Too Many Preferences, Too Little Evolution: Multivariate Mate Choice and the Limits of Mate Preference Evolution

– by Kaitlyn Harper

As evolutionary psychologists, we love a good story about sexual selection. They usually go like this: humans have evolved to prefer a given trait (e.g. intelligence, facial symmetry, waist-to-hip ratio) because it offers an evolutionary benefit, either directly or indirectly. Over time, people who prefer and therefore mate with partners who have these advantageous traits are more likely to pass on their genes and as a result these preferences become more and more common. Trait confers fitness benefit to partners. Corresponding mate preference evolves. Everyone goes home happy. But there’s a key assumption in this narrative that’s often taken for granted: that people actually end up with partners who match their preferences. And that’s where things start to get messy.

Empirical research paints a more complicated picture. Across many studies, the link between what people say they want in a partner and who they actually choose tends to be surprisingly weak. In many cases, preferences simply don’t map onto real-world mate choices. This inconsistency creates a fundamental tension: how can mate preferences be driven by sexual selection if they don’t consistently guide mate choice in reality? One suggested approach to reconcile this inconsistency is to consider mate choice as a multivariate process. That is, people aren’t picking partners based on a single trait, they’re juggling many preferences at once, making trade-offs in the process. This means that any single preference might only have a small effect on mate choice in reality, swamped by other competing criteria. It’s a tidy solution: mate preferences which have been observed in isolation can still exist, but diluted effect sizes could explain the lack of observed association between preferences and partner outcomes in more realistic settings. But this raises another important question: if the effect of each preference on partner choice is so small, are they still visible to selection?

To test the multivariate limits of the classic sexual selection narrative, we built an agent-based model that simulates a population of individuals who choose mates based on heritable traits and preferences and pass on these genes over 100 generations. We asked an overlooked but crucial question: what happens to mate preference evolution when individuals are choosing partners not just based on one or two preferences, but up to ten? Does increasing multivariate trade-offs in mate choice weaken the evolution of preferences? We ran the model under two scenarios – one where preferences evolve through indirect selection (i.e. fitness benefits are conferred through good genes passed on to offspring), and one where we added direct selection (i.e. fitness benefits are also conferred immediately through fertility benefits).

The results, which we unpack further in our full article, revealed that as the number of preferences used in the model increased, the evolution of those preferences slowed dramatically. With just one or two preferences, the model worked much like classic sexual selection theory predicts, where preferences and traits evolved hand-in-hand. But once we asked agents to consider many preferences, things started to break down. Trait evolution was largely unaffected, but preference evolution hit a wall. In our models, once mate choice involved ten preferences, indirect selection barely moved the needle of preference evolution. Direct selection held up better, proving more robust under the multivariate constraints.

So, what do our findings mean for the way we think about human mate preference evolution? They suggest that the heavily relied on approach, where each mate preference is assumed to have evolved by offering unique indirect fitness benefits, needs to be reconsidered given the many human mate preferences that have been observed. That’s not to say that these observed preferences aren’t real, or that they didn’t evolve – it just means the standard one trait, one fitness benefit, one preference approach may not hold up in the context of multivariate mate choice. Instead, we may need to consider alternative frameworks: perhaps preferences reflect mostly direct benefits, function as redundant signals, or arise from cognitive mechanisms. One idea is that only a few core preferences – such as for health, status, or resources – are truly evolved, while the traits we find attractive are indicators of these broader qualities that we learn implicitly via associative learning (e.g. finding clear skin attractive because we’ve subconsciously learned it is a signal of health).

Ultimately, our findings invite a rethinking of the classic sexual selection story, urging us to zoom out from individual preferences and develop more comprehensive models that reflect the multivariate nature of mate choice.

Read the original paper here: Harper, Kaitlyn T., & Zietsch, Brendan P. (2025) Multivariate mate choice constrains mate preference evolution. Evolution and Human Behavior, 46(4), 106694.