The “Red Pill” manosphere has an acronym: AWALT. This stands for All Women Are Like That. It is the assertion that you should accept a given stereotype about women — whatever the ideology claims is true about women in that circumstance.
A lot of effort has been dedicated to explaining why specific claims about female nature are untrue. Indeed, many claims about women in the manosphere are unsupported. However, this article is not going to debate or debunk specific claims. Instead, I have singled out AWALT as a stepping point to a few statistical concepts that I think will be much more valuable for you to understand.
In fact, understanding these statistical concepts may profoundly change the way that you read and interpret research.
(Almost) All Statistics Are Group Figures
Statistics are group measures. The core concept of statistics is generalizing from (seemingly) small sample groups to a whole population. The key word here is group. Statistics are very good at determining population parameters from samples, even seemingly small samples.
A common error people make is attempting to generalize from sample groups to individuals. Basically, the opposite of generalizing from samples to groups. The common statistical methods used in research are not good for this. In fact, there is a term for this error within the field of statistics: the ecological fallacy.
The ecological fallacy
What is the ecological fallacy? This refers to making the incorrect assumption that a group statistic will apply to a given individual within the group.
Most people make this assumption. Even highly educated people, people who have advanced degrees and who perhaps should know better, commonly make this mistake. (This is likely a symptom of limited statistical training, even through advanced degree programs.)
It’s understandable to make this mistake, because it’s actually intuitive. This is how the mind works. We cognitively group things together by similarity. We tend to assume that traits of a group describe individuals within that group.
After all, isn’t that the purpose of research? Especially in psychology. Especially in a subfield like individual differences. Young people enter psychology in large part, if you ask them, because they want to understand themselves better. This is a common reason cited for entry into the field. We want to take the results of research as insights into the minds of individuals. As insights into our own mind.
Unfortunately, the statistical methods we use are not designed to do this.
Let me give you an example below to illustrate the ecological fallacy:
The ecological fallacy: an example from personality psychology
Hypothetically I sample 100 men and 100 women. I administer a psychometric device; a personality test. Let’s say it’s a Big Five measure.
After I run the results, I find that there is a correlation between sex and the personality trait agreeableness. Let’s say it is .48, or 48% (Weisberg et al., 2011).
The correlation is a relationship between the mean, or average, scores of each group.
However, most individuals within the group do not share the same score as the mean. As usual, in statistics we see a distribution that follows a normal curve.
Here is a chart of the distributions from the study referenced above:
Look at how much the male and female distributions overlap. There is more overlap than disconnect in the two distributions. Mean or average values, which are used in our common statistical tests, give us the illusion of two discrete populations. This is how we treat them; our tests are comparisons of means. We are checking to see if the means are different beyond statistical significance.
In fact, most individuals in samples of men and women are closer to one another than the .48 or 48% correlation coefficient would imply (or, more accurately, the way people often misunderstand correlation coefficients).
A Quick Quiz
Knowing that the correlation coefficient is .48, if you pick a random woman out of the population how much more likely is she to score high in agreeableness than a random man?
Many people would say .48, or 48%
How much higher would a given individual woman in the samples score then an individual man?
Again, some people would say 48% higher.
Both of these responses would be incorrect.
48% is the group difference. If you select an individual from the group, you cannot tell how much higher or lower they will score based on the correlation coefficient.
A correlation coefficient is not the probability that a given woman will be higher than a given man in agreeableness.
It is not the degree to which an individual woman is higher than a given man.
In fact, you can’t even know if most women would score higher than most men from a correlation coefficient alone. Thanks to the potential for skewness of distributions, a mean can be higher than most individual scores in a sample!
Beyond the ecological fallacy
If you have formally studied statistics, up to this point none of this may be new information. If you haven’t, this already may be quite a revelation. Most people on the street are inferring facts about individuals from studies. Learning that it doesn’t work can come as quite a shock.
But it goes deeper than that. What if I also told you that not only can you not make these kinds of individual inferences, but if you did you would be probabilistically wrong most of the time? That a guess based on chance would often be better than a guess based on a correlation from established research?
That’s right. When correlation coefficients are small, less than .5, you would be better off making a random guess then using the correlation coefficient to infer traits of an individual!
I want to show you a recent paper by René Mõttus (2021), Editor of the European Journal of Personality. In this paper he describes the math. He also shows you how to determine the probability a correlation will apply to a given individual based on a correlation coefficient, by using a software package in the R programming language. (R is a language primarily used in psychology research and data science.)
TACT; trisecting and cross-tabulating to determine what correlations mean for individuals
TACT describes a methodology of putting individuals into “low,” “medium” and “high” groups across both dependent and independent variables that show a correlation between groups.
TACT seems, to me, like it is intuitively easy to grasp as a concept. Take your scatterplot — the plotted individual points of data in a sample — and superimpose a 3X3 grid on top of these. See figure one from René’s paper:
These are four separate correlations, representing correlation coefficients 0, .25, .50, and 1. No correlation (0) to a perfect correlation (1).
When no correlation exists, all nine slots of the grid will contain the same number of individuals. If a correlation is perfect, at 1, the three diagonal slots from left to right upward will contain all individuals. If the correlation were negative, it would be the three diagonal slots downward.
What this does is tell you precisely how many individuals in the sample fall into a category of low, medium or high.
This lets you know the probability for where a given individual will fall. It shows you how many individuals in the sample match the two correlated variables.
As René points out in this paper, most correlations in psychology will not exceed .3 or 30%. This is why we refer to a correlation of .3 as medium (Cohen, 1988), even when across other sciences .5 is the convention for what a medium correlation is. This means that most correlations in psychology will look like the two plots on the left of Figure 1.
How correlations poorly predict individual outcomes
Let’s look at some real data reported in this paper. This is a scatter plot of the correlation between the Big Five personality trait conscientiousness and general health, derived from a recent meta-analysis.
The difference between someone high and low in conscientiousness scoring high or low in general health is only 3.9% (35.3 – 31.4 = 3.9), given a small correlation of .05.
This is barely above or below chance; chance in this case being approximately 33.3%, given the split into three categories of low, medium and high.
In other words, knowing that there is a correlation between conscientiousness and health at .05 only lets us guess that any given individual is more or less healthy based on their conscientiousness by about 1% greater than chance!
An example from sociosexuality research
Let’s look at an example from some research related to sexual behavior; how partner count and sociosexuality predict infidelity. (Rodrigues et al., 2017)
This study has a measure of sociosexual behavior and sociosexual desire. (Sociosexuality, by the way, refers to a willingness to have sex outside of a relationship.) The correlation coefficients of each of these on past sexual infidelity are, respectively, .8 and .2.
Let’s say your potential date takes a similar test intended to measure their sociosexuality. You find that they score high in sociosexual desire, for example one standard deviation away from the mean. Does this mean that they are more likely to be unfaithful? As you may have guessed from a correlation coefficient less than half of our previous example, the probability that one individual has been unfaithful in the past based on the measure of sociosexual desire is no greater than chance.
What if they score high on sociosexual behavior, which had a correlation coefficient of .8 with infidelity? Now you are looking at a correlation that is high enough where you may be able to predict something meaningful about an individual based on their own score. Assuming a distribution similar to our examples above (a reasonable assumption given normal distributions), this would let us predict an individual’s probability for past unfaithfulness at ~50% above chance.
The probability that an individual will match
Figure 2 from René’s paper shows us the probability that an individual will match for a given correlation coefficient:
For .3 and below, correlations between groups predict individual scores very poorly. We see that the likelihood that high scores match low scores, as well as other outcomes, cluster together.
We really don’t start getting a reasonable predictor until correlations of around .5. Correlations don’t start becoming good predictors of individual scores until as high as .8.
Remember our previous correlation of .43 for Big Five trait agreeableness? Knowing that someone is a woman versus a man won’t predict their agreeableness much better than chance.
Implications for AWALT
Only a sliver of AWALT ideology is actually based in sex difference research. For this small portion of AWALT, you should be mindful of effect sizes. It is not sufficient to say that a difference between men and women has been established, nor that women do X more. Unless the size of the correlation is very high it’s not going to tell you anything beyond chance about a given woman that you are interacting with.
AWALT as dating advice; an objection from the Red Pill
Most of AWALT was never based on behavioral research to begin with. Rather, the stereotypes associated with AWALT were designed as dating rules for men to follow. AWALT represents a collection of expectations that men are encouraged to have for female behavior.
The primary intent may not be to accurately describe female behavior, although I’m sure AWALT proponents will say that it does. It is to give men a set of tools for protecting themselves emotionally.
By expecting the worst behavior from women, men are able to emotionally brace themselves in the case the behavior does occur.
Think of it as Pascal’s Wager for dating. It doesn’t even matter if all women truly are like that. The assumption that they are, it is claimed, will result in a better outcome for men.
The “facts and logic” crowd might reflect and admit that this is an emotional position. It is strategic, it could arguably be beneficial, but it is divorced from the truth about how individuals behave — complex, varied and disuniform more often than similar.
Women have their own version of AWALT as well. Any time we see “all men” in the discourse it is the same thing. Female dating advice often encourages women to be cautious of men, to treat all men as potential aggressors.
Men in Red Pill spaces hate the “all men” discourse, especially when it implies men may be physically violent. They want to be seen as the individuals they are. They don’t want to be grouped unfairly with antisocial people.
Is this one of the times where “both sides” is true? Are both sides just a mirror of one another in this style of discourse? You can debate that among yourselves. There may be qualitative differences between what men and women are accused of. However, neither most women, nor most men, engage in the most extreme antisocial behavior that they are accused of as a group.
Outgroup homogeneity bias
There is a cognitive bias called the outgroup homogeneity bias. This is exactly what it sounds like — a tendency to overestimate how similar members of an outgroup are. When men generalize all women, or when women generalize all men, it is likely that this cognitive bias plays a role.
We also have an ingroup heterogeneity bias. This is the opposite. It is the tendency to overestimate how different, unique and special we are as individuals within our ingroup. Additionally, how unique members our own ingroup are. Truly, cognitively, we as ingroup members are the special snowflakes.
These cognitive biases may have emerged, in particular outgroup homogeneity bias, because they protect us. Even if only 1% of a group is dangerous, if we avoid that entire group, we also avoid that 1%. It is a highly risk averse strategy.
I think this explains some of the appeal of AWALT. It would be consistent with my observation that the manosphere appeals to men who are very risk averse. If you don’t ever want to get cheated on, never dating a woman (ie MGTOW) is 100% guaranteed to prevent that.
The flip side is that highly risk-averse strategies don’t get the goods. If you never date a woman, you will not have a relationship (obviously). If you don’t get emotionally close to women, if you don’t trust your partner (because of AWALT), you will never have a good relationship.
The unwillingness to be vulnerable is not a logical position, it is an emotional position. It is fear.
The unwillingness, or the inability, to treat women as individuals, the reliance on a strategy like AWALT, is also cognitively lazy. You need to develop the skills to assess individuals, to predict individual behavior from one-on-one interactions. AWALT is a concession; you understand women so little that even when interacting with women in individual intimate environments you are unable to tell what they are going to do.
This applies to women, too.
Everything in the previous section, as well as everything I have written in this article, applies to women too. That is all for this section.
Dating advice; the part where I share some opinions
First, if you are a man or a woman and your opinion of the opposite sex is mostly negative (AWALT, “all men,” etc.) — you need to fix it. I’m not going to tell you how to fix it here. But I will tell you that it makes you an unfit partner to be with. You can’t be in a relationship with someone that you view negatively, as less, aggressive, evil, or what have you. Classically the psychologist John Gottman identified the attitude of contempt as being the strongest predictor of relationship failure. Ideologies like AWALT are basically contempt factories, churning out dysfunctional individuals and relationship failures.
Second, don’t tell yourself that you care about “facts and logic” if you subscribe to AWALT. I have covered in this article how it crosses over with the ecological fallacy and how it is rooted in cognitive biases. I don’t need to go over that again. It is simply lazy thinking.
Third, you need to be okay with vulnerability in a relationship. You need to be okay with the fact that you will get hurt. You need to be resilient enough to deal with that hurt when it happens. Building emotional walls around your relationships, which is what AWALT is, will limit them. In its most extreme, perhaps if taken to its most logical conclusion, it will make you not want to enter relationships at all. Basically MGTOW.
Fourth, you must develop the skills to assess individuals through individual interactions. This makes AWALT unnecessary. AWALT was designed as a crutch for men who were unable to do this. This is understandable, because this appeals primarily to very young men who do not have a large amount of experience in seeing red flags or predicting individual behavior. But you can’t live with a crutch forever. At some point you have to learn how to discern good people from bad people. Research can help with this, despite the limitations that I listed in this article. Importantly, experience is what will help you with this most. You need to date. You need to make mistakes. You need to learn the patterns that lead up to bad outcomes so that you can avoid them in your future relationships.
- The ecological fallacy means that you cannot generalize research results to individuals.
- Even large correlations in psychology (.5) often will not predict an individual trait beyond chance.
- Only very large correlations (>.8) begin to look like good predictors for an individual score within a sample.
- Outgroup homogeneity bias will make you overestimate the similarity of people not in your in group (for example, women if you are a man).
- Embrace cognitive complexity and nuance; learn the actual skills needed to assess individual behavior so that you do not have to rely on a crutch like AWALT.
Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers.
Mõttus, R. (2021, August 16). How correlations can (not) be applied to individual people: A tutorial for researchers, students and the public. https://doi.org/10.31234/osf.io/bpm9y
Rodrigues, D., Lopes, D., & Pereira, M. (2017). Sociosexuality, commitment, sexual infidelity, and perceptions of infidelity: Data from the second love web site. The Journal of Sex Research, 54(2), 241-253.
Weisberg, Y. J., DeYoung, C. G., & Hirsh, J. B. (2011). Gender differences in personality across the ten aspects of the Big Five. Frontiers in psychology, 2, 178.