A physically weak partner who tried to protect someone from an attacker was rated as far more desirable than a strong partner who saw the danger and stepped away. That is the clearest result from a 2025 study that gave physical strength and willingness to help separate switches, rather than assuming one stood for the other.
Michael Barlev and his colleagues report the finding in Evolution and Human Behavior. Across seven vignette experiments, 4,508 US adults recruited through Amazon Mechanical Turk rated imagined dates and friends after reading tightly controlled accounts of a violent confrontation.
This is one study, not settled consensus. Nobody faced a real attacker or chose a real partner. The experiments measured reactions to hypothetical stories, but their design lets us ask a more exact question than “does strength attract people?” It asks whether ability still matters once people know what someone is willing to do.
Strength and intention became separate variables
The main experiment recruited 712 primarily heterosexual adults, 57 percent of them women, with an average age of 37. Each participant imagined leaving a restaurant with a date. A drunk stranger approached from behind and swung a fist toward the participant’s head.
The date was always described as average in physical attractiveness. The researchers varied two other details independently. The date was much weaker than average, average, or much stronger than average. The date also either saw the threat and intervened, saw it and stepped away, or failed to notice it in time. The last version supplied no information about willingness and served as the control.
Crossing those details created nine versions of the same scene. It meant a weak but willing date could be compared directly with a strong but unwilling date, while appearance and the rest of the story stayed fixed. Participants then rated the person from zero to 10 as a possible long-term and short-term romantic partner.
The paper concentrated on the long-term rating. Ratings for the two relationship horizons were strongly correlated, and analyzing them separately did not change the authors’ main conclusions.
The weak protector beat the strong bystander
Behavior moved ratings much further than the strength description did. Compared with the no-information control, seeing a date intervene raised long-term desirability. Seeing the date knowingly step away lowered it.
For women evaluating male dates, the unwillingness penalty was particularly steep. The authors called an unwilling male date a “deal-breaker” within this scenario. Their results also confirm the headline comparison directly: women found a much weaker man who stepped in substantially more attractive than a much stronger man who stepped away.
Strength was not irrelevant. Women gave a modest advantage to stronger male dates, while men did not show a statistically reliable strength preference for female dates. But the study also asked participants how able and how willing they believed each date would be. People assumed stronger dates would be more capable and more willing to help.
In the authors’ mediation model, most of women’s apparent preference for stronger men ran through that inferred willingness. After those indirect paths were included, the manipulated strength description had no detectable direct link to attraction. A mediation model does not reveal a private thought process with certainty, but it shows how easily “strength is attractive” can conceal an expectation about behavior.
An unsuccessful attempt still carried weight
The initial scene made intervention succeed. Later experiments removed that convenience. A willing date or friend tried to help but was overpowered and pushed to the ground. Participants understood that failure as evidence of lower ability, yet the person who tried remained more desirable than a target whose willingness was unknown.
Among dates, success versus failure was not a statistically significant predictor of attraction in that experiment. The attempt mattered even when it did not produce a rescue.
The researchers then wrote a harsher version in which the participant was harmed. In some versions the attacker landed the blow after overpowering the date; in others the participant fell and hit their head. Harm reduced certain ratings, especially when women evaluated male dates. It still did not erase the willingness effect. A date who tried and failed was rated above one whose willingness remained unknown.
That makes the study about more than protection as an outcome. The action also communicated that the imagined partner was prepared to accept a cost on someone else’s behalf.
The pattern extended beyond heterosexual dating
The seven experiments did not only ask heterosexual women to judge men. They included women and men evaluating opposite-sex dates, same-sex dates and friends. One experiment specifically recruited bisexual and homosexual adults. Across those settings, willingness generally raised desirability and unwillingness generally reduced it.
The size of the response was not uniform. The penalty for stepping away was larger when the rater was a woman, the target was a man, and the relationship was framed as a date rather than a friendship. When an unwilling man was judged as a friend, however, ratings stayed below the midpoint regardless of whether the rater was male or female.
“Attraction” also needs careful reading here. In the dating experiments it meant a rating of an imagined person as a possible romantic partner, not a judgment of the person’s face or body. In the friendship experiments it meant desirability as a friend. The common element was partner choice in a broad social sense.
Earlier research on strength asked a different question
A well-known 2017 finding can sit beside this one without contradiction. Aaron Sell, Aaron Lukaszewski and Michael Townsley showed people photographs of men’s bodies with the faces obscured. In their Proceedings of the Royal Society B paper, perceived strength accounted for about 70 percent of the variation in bodily attractiveness ratings.
Those raters saw bodies, not decisions. The 2025 vignettes held physical attractiveness constant and supplied behavioral evidence at a costly moment. Looking strong may carry considerable weight when appearance is almost all a person knows. Once an action reveals commitment or indifference, it can become the more diagnostic fact.
A separate relationship literature offers a useful parallel, although it does not test physical protection. In four longitudinal studies published in 2024, emotional reactions to a romantic partner’s daily stress were associated with higher partner-rated relationship quality in newer couples. Perceived responsiveness helped account for that association. The measures and outcomes were different, but both lines of work point toward a similar social judgment: people notice whether another person’s welfare seems to matter to them.
The scenario removes choices that matter in real life
The experiment forced a narrow decision. The date noticed the danger only when the attack was already underway, leaving no time to warn the participant, seek help, talk the aggressor down, or leave. That constraint helped isolate willingness to intervene physically. It also makes the vignette a poor guide to what anyone should do during an actual confrontation.
Nothing in this research says a person should enter a fight. In real situations, physical intervention can worsen the danger. The study tells us how participants rated a choice after safer alternatives had been deliberately written out of the story.
There are other limits. All participants were US adults recruited from one online platform. They rated hypothetical people after reading a few paragraphs. The experiments did not observe dating behavior, real acts of support, or relationships over time. They also cannot tell us whether the response reflects an evolved preference, cultural expectations, a general preference for caring partners, or some combination of these.
Protection in this paper has one specific meaning: responding to an immediate physical attack. The finding should not be extended to jealousy, possessiveness, monitoring, dominance, or control.
A stronger next test would compare cultures and use ordinary situations in which partners can choose among several safe forms of help. That would show whether willingness continues to outrank capacity when people judge behavior they might plausibly encounter, rather than a single forced choice in a violent vignette.