You surely noticed men getting all angry recently that women were reporting they’d rather be found alone with a wild bear than a random man. The men are pretty much in the wrong there. Multiple converging studies suggest between an tenth and a sixth of men are (actually are) rapists, and between a quarter and a third of men have committed some form of legally defined sexual assault (study, study, studies). By contrast, bears generally leave you alone if you don’t mess with them. Indeed, evidence suggests the probability of an unprovoked attack is less than 1% per encounter (evidence, evidence, evidence, evidence, evidence; discussion, discussion, discussion). That’s way lower than slotting in a random man who is 10% likely to actually be a rapist, and over 25% likely to be some kind of sexual predator (even if just the “getting handsy” type).

Of course, that does not mean women are attacked 25% of every time they are alone with a man. But that’s because even assaulty/rapey men know they have to pick their battles. But the risk remains. All they have to believe is that they can get away with it. And that’s too likely in our present rape culture to count on. It’s not something women can control. Whereas “if I just leave that bear alone I’ll be fine” is a far more reliable strategy—one that doesn’t prevent rapes or sexual assault at all, yet does reduce bear attacks to near zero. Because raging cocaine bears aren’t a thing. In the real world bears have no interest in eating people and would rather you just bug off. Overall lifetime incidence of rape or sexual assault for any given woman is between 1 in 10 and 1 in 5. That is way over the incidence of hikers or rangers being attacked by any animal, much less bears.

The 95.1% Study

Men are needlessly, excessively dangerous. And there is no way to tell which men. Contrary to myths and stereotypes, they all look the same. That is all simply a fact. And you shouldn’t get over it. You should be concerned to change it. But you will be undermining any effort at that if you publish unreliable science that is too easily debunked. We need real, defensible facts. If we develop and promote exaggerated or erroneous results, the alarmingly growing number of people who are ideologically predisposed to deny anything is wrong (or even keen to keep things as they are) can maul your cause worse than a mythical coked-up bear. And if trolls find you promoted false information, they can go around denying there is any true information. And you will struggle to get ahead of that. False headlines hit far more eyes and travel far faster than retractions or corrections.

I’ve made this point before regarding climate activism and animal welfare activism, which are real issues that matter, yet activists exaggerate so much as to end up convincing average people they’re entirely full of shit and thus there is no real issue to be concerned about. Contrary to your mistaken intuition, exaggerating dooms your cause. It does not kick up any revolution. It kicks the legs out from under it. So if you care about creating change, you need to care about being accurate. For an example, see Debunking CNN’s Viral 62 Million Men Hoax and how false information was used by an ultraright Christian organization to dupe you, and most people fell for it—yet after melting away all the lies and distortions, there was still a real fact to be concerned about. That not millions but still thousands of men were happily exchanging rape tips with each other online is not trivial. But it got lost in the fake news, so even the real problem was dismissed as fake (though fortunately not by the authorities).

I recently encountered an example of this that I want to talk about today because it’s a study so full of errors as to be functionally useless, yet its headline keeps getting cited as some powerful new “scientific fact.” And it’s hurting the cause of reducing sexual violence. The paper is “Isolate, Inebriate, Intimidate, Repeat: High Rates of Sexual Force against Women Are Reported When Young Men Given Anonymous Surveys” by Lucia O’Sullivan and Scott Ronis, for the Journal of Interpersonal Violence. And it’s thankfully open source; paywalls magnify misinformation by rendering it impossible for most people to fact-check internet propaganda. And here that matters because the abstract’s most shocking line became headlines everywhere it was reported on: “Overall, 95.1%” of men “reported having recently used at least one of the strategies to get a woman to have sex who they knew did not want sex and had not consented.” This is not an accurate or correct description of what their study found. But that only adds to the problem that even what their study did find is unusable scientifically because of a catastrophic stack of errors in their study design. And most people won’t ever know that. Because most people—even people reporting on the study—won’t actually read it, or notice or understand its methodological defects even if they do.

To be fair, the study isn’t crank. It’s not AI slop or specious fake science generated to push an agenda. It’s just messed up. And this is a growing trend across real science now. Approximately two out of every three papers fails to replicate, indicating single-article studies tend to be false, a problem called the Reproducibility Crisis. And there is a reason for that. It’s often obvious in the study itself. Which indicates a Peer Review Crisis. This paper is one of the better examples of that, because a lot of that phenomenon is due to fraudulent or agenda-driven science getting past peer review (see Retraction Watch); plus, now, an AI Crisis of entirely fake papers. I recently discussed an example of that in physics, but I’ve covered examples of traditional fraud, too—and not just in EvoPsych (where I say a lot more, with links, on the reproducibility and peer review crises in science generally). But it’s important to understand that that is not even the bulk of the problem. Even honest science is becoming degraded by the same inattention to methodology and accuracy in study design that lets fraud slip through by the thousands. Scientists are trying to do science on the cheap, and peer reviewers are underpaid and exhausted, and the result is a ton of even well-meaning science that is bad—to the level of a universally acknowledged crisis.

I won’t run down the related problem of telephone game sensationalism in reporting science, where the rather convoluted thesis statement of the O’Sullivan & Ronis paper gets “converted” into versions they would not (I hope) endorse, like that 95.1% of men have coerced women into sex, or are even rapists. I hope my readers are aware not to trust clickbait headlines, or the media, and always to go and check the study itself (and not just its abstract) before believing anything claimed about it anywhere. Even prestigious journalism cannot be trusted. Check. Yourself. Always. You will find 95% of the time (my own made up statistic) that something about the study was misreported.

Here I’m going to ignore the media’s failure of society and just take the scientists here seriously at what they said. The skills you can apply to the media I cover in A Primer on Actually Doing Your Own Research and A Vital Primer on Media Literacy. The skills you can apply to the science paper itself I cover in Dumb Vegan Propaganda: A Lesson in Critical Thinking and Bad Science: No, Atheism Does Not Cause Suicide, those being previous examples of the kind of analysis I am doing here.

Online Survey Systems Cannot Be Trusted Anymore

Of course the title O’Sullivan & Ronis chose for their paper, and the inaccurate abstract they wrote for it, do not commend them as reliable. But they did at least write this paper clearly enough that its errors are visible, and so they weren’t trying to snow anyone. They really don’t know what mistakes they made. But they are honest and competent enough to have followed protocols in writing up their study so we can see the mistakes they made. And that’s at least an important merit to their paper, and what makes it just a congeries of errors and not a deliberate effort at obfuscation. But it’s still a congeries of errors, so bad and so large as to render the study essentially unusable. To get anything useful, it would have to be completely redesigned, and completely redone.

The first overall problem I will discuss—not at all the only or even worse one, but a very important and pervasive one—is that they did not generate anything we can reliably call a random sample of the population. Or, I should say, of either population, as their respondents hailed from either America or Canada—or so the researchers gullibly think, not knowing about VPNs and foreign survey farming. Because, yes, most of your “subjects” might be lying about what country they are in and everything else, just to net the ten dollars—which from hundreds of surveys adds up to thousands of dollars of income; and from ten thousand surveys annually now, it’s a hundred thousand dollars a year, a rather fine income in any country, much less the third world.

But let’s set that aside (we shouldn’t; but there are more problems to get to). Because that’s not the only problem, as you’ll find perusing the some four million studies now of “the problem with using online survey subjects.” But for an example see Djin Gie Liem, “The Future of Online or Web-based Research: Have You Been BOTTED?” in Appetite (2025). Even Americans and Canadians can run survey farms. And this is the problem with pay-for-play survey systems, well known now for generating biased or even fake participants. For example, people create bots to run these surveys to collect the cash, without even looking at what answers the bot gave; and even people who fill them out themselves are often “professional survey takers,” doing thousands of surveys for money, and thus not representative of most people. Indeed, this process is inherently dishonest, which thus selects a certain type of person, a type who might more likely be sexually manipulative or predatory. Consider that most American and Canadian men would never think or want to do surveys for money, or even know that exists as a thing. So the motive economy that is funneling respondents to these surveys is an enormous source of bias.

And that may especially be a problem here, as O’Sullivan & Ronis report having previously run interactive studies on manosphere forums, and those are more likely to have virally spread the link to take this study, and thereby give it an enormous overage of manosphere-channel participants. They may even have plotted to bomb the survey just to groyp it (and no, tools like CAPTCHA do not prevent this, from human swarms or bots). The researchers’ ignorance of these inevitabilities is catastrophic.

And this alone destroys the study. As none of this contamination (survey farming, manosphere gaming, and the whole “flying monkey problem”) was controlled for. Since it cannot be ruled out, or even rendered improbable, its statistical results are worthless. The title of the paper might as well have been “95.1% of bots, survey thieves, and ideological misogynists claim to trick or force women into sex.” Which is a pretty useless result, as we don’t know how many of each category, or of honest respondants, there were, so we can’t even sort this stat out among groups. Who cares what bots say? They aren’t even humans. Who is surprised that ideological misogynists are abusers and predators? We already knew that. And yet we can’t even tell how much of that 95.1% is bots or misogynists—much less “random honest North American men.”

The Stanford Problem

But that’s not the only problem. One of the most famous points made about the fatal methodological flaws in the infamous Stanford Prison Experiment is that its subjects were not randomly selected: they were funneled with advertising that disproportionately attracts sociopaths. As discussed in Thomas Carnahan and Sam McFarland, “Revisiting the Stanford Prison Experiment: Could Participant Self-Selection Have Led to the Cruelty?” in the Personality and Social Psychology Bulletin (2007), “If the traits studied … induce volunteering for prison life, either as a simulation or for a role as a real prison guard, the individuals who volunteer arrive as individuals with qualities related to abusive behavior.” So the statistical results will not be reflective of the general population, but of the population of men already inclined to be abusive.

It is therefore fatal to O’Sullivan & Ronis that they advertised (and thus funneled participants into) their study with wording even more explicitly attractive to abusers than the Stanford experiment did. Which is an error no social scientist should be committing today, as the SPE is standardly taught. Its mistakes are methodology 101 now. So this was a huge screw up and I cannot account for it.

As Carnahan & McFarland observed of their own study of the problem with the SPE:

[V]olunteers who responded to a newspaper ad to participate in a psychological study of prison life, an ad virtually identical to that used in the Stanford Prison Experiment, were significantly higher on measures of aggressiveness, authoritarianism, Machiavellianism, narcissism, and social dominance than those who responded to a parallel ad that omitted the words of prison life, and they were significantly lower in dispositional empathy and altruism.

And that was just because of the words “prison life” being in the ad. That alone. Take note.

Because here is how O’Sullivan & Ronis tried attracting participants (emphasis mine):

The consent form indicated that the survey was men’s opportunity “to provide their side of the story given that we have heard so much from women” about male–female sexual interactions, repeatedly assuring them of their guaranteed anonymity. We wrote “We realize that this survey addresses sensitive information. We realize some people might not like us asking these questions. We want this information to know more about how men operate and how they overcome barriers to get what they want. NOTHING you report will be or can be linked back to you. All responses are anonymous.”

Setting aside the problem that most people no longer trust authority and thus are very disinclined now to believe claims of preserved anonymity (another indication of researcher gullibility), thus ensuring another line of bias (e.g. men who are proud of their manipulative or coercive behaviors will be more likely to participate than men scared of being judged, not knowing yet what the questions will be), the wording of this form is even more disposed toward attracting sociopaths and misogynists than the SPE.

• It frames the matter as men getting back at women, giving them a chance to be “heard,” implying we are hearing “too much” from women, a framing exactly aligned with misogynist ideologies. Yes, they say “so much,” but what participants will hear there is “too much,” given the context. This will be offputting to many men; but alluring, even highly motivating, to certain men.
• The use of “male–female” (if that was the wording used) instead of the more humanizing “men and women” is also incel vocabulary, offputting to ordinary men and feminists, but alluring to misogynists.
• They pose their study as daring to be honest and fight against feminist-controlled institutions: “some people might not like us asking these questions” lets the reader wonder who those “some people” would be, and the only answer anyone would derive is “feminists,” and the only people who would be attracted by this framing is anti-feminists.
• And finally, framing the study as learning “how men … overcome barriers” to getting what “they want” is entirely aligned with the PUA/MRA perception of “male-female sexual interactions.” It is specifically attractive to men seeing sex as something they want and deserve and can get with better strategies (which they are now being given an opportunity to teach or boast of); and specifically repulsive to men who don’t think that way.

These effects are not just creating an even worse attraction/repulsion effect than the SPE’s mere mention of the word “prison,” but it is all creating an even greater amplifier by avoidance taboo: men won’t want to be associated with the kind of men they perceive this form is appealing to and attracting. So they are actually de facto being asked not to take this survey; while predators and misogynists are practically being goaded to. If the mere presence of the word “prison” created the rather large effect sizes discovered by Carnahan & McFarland (some twenty to forty percent deviation against a randomized norm), we should expect this far more explicit and egregious pipelining to have had an even greater effect than that. And since the effect discovered for the SPE already circular-filed its results, the effect we can expect here must do so with greater force.

This study was designed to attract abusive and manipulative men, and turn away everyone else. It therefore is not giving us any statistical result for “men,” but is skewed heavily toward measuring the worst kind of men. This means that that 95.1% statistic is not true of “young men.” It is true of “a sample of young men who disproportionately lean toward manipulative or misogynistic personalities.” And yet becaue it failed to control for this, it isn’t even showing us usable stats for them, because it indiscriminately mixed them in with other men (as well as the bots and farmers). So the stats aren’t even for “the manosphere,” much less “men in general.” Nothing can be teased out here, so none of its results can be used for anything.

Survivorship Bias

We should all know by now what Survivorship Bias is. Especially scientists. The short version: if you decide how to up-armor fighter planes by where all the bullet holes are in surviving planes, you are doing exactly the wrong thing. Where the bullet holes aren’t is where the armor needs to be. This effect can work in multiple different ways. So you have to make sure your study isn’t doing this. This study did.

Besides limiting by age (“18 to 34”), they limited participants to men who “reported having had a sexual encounter with a woman in the past 2 years.” Worse, they framed this elsewhere in questions as women they “have recently met—no sex or dating history with them beforehand.” So they have excluded men in committed relationships with wives and girlfriends (monogamous or not), and are counting only men who had one-night stands in the last two years—which statistically is looking for a rather high rate: barely more than a fifth of men qualified for this study. If we include men who are in committed relationships that started with sex on a first date within the last two years (who would technically also qualify), maybe that would get to a third of men. But this is already excluding two thirds of all “young men.” And it’s surely more, for while a third of men maybe had sex on the first date with someone they are in a relationship with now, the number of men who have only been in that relationship for fewer than two years is going to be significantly smaller. I would expect their design has ruled out some three quarters of young men at least, rendering it rather useless as a result for “young men.”

And what kind of men are going to be overrepresented in that subsample? Well, obviously, predators and players. Hence even their explicit selection methodology was designed to rope in all rapists and players, while excluding most men who don’t commonly have one-night stands or first-date sex, or have sex with new people “bi-yearly” at all. In other words, most of the ethical men. This is another example of collecting surviving planes with bullet holes, and then trying to claim the sections of the plane free of holes are the same as the sections riddled with bullets. When obviously the opposite is true.

So even their explicit methodology enormously biased their sample toward unethical men, and therefore cannot be a statistic regarding “all” young men. It’s not even a statistic true of most young men, since it has already pre-eliminated some three quarters of them. And it didn’t eliminate them randomly or irrelevantly. It eliminated precisely the men who would have brought down all their percentages. And it didn’t even do that in any useful way, because it did not control for ethical behavior among the remainder who qualified for the study (a point I’ll get to later). The two groups can’t be teased apart here.

But a method that is designed to attract players and predators and exclude ethical operators cannot sustain the title, abstract, results, or headlines of this study. Its results are simply useless. They tell us nothing about “men ages 18 to 34.” And this just compounds with the problems surveyed above and below.

The Loss Problem

Making it all worse, they even admit (but do not state the relevance) of an unusually high error rate in their study, removing over ten percent of the sample for various vague “quality” reasons, which is a worryingly large error rate. We don’t know the bias in the removal set, which would in turn bias the inclusion set. The general view of the Loss Effect is that a study starts to get unreliable when these “dropped from the study” percentages exceed five percent, and completely obliterate a study’s reliability when they exceed twenty percent. O’Sullivan & Ronis are nearer the latter than the former. Normally this would just red flag its results as “probably to some degree unreliable,” and we’d have to call for a replication study before trusting it. But when we combine this with all the other methodological errors, it sinks the study altogether. More importantly, it’s just one more example of how this study is functionally incapable of succeeding. It needs to be completely redesigned and redone before any of its results can be trusted at all.

The First Questionnaire Problem

That’s all bad. And enough to circular-file this study and its results. But it gets worse. Because the most common flaw in any survey-based study is badly designed wording, framing, and ordering of questions. And this study is so bad on this metric that it’s a teachable example of how not to compose a questionnaire. And as I go through this fatal design error, keep in mind every other error I have documented up to now: this error destroys the study even without them; with them, it makes this study an absolute face-palming travesty.

Okay. Ready? Here we go.

• They say they can conclude the participating “men knew that the woman was unwilling and not consenting through her ‘making excuses’ (41.9%), directly saying she was not interested or did not want to (24.1%), and physically moving away/trying to be out of reach (15.7%). Men also responded that, in some circumstances, women were too drunk or high to say or do much of anything (9.7%).”

This list indiscriminately mixes together wildly different things, and they get more ambiguous and unclear the higher percentage of response, which signals a fatal flaw in their design. It is absolutely clear that trying to having sex with a woman “too drunk or high to say or do much of anything” is not only immoral, but literally rape in most jurisdictions and by any moral compass. Key to this is their question’s precision: they did not ask whether someone was drinking or high (an ambiguous circumstance), but so drunk or high as to be unable “to say or do much of anything,” which is entirely clear as to what we are measuring here. Yet observe: fewer than 10% of respondents reported this, which aligns with all the other data regarding the frequency of rapists among men. And that’s in a study we already established has highly overrepresented rapists in its test sample (no doubt because not all the rapists in their sample just happened to use that tactic in the last two years).

But in what universe is that the same thing as trying to overcome a woman’s stated or signaled disinterest with persuasion? That is an effort to secure consent, and is in fact the thing we should want men to do rather than going after drunks. The stated reasons do not indicate any difference between honest and welcome vs. dishonest or unwelcome persuasion here, and thus this study is blind to whether these men actually tried to have sex against consent. Yet it claims in the title and abstract that that is what it is measuring. It isn’t. Their study has no controls for this metric at all. They just assumed all seduction is against consent. That is both conceptually and factually false. And their title and abstract and stated results should have been rewritten to reflect this—and their peer reviewers should have spotted and told them this.

This flaw is multiply bad, because not only is their wording too ambiguous to actually measure the thing they want, but it is even designed to frame the respondents. For example, they put “making excuses” in scare quotes implying distasteful attitudes towards a woman having “reasons,” but they put those words in their mouth. The respondents did not volunteer this. They were forced to tick a box that said this, and were given no box that said something less contemptuous. That the researchers ironically violated the consent of their subjects by forcing them to endorse wording they might have found distasteful or inaccurate is not lost on me. But its relevance to the science is that they have already skewed the outputs by their own manipulation of the test design, thus making it even harder to tease out what the men ticking this box actually thought or experienced, despite that being the one thing they claim to be doing.

To get out of the abstraction and center you with a real world example of the problem here, I will relate an occasion of my own experience, which would have compelled me to tick the box ‘making excuses’ (if it had happened in the last two years) even though I don’t agree with that misogynistic mischaracterization of women’s behavior (a misogynistic mischaracterization that is coming from these researchers, not their subjects). Reasons are not “excuses.” And my example illustrates this, as well as the rest:

Once upon a time I met someone at a conference. There was open flirtation between us for quite a while. When eventually I expressed interest in spending the night with her, she said she was interested but disinclined for reasons she said had nothing to do with me. I then asked her if she would permit me to persuade her, making clear it was fine to say no and leave it at that. This happened in a flirtatious bout of texting while she and I were in a conversation with other people unaware of our side-banter. She said yes. Indeed she said she was very curious to find out what such persuasion could consist of. I asked her for her objections, and they consisted solely of embarrassment over scars she has. Once I assured her (entirely honestly) that I would have no problem with that, she was delighted, and asked me to spend the night. And we did. The sex was enthusiastic and fun. And she expressed gratitude for helping her get over her embarrassment and enjoying sex again. And we’ve been friends ever since.

This was entirely consensual and in fact how such encounters should go. It is in no way “Sexual Force against Women” (as the study title claims it was measuring), or having sex despite “the woman [being] unwilling and not consenting.” Yet they claim this is what their questions are measuring. They aren’t. Encounters like mine, respectful and reading the mood and ensuring of consent, are being lumped in with physically forcing women to have sex without consent. Which pretty much destroys their study’s results. They are simply not counting what they claim.

That occasion in my past would also have compelled me to tick the box “directly saying she was not interested or did not want to” because that is vague as to when that statement occurred, what it consisted of, and how I proceeded after it—all the things that actually determine whether the ensuing sex was consensual or not. Yet these researchers are completely blind to ethical moves here, and simply counted everything as nonconsensual, because they included no way to measure whether that is what ticking these boxes even means. Their instrument is blind to encounters like mine and therefore useless.

One can also query “physically moving away/trying to be out of reach.” Although persisting after that could be rapey, it is still not as clearly so as pursuing knocked-out drunks. This got a 15% response, which is close to the expected percentage of rapists. But we still have to ask: how many men ticked this box meaning actual flirtatious evasion, or their own mistaken perception of it? Or, indeed, behavior prior to changed sentiment—in other words, how many of the indicated women later voluntarily cozied up to them before any moves or conversation toward sex? What one needs to know is, after they believed this behavior occurred, had there been any prior flirtation to muddle its meaning, or posterior conversation about what it actually meant, or any indicative change in the behavior? In other words, did this observation result in pursuing nonconsensual or still consensual sex? Their question by itself can’t answer that, and thus it fails to measure what they claim.

And to be clear, I suspect most of the ticks of this box were not ethical occasions like that. But the stat depends on all of them being unethical; otherwise it has been watered down with erroneous responses. So their vague questioning taints the result. Which is especially embarrassing as they could easily have more clearly worded this question, just as they did the question about setting on a drunk. But they didn’t. And that mistake, along with all the others, ruined their study.

They claim all these questions successfully counted only men who “tried to get a woman to engage in sex that they knew she did not want and to which she had not consented,” but their questions do not do this. Except the one about setting on a drunk, which was correctly worded to purpose, in every other case they never address whether consent existed at the time the sex occurred. Miscounting “she wasn’t down but after some communication she was” as “she did not consent to the sex” and “she did not want the sex she had when she had it” is a catastrophic mistake that dooms all the results of this study. Worse, this is exactly the kind of false framing the triggers and energizes ideological misogynists and their propaganda. This study’s mistakes are therefore damaging and dangerous.

The Second Questionnaire Problem

So the O’Sullivan & Ronis study is trash by now. And I haven’t even gotten to all the flaws in it.

The final problem with it is a conflation of different kinds of unethical behavior.

For example, they say “the most common strategy” men reported using “was telling a woman whatever she wanted to hear and this was used by the majority (78.1%) of the 2,557 men reporting any history of forced or coerced sex,” which seems to be skewing all their stats a lot. It seems misleading to lump that in with drugging, forcing, intimidating, or pressuring women (more overtly coercive actions). It could still be unethical, but again, their wording is so vague as to include true as well as false information, and thus cannot discern whether “telling a woman whatever she wanted to hear” meant lying, or telling the truth. The misogynistic wording is again theirs, not their subjects’. No option was given for men to choose between “even if it’s false” and “as long as it was true.”

But that’s the same problem of lousy questionnaire design I already called out. What I want to focus on here is that even if they had correctly demarcated “telling a woman whatever she wanted to hear—even if it was untrue,” that is still not the same thing as “drugging, forcing, intimidating, or pressuring.” That men might lie to women to seem more attractive or available is certainly an ethical problem. But it’s one women can more easily manage than “drugging, forcing, intimidating, or pressuring.” Rape by deception requires more than merely shining someone on, such as tricking someone in the dark into thinking you are their partner, or fraudulently misstating your HIV status. See, for example, California’s Rape by Fraud statute.

By collapsing all bad acts into “rape,” you dilute the meaning and seriousness of that word and its indicated offense. Which plays into misogynistic ideologies by suggesting rape is a made-up word and therefore all rape statistics are meaningless because scientists count literally anything as rape. That is a false narrative. But it is hard to dispel or prebunk if scientists are actively misdefining rape to goose their counts of it. “He lied to her about his job, therefore he raped her” is precisely the kind of nonsense that makes our battle against rape culture harder, not easier. Which also does not justify a black or white fallacy, whereby if mere lying isn’t rape, then rape doesn’t exist or still isn’t alarmingly common, or lying to procure sex isn’t itself a problem we need to denounce and combat.

Words have to mean things, and distinctions matter. So “she consented to have sex with him after he lied about his job” is certainly censurable behavior we should assess the real scope and nature of—and women all well know this, and that it is easier to defend against liars than rapists. The recent example of women getting savvy to conservatives pretending to be liberals makes this point. But when that happens it is still not a woman who “did not want and had not consented” to sex, which is the definition of “rape” pretty much everywhere. So even though O’Sullivan & Ronis avoid ever explicitly converting “did not want and had not consented” to “rape,” their study is still semantically saying all of this is “rape.” And that’s simply false—and that wreaks havoc on our efforts to promote consent and end rape culture. Scientists should not be doing this. It is, IMO, even unethical of them to do this, and not just unprofessional or mistaken.

Because the problem here is worse than I have even explained so far. Because they claim their study found “95.1%” of men under 35 “reported having recently used at least one of the strategies to get a woman to have sex who they knew did not want sex and had not consented. Most of these occasions (65%) resulted in successfully forcing the woman into sex.” Notice they say “forcing.” That is not happening in most of the occasions they count—even if they had worded their questions correctly, and they didn’t. Their study effectively misstates them as finding 95.1% of men are rapists—without explicitly using the word rape, but playing semantic games over what counts as “force” or even “consent” or “desire,” and thus de facto accusing 95.1% of their respondents of rape. That’s worse than bad study design. It’s immoral. Indeed, I have to wonder if they actually violated their own Research Ethics Boards’s principles here by defaming most of their subjects, who upon reading this study will be shocked to discover that they are now supposedly rapists. Those boards approved the study, but I can only conclude they did so not understanding the severe ethical violations its poor design would produce.

Again, this does not warrant that black or white fallacy. I am not defending the liars in their study. I am, rather, defending them against the hyperbole of calling them rapists, which is to misunderstand (and thus miscommunicate) what constitutes desiring and consenting to sex. I would do the same if they were accusing thieves of murder. That in no way defends them against being thieves. That men may have lied to get sex (though the study’s questions were so badly worded as to be incapable of counting how often) does not mean they “forced” women to have sex, or even that those women had no desire to have sex or didn’t agree to. It’s simply misdescribing the problem—and in result fueling misogynist talking points, and creating increasing distrust of science when we need to be doing the opposite.

Pretending to be wealthy or single or hiding one’s political beliefs are a problem women should not have to deal with. But it’s not attempting to force women to have sex against their will. How many of those 78.1% who “told women whatever they want to hear” were just men who pretended to be divorced or poly but were actually cheating, for example; or who pretended to be interested in crochet or a particular genre of movies they weren’t. That’s still censurable. But I would rather see deception separated from coercion. Because those are two different education challenges to solve. And we need to know the difference in their frequency, as well as of the exact deceptions used, if we are to understand and combat them.

Consider, for example, their lousy attempt to get at this with awfully described options like “had a friend, partner, or group of friends help you get what you want (46.6%), had a female friend make the woman feel safe and convince her (43.8%), told her you knew she wanted it (39.3%), focused on a stranger to have sex with (37.9%), had a female friend bring her to you (37.6%),” and many others which are so badly worded as to make this survey almost useless; it is not inherently the case that any of these behaviors is even deceptive much less coercive. The authors of the study seem to be operating on certain stereotypes here rather than real-world examples—they just “assume” these “manipulate” which isn’t necessarily the case. I can think of hundreds of perfectly ethical and honest scenarios involving these things, and thus they are over-counting noncoercive and honest behaviors as coercive or deceptive. We need more carefully worded questions that distinguish coercive/deceptive uses of these approaches from noncoercive/honest ones.

For instance, “had a friend, partner, or group of friends help you get what you want” describes even just asking a friend to mediate an expression of interest (a perfectly respectable, safe, and honest thing to do) or just asking someone what she likes before approaching her, or even asking them for her number after she already told them she’s receptive to their giving men her number that they think would be suited to her; or asking where she will be, or how she likes to be approached, or asking them to vouch for him. These are not even deceptive much less coercive moves. Yet O’Sullivan & Ronis force respondents to tick “had a friend, partner, or group of friends help you get what you want” no matter what they actually meant. Consequently, they have framed the study in the form of stereotypes rather than an actual instrument for measuring consent; and then gave respondents no ability to signal what they actually mean when choosing the researchers’ vaguely worded options.

Likewise “had a female friend bring her to you” and “had a female friend make the woman feel safe and convince her,” which invokes all the same possibilities. How is that supposed to be dishonest or coercive? Sure, I can imagine scenarios where it would be. But I also know a lot of real-world scenarios in which it isn’t—especially (but not only) in the poly, alt-sex, and hookup communities (where a lot of respondents here could have been funneled from, as they are more likely to have had a new sexual partner “in the last two years”). They literally think this describes “female peers’ complicity in acts of sexual aggression,” but their questionnaire never connects any of this to “aggression” at all, sexual or otherwise; nor deception. This is a failure of Logic 101. And it is damning of their peer reviewers that they missed this.

The phrase “told her you knew she wanted it” is also negatively worded without any option for more neutral framing of acceptable behaviors that would fall under it, forcing respondents to choose it even though they don’t engage in the negative versions of it. For example, would my exploring a hesitant lover’s conceptual resistance count as “telling her she knew she wanted it”? I never said that. But some respondents may have imagined things like that count. Any suggestion that maybe she really would like to would count. Which is pretty much all hookup culture. For instance, they aren’t distinguishing “just telling a woman she wants it” from saying that in the midst of ongoing mutual flirtation and teasing—such as (but not only) in a dating app. These researchers needed to ask something much more specific than this to tease out pressuring or harassing examples of “telling her she knew she wanted it” from more respectful and considerate (or even just eyerolling) approaches that also involve suggesting that, duly informed, she may be interested. This is, after all, such a stupid move when literally enacted (as these researchers must be imagining) that no woman is going to be deceived by it. So even in its worst imaginable case it doesn’t really even count as deception, much less coercion.

Finally, “focused on a stranger to have sex with” is a completely bizarre thing to count as coercing a woman against her will. It has no intelligible place in this study as-worded. Since they are asking only for occasions of attempted sex with a person they had “no sex or dating history” with, who even would count for that but “a stranger”? Friends they already know well would count. But men will already know the sexual interest of their friends, so ethical men won’t as likely have hit on them in the last two years; they can mostly only have pursued “strangers” (hookups; one night stands with newly met people; etc.). That does not tell you whether that pursuit was ethical or not, or whether the “stranger” they have thus so callously referred to was into it and consented.

So their instrument is just so incompetently constructed as to be effectively useless—even when trying to measure deception rather than coercion, because it can’t even tell the difference between them. It could thus never have generated the conclusions they declare in their title and abstract, even on its awful questionnaire design alone.

The Ordering Problem

They have some questions that are floating free of these and thus not qualifiers, and so, not helpful for fixing the problem. But one of the most telling indicators of their study’s flawed design is when we look at their most important aggregation of data: Table 2, which is (mostly) ordered by how common a “method” of persuading a woman to have sex was (column 2 has the percentage of respondents who “never” used a strategy, the converse of which is how many who did). That may be the most useful thing here, even if the stat is mostly made by bots, since bots tend to think of what the internet does, and the internet is mostly real people (or so we think—some observations are starting to suggest even that isn’t true anymore).

So it matters that the most common strategy selected was “told her whatever she wanted to hear,” at 78.1% of respondents, but as noted that’s too vaguely worded to identify even deception much less coercion; while next after that is ’“asked her repeatedly” which was fewer than 50%, which is a lot more telling: even in their biased sample of overabundant rapists and players, still most of even them recognize repeated asks are bad, and claim not to do it. If even the worst men thought this, even more so we can expect will “men in general.” And that’s even assuming that question was clear. If you even just revisit the issue with someone after things have developed, that’s also “asking repeatedly,” whether she welcomed it or not. Thus many answers here might not have been cases of harassment. We can’t tell. Because it’s too vaguely worded. It cannot distinguish honest, respectful, welcomed moves from unwanted ones.

This also shows up when they actually stupidly frame a question as “gradually pushed verbally or physically,” when we want to know which, verbally or physically. That’s the only way we can discern from answers there what things the respondent was thinking of. Did they think “gradually pushed verbally” meant the same thing as “asked her repeatedly”? If so, they should not have folded the same ask in a separate question with “or physically.” That completely hoses the response relevance here. Or did respondents think “gradually pushed verbally” did not mean “asked her repeatedly” but something else? Then they should have said that—actually specified what they wanted the respondent to confirm, and not something so vague as to include any courting or wooing whatever. For example, what does “push” mean? Continuing to talk about it? Even if she is positively responsive and consenting to that? Or does it mean pushing past an unambiguous “stop” or “no”? The question makes no distinction between these. So its responses have no value for their study.

Their study is full of these terribly worded, misogynistically biased questions, like “Got her away from everyone to somewhere private and under your control,” makes a perfectly normal and ubiquitous dating behavior sound ominous. “Hey guys, get a room” or “Do you want to come up to my place?” or “Do you want to talk in private?” are not “I am now taking you to a second location whether you want to or not.” Yet they are counting all of this as trying to have sex without her consent. That’s nonsensical. It thus matters when you go down the list looking for any strategy that would clearly count as trying to force a woman to have sex without consent, and the first one that qualifies is “gradually pushed verbally or physically,” which as just noted, is so badly worded it does not even distinguish physical coercion from persuasion. So its responses are useless. You have to throw them out. And yet even their useless result was that only a third of men did this. Not 95.1%.

The first time you get in this order to something seemingly nonconsensual is “kept touching and kissing her,” which reports a 59% or so response, but it is out of order, which suggests to me someone entered this data incorrectly. The rest of the list is consistently ordered by frequency, which would suggest this response was actually somewhere between 25% and 30%, not 59%. But assuming the numbers are right and the order is just wrong somehow, the question does not ask whether any of this touching and kissing was consensual, the only useful datum here. Otherwise, this is just obvious: if you are consensually touching and making out with someone, that’s the most accepted onramp to escalation—and everyone, men and women, know this. It’s normal.

They are here describing almost every affectionate date ever. A man (or the woman, as either can happen) asks or tests a small thing like holding hands, or a hand on her waist or thigh, and she’s a yes; then the next move is something more intimate, heeding a yes or no; then asks or moves to kiss, heeding a yes or no; and then to making out; and she’s deciding at each level whether to continue and does. I have described every consensual one-night-stand or first-date-sex in history. I am not describing forcing a woman to have sex against her will. So their question, correctly scored or not, is useless data for their thesis. Because it fails to demarcate the thing they claim to be measuring, by converting normal human dating behavior into rape. Indeed they illicitly change the question to “Kept touching and kissing the woman despite her known unwillingness and lack of consent” when describing these responses—but that is not the wording that was presented to the respondents and that they actually answered.

A similar defect shows up when they decide which startegies to count as “forceful.” They report that, for those strategies, “the modal number of forceful strategies used was zero,” meaning the most common behavior was none of the things they listed, “with 62.6% of the men reporting none,” thus pegging about a third of respondents (bots and humans) as forcible rapists (or at least attempted rapists). But their “forceful strategies” list is heavily populated with mere verbal abuse, not physical coercion. Which definitely pegs those respondents as awful and abusive people a woman won’t want to date, but it’s still not the same thing as intending or attempting rape, much less “by force.”

They conspicuously give us no total figure for how many men in their biased sample actually resorted to force (or drugging or other unquestionably consent-bypassing sexual acts). I doubt it’s 37.4%. Especially since the most common “using force” strategy they list is “Had her come to your house or apartment when she did not want to,” which is hopelessly vague, and unless they mean kidnapping (they do not describe it that way, so I doubt respondents read it that way), does not belong on a list of using force. Yet by being the most common behavior on their “used force” list, it may have padded out that 37.4%. What we want to know is whether the vague “had her” means actually against her will (which would be felony kidnapping), or just persuading her after an initial alternative preference—and that still does not get us to “then I tried to force her to have sex.”

Next, at 24.5% of respondents, was “started undressing before she was ready,” which is also clumsily worded. Ready how? Yes, in the stereotype, suddenly he’s undressing and she’s like, whoah nelly, we didn’t agree to this. But also, just, “Hold on I’m not ready yet” answers to this. I think at least this question implies this was a strategy, so should have elicited responses only if a man tried this as a way of persuading a woman to have sex, but even that is eyerolling more than coercive. The thing we want to know is whether she was okay with the move, or if the man kept undressing or remained partially undressed even after she objected. “Hey let’s all go skinny dipping!” followed by enthusiastic cooperation also answers to this “strategy,” yet has nothing dishonest or coercive to it. If she’s down, it’s ribald flirtation at best. So this isn’t the worst case of this vagueless, but it is still a case. Likewise “kept her drunk/high if she already was” (22%) is also somewhat indicative but still too vague to get entirely clean data from.

In the end, all of their most potentially rapey strategies have less than 20% of respondents doing it, which agrees with what we expect for a biased sample weighted towards rapists and misogynists. All those questions also have wording issues, but less and less so, as there are fewer and fewer ways someone could do any form of what they describe without at least considering raping someone. For example, only 15% reported pursuing a woman “unlikely to tell anyone if you kept going,” which is clumsily worded but at least hard to peg to any behavior that isn’t an intention to commit rape. But that suggests a much lower rate of predation than 95.1%. As do the only explicit references to physical force, which each have a response rate below 10%—or below 15% for intimidating moves. That does not exhaust the conditions qualifying as rape, but none of the others got that much higher, all between 10% and 20%. Especially in a rapist/misogynist biased sample of men, that’s more in line with other studies, and not a surprising result. But it’s so tainted by all the flaws I have surveyed here that this study just isn’t usable for arguing that.

Are There Any Merits to This Study?

The only useful thing I think we can get out of this multiply wrecked study is the data they didn’t publish: the open-answer descriptions by the respondents themselves describing things they had done in their own words. Although a lot of that could still be AI or farmer fiction; so even that is of very limited value, and probably less value than the data like that that they already acquired in previous studies from just hanging out on manosphere forums and asking them open-ended questions about this. Although a lot of that might be fantasy or false narration, too, I would suspect not as much. But since they did not publish those results, their actual publication is entirely useless.

Their 95.1% figure of course meant the count of respondents (awash with possible bots and misogynists) who tried at least one of the strategies they listed. Which means the 78.1% who “told women whatever they want to hear” is doing most of the heavy lifting in that result. What happens when you remove it? Indeed, what happens when you remove all the options too vaguely worded to ascertain whether consensual or nonconsensual, or honest or dishonest, moves are meant? Because the more certainly rapey questions did not get high response rates—though again, they were still alarmingly high. That 1 in 10 respondents are actual rapists is certainly a point of concern. But it’s not “95.1%.” And we already knew the 1 in 10 from many other studies. Which does make men more dangerous than bears. Just not “95.1%” dangerous. And that’s even if this study counted actual randomly selected men and not an overabundance of bots and misogynists. And I have no confidence it did.