Some estimates suggest that men are more likely than women to cheat on their spouses—at some points in time, more than twice as likely. Why?

The answer seems obvious. The more sex a man has, the more likely he is to get a woman pregnant. The more sex he has with more than one woman, the more likely he is to get more than one woman pregnant. The more women he gets pregnant, the more children he will have, and the more of his genes that will live on in future generations. Our selfish genes in action.

The poor woman is not so lucky, as getting pregnant once removes her from the system for at least nine months, practically speaking. But why, then, do at least some women cheat, too? With a little imagination, we can make sense of this. It turns out that cheating provides an opportunity for the woman to get better genes from fitter donors. She can maximize her limited opportunities to get pregnant and give birth. It is a matter of quality, then, not quantity.

So, if we believe the statistics, it seems the problem is in our genes. Or is it?

Speculations of this kind about the evolutionary origins of human behavior are popular and are the bread and butter of so-called Evolutionary Psychology. Evolutionary psychologists contend that much of our behavior has been hardwired into our brains by natural selection. They propose underlying psychological programs, evolved from the adaptive problems faced by our ancestors, that generate different actions in different environments. Things can go very wrong, though, when our environments differ in important ways from those of our ancestors.

The field has given off much heat but very little light. This is unfortunate, as the heat stands to keep people away from the light produced by more rigorous experimental analyses of behavior. Inventing inferred psychological programs haunting us from our Pleistocene past does not explain how the specific behavior observed is acquired, maintained, or brought under the control of present circumstances. (Never mind that the evolutionary history attributed to the program is often inferred from the very behavior it was invented to explain. Such accounts are just-so stories.)

Selection by Consequences

Learning seems a much more powerful and plausible explanation for much of what evolutionary psychologists attribute to our stone-age genes. Learning does not require our brains to be hardwired with many special-purpose modules guiding our behavior like Pleistocene zombies. But evolution certainly plays a central role. What has evolved is an exquisite behavioral sensitivity to our environment, allowing our behavior to be shaped by the pressures we face rather than the pressures our ancestors faced. More than 100 years of rigorous experimental evidence, in the laboratory and the field, supplies us a rich framework with which to understand human behavior (and the behavior of other animals).

There are many parallels between biological evolution and behavioral evolution, just not the kind usually proposed by evolutionary psychologists.

Changing Fast, Changing Slow

There are three processes critical to biological evolution: variation, selection, and retention (via replication). When individual members of a species vary in some way, in some place, and at some point in time, one or more of those variations might prove beneficial in the competition for survival. In this way, certain traits or characteristics are favored or “selected.” Those advantaged individuals are more likely to reproduce and to do so more often, meaning more of their genetic material gets into the next generation. If the advantage is heritable, there will be more of their descendants with that advantage. The beneficial trait will be retained because the corresponding genes will replicate.

And the cycle will continue, leading to more of that characteristic in future generations. Unless, of course, the environment changes. An evolved characteristic is beneficial only in environments similar in some critical ways to the environment prevailing at the time of selection. This is where learning comes in. A chief benefit of learning over natural selection is that when an environment changes rapidly, the individual can change, too.

The Darwin of Behavior Science

As it turns out, the processes of variation, selection, and retention are common features of both biological and behavioral evolution. At the behavioral level, behavior varies from one situation to another and from one time to another in the same situation. Certain types of behavior will produce beneficial outcomes, whereas others will not. The beneficial outcomes “select” the behavior that produced them, leading to that behavior being repeated (due to changes in the nervous system) in similar circumstances in the future. This is selection by consequences during the lifetime of individual organisms.

We know about this process of behavioral evolution, and the details of how it occurs, chiefly from the work of the late American psychologist B. F. Skinner and his many students and colleagues. In the 20th century, Skinner was the most notable champion of such selectionist principles as the basic mechanism of learning, but he was by no means the only psychologist to suggest something like this.

At the turn of the 20th century, Edward L. Thorndike, one of the most famous and influential psychologists of all time, was conducting his research on animal intelligence that resulted in his formulation of the Law of Effect—one of the very few laws in psychology.

Thorndike placed hungry cats in a handmade wooden “puzzle box,” with a dish of food placed outside and in view. With cats, this could be enough to get you killed. But Thorndike was very brave. The boxes differed from one to another, and they weren’t pretty, but they all had some way that the cat could open the box and escape, such as a lever that opened a door when pressed.

Thorndike observed that, at first, the cats took quite a while to accidentally make the response that opened the box, essentially stumbling on the right solution not by some flash of insight, but by trial and error. After a cat finally did escape, it was placed back in the box (presumably before it could exact revenge) and the process was repeated many times, across many days. Over time, and as a result of their experiences, the cats were quicker and quicker to escape, ultimately pressing the lever as soon as they were put into the box.

This work is transparently selectionist, even if this was not necessarily recognized by Thorndike. He reported an ever-decreasing variety of behaviors exhibited by the cats as they were repeatedly exposed to the puzzle box. Over time, the wide range of behavior originally seen (e.g., scratching, sniffing, whining, plotting to kill Thorndike) narrowed so as to be dominated by whatever behavior most reliably opened the box and resulted in escape and access to food.

But it took B. F. Skinner to uncover the key working features of operant conditioning, a term he used to collectively refer to the basic processes governing much of human and nonhuman behavior. Operant conditioning encompasses a relatively small number of very powerful principles, including reinforcement, punishment, extinction, and stimulus control. Like evolution by natural selection, operant conditioning is a simple mechanism that gives rise to great complexity. In short, reinforcement selects behavioral characteristics of the individual in much the same way survival selects other (e.g., physical) characteristics of a species.

In Skinner’s terms, behavior that produces beneficial outcomes (reinforcers) is more likely to occur in the future (an outcome called reinforcement). Additionally, the behavior is most likely to occur in the future in situations that are like situations in which the behavior has been reinforced in the past. This acquired effect of the environment is referred to as stimulus control, and the controlling stimuli are called discriminative stimuli. In this way, the reinforcer selects a relationship between a set of environmental circumstances (discriminative stimuli) and certain types of behavior. That is, the reinforcer reinforces (selects) the relationship between the circumstances and behavior, meaning that the behavior is more likely to occur when those circumstances present themselves. Essentially the opposite effect is seen with punishment, whereby certain behaviors are less likely to occur in certain situations because, in the past, those behaviors have produced unfavorable outcomes (punishers) in those situations.

We Were Born to Be Made

Learning requires, at the very least, a behaving organism that is the product of natural selection. Even more than this can be granted to evolution, however. Our genes influence our sensitivity to certain aspects of our environment and thereby play a crucial role in determining the extent to which those things influence our behavior. In this way, our evolutionary past most certainly plays a role in determining our behavior, but not in any direct way, at least in most circumstances.

As for our unfaithful men and women, it is unlikely that genes control the complex sequences of behavior involved in promiscuity, especially as the specific actions will necessarily be quite different across societies and across the ages. Renting a hotel room or pretending to have to work late is not a pattern of behavior selected by evolution. But heightened sensitivity to sexual reinforcement could itself be an evolved trait, and that sensitivity would influence what is learned as a result of life experiences. If sexual contact is a powerful reinforcer, any behavior that produces it will be more likely to occur in the situations where there are signs it is available. And if sexual contact is an especially potent reinforcer, it will select behavior more readily than other, less potent reinforcers, like social approval. As a result, you might end up spending more time at the bar and less time at home. (Of course, there could be many other selecting consequences, for and against, at work in both settings.)

Genes play a crucial role in the ability of the organism to learn and behave, but the main genetic role seems to be to provide the raw material on which our environment works. Even our genes—or at least what they do—can be changed as a result of our experiences. To quote science journalist Sharon Begley, “Evolution indeed sculpted the human brain. But it worked in malleable plastic, not stone, bequeathing us flexible minds that can take stock of the world and adapt to it.”

A better evolutionary psychology would begin there: not with speculative stories in which genes encode the complicated behavior of modern people, but with an evolved organism whose remarkable susceptibility to present consequences allows new behavior to be selected within a single lifetime. We have inherited an ability to learn from our experiences because doing so allows us to survive in an ever-changing world.

Will striking a power pose imbue you with confidence an compel you to success? Do babies have a moral compass? Can you predict the future with porn? No, no, and most certainly no.

These schizophrenic headlines reflect a schizophrenic discipline: modern-day psychology. And modern-day psychology is a discipline in crisis. Every day it seems a new study surfaces that fails to replicate a previous study. Even the most heralded of psychology findings have fallen down. It is a replication crisis, if you believe the headlines.

Why is psychology in this crisis? Theories abound. The most promising candidates include an over reliance on statistical inference and statistical significance, along with a wonky scientific system that incentivizes the wrong things. Statistical significance, rather than real significance, is the holy grail, and research funding and publications in premiere scientific journals go to the sexiest and most surprising findings. Not to mention that in the publish-or-perish world of academia, quantity matters more than quality.

The system is sick and we are all tired. A cure is unlikely in the near future, so the best we can do is learn to live with the symptoms. Here are five tips to help you do just that.

1. Statistical Significance is Not Significant

Statistical significance is not the same thing as actual significance. Not even close. Something can be statistically significant but absolutely unimportant. Despite what many researchers think, statistical significance doesn’t even tell you how likely it is that your results are due to chance, or how likely it is that you would get the same results if you did the experiment again. A statistical significance level says something much less interesting: How rare your results are given that the null hypothesis is true.

So what is a null hypothesis? Well, a null hypothesis is not that exciting, either. It is a simple statement that the experimental group does not differ from the control group with respect to what you’re measuring. A statistically significant finding means that you would expect to get your data rather infrequently if there is no real difference between your groups. If the results are statistically significant at the .05 level, you would expect to get those results about 5 out of every 100 times even if the groups are not different. Two things are very important to remember about this sort of finding. First, the math assumes that the null hypothesis is true, it is not a way to decide if it is. Second, it assumes that the data you have are real and are valid. That is, the statistical significance really just tells you how rare your data are given that the groups are the same. It does not tell you how likely the groups are to be the same or how trustworthy your data are.

Consider the following study that evaluated the extent to which an over-the-counter pain reliever might also relieve the pain of rejection. All subjects were asked to take two pills, provided by the experimenter, each day—one in the morning and one at night. Half of the subjects took two doses of acetaminophen, while the other half took two placebos. All subjects also were asked to rate their “emotional pain” on a scale each day. The researchers reported that the acetaminophen group experienced less emotional pain across weeks than did the placebo group. But how much less pain? Well, remember that scale? It was a 5-point scale, ranging from 0 (not at all) to 5 (extremely). What was the actual average difference between the acetaminophen and placebo groups on this scale? The acetaminophen group reported, on average, a 2.2, whereas the placebo group reported, on average, a 2.5. Only about a .3 difference. And not on a blood test or time in a 50-yard dash. On a 5-point scale that simply asked them to rate how characteristic of them each of six statements (e.g., “Being teased hurts my feelings”) was that day. (And, by the way, neither score suggested emotional pain based on the 0-5 scale.) The difference might be statistically significant, but it is far from actually significant.

2. The File Drawer is Deep and Full

Because of the emphasis on larger group designs, independent replications are essential. No single study can tell us much of anything. Statistical significance (see above) does not tell us anything about how likely an effect is to be replicated because only replications can tell us that. Unfortunately, the scientific “system” (publication outlets, funding agencies) incentivizes shiny new findings and discourages “mere” replication. Science pays a great deal of lip service to the role of replication as the driving force behind the incremental progress that characterizes good science. Having passed the lips, however, replications struggle to be heard. The most prominent journals don’t like to publish replications and, even more, they rarely publish findings that fail to reach the heralded level of statistical significance. This means that the publication record can be misleading because it is more likely to feature papers reporting statistically significant effectsthan those that do not. But we cannot assess the veracity of the evidence unless we know about all of the research, both hits and misses.

Now let’s return to the heralded .05 significance level. This tells you that you would expect to get your dataset 5 out of every 100 times even if there is no difference between your experimental group and your control group. This is informative only if you know how many such datasets there are. If 10 similar experiments have been conducted and yours is the only one showing a statistically significant difference between the experimental and control groups, you would reasonably conclude that your dataset is likely one of the 5 out of 100—that is, there is no effect. However, if 8 of the 10 similar experiments had revealed statistically significant differences, you might reasonably suspect you are on to something—that is, there might well be an effect. But you need to see the whole picture. You need the hits and the misses.

Consider the study described previously, in which subjects who took acetaminophen reported less emotional pain than did subjects who took a placebo. The actual differences are quite small and, even though they were statistically significant, might very well be one of the 5 out of 100 datasets you would expect even if there were no actual differences between the two groups attributable to the drug. The only way to know is to conduct other identical or similar studies and see how many produce similar results. If other attempts do not produce similar results, for example statistically significant differences between acetaminophen and placebo groups, then it probably is the case that the significant effect was one of the five expected by chance.

3. Bigger is Not Better

Numbers tend to impress, and large numbers tend to awe. This is especially true when it comes to the numbers of people studied in psychology research. The more the better, it seems. But not necessarily. It seems reasonable to assume that a study with 100 participants would be much stronger than one with 10 participants. However, the extent to which this is true depends on the question being asked and the methods used to get the answer. If a grocer wants to know how many people will buy more of an item if the item is placed near the cash register, studying many people probably is better than studying just a few, because what matters is average sales, not any particular sale. However, if a researcher wants to know which people will buy healthier food if that food is placed near the cash register, and why they do so, a large group study probably isn’t the best approach. But, again, if the goal is simply to find out how to sell more fruit to more people, studying more people is fine.

If the goal is to understand what will lead people to buy healthier food, however, we would be missing the mark with a large group for several reasons. First, studying large numbers of people will reveal how many will buy the fruit when it is near the register as opposed to when it is placed somewhere else. But it won’t tell us is whether this is actually reflecting a change in the purchase patterns of the people who do buy the fruit near the register. For example, we might be seeing an effect on people who already buy fruit, who end up buying a little more, and no change in the purchases of people who already don’t buy fruit. Second, because of the practical difficulties with studying large numbers of people, we are likely to get only a one-time snapshot of each person’s reaction to the fruit placement. We will not be able to see how they respond over time as they repeatedly encounter the fruit. Will they buy the fruit initially because of the novelty of the fruit placement and then return to their previous habits? That is, we won’t know if moving the fruit produces only a temporary disruption in a person’s habits or if that disruption will endure. These are important things to determine as we evaluate the effects of our manipulation.

Going back to the acetaminophen as emotional panacea study, we can’t tell much about individual effects from the grouped data. This is especially important when the differences in the averages are so small, because just a few subjects could dramatically alter the outcomes. For example, suppose that just a few subjects in the acetaminophen group responded in some way to the drug, and that the other subjects had no response. Those few subjects could be enough to change the average rating such that a small difference between groups is observed. But the result would be due to only a few people changing and it should lead you to ask, what about the others? Which brings us to the next important point.

4. An Average Effect is Not Really an Effect

Suppose that five students take an exam, with the resulting scores (out of 100) as follows: 20, 30, 30, 80, 80. The average score is 48. Now look at the actual scores again. None of the students scored even near 48, so that score does not reasonably describe the performance of any student. Now suppose that, because of the “overall” low scores, a new teaching strategy is introduced. The same students are tested again and score as follows: 15, 25, 100, 70, 85. The new average is 59. Still not great, but it is an improvement. However, look at the individual scores again. Only a single student improved substantially, going from a 30 to a 100, with three of the other students scoring lower. And moreover, just like with the first set of scores, the average does not reflect any actual student score. The average score is not a real thing because it never happened.

But psychology lives by the average. Compare this to science like physics. How useful would it be to show that, on average, 6 out of 10 rockets you launched made it to orbit? Not very. When a rocket doesn’t work, it’s a big deal and the source of the problem is diagnosed so that it can be fixed. The hard sciences are successful because they do what science is supposed to do: predict and control. The standards for knowing seem to be far less in psychology. When a building falls down, people demand to know why. The assumption is that it is due to faulty design or faulty construction. (Or a natural or man-made disaster.) People do not assume that sometimes buildings just fall down. But this seems to be the assumption when it comes to human behavior. The average is good enough because it is all we can expect to achieve. Sometimes behavior doesn’t follow the rules and you can’t know why. But actually we can, and we should, at least if we want to progress as a science and in useful practice.

5. Sexy Sells, But Don’t Buy It

The more surprising a finding, the more likely you are to see it in a prestigious journal and in headlines across news outlets and on social media. However, the reality is that the more surprising the finding, the more likely it is to be wrong. As the philosopher David Hume cautioned, extraordinary claims require extraordinary evidence. Unfortunately, what passes as evidence for extraordinary claims is often less than ordinary. The lower the likelihood of a finding prior to the experiment, the lower the likelihood that finding is true, if it is found, after the experiment. That is, it takes more to prove an unlikely finding because, well, the finding is very unlikely to be true. This is both common sense and an issue of basic mathematical probability. If the lights flicker in your house, you might wonder if it’s a ghost. If they keep flickering, you might call out an electrician to look things over. If that electrician finds nothing wrong with the wiring, should you conclude that it is, indeed, ghosts? No, you would need much more evidence than that.

We Can Do Better

Statistical significance can be significant, the file drawer can be emptied, bigger can be better, averages can be helpful, and sexy is sometimes worth buying. It’s just that none of this is typically true for psychology research. More often than not, statistics, publication bias, and misplaced incentives seem to lead many researchers and readers astray. I think we can do better.

BEHAVIOR, AND NOTHING MORE

To be a behaviorist is to hold that psychology is the study of behavior, and nothing more.

If I demonstrate that the rat presses the lever more when food immediately follows, and it stops pressing the lever when food no longer follows, and if I can repeat this demonstration, at will, there seems nothing controversial about what I would say. The food following the lever pressing is the reason the rat presses the lever. For convenience, we can call this reinforcement, and we say that the food reinforces lever pressing. And nothing more.

But if I say that a rat presses the lever more often when food follows each lever press because it “knows” or “expects” or “remembers” that lever pressing produce food, I am saying something more.

THE BURDEN OF PROOF

You cannot argue with the statement that food following the lever press increases lever pressing when that is the very thing that is seen. You can argue that there has to be more to it, but then the burden of proof is on you. As a radical behaviorist, I consider myself explanatorily conservative. That is, I tend toward straightforward description and parsimony, perhaps at the expense of mystery and intrigue and all sorts of fun.

So where is that evidence for the something more, such as the knowledge or expectation? We cannot remove the knowledge or expectation and stop lever pressing, except by stopping the food. Because if we don’t stop the food, the rat will know that food follows and will continue to expect it. So we’re back to the food, added or removed. So what role does the knowledge or expectation play? And, more importantly, what is the evidence for those things, at all? They are something more.

There must be independent evidence for the “something more.” The evidence can’t be the rat pressing the lever more when food follows. That is circular. With our labeling of the relation between lever pressing and food as reinforcement, we are not locked in a circularity. We can stop the food and stop the lever press. We can bring the food back and bring back lever pressing. And so on. And nothing more.

THE GHOSTS IN THE MACHINE

Unquestionably, the complete picture, of the lever press or anything else, will be seen only when we also know about the physiological changes that take place as the organism behaves. The physiologist might observe that a certain collection of neurons fire in a certain pattern as a behavioral episode unfolds, with traceable effects on the musculoskeletal system of the behaving organism. But that is a different level of analysis, belonging to the physiologist. And it doesn’t replace a behavior analysis, it complements it.

So where does the psychologist, who is not a behaviorist, work? Somewhere in between the physiology and the behavior. Knowledge, expectations, associations, representations, memories and so on, all are things other than the observed physiology or the observed behavior. There needs to be evidence of their existence, because they are something more.

For example, if I use the observed activation of a network of neurons, and the corresponding chemical happenings, as evidence of an expectation, I am saying something more. But there is no evidence of the expectation, there is only the observation of the electrical and chemical actions of the nervous system and the corresponding behavior occurring in a certain circumstance. And nothing more.

The same for memory. How do you know somebody has a memory? They can do something now that they learned before. They can open the combination lock. They can greet you by name. They can answer a question on the midterm exam. They press the lever. All behavior. How else could you possibly know someone remembered? They have to do something, and that is what you observe. You don’t ever observe a memory, you see the person doing things we call remembering. And nothing more.

Physiologists might identify the neurons that fire when they do (“remember”) something now, and they might even determine that they are the same neurons that fired when they were learning it before. But that’s the behavior of the nervous system, not memory. That’s like measuring your heartbeat while you are running and calling it your fitness. That is something more.

Anybody, with the proper tools and some time, can see the environment-behavior relations and the environment-physiology-behavior relations. But nobody has seen a so-called mental process, memory or otherwise. Extraordinary claims demand extraordinary evidence. They require something more.

OUCH! MY QUICKNESS HURTS

Even finding that damage to some area of the brain impairs some particular performance is not evidence of anything other than the correlated actions of that part of the nervous system and that part of behavior. If there is brain damage and a resulting loss of the ability to say the names of people you just met, that is something, but it is not evidence of a memory.

Assuming it is evidence of something more is like seeing that an athlete runs slower when they injure their hamstring and concluding that the hamstring must be where their “quickness” is stored. Later, when the hamstring is healed, the athlete runs faster. But then they sprain their ankle and, wouldn’t you know it, they run slower again. Did the quickness migrate to their ankle?

With something like memory or any other hypothetical construct, we might see that damage to some specific area of the brain reliably results in an inability to say people’s names after we meet them, like the athlete reliably runs slower after an injury to their hamstring. But, of course, we might observe a lessened ability to say people’s names later following hearing loss, too. Or, for that matter, when traveling in a country where the names sound very different from what we regularly hear in our own country. And so on.

ALL PSYCHOLOGISTS ARE BEHAVIORISTS

Now, some might argue that “memory” simply describes the occurrence of these two types of events—behavioral and physiological. But I worry about this because we simply can’t resist reifying terms once we use them. So memory inevitably becomes a “thing” that we try to study, even though we can never do so directly or, I would argue, fruitfully.

True, we might see changes in behavior as a result of changes in the nervous system. But we have to ask, why did we see those changes in the nervous system? That inevitably leads us back out to the environment and something that the organism experienced.

We have behavior of the whole organism occurring in certain circumstances, and we have the behavior of the nervous system occurring in certain circumstances. And that’s enough, as far as I am concerned. Invoking “memory” adds nothing to the account. There’s no room, and more importantly, no evidence for it.

Pick your poison: memory, cognition, awareness, expectation, and so on. Maybe there is more than behavior for psychology, but I don’t think we’ve gotten to a point where we should feel compelled to look for it. At least not yet.

At the end of the day, all psychologists are behaviorists, whether they know it or not, and even whether they like it or not. That’s because they can observe behavior, and nothing more. But most psychologists ignore this and talk, almost exclusively, about something more.

The 2026 annual meeting of the Association for Behavior Analysis, International (ABAI) will soon be held in San Francisco. As far as I know, it is the largest gathering of behavior analysts in the world. Thousands of us descend on some part of the U.S., en masse. It is a sight to behold, and I look forward to it all year. It’s the only conference where you can still find enough talks that showcase the depth and diversity of the field to keep you busy for several days of conferencing. And by that I mean that it is the only conference where you can, just barely, spend all your time attending talks that are not about autism.

Indeed, an anthropologist studying the odd creature called Behaviorist Analyticus would have no choice but to conclude that it was a practitioner in a niche field within special education, concerned primarily with autism spectrum disorder. This conclusion would be borne out by several key facts.

First, most credentialed (RBT, BCaBA, BCBA) Behaviorists Analyticus say they primarily serve autistic individuals (83%) and/or individuals diagnosed with intellectual or developmental disabilities (3%). Of the few remaining behavior analysts who identify otherwise, such as practicing organizational behavior management (.39%) or working in higher education (.4%) or professional supervision (.16%), many still end up in or around the special education classroom. The organizational behavior is what is happening at the autism service agencies, the professors teach those who will serve in special education settings, and the supervisors supervise those providing special education services. I suspect even many who say they work in (not higher) education (6.34%) probably at least dabble in special education.

Second, the journals written and read by Behaviorist Analyticus tend to lean heavily on research focused on autism or intellectual/developmental disabilities, and on conceptual and professional matters related to the same. Of the 9 articles published in the most recent issue (Spring 2026) of the Journal of Applied Behavior Analysis, the flagship applied research journal in the field, 6 were focused on autism or developmental disabilities, and 5 of those included autistic participants, specifically. It wasn’t a special issue, just a regular old issue. Even Perspectives on Behavior Science, the nondenominational mouthpiece of ABAI, recently published a special issue devoted to “Challenges to applied behavior analysis,” although it was really devoted to papers about autism and service delivery. Of the 17 articles, 8 were solely or mostly about autism service delivery and autism was at least mentioned in 6 others. A few years before that issue, in the same journal, Bill Heward, Tom Critchfield, Derek Reed, Ronnie Detrich, and Jonathan Kimball published what, at first blush, was an encouraging review of the many (350) research areas that applied behavior analysis has touched. A closer reading reveals something less encouraging. Almost all the areas were touched only once or twice. Love ‘em and leave ‘em, apparently. Few programmatic research lines developed. No big problems solved.

Third, the sessions presented and attended by Behaviorist Analyticus at regional and national conferences revolve almost entirely around issues related to service delivery in special education. In the program for the upcoming meeting of the Association for Behavior Analysis, International in San Francisco, CA, autism dominates. The program is divided into 15 areas, with the top 2 entries being Autism (AUT) and Developmental Disabilities (DDA), with the third being Education (EDUC). There are 170 sessions in the AUT area and 88 in DDA. The next largest is EDUC at 74 sessions. There is a lot of overlap across those three areas, with multiples sessions being categorized in two of the three areas, so there are, conservatively, about 300 sessions just in those areas, focused on Autism and/or DDA. And here’s the kicker, close inspection reveals that even many sessions in the other large areas, such as Verbal Behavior and Cognition (VBC) and Clinical /Family/Behavioral Medicine (CBM), are cross-listed with AUT or DDA. So the total number of presentations and posters is probably closer to 400 or more. By contrast, there are 54 sessions categorized as the Experimental Analysis of Behavior (EAB)—and even 8 of those are cross-listed with the AUT category. Looking at the programs of recent or upcoming state conferences reveals an even heavier, sometimes exclusive, focus on autism.

Fourth, the professional organizations populated by Behaviorist Analyticus are run by those working in autism service agencies or in jobs adjacent to them. Of the 47 state and regional organizations affiliated with ABAi that had leadership information available online, 97% (of 422 people who could be identified) worked in the areas of autism and developmental or intellectual disabilities, mostly in practice at service agencies, and many fewer in academic and research positions.^[1] This is a staggering statistic.

The largest professional organization for behavior analysts, the Association for Behavior Analysis International, has done an admirable job populating its Executive Committee with people representing different interests.^[2] They are drawing from the largest pool of behavior analysts possible, though. State and regional organizations can’t draw from that far and wide. Finding someone to serve the organization that doesn’t also serve the autism community is like finding Waldo. Hence the 97%. Absolutely staggering.

THE RISE OF MISPLACED POWER

When he left office in 1961, President Dwight Eisenhower delivered a farewell address that warned the American people about a growing military industrial complex. If you’re unfamiliar, the term was used to describe the intersection of the military, which serves the (safety and security) needs of the nation, and the arms industry, which serves the needs of the military and gains financially by doing so. In his words:

This conjunction of an immense military establishment and a large arms industry is new in the American experience. The total influence-economic, political, even spiritual-is felt in every city, every state house, every office of the Federal government. We recognize the imperative need for this development. Yet we must not fail to comprehend its grave implications. Our toil, resources and livelihood are all involved; so is the very structure of our society. In the councils of government, we must guard against the acquisition of unwarranted influence, whether sought or unsought, by the military-industrial complex. The potential for the disastrous rise of misplaced power exists and will persist.

The concern was that the defense industry will become too large and too powerful an economic force, employing so many and driving infrastructure development around it. The consequences being many, including fear mongering to keep support for defense funding, so much money and other resources pouring into the complex that much less is left for other, equally or more important areas. And here we are. Welcome to today.

Now, I am not the first to think something similar is going on in the autism world, or to invoke the phrase “Autism Industrial Complex,” but those coining it before me have mostly been referring to the broader capitalization of an autism industry within the larger society, not specific to behavior analysis.^[3]

In behavior analysis, the Autism Industrial Complex involves the major professional organizations and the many individuals and clinical organizations (including the private equity firms behind many of them) that provide autism services. The behavior analysis profession—through its associations, credentialing bodies, and training programs—promotes, if not defines, the science of behavior analysis, as well as the standards and scope of practice. The autism services industry operationalizes those standards through service delivery, which necessarily involves staffing and billing. So the profession provides legitimacy, workforce pipelines, and regulation, while the industry provides employment, funding, and professional and political influence.

Many behavior analysts move between these domains, and, over time, this arrangement can embed autism-focused decisions into research, training, credentialing, and funding structures. The big risk is what we are already seeing: the autism focus can come to dominate, and therefore define, the science and practice of behavior analysis, essentially starving out other interests.

DISCLOSURES, RECUSALS, AND GUARDRAILS

Before you flame me in the comments, I realize that it only stands to reason that, if most people in the field have connections to autism service delivery, most people willing to participate in state organizations will be from that group, most people served by those organizations will be from that group, and those people will need professional supports such as specialized research journals and conferences, not to mention professional organizations that will go to bat for their economic interests.

No matter. The fact that this all presents clear and present conflicts of interest still cannot be ignored. The people making decisions about the activities of the organizations have a vested interest in those activities being aimed at autism research and service delivery, and if those organizations spend money on things like continuing education, grants and scholarships, and lobbying state legislatures and other entities for funding for services, they stand to directly benefit.

To be clear, I am not trying to paint those working in autism as power-hungry villains bent on destruction. But even the noblest of the noble can make decisions that are not in the best interest of the larger field, whether they realize it or not. Eisenhower noted as much in his famous address, saying, “In the councils of government, we must guard against the acquisition of unwarranted influence, whether sought or unsought, by the military-industrial complex.” (Emphasis added by me.) So, too, must behavior analysts stand guard.

I am mostly just stirring the pot. We all know it’s much easier to identify problems than offer solutions. Still, I can suggest a few general areas for improvement, fully aware that they are easy to say, harder to do. Behavior analysts should, at the very least, take conflicts of interest more seriously, make concerted efforts to recruit diverse (in terms of specialization) professionals into leadership positions, and invite diverse (in terms of specialization) professionals to present at our many conferences. And let me throw a bomb into the building. Maybe those in organizational leadership positions who work in service delivery should recuse themselves from decisions that could result in organizational money being spent on efforts that would directly affect their bottom line, like lobbying for service reimbursement, etc. (Now you can flame me in the comments.) The problem is that if they do recuse themselves, there will be nobody left to vote.

THE GOOD, THE BAD, THE UGLY

But things are not all bad. The military industrial complex provides many jobs, many of which are high paying, and it provides important infrastructure to many communities. And those employed help to provide an important function: keeping us safe from harm inflicted by outside aggressors, at least in the best of cases. The autism industrial complex does the same for the field of behavior analysis. It provides many jobs, many of which are high paying, all the while performing an important function: helping autistic individuals and their families, at least in the best of cases.

Moreover, on some level, it is entirely appropriate that many of the people making decisions in our professional organizations typify the people they are representing. If the majority of those decisions and resulting actions are to the benefit of those providing and receiving autism services, that seems defensible. They are, far and away, the largest segment of our professional field. Without them, well, there really is no behavior analysis to speak of. The rooms and hallways of the universities and conference centers would be empty, as would the pages of the journals.

So we find ourselves here. Most of the jobs in behavior analysis are in the Complex, so most students are taught to work in the Complex, so most academics and their labs are in or near the Complex, so the books and articles published are mostly about the Complex, so when behavior analysts gather they talk mostly about the Complex. It is a self-perpetuating system. Behavior analysts press the levers associated with the richest schedules of reinforcement. But that means they aren’t pressing other levers. And that means the science doesn’t advance the way it could, and the scope of our application and practice is not expanded. What if they someday “cure” autism or, more likely, stop funding hours and hours of intervention services? Do we risk implosion as a consequence of our specialization.

^[1] I would like to thank three of my graduate students, Annika Cook, Sophia Livingston, and Ana Prieto Razo for poring over the websites of the many state behavior analysis organizations to identify the people in leadership positions and then scouring the web to find their affiliations, publication records, etc. The opinions presented here are mine, though, and should not be unfairly attributed to them. Blame me.

^[2] The Association for Professional Behavior Analysts (APBA), not so much. Perhaps that is to be expected. APBA represents behavior analysts in practice and almost all of those practice by providing services to people on the spectrum. So, to best represent their constituents, if not the field, APBA should probably fill their board with like-minded people.

^[3] When aimed at the field of behavior analysis, the term misses the mark, to some extent, because to square with Eisenhower’s catchy phrase it would need to be something like “Behavior-Analysis Autism Complex.” But that is just clumsy and obscures the inspiration. So “Autism Industrial Complex” it will be.

“If you can’t say something good about someone, sit right here by me.”
Alice Roosevelt Longworth (stitched on her tea table pillow)

Exercise is important. Do you run? Based on the data, you probably don’t. Do you walk? Based on the data, you probably don’t. But you’re more likely to walk than run. Only about 15% of us run, but somewhere around 60% walk, suggesting we admire the idea of movement more than the practice of it. Is running a better cardiovascular workout than walking? Not if you don’t do it. The walk you take will consistently outperform the run you intend to take.

The best treatment is one that works and that people will use. It improves something that we value, does it in a way that we like or will at least tolerate, and it gets results that are meaningful to us and those around us. In the late 1970s, Montrose Wolf popularized the term social validity to describe these three different aspects of behavioral interventions. And the real kicker was that he argued all of this could be assessed *gasp* subjectively. Like, by asking people what they thought. Can you imagine?

It has been a slow burn. Social validity is now regularly discussed, but it is not yet a fire. And most researchers who assess social validity do so by asking people what they think, meaning behavior analysts have seemingly embraced self-reports, at least for this purpose.

But is social validity what people say to us about us? Asking people what they think about what we did might not be the best way to go, with all due respect to the late, great Montrose Wolf. Or at least the way we’ve been asking them might not be the best way.

WHAT WE CAN LEARN FROM THE CAR LOT

For anyone lucky enough never to have encountered one, the car-dealer satisfaction survey is a questionnaire the salesperson asks you to complete after taking your time and money. Although “asked” is a generous term, in the same family as “enhanced interrogation technique.” Begged or prodded, is more like it. Bullied. While still clutching the survey, the salesperson explains just how important it is to them (and their job and their kids and the esteem of the parents) that they get perfect ratings. Any shred of a life they are eking out and any minimal happiness they have or will ever have depends on you giving them positive marks.

The car-dealer satisfaction survey is presented as a request for honest feedback, but by the time you get it, you’ve already been informed that honesty is a hostile act. The salesperson, with admirable efficiency, explains that a 10 means “satisfactory,” a 9 means “career damage,” and anything below that is reserved for animal cruelty and public urination.

The survey really asks what you are willing to say after a brief explanation about the consequences of sincerity. Tell the truth, provided the truth is first cleared through the person whose mortgage depends on suppressing it. This isn’t customer feedback, it’s witness tampering. But at least you get to leave when you’re done.

The car lot is just an extreme example of most satisfaction surveys, including the social validity surveys in applied behavior analysis. (We call them social validity surveys because Mont Wolf did, but they are essentially consumer satisfaction surveys.) The responses on a satisfaction survey are not necessarily a direct measure of anybody’s performance or the acceptability or effectiveness of what that person did. They are verbal responses occasioned by the printed questions on the questionnaire, under the control of a broad, messy, and transient array of discriminative stimuli, motivating operations, and consequences. At the moment of questioning, there are powerful, but maybe fleeting, controlling variables in play.

In my own research, we have seen, over and over, that participants will rate an intervention and their overall experience highly, even though nothing measurable improved and, in some cases, they no longer wish to continue. In four consecutive studies (on the effects of different health-coaching practices), we saw no clear effects of the interventions. Yet only a single participant (out of 12) in any of the studies ever rated a social validity survey question negatively—and that one person did so on only one of the questions. All of the other ratings from all of the other participants were positive. So what are we to make of that survey data? Probably not much.

WE DON’T KNOW WHAT WE DON’T KNOW

How seriously should we take our social validity data? I think right now, it’s hard to say. For one, we don’t know enough about how social validity data are being collected. We know some things, like most studies (82%) don’t report social validity data, and those that do collect it using Likert-type rating scales (81%) at the very end of an intervention (83%).

But behavior, including answering survey questions, occurs at a certain time in a certain situation, and that situation matters. I can’t recall ever seeing a published article, including any of my own, that described the actual circumstances in which social validity data were gathered. Were the researchers in the room? Were responses anonymous? What kind of instructions were provided beforehand? Was the questioner wearing pants?

How can we hope to meaningfully interpret social validity data if we don’t pay attention to the situations in which they are collected? A behavior analysis journal would never publish a study that didn’t describe the conditions under which behavior was assessed or the way the independent variable was manipulated. And for good reason.

The point here is that social validity is about more than just giving a generic survey. Among other things, we need to learn how to better develop and deliver surveys, if that is the way we decide to go about collecting our data. Mont Wolf suggested as much at the very beginning.

[W]e know that social validity measures can be manipulated and abused, but we cannot allow this to lead us to neglect them. Rather, we must establish that set of conditions under which people can be assumed to be the best evaluators of their own treatment needs, procedural preferences, and posttreatment satisfaction. True, we know little about the proper set of conditions, but we must attempt them anyway. We can expect that they will involve education about options, lack of coercion, anonymity, and so on. We can study the effects of these conditions on subjective data…And then we will be better able to control for them… Therefore, we may have to try to develop better ways of teaching people to observe their behavior and their conditions and to make more accurate decisions about their improvement. (Wolf, 1978, pp. 212-213)

IT’S NOT WHAT PEOPLE SAY TO US, IT’S WHAT PEOPLE SAY ABOUT US

I think we all appreciate that what people say to our face might not be what people say behind our back. For social validity purposes, the trick is finding out what people say behind our back.

What I would really like to see is something like a secret shopper. Let us do our typical thing, then release the participants into the wild. After a little time has passed, send a confederate out to chat them up about their experience, maybe posing as someone who also participated in the same intervention. What do our former consumers say then?

This is pie in the sky. I don’t even know how you would get around the logistics, never mind the ethics, of that approach. But we might be able to figure out some approximations. I think the key is to get someone else to collect the data in some other place, and to anonymize the respondents. At least as part of the process. So maybe an online follow-up that comes from someone who collects the social validity data and is not otherwise involved in the research.

Whatever we do, it’s probably best to not be there with the participant right as things are wrapping up, asking them if you did a good job as you stare them in the face. We’re not on the car lot, after all.

The point of collecting social validity data is to gauge if we are doing things that are important, in ways that people will accept, to produce changes that actually improve people’s lives in meaningful ways. If we don’t, people are unlikely to use, or recommend others that they use, the interventions. We might not have a problem with our social validity data, at all. I don’t know. But neither do you. So we should at least be paying more attention to how we collect those data.

At the end of the day, if we’re honest, what we want is people to tell us that they like us. But what we should want is people to tell us what they really think, even if they don’t like us. I mean, all you see in published articles are great social validity data. Everybody says they love what we do. But behavior analysts are always complaining that almost none of our interventions are widely used. Hmmm.

The real question is not whether machines think but whether men do.
B. F. Skinner

Have you ever wondered what would happen if everyone on earth jumped at the same time? Or how many times you can return items to Amazon before they ban you? Or have you ever wished you could have someone create a Spotify playlist for you of songs from the 90s with the names of cities or people in the title?

Well, it turns out there’s an app for that. But you probably already know that. If you’re under 30, odds are you are using ChatGPT or something like it. If you’re over 30, odds are that you have at least used it or something like it, even if you didn’t know it.

ChatGPT, Claude, Gemini, and other programs like them, are AI chatbots that can reply to emails for you, screen your calls, plan your next vacation, and validate your worst ideas. (According to OpenAI, who makes ChatGPT, about half of the time people ask for information, and about 40% of the time they want to perform a task—usually writing something.)1

The hope is that AI will make us smarter and better at what we do. We’ll be more accurate, insightful, reliable, and efficient. But there are reasons to be skeptical, if not worried. The consequences of AI could be significant, upheaving our economy and eroding our society. And the machines might rise up and kill us.

But even if AI doesn’t kill us, it will make us dumber.

How? By turning us into readers.

As a college professor, I warn my students that reading is not studying. Studying is practicing. If you want to learn to hit a baseball, you don’t just watch baseball games. You have to stand up and swing a bat at a ball. Over and over. If you want to do well on an exam, you can’t just read the book and your notes. You should sit down and practice answering the questions that might be asked. Reading is important, but it will only get you so far, unless your professor tests you by asking you to read out loud.2

What does this have to do with AI chatbots?

We might read the answer an AI chatbot gives us but do little more after that than cut and paste it into a new place. But reading is not thinking. We can read the answer, slowly and carefully, but that doesn’t mean we can now answer the original question or any new ones. I mean, can you read the following sentence?

Neverland damn the tam about the pig around the room under the definitely upward and bound.

Great. Can you explain it to me? Didn’t think so.

But you probably think I am setting a strawman. That was gibberish. How about this one?

We investigated ultrafast defect-lattice dynamics in diamond using the Ns:H−C0 defect, an analog of bond-centered hydrogen in semiconductors.

I am sure you can read it, say it, and copy it down. But I doubt you can explain it, even though some people can.

My point is that reading is not the same as thinking. Reading an answer to a question is not the same thing as knowing something.

From the standpoint of a radical behaviorist, my primary concern is that AI is changing verbal repertoires, and maybe not for the better. The new verbal sequence looks like this: A question is asked, dictation is taken, then ChatGPT responds. At this point, the listener engages in simple textual behavior. They read the answer.

What’s the problem?

This is where it gets technical and aimed at behavioral psychologists. (But here’s a cheat sheet.)

Questions should evoke echoic (repeating) and intraverbal (saying different but related things) behavior. We repeat the question to ourselves, then respond verbally in ways that are controlled by the verbal stimulus but that have no point-to-point correspondence with it. When someone asks, “How do I,” we don’t just respond, “How do I?” That would be an echoic relation, because each part of our response corresponds (point-to-point) to the question. Instead, we answer, “You…” Here, the form of what we say is determined by what was asked, but there is no point-to-point correspondence—the answer is something different from the question.

Questions, being verbal stimuli, start to boil the cauldron of our behavioral repertoire, to invoke Dave Palmer, bringing various members of various response classes to strength and, ultimately, one or a few responses percolate and jump past the prepotency threshold to be emitted.

If we’ve never answered a question like the one we were just asked, there might be no responses to come to strength. Then we have a problem to solve. The question evokes verbal responses that are not themselves answers to the question, but that might alter the environment in such a way as to evoke the answer.

But with AI, the question is mostly likely to evoke dictation—we type the question into a chatbot. Then the chatbot responds with an answer and we engage in textual, not intraverbal, behavior. That is, we read the answer provided by the chatbot. Probably the strongest response evoked in the future (by the question) will be the echoic and a fragmentary textual response.

Or maybe not even that. Recent research suggests that students who use ChatGPT to respond to writing prompts cannot recall a single sentence they “wrote.” Probably because they never said, wrote, or typed anything—they struck some key combination to cut and paste the text from one place to another. They never engaged in behavior relevant to the question, so there is no relevant behavior to be evoked later by the same question or some other prompt. Students who did not use ChatGPT could recall some or many sentences, presumably because they had actually written them.

This illustrates the crux of the problem. Using AI, our repertoires are not changed in ways that allow us to behave more effectively with respect to the subject matter absent the continued use of AI. But, even then, the actual “answer” to the question is not part of our verbal repertoire in any meaningful sense. It exists only in the physical memorandum of the AI text. And reading cannot occur without a text, meaning the behavior cannot occur without environmental supports.

AI is a crutch. Great if you need it, but better to do without it, if you can.

The stimulus control established in the textual relation is not very useful in real-time interactions, provided anybody actually still has those. To behave effectively, you need an additional, and substantial, source of stimulation—a text or audio or video clip generated by AI that you can see and repeat in some way. The relevant circumstances alone will not occasion behavior appropriate to the situation absent such a technological middleman.

So I worry that the more we let machines think, the less we will.

“Psychology as the Behaviorist Views It” was the title of John B. Watson’s infamous behaviorist manifesto, as it is sometimes called, of 1913. Watson called for a natural science of psychology that focused on the prediction and control of behavior as the primary subject matter, as opposed to introspective analyses of fuzzy hypothetical constructs such as consciousness or, in the modern parlance, cognition. These were fighting words and they still are. According to Watson, “Our minds have been so warped by the fifty-odd years which have been devoted to the study of states of consciousness that we can envisage these problems only in one way.” It has now been over 150 years. We have, I think, even more to say. Circling back, the formal title of Watson’s manifesto was appropriate, as Watson was the behaviorist at the time. But others would follow, of course, myself included. The title of this Substack is an homage to Watson that recognizes that I will be speaking as a behaviorist, among many, and that my words are not necessarily representative of all behaviorists.

Subscribe now