Another failure to replicate a much-cited study on free will and cheating

September 2, 2015 • 9:45 am

There’s been a lot of press about a study just published in Science in which a large consortium of researchers tried to replicate 100 studies published in psychology journals, and managed to get significant results in only 36% of the replications. Further, investigators who repeated the earlier studies judged subjectively that they had replicated the original results only 39% of the time.

There has been a lot of analysis of these results in the press, with conclusions ranging from “psychology can’t be trusted” to “this is just normal science.” I’ll talk more about that, and summarize the Science paper, tomorrow, as it’s complicated and I need to read it for the third time. But in the meantime, and to save space in tomorrow’s post, I’ll cite one study that failed to replicate when redone by the large group of investigators constituting the “Open Science Collaboration” (OSC). That study is of special interest to me because it involves free will and its supposedly salubrious effects on society.

The paper chosen was from Psychological Science, one of the three well-known journals chosen by the OSC as sources of the 100 replicated studies. And this one was Vohs and Schooler’s 2008 paper (reference below) which I discussed in 2014, reproducing its abstract from the journal:

Does moral behavior draw on a belief in free will? Two experiments examined whether inducing participants to believe that human behavior is predetermined would encourage cheating. In Experiment 1, participants read either text that encouraged a belief in determinism (i.e., that portrayed behavior as the consequence of environmental and genetic factors) or neutral text. Exposure to the deterministic message increased cheating on a task in which participants could passively allow a flawed computer program to reveal answers to mathematical problems that they had been instructed to solve themselves. Moreover, increased cheating behavior was mediated by decreased belief in free will. In Experiment 2, participants who read deterministic statements cheated by overpaying themselves for performance on a cognitive task; participants who read statements endorsing free will did not. These findings suggest that the debate over free will has societal, as well as scientific and theoretical, implications.

Since then the study has been touted widely (see below), often as proving that determinism is bad for society. It’s has been used—by Dan Dennett and Eddy Nahmias among others—to show that unless people believe in some form of free will, they’ll behave badly and society will fall apart. I find that argument very odd, for when a similar argument is applied to God (“It’s important for society to be religious, for without religion, the moral glue that keeps us harmonious will dissolve”), it’s rejected by people like Dennett and me. In this way, belief in free will has come to resemble “belief in belief” (as Dan calls it) in a religious sense. But why accept one argument for societal harmony but reject the other?

But as a friend wrote me about this study and how it’s used to promote free-will compatibilism:

A huge proportion of factual statements could, when read immediately prior to an opportunity to act,  nudge the percentage of alternative behaviors up or down by a few points (particularly in experiments where the subjects can infer that the statement was there for a reason and the point of the experiment is to see what they’ll do in response – but let’s put that aside). The response isn’t to disseminate falsehoods that will nudge people to behave better; it’s to disseminate explanations as to why one shouldn’t behave badly even though certain propositions are true.

That last sentence is full of wisdom, and we should remember it whenever we feel tempted to foster “belief in belief”—whether that belief be in God or free will.

At any rate, the New York Times, in a subsection of its blurb on the OSC paper, describes three famous studies that were not replicated but were still widely cited in the press. You can see them all at the piece called “Three popular psychologies studies that did not hold up”. I’ll reproduce only what the Times says about the Vohs and Schooler study.

Free Will and Cheating:

In 2008, a paper in Psychological Science found that people were more likely to cheat on a test after they had read an essay arguing that behavior was predetermined by environmental factors. The authors suggested from their findings that belief in free will had societal implications.

The redone study found an effect pointing in the same direction as the original, but far weaker. One possible reason, the authors suggest, had to do with how subjects’ opinions about free will were manipulated. Participants read an essay, and it’s plausible that they were not as engaged in reading and thinking about it as were those in the first study.

The study was cited 341 times in other journals, the most of any of the 100 studies that the Reproducibility Project tried to replicate. There are 24 citations listed in the PubMed database.

In popular news media at the time, the study was covered with a focus on what it meant for societal belief in free will. A Scientific American report in August 2008 called the study clever and added, “The results were clear: Those who read the anti-free-will text cheated more often!” In Psychology Today in March 2008, a reporter wrote, “Reducing belief in free will might also make people exercise less and drink more.” A New York Times story in February of the same year said that the researchers interpreted their findings to raise, “questions about how human behavior might change if the belief in free will continued to decrease.” However, it added that the researchers, “cautioned against reading too much into the results.”

I’m amazed—and appalled—that the Vohs and Schooler study was not only cited uncritically (doesn’t anybody care if there are long-term effects?), but was even cited improperly, as if denial of fee will would affect exercise and drinking.

I’ve also written that Rolf Zwaan at the University of Rotterdam has failed to replicate the Vohs and Schooler result, so that makes two failures to replicate, even if you accept that some truth is conveyed in the kind of experiment that was done originally. (I have serious doubts about whether cheating immediately after reading a deterministic passage—even if it’s a real effect—says anything about long-term behavior.) All I know is that I haven’t been tempted to cheat more often since I’ve become a hidebound determinist and incompatibilist!

But in the end, these are the words to remember: “The response isn’t to disseminate falsehoods that will nudge people to behave better; it’s to disseminate explanations as to why one shouldn’t behave badly even though certain propositions are true.”


Vohs, K. D., and J. W. Schooler. 2008. The value of believing in free will: encouraging a belief in determinism increases cheating. Psychol. Sci. 19:49-54.

52 thoughts on “Another failure to replicate a much-cited study on free will and cheating

  1. :I have serious doubts about whether cheating immediately after reading a deterministic passage—even if it’s a real effect—says anything about long-term behavior. :

    There might also be an implicit suggestion to the subjects that the passages are supposed to influence their behavior, merely by their placement prior to decision-making.

  2. If I’m understanding it properly from Jerry’s synopsis, the Vohs and Schooler results, if correct, still undermine the existence of free will, as study participants seemingly altered their behavior in response to environmental cues.

        1. Unintended, something was up with my device. I apologize. The comment also did not appear until long after I posted it. I had the same problem below.

    1. Correct, but the study was not aimed at determining the existence of free will, but to determine if belief in free will affected behavior.

    2. I think you are confusing determinism with fatalism. This common confusion is precisely what this study was investigating. In other words, if we convince people of the fact that free will (in the commonly understood “spooky” sense) does not exist, and they come to believe in determinism, does that make them fatalistic? In other words, if people believe that their decisions are fully determined by circumstances (including both external stimuli and their state of mind), does that then make them fatalistic? Does that mean that they stop trying to make “good” decisions, because it’s pointless?

      (When most people think of free will, they are referring to a “spooky” version – that a precisely identical agent could have acted differently under precisely identical circumstances – excluding random effects. This is not really coherent, so free will per se is not amenable to empirical investigation.)

  3. I’m wondering if the study only exhibits what people believe about determinism. If people believe determinism means they are no longer responsible for their actions, and more relevant, their consequences, then why not choose to cheat? Similar to using any other belief to justify trying to gain advantage or power.

    Failure to replicate may also be confounded not just by the study, but by changing beliefs about determinism itself.

  4. I’m wondering what practical distinction exists between believing in determinism, and believing in something less lofty and more prosaic — that the way our parents raised us shaped our character, say.

    That’s true enough. And yet there are people who might be said to abuse it. They deliberately use a bad childhood as their all-purpose excuse to get out of personal responsibility and the possibility of changing their behavior. “I beat my wife because my dad always beat my mom: not my fault, then. Don’t blame me, blame my parents.” Sometimes called the Officer Krupke Defense, I think.

    Examples like the above are much more familiar to us than esoteric arguments on the philosophical level of the cosmos. The different results then might reflect subtle changes in wording or situation, with some respondents framing the question correctly, and others framing it as an Officer Krupke Defense and then either buying it — or not.
    Would it make a difference? I don’t know.

  5. “to disseminate explanations as to why one shouldn’t behave badly even though certain propositions are true.”

    Isn’t that the point of compatibilism? To my understanding, hard determinists want to make reductive conclusions about morality, desert, and punishment based on metaphysical propositions. Compatibilists want to insulate those subjects from reductive metaphysical arguments. It seems impossible to achieve agreement about the logical implications of determinism, so maybe it would be better to study morality and punishment without muddling the subject with the ambiguous implications of determinism.

    1. I think determinism could be dropped if all parties rule out “spooky events”.

      However, I doubt studying morality and punishment will bring us any further in the free will discussion. Probably only neurosciences can show us how free our choices really are.

      We have similar problems and discussions in robotics: do we have to blame the google car or the programmer or the improperly trained neural network when something goes horribly wrong. Do we have to punish this car or can we simply repair it immediately?

      1. Both compatibilists and hard determinists reject “spooky events”. The controversy is all about how determinism should impact our understanding of moral responsibility, punishment, and whether it’s possible for a person to really deserve things (good or bad). Hard determinists think there are big impacts, and we therefore should make big changes in our moral understanding; compatibilists are skeptical of those alleged impacts and prefer not to draw any big conclusions based on determinism.

      2. Okay, now I see what you mean. Let me clarify: lots of people are already studying morality and criminal punishment without a very detailed consideration of free will and determinism. The hard determinists want those people to take notice of determinism and alter the way they approach their subjects. Compatibilists don’t want fields of pragmatic study on morality, ethics and public policy to get hung up on metaphysical musings.

  6. I think a better study would be to test randomly selected, unprimed people. Then afterwards ask them about their views on free will. Then see if there are any correlations between degree of cheating and free will beliefs.

    1. The original study and both replication attempts included a variant of that. After subjects were read the statement (“primed,” using your term) according to the condition to which they were randomized, they completed a standard questionnaire assessing their belief in free will/determinism. The original study found a significant correlation (r=.53) between belief in determinism and amount of cheating, whereas no correlation at all was found in either replication attempt, despite both replication attempts having much greater statistical power than the original study.

    2. The verification done by other researchers, of course.

      Have you heard about the double-blind study on vision acuity?

      Me neither, it appears they couldn’t see any result.

    3. Even with such a correlation, it wouldn’t necessarily be clear which is the cause and which is the effect, right?

      It seems plausible, at least, that people with poor impulse control might be more inclined to believe in determinism, while people with strong impulse control might be more inclined to believe in free will.

      In other words, individuals might have a tendency to adopt whichever worldview casts their behavior in the best light: those inclined to behave well want to receive full credit for their actions; those inclined to behave badly don’t want to receive full blame.

      1. That’s a great point.

        It brings to mind Valmont in Dangerous Liasons. “It’s beyond my control.”

        Link broken with a space so that it’s doesn’t embed (I hope)

  7. The majority of people on the planet have no idea what free will is and if you told them that they were just sacks of meat robots it would not be surprising that they immediately begin to act like white-fanged baboons.

    In a few days they will forget and go back to pedestrian moral behaviors, the ones that are regulated by either tradition or reason or religion. Determinism will matter to them as much as the 3 billionith star on the edge of our galaxy.

    The only study that makes sense would be one that incorporates people who have actually thought about determinism for more than a heart beat and know the difference between compatibilism and dualism. My guess is these people will very likely not cheat more or less based on what you tell them, because they will continue to believe what they believe anyway.

  8. “The response isn’t to disseminate falsehoods that will nudge people to behave better; it’s to disseminate explanations as to why one shouldn’t behave badly even though certain propositions are true.

    I think this framing of the issue is on point. It also, to me, illuminates how the position of “people need to believe in free will in order to behave well” seems to reduce to a “little people” argument. It seems to include a prior claim that people will not be capable of a reaonable level of understanding of the explanations

    Now, I don’t think little people arguments are always necessarily inaccurate. But I can’t see any difference between “people need to believe in free will in order to behave well” and religious accommodationism.

    I think it is important to counter conceptions of magical, contracausal free will. But compatibilism vs incompatibilism? Both reject those kinds of free will and are virtually identical regarding the cognitive processes of humans, despite the arguments between the two camps.

    1. Not that you’re making this claim, but I just wanted to take this opportunity to clarify that “people need to believe in free will in order to behave well” is not an argument all compatibilists make. I know there’s that (in)famous quote from Dennett, but none of the folks arguing for compatibilism here at WEIT have included that appeal to consequence in their arguments.

  9. I briefly scanned the Vohs and Schooler paper and I was struck by the absence of a measurable definition of Free Will.

    In the abstract they talk about “The belief that one determines one’s own outcomes is strong and pervasive.” Well I share that belief but recognise that my determination is determined by many things, some of them hidden from me.

    If the paper had been titled ‘Beliefs – can they be affected by priming?’ it would have been more accurate.

    1. Note that that is just the definition of plain old regular “will”.

      Calling it “free will” is simply question-begging. Really, people need to start getting called out for using the two interchangeably.

  10. Free Will a concept. Is it free if we let reasoning away from it allow for those who otherwise would not cheat, will cheat under what they read that absolves them of their own choices in the matter? Or their propensity in the matter?

    Without a firm base line the tests will be useless.

  11. Something not mentioned above that I’m wondering about.

    Let’s assume for the sake of argument that the essays really did prime the subjects to cheat more or less depending on which essay they were given.

    How do we know that the priming was the result of the actual content of each essay? Could there have been something in the “no free will” essay giving a “wink, wink, nudge, nudge” suggestion? Maybe one had a significant number of stylistic errors causing people to be more or less diligent?

    That is, even if they actually did measure something, how do we know that what they measured is what they think they measured?


    1. Well, in a good study design, one tries to control extraneous differences in stimulus materials, although I have no idea how well that was done in the original study.

      That said, I think the two auxiliary tests in the experiment were intended to show that the determinism stimulus had the intended effect by the intended mechanism. That is, if belief in determinism causes cheating behavior, then there should be a correlation between a subject’s degree of belief in determinism and the amount they cheat. Moreover, if the determinism essay increases belief in determinism, then the mean degree of belief in determinism, after reading their respective essays, should be higher among subjects randomized to the determinism essay than to the control essay. Together, these two observations, if found, would suggest that the determinism stimulus increased subjects’ degree of belief in determinism, which in turn caused them to cheat more.

      In fact, both auxiliary results (the correlation between belief and cheating, and more cheating on average in the determinism group) were found in the original study, whereas neither was found in either of the two replications. Curious, eh?

      1. Thanks for the elaboration; makes sense.

        And I’m reminded again of what’s becoming one of my favorite XKCD editions:

        I think many sciences could benefit from something more rigorous than p < 0.05. Maybe not the several sigmas of particle physics standards…but the softer sciences are just now recapitulating a lot of what got the particle physicists to where they are today.


  12. “The response isn’t to disseminate falsehoods that will nudge people to behave better; it’s to disseminate explanations as to why one shouldn’t behave badly even though certain propositions are true.”

    Excellent advice to be sure. ( Every compatibilist I’m aware of
    would agree.)

  13. I find this interesting in the sense of conducting experiments to prise responses by individual free will.
    I do not hold to the free will concept
    I also entertained the idea, had, or if they were of some faith, could the foolish also consider this a connection to their soul. The thing that departs after death, but like I say, it was for amusement only.
    But my sticking point is, how do you show responses for something that does not exist?
    The brain is only going to give out what it’s genetic and environmental makeup has to respond with, and if it has been manipulated by external stimulation, words, a warm breeze, gun to the head.. then there you have it, an individual response.

  14. As silly as I think most of the discussion around free will is in a pragmatic sense, the assertion that believing in determinism causes less exercise and more drinking is sillier still. Why wouldn’t it have an equal chance of causing less drinking and more exercise? After all, if there’s no choice involved, you can’t help but exercise more. And to clarify, by silly I mean that I find the more useful discussion to center around what determining factors modify behaviors (even though the goal behaviors desired are themselves determined); i.e. we are gaining ground in figuring out how to modify behavior through environmental stimuli and this is a much more clear cut case of determinism to demonstrate to the average person.

    Of course, by silly I also mean the discussion of religious people’s idea that free will entails infinite degrees of freedom to control choice outside of the constraints of Physics; e.g. why would this ability only be limited to our thoughts, surely there would at least in principle be a way to extend it beyond thoughts and make a free choice to say, jump over the moon. Perhaps silly is the wrong word here, I thought about using different adjectives, but in the end there really was no other choice.

  15. I find it intriguing how typical it is for people to react to the lack of free will as though something has changed and they lost something rather than truly acknowledging or at least truly considering the lack of free will, that it was never there and conditions haven’t actually changed. To react as though you lost something misses the point entirely.

    1. I certainly try to convey to others that responsibility doesn’t go away but culpability should. Environment influences behavior, so deterrents to bad behavior shouldn’t be abandoned but people should stop being demonized. Use effective methods to improve behavior and don’t just blame people for not acting better in the first place.

      1. I find no reason to doubt determinism. But when I reflect on the subject I also really don’t understand the type of assertions you make here Benjamin. If all outcomes in behaviour, including language expressions and what the folk call intentional thought, are determined then what does ” responsibility doesn’t go away but culpability should” mean? If causal forces at work long before I existed determined that at time t my body would move in ways X, perhaps behaving in ways we tend to call “cheating”, then that is what happens at t and things could not have happened differently. Yet you suggest “Use effective methods”. But on determinism it was determinied already before the existence of you or me if I would by reading your suggestion be caused (together with whatever other causal processes in my body) to behave in ways that fit the description “use effective methods” or not. Is not all moral talk an instance of “belief in belief” if determinism is true?

        1. Brains identify patterns, whether free will exists our not. Identifying factors that alter someone’s behavior is also possible without free will. Effective methods would be environmental influences that have been identified as improving behavior, such as education. Determinism isn’t defeatist by default. Individuals commit acts whether they can help out our not, so they are responsible for them but shouldnt be treated as if they could have done something different at the time. Our internal knowledge base doesn’t cease to exist just because we aren’t “in control” if it. All the things that have happened and our preferences still exists whether free will does or not. Morality doesn’t go away just because there’s no ghost in the machine.

          1. Benjamin: “Identifying factors that alter someone’s behavior is also possible without free will”

            Yes, but it was determined before you and me began to exist which factors each of our bodies would express and how that in turn would causally determine what happens next. Likewise it was predetermined which factors would be identified or not by which person. That is why I when reflecting on determinism find it hard to see any room left for responsibility. I don’t say that in an attempt to bash determinism, as I said I don’t have any reason to doubt determinism. Though thinking about determinism does give me unpleasant feelings. In everyday life I of course have moral and responsibility intuitions and like for most others (I guess) it feels like I have free will to for example control where I focus my vision as I type this sentence. But when reflecting on determinism I realize that it was determined already before I came to exist just how my vision would come to be focused at that moment.

            “… so they are responsible for them but shouldnt be treated as if they could have done something different at the time” … “Morality doesn’t go away”

            But morality without free will in a deterministic universe only means that it was determined before any humans existed which human body would in 2015 move in ways that are seen as morally wrong and which human body would move in ways that are seen as morally right. It was also predetermined which body would, when other bodies in equally predetermined ways emitted the linguistic pattern “doing X is immoral”, be caused by that emission, and the predetermined ways in which the brain mechanically processes it, to not behave in way X and which bodies would continue to be caused to behave in way X. To say that the former bodies are “responsible” or “moral” only puts a descriptive tag on what was bound to happen.

          2. I don’t see how the fact it’s all predetermined is relevant. Attitudes, opinions, etc. don’t cease to have relevance just because they obey the laws of physics. Just like our exchange right now. Our ideas are influencing each other to one degree or another despite the fact they are predetermined.

          3. It itches in my brain when I try to spell out my thinking further on this topic but here is another shot.

            “Attitudes, opinions, etc. don’t cease to have relevance just because they obey the laws of physics.”

            What do you mean with “have relevance” here? I see two possibilities:

            A. You, when making the “morality has relevance” claim, mistakenly slip back into believing in some free will and believe that people do have the power to choose their future behaviour, including how they will signal and set up incentives that will shape the behaviour of others. I think you then make a false claim, if determinism is true.

            B. You only mean to say that it is a fact about causality for biological structures like us that if person1 in a situation is caused to emit “doing X is wrong” language signals then some person2 will be caused to in the future not behave in ways X whereas if person1 had not been caused to emit those language signals then person2 would have been by other factors been caused to behave in ways X. Morality talk then “has relevance” in the sense that it is a predetermined fact that our kind of organisms will output morality talk and that our kinds of brains (and in the next step, bodies) will be causally affected by receiving such morality talk.

            Two reactions to interpretation B of your claim:

            1 First note that it is with B possible that the cause and effect details of such morality talk include, in some/many/all sender/reciever brains, as a causal step the formation of a false belief that people can (freewillily) choose to do X or not.

            I suspect it is like that. The evidence I see for that is how people in everyday life and looking forward seem to think that they and others can have responsibility and control over certain kinds of things but not other kinds of things. E.g. we tend to think a person has a responsibility to not go up and hit someone in the face but we do not tend to think a person has a responsibility to stop an unpredictable and unstoppable wind from lifting the same arm and doing the same causal impact on the other persons face. But if we in our everday lives and actions believed in determinism then that distinction would make no sense since each and every movement of our bodies is caused by an unstoppable wind, so to speak, including the instances where someone goes up to someone else and hits that person in the face.

            If there is such a causal step in some/many/all cases then the claim that morality has relevance in the sense B involves a kind of “belief in belief”: belief that some good things come about as an effect of a process that includes people forming false beliefs about reality.

            2 Second if we think morality talk “has relevance” in the determinist sense B above we can also see that the same might be true for free will talk, or God talk for that matter. It may be a fact about causality for biological structures like us that if person1 is caused to emit such talk then it will have certain effects in person2. It is of course up to science to investigate which such causalities are actually out there.

            Let me now get back to the question if anything is lost when one comes to think that there is no free will. Of course if it is true that there is no free will there has never ever been any free will and so nothing is lost in the way that e.g. a house that once existed but then burned down is now lost. However I think we (or I can only think for myself here, so I) in much everyday activities assume the belief that, and feel like, there is free will control over our actions e.g. on where to focus our vision or on in making an effort to do the right thing. When I on reflection find no good reason to doubt determinism I feel an unpleasant and at times crushing angst. I didn’t feel that particular angst prior to when I came to accept determinism – my unreflected free will beliefs were more pleasant. So in one sense something is lost, at least for me, when accepting determinism. I have lost the absence of those particular unpleasant feelings.

Leave a Reply to eris Cancel reply

Your email address will not be published. Required fields are marked *