond to humour. Men, more so than women, should attempt to produce high-quality humour and value receptivity towards it in a potential mate, whereas women, more so than men, should value high quality humour production in a potential mate.

Previous studies that sought to investigate the fitness indicator hypothesis have primarily relied on captioning tasks (e.g. writing an amusing caption to a cartoon), retrospective reports (e.g. having participants recall how funny they or other people are), or stated preferences (Bressler et al., 2006; Bressler & Balshine, 2006; Greengross & Miller, 2011; Hone, Hurwitz, & Lieberman, 2015; Martin & Sullivan, 2013; Weisfeld et al., 2011). While many of these studies found results consistent with the fitness indicator hypothesis, the measures are problematic. Captioning tasks have little in common with live interactions, and stated preferences assume that participants have sufficient insight into their own preferences. Additionally, both stated preferences and retrospective reports are likely prone to memory or social desirability biases. In practice, stated preferences often fail to predict individual's evaluations of potential partners (i.e. their revealed preferences; Eastwick, Luchies, Finkel, & Hunt, 2014). Therefore, preferences for humour production and humour receptivity found using captioning tasks, retrospective reports, and in particular, stated preferences, may not generalise to live interactions, which have characterised mate evaluation throughout our evolutionary history.

Speed-dating paradigms, which involve participants having brief one-on-one interactions (e.g. 3 to 5 min) with multiple partners each followed by a measure of attraction or willingness to date, provide a more ecologically valid way to test the fitness indicator hypothesis in that they involve face-to-face interactions. To be sure, speed-dating does not fully capture the mating process in our environment of evolutionary adaptiveness, given that they are short, one-off interactions. But attraction in speed dates predicts later romantic interest and dating (Baxter et al., 2022), suggesting that they to some extent capture processes of natural mate choice. Importantly, speed-dating results reflect revealed as opposed to stated preferences. Other studies such as Wilbur and Campbell (2011) measured revealed preferences using online dating profiles, but not with interactions or a behavioural measure of humour in the form of laughter. Despite these advantages, only a single study testing the fitness indicator hypothesis has employed a genuine speed-dating paradigm; McFarland, Jurafsky, and Rawlings (2013) found that neither laughter in response to an interaction partner's humour (at-partner laughter) nor laughter at oneself (at-self laughter) predicted a participant's sense of connection with their partner. However, McFarland et al. (2013) collapsed both measures of laughter across men and women, meaning they could not examine whether the sexes differ in the value they place upon humour production.

Hall (2015) employed a similar paradigm to McFarland et al. (2013) (the pretence, however, was a study of how first impressions are formed) while looking at possible differences between men and women. They found no significant sex differences in the relationship between at-partner laughter and dating interest. However, in Hall (2015) participants had only one interaction partner, making it impossible to isolate participant- and interaction-related variance. Additionally, the interactions were partially scripted, reducing the experiment's ecological validity. Further, neither Hall (2015) nor McFarland et al. (2013) had participants rate their partners' funniness or humour receptivity. While post-interaction ratings such as these can be subject to a halo effect (Nisbett & Wilson, 1977), testing for an association between ratings of humour and attraction would still have allowed for an additional evaluation of the fitness indicator hypothesis. As well as the studies not collecting ratings, neither study assessed humour receptivity (overt displays of amusement in response to another's attempt at humour, e.g. laughing). Finally, in both studies, participants were made aware that they were being recorded and studied, due to pre-approval of recordings and/or worn audio recording devices. If assessment and recordings are salient during interactions, participants may be less likely to act naturally, reducing the generalisability of findings to interpersonal interactions outside of an experimental setting.

Our study advances upon previous tests of the fitness indicator hypothesis of humour by using a behavioural measure of both humour production and humour receptivity (i.e. laughter) in conjunction with unscripted interactions. We also asked participants to rate both funniness and humour receptivity, used inconspicuous recording devices, and collected a larger sample (860 interactions leading to 1720 rating observations). First, we tested predictions from the fitness indicator hypothesis using stated preferences: that women value humour production in a partner more than men, and that men value humour receptivity in a partner more than women. We then moved on to testing predictions in a more ecologically valid setting by assessing revealed rather than stated preferences. We measure humour production and humour receptivity in a speed-date partner first by asking participants to rate these characteristics in each partner, as well as through measuring laughter frequency during the interactions. Preferences are revealed by how these measures are associated with ratings of overall attractiveness. As well as testing whether these preferences for humour production and receptivity exist overall, we also test for the predicted sex differences: we expect that women would have a stronger revealed preference for humour production compared to men, while men would have a stronger revealed preference for humour receptivity compared to women.