ArticlePDF Available

Motivated free will belief: The theory, new (preregistered) studies, and three meta-analyses

Authors:

Abstract and Figures

Clark and colleagues (2014) proposed a theory of motivated free will beliefs, according to which at least part of free will beliefs and attributions are caused by a desire to hold moral transgressors responsible. Recently, this theory has been challenged. In the following article, we examine the evidence and conclude that, although not dispositive, much of the evidence seems to support the motivated account. For example, in 14 new (7 preregistered) studies (n=4,014), results consistently supported the motivated theory; and these findings consistently replicated in studies (k=8) that tested an alternative (counternormative) hypothesis. In addition, three meta-analyses of the existing data (including eight vignette types and eight free will judgment types) found support for motivated free will attributions (k=22; n=7,619; r=.25, p<.001) and beliefs (k=27; n=8,100; r=.13, p<.001), which remained robust after removing all potential confounds (k=26; n=7,953; r=.12, p<.001). However, the size of these effects varied by vignette type and free will belief measurement. We discuss these variations and the implications for different theories of free will beliefs and attributions. And we end by discussing the relevance of these findings for past and future research and the significance of these findings for human responsibility.
Content may be subject to copyright.
Running Head: MOTIVATED FREE WILL BELIEF
1
Motivated free will belief: The theory, new (preregistered) studies, and three meta-analyses
Cory J Clark
University of Pennsylvania
Bo M Winegard
Marietta College
Azim F Shariff
University of British Columbia
© 2020, American Psychological Association. This paper is not the copy of record and may
not exactly replicate the final, authoritative version of the article. Please do not copy or cite
without authors' permission. The final article will be available, upon publication, via its
DOI: 10.1037/xge0000993.
Corresponding author: Cory J Clark, University of Pennsylvania, cjclark@sas.upenn.edu.
These studies were funded by the Social Science and Humanities Research Council of Canada.
There are no known conflicts of interest.
MOTIVATED FREE WILL BELIEF
2
Abstract
Clark and colleagues (2014) proposed a theory of motivated free will beliefs, according to which
at least part of free will beliefs and attributions are caused by a desire to hold moral transgressors
responsible. Recently, this theory has been challenged. In the following article, we examine the
evidence and conclude that, although not dispositive, much of the evidence seems to support the
motivated account. For example, in 14 new (7 preregistered) studies (n=4,014), results
consistently supported the motivated theory; and these findings consistently replicated in studies
(k=8) that tested an alternative (counternormative) hypothesis. In addition, three meta-analyses
of the existing data (including eight vignette types and eight free will judgment types) found
support for motivated free will attributions (k=22; n=7,619; r=.25, p<.001) and beliefs (k=27;
n=8,100; r=.13, p<.001), which remained robust after removing all potential confounds (k=26;
n=7,953; r=.12, p<.001). However, the size of these effects varied by vignette type and free will
belief measurement. We discuss these variations and the implications for different theories of
free will beliefs and attributions. And we end by discussing the relevance of these findings for
past and future research and the significance of these findings for human responsibility.
Keywords: punishment, free will belief, motivated reasoning, attribution, replication
MOTIVATED FREE WILL BELIEF
3
Motivated free will belief: The theory, new (preregistered) studies, and three meta-analyses
In the United States, political conservatives believe that homosexuality is more of a
choice and more controllable than liberals, and correspondingly, conservatives find
homosexuality more morally wrong (Haider-Markel & Joslyn, 2008). It is possible that liberals
and conservatives did not originally differ in their condemnation of homosexuality but then came
upon different information regarding the controllability of homosexuality, and only then did
conservatives decide it was more morally wrong. But it seems plausible that because many
religious conservatives morally condemn homosexual behavior, their beliefs about the
controllability of homosexuality were at least partially caused by their desire to denounce it. That
is, they began with an intuition that it was wrong and only then developed a belief about how
controllable it was.
In 2014, Clark and colleagues presented evidence that beliefs in human free will were
similarly (at least partially) driven by moral disapproval. Specifically, they argued that desires to
punish a transgressor (at least partially) drove belief in free will, which, according to this theory,
likely operates as a justification for punishment. In four experiments, exposure to immoral
actions led people to attribute more free will to a perpetrator (motivated free will attributions)
and to believe more in free will in general (motivated free will beliefs), which was consistent
with theoretically derived predictions.
As we explain below, this theory is consistent with observations and consistent with
comprehensive accounts of human evolution and reasoning. It also makes novel predictions that
have been confirmed and explains a disparate array of data parsimoniously. Therefore, we
believe that it is a powerful and plausible account of part of the source of free will beliefs.
MOTIVATED FREE WILL BELIEF
4
However, Monroe and Ysidron (2019) mounted a spirited argument against this theory,
contending that an attribution-based account of free will is more parsimonious. According to this
theory, free will beliefs and ascriptions are based on the information that a person’s behaviors
reveal about his or her character. Behaviors that appear to come from the person’s character such
as wearing a Pink Floyd T-shirt or spontaneously skipping and jumping are judged as relatively
free, whereas behaviors that are compelled by social norms such as following a teacher’s order or
doing something one’s parents tell them are judged as less free. Furthermore, Monroe and
Ysidron failed to replicate motivated free will beliefs in three out of four studies.
In what follows, we first articulate our most current thinking about the theory of
motivated free will beliefs. We then carefully, and we hope charitably, state Monroe and
Ysidron’s challenge to the theory. Then we respond and forward data from new, preregistered
studies and meta-analyses that show that (1) both motivated free will attributions and motivated
free will beliefs reliably replicate and (2) Monroe and Ysidron’s challenges cannot account for
these effects. Thus, we contend that there is almost indisputably a real (though small-medium
[Cohen, 1988]) effect that supports the theory of motivated free will beliefs.
The Theory of Motivated Free Will Beliefs
Over a hundred years ago, Friedrich Nietzsche argued that free will was an artifice, a
delusion foisted upon humans by theologians to make them “responsible” for their behavior and
for their very being (1889/1954). Spurred by this provocative but plausible claimand the
consistency of this account with modern views on human reasoning (e.g., Haidt, 2001; Mercier
& Sperber, 2011)Clark and colleagues (2014) empirically tested and found support for the
claim that free will beliefs are partially motivated by desires to punish other people. That is,
people believe in free will at least partially to justify punishing other humans; and they attribute
MOTIVATED FREE WILL BELIEF
5
more free will to others whom they wish to punish than to those whom they are indifferent
toward. From this perspective, then, the desire for punishing or holding other people responsible
is the part of the impetus that leads to the conscious belief in free will and this is why the theory
is a “motivated” account of free will beliefs: It argues that free will beliefs are not arrived at
solely by dispassionate reasoning but also by strong motivations to hold moral transgressors
responsible.
Originally, Clark and colleagues argued that this might be an evolutionary (or cultural)
adaptation. Punishment is crucial for protecting one’s self and the social group from exploitation
and for maintaining social order and cohesion (Fehr & Gächter, 2002; Henrich et al., 2006). But
people are averse to causing harm to others (Cushman, Gray, Gaffey, & Mendes, 2012). They are
often pained if compelled to inflict suffering on the innocent. Therefore, perhaps they need a
belief that justifies harming others, allowing them to punish others without being wrecked by
guilt (Clark, Baumeister, & Ditto, 2017). Recently, however, our thoughts have changed. It
seems unlikely that inflicting pain on other people was much of a problem for most of human
history. History books, after all, are filled with grisly torture and retaliation. However, in post-
Enlightenment societies, people strongly discourage violence, emphasize rational and
proportionate punishment, and are generally reluctant to punish people who are mentally
compromised or otherwise not responsible. In these settings, belief in free will might serve an
important rationalizing function, justifying one’s desire to punish not primarily to one’s self but
to others. The best way to persuade others is often to believe something one’s self (von Hippel &
Trivers, 2011), so our claim is not that people cynically endorse free will beliefs to hoodwink
others into supporting punishing people; rather, our claim is that one truly and earnestly believes
in free will but at least partially because such a belief allows one to justify punitive desires.
MOTIVATED FREE WILL BELIEF
6
Suppose, for example, that a person’s friend is murdered by another person. Immediately,
desires for revenge might arise. And then desire for punishment. But what if the killer did not
freely choose his or her actions and was not fully responsible for his or her deeds? That is, after
all, a common criminal defense strategy and is also an argument sometimes forwarded by
philosophers: Without free will, a person is not ultimately responsible, and therefore punitive
desires, although perhaps understandable, are somewhat misguided and should be subordinated
to reason and the need to create a fair society that only punishes as a deterrent and to keep
dangerous people off of the streets. Although most people thankfully never suffer the loss of a
friend to violent crime, they do read and hear about such cases often. And those cases likely
trigger a desire for justice. The feeling that at least some immoral deeds should be greeted with
unpleasant, painful punishment is probably close to universal. And it is this feeling, then, that
creates a need for justification. And free will serves this purpose quite well, because it suggests
that the moral transgressor “deserves” his or her punishment because he/she is a fully
autonomous and responsible actor who actively chose to do wrong.
Monroe and Ysidron, in contrast, favor an attribution-theory of free will (we’ll also call
this a rationalist account, because it suggests that free will ascriptions are compelled by a desire
to understand a person and his or her behavior rather than a motivation to justify punishment).
According to this view, free will ascriptions reflect information revealed by an actor’s behaviors.
Acts that are presumably more informative of a person’s internal character (spontaneous acts that
aren’t expected, for example) are judged as more free than acts that are not (following orders, for
example). So, the big difference is the explanation for the relative causes of free will attributions
to morally bad actors. Imagine, for example, that people read a story about a man who stole
pearls from a house. The attribution theory argues that people attribute more free will to this act
MOTIVATED FREE WILL BELIEF
7
than they attribute to a more neutral act because it is more counternormative; therefore, it seems
to stem from something unique about the person (and cannot be attributed to norm adherence).
The motivated free will theory, on the other hand, contends that the greater belief in free will in
the pearl robber case is likely caused (at least partially) by a desire to see the robber punished
and to justify that desire to others.
It is crucially important to note that we do not claim that a desire to punish explains all,
or perhaps even most, of the belief in free will. In fact, we think it is quite likely that the concept
of free will arose for a variety of reasons (e.g., the strong subjective experience of personal
causal responsibility [Wegner, 2002; 2003]), perhaps some consistent with Monroe and
Ysidron’s account, and was only later co-opted to rationalize punitive desires. Our claim is
simply that one cannot fully understand belief in free will (and, more specifically, variation in
belief in free will) in modern, post-Enlightenment societies without considering motivated
reasoning that is driven by the need to justify punishment.
This motivated account is quite consistent with contemporary theories of reasoning and
congruent with well-established asymmetries in responsibility attribution (Alicke, 2000), which
social psychologists have documented in the past 50 years. People, for example, attribute more
causality and responsibility to bad actions and actions with bad outcomes than to closely
matched (1) good actions and actions with good outcomes (e.g., Alicke, 1992; Knobe, 2003;
Leslie, Knobe, & Cohen, 2006; Reeder & Spores, 1983), (2) neutral actions and actions with
neutral outcomes (e.g., Cushman, Knobe, & Sinnott-Armstrong, 2008; Knobe & Fraser, 2008),
and (3) less bad actions and actions with less bad outcomes (e.g., Walster, 1966). To our
knowledge, few if any scholars contest the existence of these responsibility-type asymmetries.
What is debated is whether such asymmetries are rational. As noted above, Monroe and Ysidron
MOTIVATED FREE WILL BELIEF
8
(2019) contend that many of them are, in fact, rational, because immoral actions are often
strongly counternormative, and counternormative actionsaccording to the attribution theory
that Monroe Ysidron promote—reveal more about an actor’s desires than do more normative
actions Therefore, it is rational to impute more responsibility (and free will) to those who
perform immoral actions than those who perform similar neutral or good actions.
However, the theory of motivated free will beliefs seems capable of solving puzzles that
are difficult to explain from the attribution theory. For example, it’s difficult for the attribution
theory to explain why people would agree more with broad statements such as ‘People have free
will’ after reading a vignette about an isolated moral action. It seems even more difficult for the
attribution theory to explain Clark and colleagues’ (2014) studies, which demonstrated that
participants were more critical of anti-free will scientific research after reading about an isolated
immoral action. These findings are, however, perfectly explicable from the motived reasoning
account (e.g., Kunda, 1990) and the social intuitionist model of moral judgment (e.g., Haidt,
2001). People’s desired beliefs influence their actual beliefs, and their explicitly stated beliefs
often serve as post-hoc justifications for their intuitive moral positions. In this case, when people
wish to hold another actor morally responsible for their harmful behavior, they increase their
broad beliefs in free will so as to justify their moral condemnation (free will is considered a
prerequisite for moral responsibility [Nichols & Knobe, 2007; Sarkissian et al., 2010; Shariff, et
al., 2014]). Although humans prefer to see their reasoning as bottom-up, from evidence to
conclusions, this idealistic view of human reasoning has been convincingly challenged time and
time again (Mercier & Sperber, 2011). People’s desired conclusions often influence their
reasoning and the outcomes their reasoning produces (e.g., Ditto & Lopez, 1992; Kunda, 1990;
Haidt, 2001). Rationalist models fail to account for many findings in the social intuitionist
MOTIVATED FREE WILL BELIEF
9
tradition (Greene & Haidt, 2002), including the influences of incidental emotions and affect on
later judgments (e.g., Lerner, Goldberg, & Tetlock, 1998; Valdesolo & DeSteno, 2006), the
tendencies for people to bring their supposedly objective beliefs in line with their moral beliefs
(e.g., Clark, Chen, & Ditto, 2015; Liu & Ditto, 2012), moral dumbfounding effects (Haidt, 2001;
Haidt, Björklund, & Murphy, 2000), and many ingroup biases (e.g., Ditto et al., 2018, 2019;
Winegard, Clark, Hasty, & Baumeister, 2018).
This is not to say that humans are completely irrationalthey are not. Human cognition
is immensely sophisticated, and we can at least partially thank human reasoning for the
(relatively) peaceful, comfortable, and technologically advanced worlds we inhabit. However,
there is little reason to believe that humans evolved to be dispassionate reasoners whose lodestar
is the truth (Haselton & Nettle, 2006). If erroneous beliefs facilitate genetic propagation (or are
associated with other cognitive traits that do), then a propensity to believe them will evolve
(Clark, Liu, Winegard, & Ditto, 2019). For one example, beliefs in supernatural entities appear
ubiquitous, and despite valiant efforts, nobody has yet forwarded reliable evidence that they exist
(Boyer & Bergstrom, 2008). Humans, in other words, are motivatedthat is, cognitively
inclined for not truth-related reasonsto posit the existence of supernatural agents. Similarly, we
contend that humans are motivated (inclined) to attribute free will to those who behave
immorally because it allows them to impute moral responsibility to those whom they desire to
punish (or see punished). Our theory then it is perfectly consistent with evolutionary accounts of
human reasoning, which assert that human cognition is a tool that facilitated the spread of genes
and is therefore prone to predictable biases.
Motivated Free Will Beliefs Have Explained and Predicted Other Phenomena
MOTIVATED FREE WILL BELIEF
10
The theory of motivated free will beliefs has been useful for predicting and explaining
various other phenomena. For example, for nearly 15 years, experimental philosophers have
struggled to explain a variety of contradictory findings regarding whether everyday people
believe free will is compatible with scientific determinism. People judged an actor morally
responsible in a deterministic universe (Nahmias, Morris, Nadelhoffer, & Turner, 2005), but only
for concrete moral actions and not in the abstract (Nichols & Knobe, 2007). Experimental
philosophers thought this might be due to high affect for concrete moral actions, but when both
high and low affect conditions involved morally bad behaviors, people generally believed free
will was compatible with determinism (e.g., Feltz, Cokely, & Nadelhoffer, 2009; Feltz & Cova,
2014). And, in an apparent contradiction, when people read of a supercomputer that could
predict future human behaviors with 100% accuracy, people asserted that it was possible for an
immoral actor to avoid performing the immoral action, but not for a morally good or neutral
actor to avoid performing their good or neutral actions (Nahmias et al., 2005). What do these
results have in common? People routinely asserted that people have free will and the ability to do
otherwise when contemplating immoral actions. This is congruent with seven studies by Clark,
Winegard, and Baumeister (2019) that consistently found support for the notion that judgments
about whether free will is compatible with a deterministic universe do not reflect consistent,
principled beliefs about the necessary capacities for human free will. Rather, these judgments are
influenced by motives to preserve human moral responsibility. This suggests that these
seemingly contradictory findings in experimental philosophy reflect varying desires about moral
responsibility. When people wish to hold an actor morally responsiblethey assert free will.
Therefore, the theory of motivated free will beliefs appears to explain and organize a variety of
hitherto inexplicable results.
MOTIVATED FREE WILL BELIEF
11
Also congruent with the theory, one (preregistered) study found that participants reduced
their beliefs that free will was compatible with determinism after reading an argument that moral
responsibility could be preserved even in the absence of free will (Clark et al., 2019), and this
same sort of argument has been found to reduce attributions of free will to specific actors (Clark
et al., 2017). Specifically, when participants read a passage that argued that moral responsibility
and punishment could be preserved for the sake of deterring bad behavior and that free will was
not necessary for preserving moral responsibility, but crucially, that said nothing about the
existence of free will, people subsequently attributed less free will to an immoral actor. This
suggests that when free will is no longer viewed as a prerequisite for moral responsibility, people
feel less compelled to assert that immoral actors have it, lending further support to the suggestion
that free will beliefs partially function to justify moral condemnation.
That same paper also found (in five studies) that believing in free will alleviates distress
about punishment. Specifically, if free will beliefs are reduced when an individual is faced with
an opportunity to punish, they report heightened distress. For example, in one (preregistered)
quasi-experiment, participants who punished an unfair partner in an economic exchange game
reported particularly high anxiety when they learned that their partner did not freely choose the
unfair offer. In other experiments, participants desiring to punish an immoral actor reported
heightened anxiety when their free will beliefs were undermined by an anti-free will argument
(and not when their free will beliefs were supported by a pro-free will argument, nor when their
desires to punish were justified with a deterrence argument). These results could mean several
things. Perhaps people experience cognitive dissonance when they wish to punish an immoral
actor who is perceived as not having free will. Perhaps people fear that condemning an immoral
actor without free will will be unjustifiable to outside observers. Or perhaps free will beliefs
MOTIVATED FREE WILL BELIEF
12
evolved (in part) for purposes of promoting ruthless condemnation. Future work is needed to
tease apart these multiple possibilities, but in any case, free will beliefs appear to comfort those
who are wishing to punish, a finding predicted by motivated free will beliefs, that presumably
would not be predicted by rationalist attribution models.
According to the theory of motivated free will beliefs, just as people increase ascriptions
of free will when they want to emphasize moral responsibility, so too they decrease them when
they wish to minimize moral responsibility and punishment. And work has confirmed this
prediction. For example, in a preregistered study (Vonasch, Clark, Lau, Vohs, & Baumeister,
2017), participants were randomly assigned to write about a time they either succumbed to an
addiction or temptation or a time they overcame an addiction or temptation and then reported
broad belief in their own free will (on the Rakos, Laurene, Skala, & Slane, 2008 Personal Will
subscale). Participants who wrote about a time they succumbed to an addiction or temptation
reported that they generally had less free will than those in the “overcome” condition, and this
was mediated by the perceived moral badness of their own behavior. In a similar study, all
participants wrote about a time they gave in to an addiction or temptation and were randomly
assigned to write about a time when this behavior had relatively harmless consequences or a time
this behavior resulted in harmful consequences (Vonasch et al., 2017). Participants in both
conditions wrote about similar behaviors at similar frequencies, but participants whose behaviors
resulted in harmful consequences reported that they had less free will and control over their
behavior than those who wrote about harmless consequences. These results suggest that when
people wish to avoid responsibility for their own harmful behavior, they reduce perceptions of
their own free will.
MOTIVATED FREE WILL BELIEF
13
Thus far, the theory that free will beliefs are motivated by desires for (or against) moral
responsibility has been very useful and productive for explaining and predicting other
phenomena. The consistency of these findings across a variety of domains and experimental
approaches (not to mention real world observations such as political conservatives’ belief that
sexual orientation is a choice) should increase confidence in the theory.
Monroe and Ysidron’s challenge
Monroe and Ysidron (2019) forwarded several cogent challenges to the theory of
motivated free will beliefs. First, they argued that the free will subscale from the Free Will and
Determinism Scale (FAD+, Paulhus & Carey, 2011) that was used in three of the five studies
from the original Clark and colleagues (2014) paper confounds free will and moral
responsibility. Three of the seven items in the scale regard moral responsibility rather than
freedom and control. We agree and believe this is an important and helpful argument. We
therefore address this by conducting all analyses with and without the moral responsibility items
and by testing the effect with various other measures of free will belief. The results demonstrate
that removal of these items has very little influence on the size of the effect, and the largest
effects are observed for free will belief measures that do not include moral responsibility
confounds (evaluations of anti-free will science, and the Stroessner and Green [1990] free will
scale). Thus, this challenge, though valid, cannot account for motivated free will belief effects.
Second, Monroe and Ysidron argued that people likely attribute more free will to morally
bad actions, not because of desires to punish, but rather because morally bad actions are more
counternormative than are neutral or morally good acts. When a behavior deviates from social
norms, people draw inferences about an actor’s intentions and desires and these are related to
imputations of free will. Monroe and Ysidron argue that this likely explains motivated free will
MOTIVATED FREE WILL BELIEF
14
attribution effects; that is, people attribute more free will to morally deviant actors not because
they desire to hold them morally responsible, but rather because their deviance more reliably
reveals underlying preferences than do other kinds of behaviors. We address this criticism
directly by testing motivated free will attributions and beliefs following exposure to immoral
actions that are perceived as equally or less counternormative than the paired control conditions.
We consistently replicate both motivated free will attributions and beliefs in these cases (which
include the original vignette from Clark and colleagues’ [2014] Studies 2 and 4). Note that we do
not disagree that higher perceived counternormativity likely increases attributions of free willit
probably does. For example, Clark, Shniderman, Luguri, Baumeister, and Ditto (2018) found that
higher perceived counternormativity predicted increased attributions of free will, but mainly for
morally good actions, whereas free will attributions (and beliefs) for morally bad actions mainly
were driven by affective desires to punish. Here, we experimentally rule out perceived
counternormativity as a potential alternate explanation for the current effects.
Third, Monroe and Ysidron failed to replicate motivated free will beliefs in three out of
four studies. The main deviation we noticed about their studies in comparison to previous work
was their use of the entire four subscale 27-item FAD+ rather than a subset of these items
(previous work has most often used the 7-item free will belief subscale only, or occasionally the
free will belief subscale paired with one or two of the other three subscales, or other measures of
free will belief). Given that this was the only major deviation, we suspected that this might
explain the diminished effects found in their studies. After replicating the effect several times
using the free will belief subscale only, we tested the effect again using the entire FAD+ (Study
4c) and indeed found an attenuated effect. This methodological difference likely explains their
failed replications.
MOTIVATED FREE WILL BELIEF
15
We address Monroe and Ysidron’s important and thought-provoking challenges first by
pilot testing and identifying vignettes in which Punish Conditions are perceived as equally or
more counternormative than Control Conditions. We then conduct several preregistered studies
of these vignettes and extend the results to additional measures of free will belief. We close with
meta-analyses of the existing work (including Monroe and Ysidron’s [2019] failed replications)
and test for moderators of these effects. We find that motivated free will attributions and
motivated free will beliefs are robust and that neither moral responsibility confounds nor
counternormativity assessments can account for these effects.
Open Science Statement
Materials are described in text sufficiently for replication. Links to preregistrations are
included in text, and these preregistrations were followed exactly. All data will be made publicly
available upon acceptance for publication.
Pilot Studies 1a-1d
As a first step to retesting our theory of motivated free will beliefs, we conducted pilot
studies with four vignettes. Three had been used in prior research, and one was new, which
attempted to completely rule out potential normative counter-explanations by holding the
behavior constant. The original vignettes from Clark and colleagues’ (2014) paper about the
home robbery and aluminum can foraging incident were specifically designed to minimize
concerns about normativity because aluminum can foraging is also counternormative. But these
vignettes were not tested for perceived counternormativity, so we tested that here. We call this
the ‘Home’ vignette. We also retested vignettes used in Studies 2a and 2b from Clark and
colleagues’ 2018 paper, which included a morally bad (embezzling from work) and a morally
good (sacrificing work bonus) action, which we call the ‘Corporation’ vignette. We also tested
MOTIVATED FREE WILL BELIEF
16
the neutral and morally bad behaviors from Monroe and Ysidron’s (2019) critique of motivated
free will beliefs, which we call the ‘Bus’ vignette. Finally, we tested a new vignette (which we
call ‘GPA’) in which the behaviors (lying about GPA on a resume) were identical with identical
(and low) stated base rates, and attempted to manipulate desires to punish only by manipulating
the consequences of the behavior.
Studies 1a-1d Method
Participants. This and all subsequent studies were approved by the University of British
Columbia Office of Research Ethics. U.S. participants (Mage = 37.13, SD = 11.72; 372 female)
were recruited via Mechanical Turk (MTurk).
1
We aimed for 100 participants per condition
within each of the four studies (800 total); 803 participated. Nine participants were removed for
failing an attention check, thus our final sample included 754 participants. We used this same
attention check across all studies (and this was included in the upcoming preregistrations). To
maximize power across studies, we conducted pilot tests with smaller samples (70-100
participants per condition), replicated these effects in preregistered studies with 200 participants
per condition, and then conducted meta-analyses of these effects.
Procedure. Participants were randomly assigned to read one of eight vignettes which
were paired within each study (one morally bad [Punish Condition], one less morally bad
[Control Condition]). Each study involved a different vignette type (Home, Corporation, Bus,
1
We used U.S. MTurk participants in all studies. Though MTurk is not perfectly representative
of the U.S. as a whole, it is quite diverse with respect to age, sex, race, education, SES, and
ideology (and certainly more diverse than most university subject pools). On average, MTurk
workers are slightly younger, more educated, and more liberal than the US population; Asians
are overrepresented and Hispanics are underrepresented (Paolacci & Chandler, 2014). We do not
know whether our results would generalize to other countries and cultures. In the General
Discussion, we discuss how investigating motivated free will beliefs across cultures could help
identify the specific motivations underlying these effects.
MOTIVATED FREE WILL BELIEF
17
GPA). These are displayed in Table 1 below. Each vignette is considered a separate study
because they include different participants, and thus provide independent effect sizes in the
upcoming meta-analyses. By keeping these vignettes as separate studies, they can be included in
the upcoming meta-analyses as independent effect sizes that can be coded for vignette type,
which allows us to test for possible differences between types of vignettes.
Table 1. Exact phrasing of each condition within each vignette type in Studies 1a-1d
Vignette
Punish Condition
Study 1a:
Home
Vignette
Clark et al.
(2014)
Sam, a special education teacher, wakes up
one morning and finds that someone
robbed his home while he was sleeping.
His window is broken and all of his
valuables are missing. After a police
investigation, he learns that the robber is
unemployed, has two children, and sold all
of his belongings on Ebay.
Study 1b:
Corporation
Vignette
Clark et al.
(2018)
John is a district manager for a large
department store corporation that employs
nearly 200,000 people nationwide. His
district includes six stores, and so he is
responsible for the jobs of nearly 720
people. A couple years ago he began
falsely reporting earnings and managed to
steal over $9.2 million dollars unnoticed.
As a result, three of the stores he managed
closed and others had to cut back
employees, causing hundreds of people to
lose their jobs.
Study 1c:
Bus Vignette
Monroe &
Ysidron
(2019)
Andrew is riding the city bus, sitting next
to an old lady. A little while later the bus
stops, and the lady slowly gets up and exits
the bus. As she leaves Andrew notices that
she has a fat wallet. Andrew pulls the stop
lever and follows her for several blocks.
When she turns a corner, Andrew runs up
to her, knocks her over, and steals her
purse.
Study 1d:
GPA Vignette
New
Jim recently graduated from university and
has spent the past few months preparing
his resume and other job application
materials. He recently saw an
advertisement for his dream job, an entry-
level, but high-paying position at a
marketing firm with excellent potential for
advancement. Unfortunately, he knows a
few of his former classmates are applying
for the position as well, and some of them
received better grades than he did.
MOTIVATED FREE WILL BELIEF
18
Jim recently read an article about how
companies rarely check on certain aspects
of resumes, including GPAs. Apparently,
research has shown that fewer than 1% of
people overtly lie on their resumes, so it is
not worth the financial cost to companies
to verify certain aspects of resumes. He
decides to increase his GPA from a 3.42 to
a 3.92. He ends up getting the job over his
more qualified classmates.
Participants rated the extent to which the actor behaved of their own free will on the same
three items used in prior research: whether the action was freely chosen, whether the actor could
have made other choices, and whether the actor exercised his own free will in choosing to
perform the action on 7-point scales from Not at all to Very much so, s = .66 - .91. Participants
then reported the extent to which the actor should be punished on a 7-point scale from Not at all
to Very severely. Participants also reported their free will beliefs on the FAD+ free will belief
subscale (Paulhus & Carey, 2011). In this and all upcoming studies, we compute separate indices
for the free will belief subscale of the FAD+ ( = .88) and a non-moralized version of the
subscale ( = .79), which omits three items related to moral responsibility (as used in Monroe &
Ysidron, 2019). When we report results for the FAD+, we refer to the full subscale as ‘moralized
free will beliefs’ and the non-moralized version of the subscale as ‘non-moralized free will
beliefs’.
The final question of the free will belief subscale contained an attention check item, “If
you are paying attention select ‘Strongly Agree’.” Participants then re-read the vignette and were
asked to make normativity assessments on four questions: how often they see or hear about the
actions in their daily life (reverse-scored), how atypical or unusual the actions are, how much the
behavior breaks a social norm, and how likely it is that someone else in the same situation would
MOTIVATED FREE WILL BELIEF
19
do the same thing (reverse-scored), on relevant 7-point scales, = .67. Last, participants
completed demographic information.
Studies 1a-1d Results
Free will judgments. As can be seen in Table 3 and Figure 1, for all four vignettes, free
will attributions were higher in the Punish conditions than the Control conditions with effect
sizes ranging from very small to very large (Cohen’s ds = .15 1.03). These differences were
significant for the Home, Corporation, and Bus vignettes, ps < .003, but not significant for the
new GPA vignette, p = .235.
For moralized free will beliefs and non-moralized free will beliefs (see Table 3 and
Figure 2), the results were somewhat more ambiguous. Free will beliefs were higher in the
Punish conditions than the Control conditions with small to medium effects for Home and Bus
(Cohen’s ds = .25-.52), ps <.001 - .072. Results were in the expected direction, but with a small,
non-significant effect for Corporation (Cohen’s ds = .11), ps < .471. For the new GPA vignette,
there were virtually zero differences between conditions (Cohen’s ds = -.06 and -.01), ps > .691.
Due to the general consistency of the direction of the effects, but the inconsistency of the
statistical significance of the effects, all results are meta-analyzed in Studies 5a-c to compute
average estimated effect sizes of punish manipulations on free will attributions and free will
beliefs and to examine the different sizes of the effects across vignette types and free will
measure types.
Desires to punish. As can be seen in Table 3, participants consistently had stronger
desires to punish in the Punish conditions than the Control conditions, though note this
difference was substantially smaller for the new GPA vignette. This is perhaps not surprising as
both versions of the GPA vignette involved an identical morally bad action (lying about GPA on
MOTIVATED FREE WILL BELIEF
20
a resume). Future studies attempted to increase the discrepancy in desires to punish between the
two conditions in this vignette.
Normativity judgments. Regarding normativity judgments, participants rated the Punish
Conditions as more counter-normative than the Control Conditions for both the Home and Bus
vignettes. This makes it difficult to interpret whether desires to punish or perceived counter-
normativity increased free will attributions and free will beliefs for these vignettes. Supporting
our motivated account, for the Corporation vignette, counternormativity judgments were
significantly lower in the Punish Condition than the Control Condition. Thus, higher counter-
normativity judgments cannot account for the significantly higher free will attributions in the
Punish Condition. Though free will beliefs were in the hypothesized direction in this vignette,
they were not statistically significant. Given that this vignette did produce small but statistically
significant differences in free will beliefs in Studies 2a and 2b in Clark and colleagues’ (2018)
paper, which had a similar design, the present study simply may have been underpowered to
detect this small effect. Future studies increased the sample size.
Last, as intended, there were no significant differences in normativity judgments for the
GPA vignette, but there also were no significant differences in free will judgments between
conditions for this vignette. Furthermore, the discrepancy in desires to punish between the
Punish and Control conditions was relatively small (compared to the other vignettes). Future
studies attempted to increase the discrepancy in desires to punish between the Control and
Punish conditions for this vignette, while preserving the equivalent counternormativity.
Exploratory regressions. After reading a version of this paper, Monroe suggested we
either test whether the effects of the condition on free will judgments hold controlling for
counternormativity or to compare mediation models with counternormativity or desires to punish
MOTIVATED FREE WILL BELIEF
21
as the mediator. We decided to take a hybrid approach to this suggestion and test how much of
the variance of the effect of the condition on free will judgments is explained by
counternormativity and desires to punish. Therefore, in this and all subsequent studies (except
additional Pilot Study 1db), we conducted a second set of analyses, regressing each free will
judgment on the condition in Step 1 and the condition plus either desires to punish or
normativity judgments in Step 2 separately for each vignette. Note that desires to punish and
normativity judgments were added in Step 2 in separate models, so they were never competing
for the same variance.
We computed the semipartial r (the proportion of variance in the outcome uniquely
explained by the condition) as an estimate of the effect size for the condition in Step 1, and
estimated how much this effect size decreased in Step 2 as a result of adding desires to punish or
normativity judgments to the model. In other words, we examined how much the condition effect
was reduced when controlling for desires to punish or normativity judgments. If motivated free
will attributions and beliefs cannot be explained entirely by differences in counternormativity (as
we hypothesize), controlling for normativity should not abolish the condition effects. If
motivated free will attributions and beliefs can be explained at least partially by desires to
punish, desires to punish should weaken the condition effects. Note that these analyses, though
reported for every study (except additional Pilot Study 1db), were conducted after a full version
of the paper was written, thus none of these analyses were preregistered. We include these
additional analyses under the subheading ‘Exploratory regressions’ for each set of studies. Note
also that we performed these analyses regardless of the statistical significance of the condition
effect on the relevant outcome, so we are merely looking at reductions in the condition effect
size when controlling for either normativity or desires to punish.
MOTIVATED FREE WILL BELIEF
22
For the Home vignette, controlling for desires to punish eliminated entirely (and in fact
slightly reversed the direction of) the condition effect on free will attributions (Step 1:
semipartial r = .220, p = .002; Step 2: semipartial r = -.053, p =.434), moralized free will beliefs
(Step 1: semipartial r = .127, p = .072; Step 2: semipartial r = -.092, p =.184), and non-moralized
free will beliefs (Step 1: semipartial r = .132, p =.061; Step 2: semipartial r = -.057, p =.416). In
contrast, controlling for normativity had very little influence on the condition effect size on free
will attributions (Step 1: semipartial r = .220, p =.002; Step 2: semipartial r = .212, p = .003),
moralized free will beliefs (Step 1: semipartial r = .127, p = .072; Step 2: semipartial r = .113, p
= .111), and non-moralized beliefs (Step 1: semipartial r = .132, p =.061; Step 2: semipartial r =
.124, p =.080).
For the Corporation vignette, controlling for desires to punish substantially reduced or
eliminated entirely the condition effect on free will attributions (Step 1: semipartial r = .314, p <
.001; Step 2: semipartial r = .138, p =.042), moralized free will beliefs (Step 1: semipartial r =
.055, p = .438; Step 2: semipartial r = .006, p =.929), and non-moralized free will beliefs (Step 1:
semipartial r = .051, p =.470; Step 2: semipartial r = -.050, p =.486), though note the condition
effect was not statistically significant in Step 1 for moralized or non-moralized free will beliefs.
In contrast, controlling for normativity actually somewhat increased the size of the effect of the
condition on free will attributions (Step 1: semipartial r = .314, p <.001; Step 2: semipartial r =
.384, p < .001), moralized free will beliefs (Step 1: semipartial r = .055, p = .438; Step 2:
semipartial r = .095, p = .183), and non-moralized beliefs (Step 1: semipartial r = .051, p = .470;
Step 2: semipartial r = .071, p =.058).
For the Bus vignette, controlling for desires to punish substantially reduced or eliminated
the effect of the condition on free will attributions (Step 1: semipartial r = .457, p < .001; Step 2:
MOTIVATED FREE WILL BELIEF
23
semipartial r = .270, p < .001), moralized free will beliefs (Step 1: semipartial r = .260, p < .001;
Step 2: semipartial r = .004, p =.949), and non-moralized free will beliefs (Step 1: semipartial r =
.253, p < .001; Step 2: semipartial r = -.039, p =.564). Controlling for normativity also reduced
the effect of the condition on free will attributions (Step 1: semipartial r = .457, p < .001; Step 2:
semipartial r = .329, p < .001), moralized free will beliefs (Step 1: semipartial r = .260, p < .001;
Step 2: semipartial r = .186, p =.007), and non-moralized free will beliefs (Step 1: semipartial r =
.253, p < .001; Step 2: semipartial r = .178, p =.010), but to a lesser degree than desires to punish
for all three outcome variables (and did not eliminate the effect for any of the three outcome
variables).
For the GPA vignette, controlling for desires to punish very slightly reduced the size of
the effect of the condition on free will attributions (Step 1: semipartial r = .086, p = .235; Step 2:
semipartial r = .078, p = .280), moralized free will beliefs (Step 1: semipartial r = -.029, p =
.692; Step 2: semipartial r = -.052, p =.474), and non-moralized free will beliefs (Step 1:
semipartial r = -.009, p = .900; Step 2: semipartial r = -.020, p =.782), but these patterns are
difficult to interpret because the condition had a very small to slightly opposite the predicted
effect on free will judgments. Controlling for normativity had virtually zero influence on the
effect size of the condition on free will attributions (Step 1: semipartial r = .086, p = .235; Step
2: semipartial r = .087, p = .225), moralized free will beliefs (Step 1: semipartial r = -.029, p =
.692; Step 2: semipartial r = -.029, p =.691), and non-moralized free will beliefs (Step 1:
semipartial r = -.009, p = .900; Step 2: semipartial r = -.009, p =.902).
Overall then, in every single case, controlling for desires to punish at least slightly or
substantially reduced the effect size of the condition on free will judgments, whereas controlling
for counternormativity had no consistent pattern of influence on the effect of the condition on
MOTIVATED FREE WILL BELIEF
24
free will judgments. Thus these findings provide fairly compelling support that desires to punish
at least partially explain the effect of the manipulation on free will judgments, whereas they are
somewhat ambiguous regarding the role of counternormativity assessments (though we do not
think they eliminate the possibility that counternormativity assessments influence free will
judgments at least in some cases).
Pilot Study 1db
Because the GPA vignette seemed potentially promising for minimizing concerns about
counternormativity, Pilot Study 1db retested this vignette with slight alterations meant to
increase the discrepancy in desires to punish between the Punish and Control conditions. We
tested two new Control Conditions to contrast with a new Punish Condition in hopes of
identifying one usable Control Condition for future studies.
Study 1db Method
Participants. U.S. participants (Mage = 36.19, SD = 11.68; 89 female) were recruited via
MTurk. Because we were only interested in identifying trends for purposes of selecting vignettes
for future use (and not particularly concerned with statistical significance), we aimed for only 70
participants per condition (210 total). Three participants were removed for failing the attention
check, thus our final sample included 207 participants.
Procedure. Participants were randomly assigned to read one of three vignettes similar to
Study 1d about a person lying about their GPA on their resume, two which were Control
Conditions (the potential victim was either an unidentified undeserving group [Control 1] or an
identified undeserving male [Control 2], but the job search ended up being cancelled) and one
that was the Punish Condition (the victim was an identified more qualified female, and the actor
ended up getting the job over her). These are displayed in Table 2 below.
MOTIVATED FREE WILL BELIEF
25
Table 2. Exact phrasing of each condition within GPA vignette in Study 1db
Control 1
Control 2
Punish
Jim recently graduated from
university and has spent the
past few months preparing his
resume and other job
application materials. He
recently saw an advertisement
for his dream job, an entry-
level, but high-paying
position at a marketing firm
with excellent potential for
advancement. Unfortunately,
he knows some of his former
classmates are applying for
the position as well. Jim
received better grades than
many of them, but some of
them have contacts in
the company. Jim is
concerned this will put him at
a disadvantage.
Jim recently read an article
about how companies rarely
check on certain aspects of
resumes, including GPAs.
Apparently, research has
shown that fewer than 1% of
people overtly lie on their
resumes, so it is not worth the
financial cost to companies to
verify certain aspects of
resumes. He decides to
increase his GPA from a 3.42
to a 3.92. Regardless, the job
posting ultimately ends up
being canceled, and nobody
gets the job.
Jim recently graduated from
university and has spent the
past few months preparing his
resume and other job
application materials. He
recently saw an advertisement
for his dream job, an entry-
level, but high-paying
position at a marketing firm
with excellent potential for
advancement. Unfortunately,
he knows his former
classmate, Ethan, is applying
for the position as well. Ethan
received worse grades than
Jim, but Ethan's father went to
university with the company
president. Jim is concerned
this will put him at a
disadvantage.
Jim recently read an article
about how companies rarely
check on certain aspects of
resumes, including GPAs.
Apparently, research has
shown that fewer than 1% of
people overtly lie on their
resumes, so it is not worth the
financial cost to companies to
verify certain aspects of
resumes. He decides to
increase his GPA from a 3.42
to a 3.92. Regardless, the job
posting ultimately ends up
being canceled, and neither
Jim nor Ethan get the job.
Jim recently graduated from
university and has spent the
past few months preparing
his resume and other job
application materials. He
recently saw an
advertisement for his dream
job, an entry-level, but high-
paying position at a
marketing firm with
excellent potential for
advancement. Unfortunately,
he knows his former
classmate, Elizabeth, is
applying for the position as
well, and she received better
grades than he did. Jim is
concerned this will put him
at a disadvantage.
Jim recently read an article
about how companies rarely
check on certain aspects of
resumes, including GPAs.
Apparently, research has
shown that fewer than 1% of
people overtly lie on their
resumes, so it is not worth
the financial cost to
companies to verify certain
aspects of resumes. He
decides to increase his GPA
from a 3.42 to a 3.92. He
ends up getting the job over
his more qualified classmate,
Elizabeth.
On the same items as before, participants then rated the extent to which the actor behaved
of their own free will, ∝s = .82 - .87 and the extent to which the actor should be punished,
MOTIVATED FREE WILL BELIEF
26
reported their free will beliefs (full subscale = .88; non-moralized subscale = .79), and
completed the attention check item.
We were concerned that the “norm-breaking” question implied immorality, so in this and
all subsequent studies, we altered this question to read “How much does the behavior differ from
typical social norms?” and combined it with two other questions from the pilot studies as our
measure of counternormativity: how often they see or hear about the actions in their daily life
(reverse-scored) and how atypical or unusual the actions are, = .64. Last, participants
completed demographic information.
Study 1db Results
Because we were interested in comparing each Control condition to the Punish condition,
we conducted independent samples t-tests between each Control condition and the Punish
condition on free will attributions, free will beliefs, non-moralized free will beliefs, desires to
punish, and counter-normativity assessments. As can be seen in Table 3 and Figures 1 and 2, in
Study 1db, participants attributed more free will and believed more in free will on both the
moralized and non-moralized free will belief subscales in the Punish condition than the Control
conditions with small to medium effect sizes, Cohen’s ds = .23-.37. The exception was the effect
of the Punish condition vs. Control condition 1 on free will attributions, which was trending in
the expected direction, but with a very small effect, Cohen’s d = .13.
We successfully increased the discrepancy in desires to punish between the conditions;
participants wanted to punish more in the Punish Condition than both Control Conditions,
Cohen’s d = .76-.91. Though the behaviors in the three vignettes were identical and were
reported to have identical base rates, participants rated the Punish condition as slightly more
counternormative than the Control conditions. This suggests the possibility that desires to punish
MOTIVATED FREE WILL BELIEF
27
increase judgments of counternormativity as well, but that is a question for future research. This
counternormativity difference was smaller with Control 1, Cohen’s d = .17, and not significant, p
= .333. For this reason, Control 1 was used in the GPA vignette in subsequent studies.
MOTIVATED FREE WILL BELIEF
28
Table 3
Free will attributions, free will beliefs, non-moralized free will beliefs, desires
to punish, and normativity ratings within each vignette type
Free will attributions
t
df
p
Cohen’s
d
1a: Home
3.18
199
.002
.45
1b: Corporation
4.65
197
<.001
.67
1c: Bus
7.24
198
<.001
1.03
1d: GPA
1.19
192
.235
.15
1db: GPA Control 1
0.75
137
.452
.13
1db: GPA Control 2
1.37
136
.172
.24
Free will beliefs
1a: Home
1.81
199
.072
.25
1b: Corporation
0.78
197
.438
.11
1c: Bus
3.79
198
<.001
.47
1d: GPA
0.40
192
.692
-.06
1db: GPA Control 1
2.28
137
.024
.39
1db: GPA Control 2
1.54
136
.125
.26
Non-moralized free will beliefs
1a: Home
1.88
199
.061
.27
1b: Corporation
0.72
197
.470
.11
1c: Bus
3.68
198
<.001
.52
1d: GPA
0.13
192
.900
-.01
1db GPA Control 1
1.91
137
.058
.32
1db: GPA Control 2
1.34
136
.183
.23
Desires to punish
1a: Home
17.39
199
<.001
2.46
1b: Corporation
32.34
197
<.001
4.60
1c: Bus
32.81
198
<.001
4.63
1d: GPA
3.08
192
.002
.44
1db GPA Control 1
4.92
137
<.001
.76
1db: GPA Control 2
5.77
136
<.001
.91
Counter-normativity
judgments
1a: Home
3.33
199
.001
.47
1b: Corporation
-6.70
197
<.001
-.94
1c: Bus
11.74
198
<.001
1.65
1d: GPA
0.22
192
.828
.03
1db: GPA Control 1
0.97
137
.333
.17
1db: GPA Control 2
2.85
136
.005
.48
MOTIVATED FREE WILL BELIEF
29
Figure 1. Free will attributions by Punish and Control conditions in Studies 1a-1db. Error bars
are standard errors.
Figure 2. Non-moralized free will beliefs by Punish and Control conditions in Studies 1a-1db.
Error bars are standard errors.
5.61
6.02
5.66
6.20 6.44 6.33
6.12
6.64 6.72
6.36 6.56 6.56
4.00
4.50
5.00
5.50
6.00
6.50
7.00
1a: Home 1b:
Corporation 1c: Bus 1d: GPA 1db: GPA C1 1db: GPA C2
Free Will Attributions
Control Punish
3.69 3.85 3.76 3.88 3.87 3.93
3.90 3.93
4.14
3.87
4.11 4.11
3.00
3.50
4.00
4.50
5.00
1a: Home 1b:
Corporation 1c: Bus 1d: GPA 1db: GPA C1 1db: GPA C2
Non-Moralized Free Will Beliefs
Control Punish
MOTIVATED FREE WILL BELIEF
30
Preregistered Studies 2a-2d
We used the results of our pilot Studies 1a-1db to select and alter vignettes for a
preregistered study. Because the Punish condition was rated as more counternormative than the
Control condition in the Bus vignette, we altered the Control condition to be more
counternormative. Because the Control condition in the Corporation vignette was already rated
as more counternormative than the Punish condition, we did not alter these at all. We suspected
our adjustment to the counternormative question in Study 1db would eliminate
counternormativity differences in the Home vignettes, so these were not altered at all either. And
we used the GPA Control 1 condition that was rated as roughly equally counternormative as the
GPA Punish condition in Pilot Study 1db.
Studies 2a-2d Method
This study was preregistered: http://aspredicted.org/blind.php?x=xa5h2e. We followed
this preregistration exactly.
Participants. U.S. participants (Mage = 37.19, SD = 12.68; 845 female) were recruited via
MTurk. Consistent with Monroe and Ysidron (2019), we aimed for 200 participants per
condition (1600 total); 1603 participated. This allowed us to detect a small to medium effect size
with >80% power (G*Power; Faul, Erdfelder, Buchner, & Lang, 2009). Following our pre-
registered criterion, twenty-four participants were removed for failing an attention check, thus
our final sample included 1579 participants.
Procedure. As in Studies 1a-1d, participants were randomly assigned to read one of eight
vignettes which were paired (one morally bad [Punish condition], one less morally bad [Control
condition]) within the four studies, each of which included a different vignette type. The Home
(Study 2a) and Corporation (Study 2b) vignettes were identical to Studies 1a and 1b, and the
MOTIVATED FREE WILL BELIEF
31
GPA vignettes (Study 2d) were identical to Study 1db (with Control 1 as the Control condition).
For the Bus vignette (Study 2c), only the Control condition was slightly altered from Study 1c to
be more counternormative, which can be seen in Table 4. All questions were identical to Study
1db.
Table 4. Exact phrasing of Control Condition within Bus vignette in Study 2c
Vignette Type
Control Condition
Bus Vignette (2)
Andrew is riding the city bus, sitting next to an old lady. A little while later the bus stops,
and the lady slowly gets up and exits the bus. As she gets off Andrew slides over to sit in
her former seat. He continues this the rest of the way home. Every time someone exits, he
moves to take over their former seat. As a result, he ends up changing seats 14 times over
the course of his ride home.
Studies 2a-2d Results
Free will judgments. As can be seen in Table 5 and Figure 3, for the Home, Corporation,
and Bus vignettes, free will attributions were significantly higher in the Punish Conditions than
the Control Conditions with effect sizes ranging from medium to large (Cohen’s ds = .54 .77),
ps < .001. Despite the seemingly promising Pilot Study 1db results, there were no differences
between conditions for the GPA vignette, p = .795.
Turning from free will attributions to general free will beliefs, people reported higher
moralized free will beliefs and non-moralized free will beliefs (see Table 5 and Figure 4) in the
Punish conditions than the Control conditions across all four vignettes, but these differences
ranged from very small to small/medium, Cohen’s ds = .04-.26. For the full free will belief
subscale, the effect of condition was significant or marginal for Home, p = .046, Corporation, p =
.093, and Bus, p = .013, but not for GPA, p = .727. For non-moralized free will beliefs, the
condition effect was significant for Home, p = .050, and Bus, p = .024, but not Corporation, p =
.222, and GPA, p = .388. All of these effects are included in the upcoming meta-analysis.
MOTIVATED FREE WILL BELIEF
32
Table 5
Free will attributions, free will beliefs, non-moralized free will beliefs, desires
to punish, and normativity ratings within each vignette type
Free will attributions
t
df
p
Cohen’s
d
2a: Home
5.42
395
<.001
.54
2b: Corporation
6.67
391
<.001
.67
2c: Bus
7.55
394
<.001
.77
2d: GPA
0.26
391
.795
-.03
Free will beliefs
2a: Home
2.00
395
.046
.20
2b: Corporation
1.67
391
.093
.17
2c: Bus
2.50
394
.013
.26
2d: GPA
0.35
391
.727
.04
Non-moralized free will beliefs
2a: Home
1.96
395
.050
.20
2b: Corporation
1.22
391
.222
.13
2c: Bus
2.26
394
.024
.23
2d: GPA
0.86
391
.388
.09
Desires to punish
2a: Home
26.95
395
<.001
2.70
2b: Corporation
50.03
391
<.001
5.03
2c: Bus
44.98
394
<.001
4.52
2d: GPA
6.80
391
<.001
.69
Counter-normativity
judgments
2a: Home
1.29
395
.198
.13
2b: Corporation
-9.63
391
<.001
-.97
2c: Bus
-8.98
394
<.001
-.90
2d: GPA
2.09
391
.038
.21
MOTIVATED FREE WILL BELIEF
33
Figure 3. Free will attributions by Punish and Control conditions in Studies 2a-2d. Error bars are
standard errors.
Figure 4. Non-moralized free will beliefs by Punish and Control conditions in Studies 2a-2c.
Error bars are standard errors.
5.67 5.96 5.85
6.41
6.25 6.60 6.65 6.38
4.00
4.50
5.00
5.50
6.00
6.50
7.00
Home Corporation Bus GPA
Free Will Attributions
Control Punish
3.77 3.74 3.84 3.81
3.92 3.85 4.01 3.88
3.00
3.50
4.00
4.50
5.00
Home Corporation Bus GPA
Non-Moralized Free Will Beliefs
Control Punish
MOTIVATED FREE WILL BELIEF
34
Desires to punish and normativity judgments. As can be seen in Table 5, participants
consistently had stronger desires to punish in the Punish conditions than the Control conditions
with medium to very large effects (Cohen’s ds = .69 -5.03), ps < .001.
As expected, counter-normativity assessments did not significantly differ between the
Punish and Control conditions for the Home vignette, Cohen’s d = .13, p = .198. This suggests
that framing a normativity question as “norm-breaking” might appear to imply something
immoral. When phrased more neutrally, there were no differences between conditions for the
Home vignette. For the Corporation vignette and new Bus vignette, the Control conditions were
rated as more counternormative than the Punish conditions (Cohen’s ds = -.90 and -.97), ps <
.001, thus higher counternormativity assessments cannot account for higher attributions of free
will and higher beliefs in free will in the Punish conditions in these vignettes.
Again suggesting the possibility that desires to punish might influence normativity
assessments, participants rated lying about GPA on a resume as more counternormative when the
behavior resulted in the perpetrator getting the job over a more qualified female than when the
job search ended up being cancelled (Cohen’s d = .21), ps = .038, despite that both the action and
stated base rate were identical in the Punish and Control conditions. Future work might test this
possibility more thoroughly.
Exploratory regressions. As explained under the subheading ‘Exploratory regressions’
in the results of Pilot Studies 1a-1d, we again regressed each free will judgment on the condition
in Step 1 and the condition plus desires to punish or normativity judgments in Step 2 (separately)
for each vignette. We report the semipartial r (the proportion of variance in the outcome uniquely
explained by the condition) as an estimate of the effect size for the condition in Step 1, and
MOTIVATED FREE WILL BELIEF
35
estimated how much this effect size decreased in Step 2 as a result of controlling for desires to
punish or normativity judgments.
For the Home vignette, as in Pilot Studies 1a-1d, controlling for desires to punish
eliminated entirely (and slightly reversed the direction of) the effect of the condition on free will
attributions (Step 1: semipartial r = .263, p < .001; Step 2: semipartial r = -.002, p =.974),
moralized free will beliefs (Step 1: semipartial r = .100, p = .046; Step 2: semipartial r = -.160, p
=.001), and non-moralized free will beliefs (Step 1: semipartial r = .098, p =.050; Step 2:
semipartial r = -.150, p =.002). In contrast, controlling for normativity had very little influence
on the effect size of the condition on free will attributions (Step 1: semipartial r = .263, p < .001;
Step 2: semipartial r = .259, p < .001), moralized free will beliefs (Step 1: semipartial r = .100, p
= .046; Step 2: semipartial r = .098, p = .052), and non-moralized beliefs (Step 1: semipartial r =
.098, p =.050; Step 2: semipartial r = .097, p =.054), as in Pilot Studies 1a-1d.
For the Corporation vignette, as in Pilot Studies 1a-1d, controlling for desires to punish
substantially reduced or eliminated entirely and reversed the effect of the condition on free will
attributions (Step 1: semipartial r = .320, p < .001; Step 2: semipartial r = .079, p =.102),
moralized free will beliefs (Step 1: semipartial r = .085, p = .093; Step 2: semipartial r = -.094, p
=.061), and non-moralized free will beliefs (Step 1: semipartial r = .062, p =.222; Step 2:
semipartial r = -.119, p =.018), though note the condition effect was marginal in Step 1for
moralized free will beliefs and not statistically significant in Step 1 for non-moralized free will
beliefs. In contrast, controlling for normativity had little influence on (or slightly increased) the
effect size of the condition on free will attributions (Step 1: semipartial r = .320, p < .001; Step
2: semipartial r = .360, p < .001), moralized free will beliefs (Step 1: semipartial r = .085, p =
MOTIVATED FREE WILL BELIEF
36
.093; Step 2: semipartial r = .073, p = .148), and non-moralized beliefs (Step 1: semipartial r =
.062, p =.222; Step 2: semipartial r = .047, p =.351).
For the Bus vignette, as in Pilot Studies 1a-1d, controlling for desires to punish
substantially reduced or eliminated the effect of the condition on free will attributions (Step 1:
semipartial r = .355, p < .001; Step 2: semipartial r = .156, p = .001), moralized free will beliefs
(Step 1: semipartial r = .125, p = .013; Step 2: semipartial r = -.068, p =.171), and non-moralized
free will beliefs (Step 1: semipartial r = .113, p = .024; Step 2: semipartial r = -.050, p =.316).
Controlling for normativity produced somewhat mixed results, slightly increasing the effect of
the condition on free will attributions (Step 1: semipartial r = .355, p < .001; Step 2: semipartial r
= .371, p < .001), but decreasing the effect of the condition on moralized free will beliefs (Step
1: semipartial r = .125, p = .013; Step 2: semipartial r = .054, p =.278) and non-moralized free
will beliefs (Step 1: semipartial r = .113, p = .024; Step 2: semipartial r = .028, p =.567), though
again to a lesser degree than desires to punish.
For the GPA vignette, the results were again difficult to interpret because the condition
had no sizable influence on free will judgments. For this reason, these results are reported only in
the corresponding footnote.
2
Replicating the results of Pilot Studies 1a-1d, in every single case, controlling for desires
to punish at least slightly or substantially reduced the effect size of the condition on free will
judgments, whereas controlling for counternormativity demonstrated no consistent pattern of
2
Controlling for desires to punish slightly reduced the size of the effect of the condition on free will attributions
(Step 1: semipartial r = -.013, p = .795; Step 2: semipartial r = -.041, p = .420), moralized free will beliefs (Step 1:
semipartial r = .018, p = .727; Step 2: semipartial r = -.048, p =.337), and non-moralized free will beliefs (Step 1:
semipartial r = .044, p = .388; Step 2: semipartial r = -.025, p =.615). Controlling for normativity very slightly
reduced the effect size of the condition on free will attributions (Step 1: semipartial r = -.013, p = .795; Step 2:
semipartial r = -.015, p = .761), moralized free will beliefs (Step 1: semipartial r = .018, p = .727; Step 2: semipartial
r = .011, p =.820), and non-moralized free will beliefs (Step 1: semipartial r = .044, p = .388; Step 2: semipartial r =
.036, p =.471).
MOTIVATED FREE WILL BELIEF
37
influence on the condition effect. These findings again provide pretty compelling support that
desires to punish at least partially explain the effect of the experimental manipulation on free will
judgments. They are again somewhat ambiguous regarding the role of counternormativity
assessments, but they do again suggest that counternormativity cannot explain entirely the effect
of the condition on free will judgments.
Studies 3a-3b
Studies 3a-3b sought to extend the present findings to additional measures of free will
belief that are not at all susceptible to the ‘moral responsibility confound’ criticism. Studies 3a
and 3b used the Bus vignette from Study 2c. Motivated free will attributions and motivated free
will beliefs were robust in this vignette, and the Control Condition was rated as significantly
more counternormative than the Punish Condition, and thus this vignette is immune to the
counternormativity counterexplanation. These studies measured motivated free will beliefs on a
new standard personality measure of free will belief and, as an indirect measure, evaluations of
science opposing the existence of free will (replication of Clark et al. [2014] Study 4).
Studies 3a-3b Method
Participants. U.S. participants (Mage = 37.32, SD = 12.60; 240 female) were recruited via
MTurk. Because this was an initial test and not a pre-registered follow-up, we aimed for only
100 participants per condition (400 total); 400 participated. Studies 4a-c are pre-registered
follow-ups and increase the sample size to 200 participants per condition as in Studies 2a-2d.
Procedure. Participants were randomly assigned to the Punish or Control Condition of
the Bus vignette used in Study 2c. Participants responded to the same free will attribution
questions as in previous studies, s = .78-.79. Random subsets of participants were then assigned
to complete one of two measures of free will belief. In Study 3a, they completed the five free
MOTIVATED FREE WILL BELIEF
38
will factor items from Rakos, Laurene, Skala, & Slane’s (2008) Free Will and Determinism
Scale, including “Free will is a part of the human spirit.”, “I have free will even when my
choices are limited by external circumstances.”, “Free will is a basic part of human nature.”,
People have free will regardless of wealth or life circumstances.”, “Life's experiences cannot
eliminate a person's free will.”, on 5-point scales from Not true at all to Almost always true, =
.80. This scale contains no moral responsibility items and thus is not vulnerable to the ‘moral
responsibility confound’ criticism.
In Study 3b, participants evaluated science (see below) arguing against the existence of
free will (very similar, but not identical to Clark et al. [2014] Study 4), on five items, How
convinced were you by the psychologist’s argument?” [reversed], “How much would you like to
read the research articles that were talked about?” [reversed], “How much do you think this
psychologist was being controversial for the sake of getting his/her name into the papers?”,
“How important do you think research on automatic and unconscious processes is?” [reversed],
How much do you think research on automatic and unconscious processes should be a central
area within social psychology? [reversed], on 7-point scales from Not at all to Extremely, =
.71. Higher values indicated more skepticism of the anti-free will science, and thus stronger
motivations to believe in free will. This indirect measure of free will belief avoids both the moral
responsibility confound concerns (indeed, the scientific argument did not mention morality at all)
and concerns about demand characteristics.
Anti-free will science argument:
Recently, there has been a debate within psychology about the existence of free will. Two
distinguished psychologists recently had a public debate about this topic. One social
psychologist argued in favor of free will, and one argued against it.
MOTIVATED FREE WILL BELIEF
39
Below, please read one side of the debate:
"Of course, we all hold dear the idea that we have free will. It sure seems like we're in
charge, and it's a very scary feeling when feel we're not.
Yet thirty years plus of research has challenged the idea that we are 'masters of
ourselves' through our reason and free will, and suggested instead that a great deal of
our judgments about and responses to the world are automatic and unconscious, and that
our reason often acts as a sort lawyer, defending the decisions we have already
automatically made.
Studies have shown over and over again how unconscious influences have strong effects
over our behavior. These studies, and others like them, shows that actions and behaviors
that we feel we have control over are actually a result of automatic reactions to stimuli in
our environment. In other words, the subjective feeling that we experience of making
choices are likely a byproduct of unconscious processes, and our feelings of free will are
just an illusion."
Both of the free will belief measures contained one attention check item asking
participants to select a particular response if they were paying attention. This resulted in twelve
participant exclusions for the Rakos free will belief measure and no exclusions for the Science
Evaluation. All other questions were identical to Studies 2a-2d (counternormativity = .70).
Studies 3a-3b Results
Free will judgments. As can be seen in Table 6 and Figures 5 and 6, free will
attributions were significantly higher in the Punish Condition than the Control Condition with a
large effect size (Cohen’s d = .80, p<.001). Both the Rakos free will belief measure and the
science evaluation free will belief measure were in the expected direction; the effect was small
(and not significant) for Rakos (Cohen’s d = .15, p=.311) and small to medium for the Science
Evaluation (Cohen’s d = .34, p=.016).
Desires to punish and normativity judgments. As can be seen in Table 6, participants
had stronger desires to punish in the Punish condition than the Control condition with a very
large effect size (Cohen’s d = 4.44), p < .001. And again, the Control condition was rated as
MOTIVATED FREE WILL BELIEF
40
more counternormative than the Punish condition with a very large effect (Cohen’s d = -.91).
Thus, higher counternormativity assessments cannot account for the higher free will attributions
and beliefs in the Punish condition (and if counternormativity assessment do influence free will
judgments, that would be working against the effect observed here).
Exploratory regressions. As in the previous studies, controlling for desires to punish
substantially reduced or eliminated the effect of the condition on free will attributions (Step 1:
semipartial r = .387, p < .001; Step 2: semipartial r = .205, p = .003), Rakos free will beliefs
(Step 1: semipartial r = .075, p = .311; Step 2: semipartial r = -.041, p =.582), and Science
Evaluation free will beliefs (Step 1: semipartial r = .170, p = .016; Step 2: semipartial r = .072, p
=.302), though note again that the condition was not significant in Step 1 for Rakos. Controlling
for normativity again produced somewhat mixed results, slightly reducing the effect of the
condition on free will attributions (Step 1: semipartial r = .387, p < .001; Step 2: semipartial r =
.379, p < .001), slightly increasing the effect of the condition on Rakos free will beliefs (Step 1:
semipartial r = .075, p = .311; Step 2: semipartial r = .079, p =.286), and slightly decreasing the
effect of the condition on Science Evaluation free will beliefs (Step 1: semipartial r = .170, p =
.016; Step 2: semipartial r = .135, p =.054). Overall then, as in previous studies, the exploratory
regression analyses produced consistent support for desires to punish and mixed support for
counternormativity assessments.
Studies 4a-4c
Studies 4a and 4b sought to replicate Studies 3a-3b, preregistered and with larger
samples. Study 4c also included the entire FAD+ as used in Monroe and Ysidron (2019) to test
whether their null findings were due to their use of the entire 27 item FAD+ rather than the 7-
item free will belief subscale only (as used in Studies 1a-2d). We expected to replicate the small
MOTIVATED FREE WILL BELIEF
41
to medium effect of the Punish manipulation on evaluations of anti-free will science and the
small effect on the Rakos free will belief scale. As stated in the preregistration, we were unsure
whether we would replicate the effect of the Punish manipulation on the FAD+ free will belief
subscale (as found in Studies 1a-2d) when the subscale was presented mixed in with the other 20
items of the FAD+ scale. Possibly, contemplation of these other 20 items distracts from the
lingering desire to punish from the vignette.
Studies 4a-4c Method
This study was preregistered: http://aspredicted.org/blind.php?x=up4rs4. We followed
this preregistration exactly.
Participants. U.S. participants (Mage = 37.98, SD = 12.87; 699 female) were recruited via
MTurk. We aimed for 200 participants per condition within each study (1200 total); 1208
participated.
Procedure. Participants were randomly assigned to the Punish or Control condition of
the Bus vignette used in Studies 2c, 3a, and 3b. Participants responded to the same free will
attribution questions as in previous studies, s = .79-.83. Random subsets of participants were
then assigned to complete one of three measures of free will belief: the Rakos scale from Study
3a ( = .85), the Science Evaluation from Study 3b ( = .76), or the FAD+ free will belief
subscale (Paulhus & Carey, 2011; full subscale = .78; non-moralized subscale = .71) but this
time embedded within the entire FAD+, which includes 20 additional questions about beliefs in
fate, luck, and scientific determinism.
All of the free will belief measures contained one attention check item asking participants
to select a particular response if they were paying attention. This resulted in four participant
exclusions for the Rakos free will belief measure, one exclusion for the Science Evaluation, and
MOTIVATED FREE WILL BELIEF
42
twelve exclusions for the FAD+ (the higher exclusion rate for the FAD+ likely reflects the longer
length and tediousness of the scale), thus our final sample included 1191 participants. All other
questions were identical to Studies 2a-3b (counternormativity = .65).
Studies 4a-4c Results
Free will judgments. As can be seen in Table 6 and Figures 5 and 6, free will
attributions were significantly higher in the Punish condition than the Control condition with a
medium to large effect size (Cohen’s d = .70), ps < .001. Both the Rakos free will belief measure
and the Science Evaluation free will belief measure were in the expected direction, with a small,
marginal effect for Rakos (Cohen’s d = .19, p=.067) and a medium, significant effect for the
Science Evaluation (Cohen’s d = .34, p<.001). Though we did not know whether to expect an
effect of the Punish manipulation on free will beliefs as measured by the full FAD+, we
replicated a small, marginal effect of the Punish condition on higher beliefs in free will on the
full free will subscale (Cohen’s d = .19, p=.064), and a smaller (but not significant) effect on the
non-moralized free will belief subscale (Cohen’s d = .15, p=.141).
MOTIVATED FREE WILL BELIEF
43
5.88 5.93
6.70 6.68
4.00
4.50
5.00
5.50
6.00
6.50
7.00
Study 3 FWA Study 4 FWA
Free Will Attributions
Control Punish
Table 6
Free will attributions, Rakos free will beliefs, Science Evaluation free will
beliefs, desires to punish, and normativity ratings in Studies 3a-4c
Free will attributions
t
df
p
Cohen’s d
Study 3 Bus 2
7.99
398
<.001
.80
Study 4 Bus 2
12.21
1204
<.001
.70
Rakos free will beliefs
Study 3a Bus 2
1.02
183
.311
.15
Study 4a Bus 2
1.84
398
.067
.19
Science evaluation free will beliefs
Study 3b Bus 2
2.44
200
.016
.34
Study 4b
5.49
399
<.001
.55
Full FAD+ moralized free will beliefs
Study 4c Bus 2
1.86
386
.064
.19
Full FAD+ non-moralized free will beliefs
Study 4c Bus 2
1.48
386
.141
.15
Desires to punish
Study 3 Bus 2
44.54
398
<.001
4.44
Study 4 Bus 2
77.68
1204
<.001
4.47
Counter-normativity judgments
Study 3 Bus 2
-9.91
398
<.001
-.91
Study 4 Bus 2
-13.66
1205
<.001
-.79
MOTIVATED FREE WILL BELIEF
44
Figure 5. Free will attributions by Punish and Control conditions in Studies 3 and 4. Error bars
are standard errors.
Figure 6. Non-moralized free will attributions by Punish and Control conditions in Studies 3 and
4. Error bars are standard errors.
Desires to punish and normativity judgments. As can be seen in Table 6, participants
had stronger desires to punish in the Punish condition than the Control condition with a very
large effect (Cohen’s d = 4.47, p < .001). And again, the Control condition was rated as more
counternormative than the Punish condition, with a medium to large effect (Cohen’s d = -.79, p <
.001).
Exploratory regressions. As in previous studies, controlling for desires to punish
substantially reduced or eliminated the effect of the condition on free will attributions (Step 1:
semipartial r = .332, p < .001; Step 2: semipartial r = .110, p < .001), Rakos free will beliefs
(Step 1: semipartial r = .092, p = .067; Step 2: semipartial r = -.028, p =.575), and Science
Evaluation free will beliefs (Step 1: semipartial r = .265, p < .001; Step 2: semipartial r = .078, p
4.06 3.99
3.49 3.52 3.60
4.17 4.14
3.86
4.13
3.71
3.00
3.50
4.00
4.50
5.00
Study 3a Rakos Study 4a Rakos Study 3b Science
Evaluation Study 4b Science
Evaluation Study 4c FAD+
Non-moralized Free Will Beliefs
Control Punish
MOTIVATED FREE WILL BELIEF
45
=.108), though only very slightly reduced the (marginal) effect of the condition on full FAD+
moralized free will beliefs (Step 1: semipartial r = .094, p = .064; Step 2: semipartial r = .062, p
=.219) and full FAD+ non-moralized free will beliefs (Step 1: semipartial r = .075, p = .141;
Step 2: semipartial r = .065, p =.200). Controlling for normativity assessments again produced
inconsistent results, slightly increasing the effect of the condition on free will attributions (Step
1: semipartial r = .332, p < .001; Step 2: semipartial r = .340, p < .001), Rakos free will beliefs
(Step 1: semipartial r = .092, p = .067; Step 2: semipartial r = .115, p =.021), and Science
Evaluation free will beliefs (Step 1: semipartial r = .265, p < .001; Step 2: semipartial r = .277, p
< .001), but reducing the (marginal) effect of the condition on full FAD+ moralized free will
beliefs (Step 1: semipartial r = .094, p = .064; Step 2: semipartial r = .018, p =.723) and full
FAD+ non-moralized free will beliefs (Step 1: semipartial r = .075, p = .141; Step 2: semipartial
r = .002, p =.974), and for these latter two measures (and for the first time in any of the studies)
to a stronger degree than desires to punish.
As in previous studies, these exploratory analyses provide compelling support that desires
to punish play some role in free will judgments, though the support is more compelling for free
will attributions, Rakos free will beliefs, and the science evaluation free will beliefs than for the
full FAD+ free will beliefs. The results for counternormativity assessments are again ambiguous,
actually strengthening the effect of the condition for free will attributions, Rakos free will
beliefs, and science evaluation free will beliefs, but mostly eliminating the (marginal) effect of
the condition on the full FAD+ free will belief measures.
Across Studies 1a-4c, controlling for desires to punish consistently reduced or eliminated
the effect of the experimental manipulations on free will judgments. In contrast, controlling for
counternormativity assessments sometimes increased, sometimes decreased, and sometimes had
MOTIVATED FREE WILL BELIEF
46
virtually no influence on the effect size of the experimental manipulation, and in only very few
cases did controlling for counternormativity eliminate the significant effects of the conditions on
free will judgments.
Discussion
The results thus far strongly support our contention that motivated free will attributions
and beliefs are not solely the result of stronger assessments of counternormativity. We do not
doubt that counternormativity assessments can influence free will judgments, but they are not the
explanation for the present effects. However, the size of the effect for motivated free will
attributions and free will beliefs varied substantially across studies, vignette type, and free will
measure. Studies 5a-c are three meta-analyses of all studies included in the present paper, and the
studies we are aware of from previous work. This should be considered a working meta-analysis
that can be expanded upon over time as more work is conducted. Ideally, more researchers (other
than Clark and Monroe) would contribute included studies.
Study 5a-c: Three Meta-Analyses
As a last step, we conducted three meta-analyses of the existing literature (the studies
reported here, plus studies from four additional papers [Clark et al., 2014; Clark et al., 2017;
Clark et al., 2018; Monroe & Ysidron, 2019], and three unpublished studies) on motivated free
will attributions, motivated moralized free will beliefs, and motivated non-moralized free will
beliefs.
Study 5a-c Method
We used procedures outlined by Goh, Hall, and Rosenthal (2016) for conducting three
meta-analyses on motivated free will attributions, motivated moralized free will beliefs, and
motivated non-moralized free will beliefs. We computed r effect sizes from the Ms, SDs, and ns
MOTIVATED FREE WILL BELIEF
47
for each Punish and Control condition in each study, which were then were Fisher’s Z
transformed to rz. We performed random effects meta-analyses using single sample t-tests.
Random effects is appropriate because the studies involved slightly different experimental
designs with different manipulations and different free will belief measures (see Goh et al.,
2016). Fixed effects (weighted by sample size) would bias the results by overweighting
particular studies (with particular experimental designs) that happened to have the most
participants.
We planned to test a number of moderators. These include:
(1) Sample: Undergraduates (U) or Mturk (M)
(2) Vignette: Home, Bus, Corporation (Corp), GPA, Judge (a corrupt judge), Field Study
(an ostensibly real classroom cheating incident), Lasers (a person stealing expensive
lasers from a hospital) and General (a broad suggestion about responsibility)
(3) Whether the Control condition was a Good, Neutral, or Bad action (G, N, B)
3
(4) Free Will Measurement Type (DV): Free Will Attributions (FWA), evaluations of
anti-free will science (Science Eval), Rakos, Stroessner (Stroessner & Green, 1990), the
free will belief subscale from the FAD+ when presented with the full FAD+ (FFAD+),
the free will belief subscale from the FAD+ when presented in shorter versions of the
FAD+ (SFAD+),
4
and the non-moralized versions of these same measures (FFAD+
[NM] and SFAD+ [NM]).
3
In studies that included both Good and Neutral conditions, we used the Neutral condition only
4
Shortened versions included one to three of the four subscales included in the full FAD+.
MOTIVATED FREE WILL BELIEF
48
The meta-analysis for motivated free will attributions included only FWA. The meta-
analysis for motivated free will beliefs included all of the free will belief DVs with the full
(moralized) FFAD+ and SFAD+ subscales. The meta-analysis for the non-moralized free will
beliefs is the same meta-analysis as for motivated free will beliefs except (1) the non-moralized
FFAD+ and SFAD+ replaced all values for the FFAD+ and SFAD+, and (2) one study is omitted
(Clark et al., 2014; Study 1) because we did not have access to the raw data in order to compute
the non-moralized versions of the FFAD+ and SFAD+ (the experimenter for that study has left
academia). Thus, the non-moralized meta-analysis is a subset of the data from the motivated free
will beliefs meta-analysis.
All Ms, SDs, ns, rs, and all coded variables for each study included in each of the three
meta-analyses are displayed in Tables 8 (free will attributions), 9 (free will beliefs), and 10 (non-
moralized free will beliefs).
Other notes on coding decisions. In one unpublished study (which we call Clark et al
2014 Study X), there were two measures of free will belief in one study with the same
participants; in this case, only one average effect size was included; however, when we report
separate average effect sizes for different types of free will belief measurements, we keep them
separate. For Clark et al 2014 Study 1, we were not sure whether the FFAD+ or SFAD+ was
used, so this effect size was omitted from the moderation by DV type analysis. For Science Eval,
higher values correspond to more negative evaluations of the anti-free will science and thus
higher free will beliefs like all other DVs. For Clark et al 2014 Study 3, there were two Punish
conditions, which were collapsed. For Clark et al 2014 Study X, which used the Home vignette,
there was also a target manipulation such that the bad behavior was directed toward a good other,
MOTIVATED FREE WILL BELIEF
49
a bad other, or the self. Because all other studies that used the Home vignette involved the good
other condition only, only this condition was included.
Study 5a-c Results
We conducted three single-sample t-tests on the rzs for each of the three meta-analyses.
For contrasts, we conducted independent samples t-tests on the rzs for two group contrasts, and
ANOVAs with Bonferroni post-hoc tests on multi-group contrasts (though note, some real
effects might not be detected with the limited sample size).
Free will attributions. There was a significant effect of motivated free will attributions
(k=22, n=7,619), rz = .26, t(21)=8.14, p<.001, 95% CI [.19, .33].
No effect size differences were detected between either Undergraduate (k=6; rz=.24) and
Mturk (k=16; rz=.26) participants, t(20)=-0.29, p=.776, or between Published (k=19; rz=.26) and
Not Published (k=3; rz=.27) studies, t(20)=0.12, p=.908.
The effect size was moderated by vignette type, F(4, 17)=9.95, p<.001. The largest
effects were found for the Bus (k=5; rz=.40) and Lasers (k=3; rz=.40) vignettes, which did not
significantly differ (p=1.00) and were significantly higher than all others (ps<.008) but the
Corporation vignette (k=4; rz=.26), p=.359, which marginally differed from the GPA vignette
(k=3; rz=.05), p=.078. The Home vignette fell between GPA and Corporation (k=7; rz=.18), and
did not significantly differ from either, ps>.541.
The effect was also moderated by whether the Control condition in the vignette was a
Good, Neutral, or Bad action, F(2, 19)=4.52, p=.025. The largest effects were observed when
control conditions were Neutral (k=15; rz=.30), followed by Good (k=4; rz=.26), with the
smallest effects being observed for Bad controls (k=3; rz=.05). Neutral and Bad significantly
MOTIVATED FREE WILL BELIEF
50
differed, p=.022. Good did not significantly differ from Bad, p=.143, or Neutral, p=1.00. These
comparisons are displayed in Figure 7 below.
Figure 7. Average effect sizes for free will attributions by each control condition type. Error bars
are standard errors.
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Neutral
Good
Bad
Free Will Attributions Effect Size by Control Condition Type
Running Head: MOTIVATED FREE WILL BELIEF
51
Table 8. Descriptives and effect sizes for motivated free will attributions by study
Paper
Study
Punish
Control
Total n
r
Sample
DV
Vignette
G, N, B
Published
Clark et al 2014
2
M
5.33
4.67
95
0.3
U
FWA
Home
N
Y
SD
1.29
1.31
n
49
46
Clark et al 2014
X
M
5.73
5.3
110
0.2
U
FWA
Home
N
N
SD
1.21
1.13
n
60
50
Clark et al 2017
5
M
5.88
5.65
388
0.1
U
FWA
Home
N
Y
SD
1.13
1.24
n
202
186
Clark et al 2018
1b
M
6.05
4.56
124
0.5
U
FWA
Lasers
N
Y
SD
1.32
1.07
n
70
54
Clark et al 2018
2a
M
8.52
7.86
451
0.3
U
FWA
Corp
G
Y
SD
0.93
1.53
n
223
228
Clark et al 2018
2b
M
8.14
7.85
784
0.1
M
FWA
Corp
G
Y
SD
1.05
1.21
n
393
391
Clark et al 2018
X
M
6.43
5.56
141
0.2
M
FWA
Lasers
N
N
SD
1.83
2.01
n
63
78
Clark et al 2018
Y
M
7.12
6.03
140
0.4
M
FWA
Lasers
N
N
SD
1.17
1.45
n
63
77
Clark et al 2019
1a
M
6.12
5.61
201
0.2
M
FWA
Home
N
Y
SD
1.1
1.19
MOTIVATED FREE WILL BELIEF
52
n
101
100
Clark et al 2019
1b
M
6.64
6.02
199
0.3
M
FWA
Corp
G
Y
SD
0.79
1.05
n
100
99
Clark et al 2019
1c
M
6.72
5.66
200
0.5
M
FWA
Bus
N
Y
SD
0.86
1.18
n
101
99
Clark et al 2019
1d
M
6.36
6.2
194
0.1
M
FWA
GPA
B
Y
SD
0.93
0.92
n
97
97
Clark et al 2019
1db
M
6.56
6.39
207
0.1
M
FWA
GPA
B
Y
SD
0.97
0.9
n
70
137
Clark et al 2019
2a
M
6.25
5.67
397
0.3
M
FWA
Home
N
Y
SD
0.96
1.18
n
197
200
Clark et al 2019
2b
M
6.6
5.96
393
0.3
M
FWA
Corp
G
Y
SD
0.92
0.98
n
196
197
Clark et al 2019
2c
M
6.65
5.85
396
0.4
M
FWA
Bus
N
Y
SD
0.77
1.26
n
201
195
Clark et al 2019
2d
M
6.38
6.41
393
-0
M
FWA
GPA
B
Y
SD
1.03
0.88
n
199
194
Clark et al 2019
3
M
6.7
5.88
400
0.4
M
FWA
Bus
N
Y
SD
0.76
1.23
n
201
199
MOTIVATED FREE WILL BELIEF
53
Clark et al 2019
4
M
6.68
5.93
1206
0.3
M
FWA
Bus
N
Y
SD
0.79
1.27
n
604
602
Monroe & Ysidron
2a
M
6.22
5.86
401
0.2
M
FWA
Home
N
Y
2019
SD
0.93
1.27
n
202
199
Monroe & Ysidron
2b
M
5.48
5.26
399
0.1
U
FWA
Home
N
Y
2019
SD
1.19
1.2
n
205
194
Monroe & Ysidron
3
M
6.63
5.78
400
0.4
M
FWA
Bus
N
Y
2019
SD
0.84
1.18
n
200
200
Running Head: MOTIVATED FREE WILL BELIEFS
54
Free will beliefs. There was a significant effect of motivated free will beliefs (k=27,
n=8,100), rz = .13, t(26)=7.39, p<.001, 95% CI [.10, .17].
There were no significant differences between Undergraduate (k=6; rz=.18) and Mturk
(k=21; rz=.12) participants, t(25)=-1.51, p=.145. The only Not Published study had a
significantly larger effect size (k=1; rz=.31) than the Published studies (k=26; rz=.13),
t(25)=2.06, p=.050, but given the single data point for Not Published studies, this result should
not be taken as meaningful.
Unlike free will attributions, the effect size was not significantly moderated by vignette
type. The effect sizes for each vignette in order from largest to smallest are Field Study (k=1;
rz=.24), General (k=1; rz=.24), Lasers (k=1; rz=.20), Bus (k=8; rz=.16), Home (k=7; rz=.14),
Judge (k=2; rz=.11), Corporation (k=4; rz=.09), and GPA (k=3; rz=.04).
Similar to free will attributions, the effect was marginally moderated by whether the
Control condition in the vignette was a Good, Neutral, or Bad action, F(2, 24)=2.83, p=.079.
Again, the largest effects were observed when control conditions were Neutral (k=20; rz=.16),
followed by Good (k=4; rz=.09), with the smallest effects being observed for Bad controls (k=3;
rz=.04), though none significantly differed in post-hoc tests, ps>.148. This is displayed in Figure
8 below.
MOTIVATED FREE WILL BELIEFS
Figure 8. Average effect sizes for free will beliefs by each control condition type. Error bars are
standard errors.
The effect was also significantly moderated by DV, F(3, 22)=3.30, p=.039. The largest
effects were observed for evaluations of the anti-free will science passages (k=4; rz=.22),
followed by short versions of the FAD+ (k=15; rz=.14), followed by Rakos (k=2; rz=.09), and
last by the full FAD+ (k=5; rz=.05). Only the full FAD+ and Science Evaluations significantly
differed in post-hoc tests, p=.036. Note the Stroessner scale could not be included in these
comparisons because only one study used Stroessner, but this effect size was largest (rz=.29).
These are displayed in Figure 9 below.
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Neutral
Good
Bad
Free Will Beliefs Effect Size by Control Condition Type
MOTIVATED FREE WILL BELIEFS
Figure 9. Average effect sizes for free will beliefs by each DV type. Error bars are standard
errors.
Note. Stroessner has no SE because it included only one study.
-0.4 -0.3 -0.2 -0.1 00.1 0.2 0.3 0.4
Stroessner
Science Evaluation
SFAD+
Rakos
FFAD+
Free Will Beliefs Effect Size by DV Type
Running Head: MOTIVATED FREE WILL BELIEFS
1
Table 9. Descriptives and effect sizes for motivated free will beliefs by study
Paper
Study
Punish
Control
Total n
r
Sample
DV
Vignette
G, N, B
Published
Clark et al 2014
1
M
4.01
3.72
154
0.21
M
*FAD+
Judge
N
Y
SD
0.66
0.7
n
77
77
Clark et al 2014
2
M
3.68
3.38
92
0.22
U
SFAD+
Home
N
Y
SD
0.7
0.62
n
46
46
Clark et al 2014
3
M
3.72
3.39
277
0.24
U
SFAD+
Field
N
Y
SD
0.6
0.64
Study
n
190
87
Clark et al 2014
4
M
2.71
2.32
212
0.18
M
Science
Home
N
Y
SD
1.05
1.04
Eval
n
106
106
Clark et al 2014
X
M
3.93
3.56
111
0.31
U
SFAD+
Home
N
N
SD
0.56
0.57
n
61
50
Clark et al 2014
X
M
5.34
4.71
110
0.28
U
Stroessner
Home
N
N
SD
1.04
1.11
n
62
48
Clark et al 2018
1a
M
3.82
3.23
137
0.24
M
Science
General
N
Y
SD
1.2
1.17
Eval
science
n
69
68
Clark et al 2018
1b
M
3.75
3.49
121
0.2
U
SFAD+
Lasers
N
Y
SD
0.56
0.72
n
70
51
Clark et al 2018
2a
M
4
3.84
450
0.12
U
SFAD+
Corp
G
Y
SD
0.64
0.7
MOTIVATED FREE WILL BELIEFS
n
222
228
Clark et al 2018
2b
M
4.06
3.96
784
0.08
M
SFAD+
Corp
G
Y
SD
0.61
0.65
n
393
391
Clark et al 2019
1a
M
3.97
3.78
201
0.13
M
SFAD+
Home
N
Y
SD
0.66
0.83
n
101
100
Clark et al 2019
1b
M
3.99
3.91
199
0.06
M
SFAD+
Corp
G
Y
SD
0.74
0.67
n
100
99
Clark et al 2019
1c
M
4.23
3.87
200
0.25
M
SFAD+
Bus
N
Y
SD
0.64
0.73
n
101
99
Clark et al 2019
1d
M
3.94
3.98
194
-0.03
M
SFAD+
GPA
B
Y
SD
0.63
0.71
n
97
97
Clark et al 2019
1db
M
4.22
3.99
207
0.15
M
SFAD+
GPA
B
Y
SD
0.69
0.72
n
70
137
Clark et al 2019
2a
M
3.98
3.84
397
0.1
M
SFAD+
Home
N
Y
SD
0.75
0.72
n
197
200
Clark et al 2019
2b
M
3.94
3.81
393
0.08
M
SFAD+
Corp
G
Y
SD
0.79
0.76
n
196
197
Clark et al 2019
2c
M
4.11
3.93
396
0.13
M
SFAD+
Bus
N
Y
SD
0.65
0.71
n
201
195
MOTIVATED FREE WILL BELIEFS
Clark et al 2019
2d
M
3.91
3.89
393
0.01
M
SFAD+
GPA
B
Y
SD
0.7
0.76
n
199
194
Clark et al 2019
3a
M
4.18
4.06
185
0.08
M
Rakos
Bus
N
Y
SD
0.68
0.76
n
89
96
Clark et al 2019
3b
M
3.86
3.49
202
0.17
M
Science
Bus
N
Y
SD
1.08
1.09
Eval
n
104
98
Clark et al 2019
4a
M
4.14
3.99
400
0.1
M
Rakos
Bus
N
Y
SD
0.75
0.82
n
202
198
Clark et al 2019
4b
M
4.13
3.52
401
0.27
M
Science
Bus
N
Y
SD
1.15
1.06
n
212
189
Clark et al 2019
4c
M
3.88
3.76
388
0.09
M
FFAD+
Bus
N
Y
SD
0.62
0.66
n
182
206
Monroe & Ysidron
1
M
3.72
3.71
406
0.01
M
FFAD+
Judge
N
Y
2019
SD
0.66
0.68
n
205
201
Monroe & Ysidron
2a
M
3.87
3.85
401
0.01
M
FFAD+
Home
N
Y
2019
SD
0.69
0.78
n
202
199
Monroe & Ysidron
2b
M
3.66
3.67
399
-0.01
U
FFAD+
Home
N
Y
2019
SD
0.62
0.66
n
205
194
Monroe & Ysidron
3
M
3.93
3.74
400
0.14
M
FFAD+
Bus
N
Y
MOTIVATED FREE WILL BELIEFS
2019
SD
0.63
0.76
n
200
200
Running Head: MOTIVATED FREE WILL BELIEFS
1
Non-moralized free will beliefs. There was a significant effect of non-moralized
motivated free will beliefs (k=26, n=7953), rz = .12, t(25)=6.96, p<.001, 95% CI [.08, .16].
There were no significant differences between Undergraduate (k=6; rz=.16) and Mturk
(k=20; rz=.11) participants, t(24)=-1.13, p=.268. The only Not Published study had a
significantly larger effect size (k=1; rz=.30) than the Published studies (k=25; rz=.11),
t(24)=2.24, p=.035, but given the single data point for Not Published studies, this should not be
considered meaningful.
The effect size was not significantly moderated by vignette type. The effect sizes for each
vignette in order from largest to smallest are General (k=1; rz=.24), Field Study (k=1; rz=.21),
Lasers (k=1; rz=.18), Bus (k=8; rz=.15), Home (k=7; rz=.12), Corporation (k=4; rz=.08), and
GPA (k=3; rz=.06), and Judge (k=1; rz=-.03).
Unlike free will attributions and free will beliefs, the effect was not moderated by
whether the Control condition was Good, Neutral, or Bad, F(2, 23)=1.79, p=.189, though the
pattern was similar. Again, the largest effects were observed when control conditions were
Neutral (k=19; rz=.14), followed by Good (k=4; rz=.08), with the smallest effects observed for
Bad controls (k=3; rz=.06). This is depicted in Figure 10 below.
MOTIVATED FREE WILL BELIEFS
Figure 10. Average effect sizes for non-moralized free will beliefs by each control condition
type. Error bars are standard errors.
The effect was again moderated by DV, F(3, 22)=4.77, p=.010. The largest effects were
observed for evaluations of anti-free will science (k=4; rz=.22), followed by short versions of the
FAD+ (k=15; rz=.13), followed by Rakos (k=2; rz=.09), and last by the full FAD+ (k=5; rz=.04).
Only the full FAD+ and Science Evaluations significantly differed in post-hoc tests, p=.007.
Note the Stroessner scale could not be included in these comparisons because only one study
used Stroessner, but this effect size was largest (rz=.29). This is depicted in Figure 11 below.
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Neutral
Good
Bad
Non-moralized Free Will Beliefs Effect Size by Control Condition Type
MOTIVATED FREE WILL BELIEFS
Figure 11. Average effect sizes for non-moralized free will beliefs by each DV type. Error bars
are standard errors.
Note. Stroessner has no SE because it included only one study.
-0.4 -0.3 -0.2 -0.1 00.1 0.2 0.3 0.4
Stroessner
Science Evaluation
SFAD+
Rakos
FFAD+
Non-moralized Free Will Beliefs Effect Size by Measurement Type
Running Head: MOTIVATED FREE WILL BELIEFS
1
Table 10. Descriptives and effect sizes for non-moralized motivated free will attributions by study
Paper
Study
Punish
Control
Total n
r
Sample
DV
Vignette
G, N, B
Published
Clark et al 2014
2
M
3.71
3.52
93
0.13
U
SFAD+ (NM)
Home
N
Y
SD
0.74
0.75
n
47
46
Clark et al 2014
3
M
3.78
3.44
278
0.21
U
SFAD+ (NM)
Field
N
Y
SD
0.72
0.76
Study
n
190
88
Clark et al 2014
4
M
2.71
2.32
212
0.18
M
Science
Home
N
Y
SD
1.05
1.04
Evaluation
n
106
106
Clark et al 2014
X
M
3.97
3.6
111
0.29
U
SFAD+ (NM)
Home
N
N
SD
0.58
0.65
n
61
50
Clark et al 2014
X
M
5.34
4.71
110
0.28
U
Stroessner
Home
N
N
SD
1.04
1.11
n
62
48
Clark et al 2018
1a
M
3.82
3.23
137
0.24
M
Science
General
N
Y
SD
1.2
1.17
Evaluation
science
n
69
68
Clark et al 2018
1b
M
3.83
3.58
126
0.18
U
SFAD+ (NM)
Lasers
N
Y
SD
0.59
0.78
n
72
54
Clark et al 2018
2a
M
4.06
3.9
450
0.11
U
SFAD+ (NM)
Corp
G
Y
SD
0.71
0.73
n
222
228
Clark et al 2018
2b
M
4.02
3.92
784
0.08
M
SFAD+ (NM)
Corp
G
Y
SD
0.64
0.68
MOTIVATED FREE WILL BELIEFS
n
393
391
Clark et al 2019
1a
M
3.9
3.69
201
0.13
M
SFAD+ (NM)
Home
N
Y
SD
0.69
0.89
n
101
100
Clark et al 2019
1b
M
3.93
3.85
199
0.05
M
SFAD+ (NM)
Corp
G
Y
SD
0.82
0.67
n
100
99
Clark et al 2019
1c
M
4.14
3.76
200
0.25
M
SFAD+ (NM)
Bus
N
Y
SD
0.68
0.77
n
101
99
Clark et al 2019
1d
M
3.87
3.88
194
-0.01
M
SFAD+ (NM)
GPA
B
Y
SD
0.7
0.73
n
97
97
Clark et al 2019
1db
M
4.11
3.9
207
0.13
M
SFAD+ (NM)
GPA
B
Y
SD
0.76
0.77
n
70
137
Clark et al 2019
2a
M
3.92
3.77
397
0.09
M
SFAD+ (NM)
Home
N
Y
SD
0.81
0.79
n
197
200
Clark et al 2019
2b
M
3.85
3.74
393
0.07
M
SFAD+ (NM)
Corp
G
Y
SD
0.86
0.83
n
196
197
Clark et al 2019
2c
M
4.01
3.84
396
0.11
M
SFAD+ (NM)
Bus
N
Y
SD
0.73
0.75
n
201
195
Clark et al 2019
2d
M
3.88
3.81
393
0.05
M
SFAD+ (NM)
GPA
B
Y
SD
0.74
0.81
n
199
194
MOTIVATED FREE WILL BELIEFS
Clark et al 2019
3a
M
4.18
4.06
185
0.08
M
Rakos
Bus
N
Y
SD
0.68
0.76
n
89
96
Clark et al 2019
3b
M
3.86
3.49
202
0.17
M
Science
Bus
N
Y
SD
1.08
1.09
n
104
98
Clark et al 2019
4a
M
4.14
3.99
400
0.1
M
Rakos
Bus
N
Y
SD
0.75
0.82
n
202
198
Clark et al 2019
4b
M
4.13
3.52
401
0.27
M
Science
Bus
N
Y
SD
1.15
1.06
n
212
189
Clark et al 2019
4c
M
3.71
3.59
388
0.08
M
FFAD+ (NM)
Bus
N
Y
SD
0.73
0.79
n
182
206
Monroe & Ysidron
1
M
3.64
3.68
406
-0.03
M
FFAD+ (NM)
Judge
N
Y
2019
SD
0.72
0.74
n
205
201
Monroe & Ysidron
2a
M
3.76
3.74
401
0.01
M
FFAD+ (NM)
Home
N
Y
2019
SD
0.74
0.83
n
202
199
Monroe & Ysidron
2b
M
3.64
3.64
399
0
U
FFAD+ (NM)
Home
N
Y
2019
SD
0.72
0.68
n
205
194
Monroe & Ysidron
3
M
3.82
3.63
400
0.12
M
FFAD+ (NM)
Bus
N
Y
2019
SD
0.69
0.84
n
200
200
Running Head: MOTIVATED FREE WILL BELIEFS
1
Summary of average effect sizes. Including all studies, there were small to medium
effects of free will attributions (r = .25), beliefs (r = .13), and non-moralized beliefs (r = .12).
These results did not differ by sample type or whether studies were published (though few
studies were unpublished). However, the effect sizes did appear to differ by vignette type,
control condition type, and DV type. There are two main takeaways from these analyses. First,
the GPA vignettes (which also were the only studies with bad Control conditions) produced very
small effects. Thus, when both the Control condition and the Punish condition involve bad
actions, motivated free will attributions and beliefs may not be detectable (possibly, because the
Control conditions are already near ceiling). Second, assessing free will beliefs within the full
27-item FAD+ produced very small effects. Future research should likely avoid using long
measures where the items of interest are mixed among various others kinds of questions.
If we omit studies that included the GPA vignettes (because studies with control
conditions that are also morally bad appear to be at ceiling and thus do not demonstrate the
effect) and the full FAD+ (which contains more distractor items than free will belief items), we
find slightly larger average effects, but still in the range of small to medium effects for free will
attributions (r = .28), beliefs (r = .18), and non-moralized beliefs (r = .17). Future work that uses
similar methods as those used in the studies presented here can use these average effect sizes for
their power calculations.
Discussion
The meta-analyses revealed significant effects for motivated free will attributions and
motivated free will beliefs (with or without moral responsibility items), with effect sizes in the
small to medium range. Thus, when taken together, the extant data reveals that motivated free
MOTIVATED FREE WILL BELIEFS
will attributions and free will beliefs are real and replicable effects (and motivated free will
beliefs are not due solely to the ‘moralized’ nature of the FAD+).
Moderator analyses revealed factors that may minimize or eliminate these effects.
Generally, these effects were largest for studies with Control conditions that involved Neutral
actions (rather than Good or Bad), though Good actions fell between Bad and Neutral (these
findings are consistent with Clark and colleagues [2018] findings that free will attributions for
morally good actions fall between those for neutral and bad actions). These effects were smallest
for studies with Control conditions that also involved bad behaviors (these were the GPA
vignettes). Thus, the effects of motivated free will attributions and motivated free will beliefs are
likely very small or even nonexistent for studies comparing a Bad Control condition to a More
Bad Punish condition. This may be due to ceiling effects in the Bad Control conditions, but
generally, we expect differences will be small to nonexistent when comparing free will
attributions and free will beliefs following exposure to vignettes including two bad actions that
merely vary in severity.
Effect sizes were generally robust across free will belief measurement, with the largest
effects observed for the Stroessner scale, evaluations of anti-free will science, and short versions
of the FAD+. Effects were small for the Rakos scale, and very small for the full FAD+ (as used
in Monroe and Ysidron [2019] failed replications). We expect then that these effects will be
largest when free will belief measures come directly after the manipulation and participants are
not distracted by numerous unrelated questions. As noted the introduction of this study, we
consider these to be working meta-analyses. We hope other researchers will continue to try to
replicate these studies using both similar and new methods to further refine our understanding of
when we observe motivated free will attributions and beliefs and when we might not.
MOTIVATED FREE WILL BELIEFS
General Discussion
Monroe and Ysidron forwarded very plausible and compelling alternate explanations for
motivated free will attributions and free will beliefs: that these effects may be explained by the
high counternormativity of immoral actions and moral responsibility confounds in the FAD+.
We agree that higher counternormativity assessments likely do increase attributions of free will
and that the FAD+ does confound free will belief and moral responsibility. The present results,
however, suggest that neither of these alternative hypotheses are sufficient for explaining these
effects. We found motivated free will attributions and beliefs when Punish conditions were rated
as equally and less counternormative than paired Control conditions and when moral
responsibility confounds were removed. Moreover, the present results replicated these effects
with higher standards of evidence than the original Clark and colleagues (2014) paper by
including numerous preregistered studies and meta-analyses of these effects. These findings
increase our confidence that these effects are real and replicable. However, we also discovered
that the effects are weak to nonexistent in studies that measured free will beliefs with the entire
FAD+ and in studies with control conditions involving bad actions (as opposed to neutral
actions), and thus we have no confidence that the effects would replicate in these kinds of
studies.
We believe that our motivated free will belief account is the most plausible and
parsimonious explanation for the observed findings. Monroe and Ysidron’s proposal simply
cannot account for all of the data, although, as noted above, we believe counternormativity likely
does influence attributions of free will (and related responsibility judgments [see, e.g.,
Mandelbaum & Ripley, 2012]). That we observed higher attributions of free will and free will
beliefs even in cases in which Punish conditions were rated as less counternormative than
MOTIVATED FREE WILL BELIEFS
Control conditions, suggests the possibility that the motivational component is capable of
overriding (and indeed reversing) the potential influence of counternormativity on such
judgments (at least in some cases).
Confirmation and Extension of Prior Work
Numerous previous studies have demonstrated asymmetric attributions of responsibility:
People attribute more control and responsibility to bad actions than to similar good or neutral
actions (e.g., Alicke, 1992; Knobe, 2003), and the present results demonstrate a similar pattern
for free will attributions. However, debates surrounding whether such findings reflected
motivated social cognition (and specifically, desires to blame and punish) or more rational
considerations (such as the counternormativity of the action) have been ongoing for decades. To
our knowledge, the present research is first to rule out this potential alternate explanation entirely
(and experimentally) by consistently replicating such findings even in cases in which behaviors
meant to elicit desires to punish are rated as less counternormative than behaviors in paired
control conditions.
Moreover, the present findings go one step further: People believe more in free will after
contemplating bad actions (those that elicit desires to punish) than after contemplating similar
good and especially neutral actions, a finding difficult or perhaps impossible to reconcile with
more rationalist perspectives. These results suggest that desires to punish at least partially
motivate broad beliefs in the human capacity for free action. This is not to say that desires to
punish are the only reason people believe in free will. However, much the same way beliefs in
omnipotent Gods likely do not reflect careful considerations about the plausibility of
supernatural agents, free will beliefs likely do not reflect careful analysis about the metaphysical
requirements for free will and moral responsibility (Clark et al., 2019). Though we suspect there
MOTIVATED FREE WILL BELIEFS
are numerous reasons people believe in free will, at least one reason appears to be that such
beliefs justify one’s punitive desires (Clark et al, 2017).
We suspect that motivated free will beliefs have become more common as society has
become more humane and more concerned about proportionate punishment. Many people now
assiduously reflect upon their own society’s punitive practices and separate those who deserve to
be punished from those who are incapable of being fully responsible for their actions. Free will is
crucial here because it is often considered a prerequisite for moral responsibility (Nichols &
Knobe, 2007; Sarkissian et al., 2010; Shariff et al., 2014). Therefore, when one is motivated to
punish another person, one is also motivated to inflate free will beliefs and free will attributions
to specific perpetrators as a way to justify punishing the person. Furthermore, punishing a target
that others deem unworthy of punishment puts one at risk of retaliation from third party
observers of the unjust punishment (Cinyabuguma, Page, & Putterman, 2006). It is therefore
critical to convince others that one’s punitive actions are reasonable and even righteous.
Claiming that the perpetrator one desires to punish had free will is an important part of this
effort.
The results of the GPA vignette also offer some new insights about what features of an
actor’s behavior influence free will judgments. In this vignette, an actor was said to have lied
about his GPA on a resume in both conditions, but in the control condition, the job search was
cancelled, and in the punish condition, the actor ended up getting the job over his more qualified
classmate. Thus the actual actions were the same, but the punish condition resulted in more
harmful consequences. Yet we pretty consistently observed little to no differences between these
two conditions in free will attributions and free will beliefs. These results may indicate
MOTIVATED FREE WILL BELIEFS
that motivated free will belief effects are more driven by an actor’s behavior than by the
outcome of that behaviors. In other words, free will judgments may be more sensitive to
indicators of an actor’s character than to the overall harm caused by a particular action. Future
research should more carefully examine whether this is correct.
Implications for Human Responsibility
These findings may be troubling to those who wish to uphold moral responsibility.
Indeed, a deep concern for the preservation of moral responsibility is apparent in Monroe and
Ysidron’s critique of motivated free will beliefs (emphases are our own):
“If true, this would represent a coup for scholars’ and laypersons’ conception of free will
and morality… Accepting this provocative claim would invite significant and disruptive
upheaval upon modern legal theory, which presumes that it is permissible and just to
punish offenders’ misdeeds because they had the free will to reasonably “have done
otherwise” (Greene & Cohen, 2004). Similarly, this account would threaten
philosophical thinking dating back to Aristotle insofar as free will is often assumed to be
a necessary condition for holding people morally responsible (Aristotle, 1985; Kant,
1953; Nichols & Knobe, 2007; Sarkissian et al., 2010).”
We wish to make two points here. First, the present results have no bearing on whether
people have the freedom necessary for moral responsibility. Scholars have been debating this for
millennia, and we suspect they will continue to do so for the foreseeable future (List, Caruso, &
Clark, 2020). To our knowledge, this debate no longer involves disagreements about empirical
reality (i.e., scholars have mostly accepted that behavior results from an interaction between
genes and environment, and that even environments that are chosen result from previous genetic
and environmental causes [though we are simplifying for space]). These disagreements are now
MOTIVATED FREE WILL BELIEFS
about whether the limited freedom humans do have is sufficient for moral responsibility, which
is a normative and conceptual debate, not a straightforwardly empirical one. The present data
show that people bolster their belief in free will when wishing to punish; they do not show that
people lack the capacities necessary for moral responsibility (if scholars could agree in the first
place on what those are).
Second, punishment appears necessary for social group functioning (Fehr & Gächter,
2002; Henrich et al., 2006). And because humans are a social species and care deeply about their
moral reputations (Vonasch, Reynolds, Winegard, & Baumeister, 2018), judgment and
condemnation are effective regulators of social behavior (Sperber & Baumard, 2012). Whether
humans ultimately deserve moral judgment is a separate question from whether moral judgment
is a useful (or perhaps even necessary) social practice (and probably the latter is empirical
whereas the former might not be). One could imagine, for example, that receiving a fine and
temporarily losing one’s license after driving while intoxicated is less of a deterrent than
receiving a fine, temporarily losing one’s license, and temporarily feeling socially shamed
because one is told one deserves to be blamed for freely behaving irresponsibly. Furthermore,
there are many behaviors that are not illegal (and thus do not receive organized punishment), but
that society might still wish to discourage, such as having extramarital affairs and being an
absentee father. In these cases, moral judgment and social condemnation might be the main
deterrents. Until empirical data have demonstrated that societies can function just as well without
moral judgment, we might preserve the concept.
This assumes that intellectual discussions about moral responsibility and free will beliefs
could eliminate such beliefs. But we doubt this is so (see, e.g., Knobe, 2014). Humans likely
evolved to judge one another morally precisely because it deterred destructive behaviora
MOTIVATED FREE WILL BELIEFS
benefit for social group functioning (and for the self and one’s kin within the social group)
(Clark & Winegard, 2019). Although this does not necessarily mean it is good that they do so, it
suggests that it served an important function that was independent of the truth value of free will.
Moral judgments are likely as inescapable an impulse as beliefs in an external world or in other
minds, and philosophical and intellectual discussions are as unlikely to get rid of them (at least
among everyday people, and possibly even among philosophers and intellectuals [Nichols,
2007]) as they are to turn the average person into an idealist or a solipsist. Our findings show that
people are likely motivated to preserve moral responsibility, and particularly for harmful
behaviors. Our findings may explain why, for example, genetic evidence and neurological
explanations have little influence on judgments about legal cases (Appelbaum, Scurich, & Raad,
2015; De Brigard, Mandelbaum & Ripley, 2009; although see also Confer & Chopik, 2019), and
that the moral valence of actions may influence whether genetic explanations impact moral
responsibility assessments (Tabb, Lebowitz, & Appelbaum, 2019). Whether these patterns of
human cognition are problematic is a separate issue. And if they are problematic, intervening on
them still may not be possible.
Limitations and Lingering Questions
The present work has a number of limitations to consider, perhaps the main one being the
artificiality of vignette studies with self-report measures (a common, though serious problem
with much of social psychology). Clark and colleagues’ (2014) paper included a field study in
which undergraduates were led to believe that one of their own classmates had cheated on their
recent midterm exam, a far more realistic context than the new studies reported here. Studies
high in realism are harder to conduct (and perhaps more ethically dubious) but likely get closer
to demonstrating how these findings play out in the real world. This may be why the field study
MOTIVATED FREE WILL BELIEFS
produced one of the largest effect sizes for motivated free will beliefs (r=.24). Despite that
cheating on an exam is likely a less serious infraction than robbing a home (as in the Home
vignette), that participants genuinely believed one of their classmates had cheated likely elicited
more powerful desires to punish (and thus stronger motivations to preserve free will) than a
hypothetical story on a computer screen.
This limitation may also explain the fragility of the effect. The effect was nearly wiped
away when measures of free will belief were mixed among twenty other questions (about, e.g.,
fatalism and scientific determinism) in the full FAD+. One possible explanation for these
findings is that our manipulations of desires to punish are fleeting. Future work might continue to
test these effects in more realistic contexts to determine whether the size of this effect is
substantially larger (and longer lasting) when people are experiencing real and powerful desires
to punish others. Even small effects could have large real-world consequences, especially for
dichotomous outcomes, such as “guilty” vs. “not guilty”. It is also possible that something about
contemplating fate, luck, and scientific determinism attenuates the effect. Perhaps desires to
punish also influence belief in fate, luck, and scientific determinism, and as those beliefs shift, so
do free will beliefs. Still, it is possible that the stronger effect in shorter and more immediate
measures is boosted by demand characteristics. This is perhaps less likely in the studies in which
participants evaluated an anti-free will scientific passage but might explain the relatively larger
effects in the more direct measures (e.g., SFAD+, Rakos, Stroessner). Future work might test
these multiple possible explanations for why we observe little to no motivated free will belief
effect on the full FAD+.
Another rather severe limitation is that the meta-analysis represented only two first
authors’ work (Clark and Monroe) and very heavily represented work by Clark. Ideally, meta-
MOTIVATED FREE WILL BELIEFS
analyses summarize work from many different research teams with different perspectives and
interests (Tetlock, 2020) because the views of individual researchers can likely bias the results of
scientific studies and even the academic literature more broadly (e.g., Clark & Winegard, 2020;
Winegard & Clark, 2020). Indeed, research teams sometimes influence effect sizes within meta-
analyses (e.g., Yen & Cheng, 2013) and perhaps even between meta-analyses testing similar
research questions (e.g., Gildersleeve, Haselton, & Fales, 2014; Wood, Kressel, Joshi, & Louie,
2014). Unfortunately, the studies included in our meta-analysis are the only studies we are aware
of that test our particular research question. As noted above, we consider this a working meta-
analysis that can be updated as other scholars participate in this research area.
It is also important to note that we do not know exactly why people believe more in free
will while wishing to punish (and why this alleviates distress about punishing). Do people
experience internal cognitive dissonance when they wish to punish someone who does not have
free will? Do people fear external judgment and punishment when they wish to punish someone
who does not have free will? Or does believing in free will increase one’s capacity to be a
ruthless punisher? These are challenging questions to answer, but two strategies would be to look
at free will beliefs historically and cross-culturally. For example, if certain cultures do not
consider free will a prerequisite for moral responsibility, but still punish ruthlessly and do not
demonstrate motivated free will belief effects, this might suggest that motivated free will beliefs
are largely about persuading other people that one’s moral judgments and punitiveness are
justified. Some evidence demonstrates that cultures with different moral systems differ in their
attribution asymmetries for morally good and morally bad actions (Clark, Bauman, Kamble, &
Knowles, 2017), but it remains unknown how these would influence motivated free will beliefs.
MOTIVATED FREE WILL BELIEFS
Future research should leverage cultural differences to test the multiple possible explanations for
these kinds of effects.
Future work should also assess how individual-differences in punitiveness might explain
individual-differences in free will beliefs. Some preliminary work suggests that political
conservatives’ tendency to moralize more actions might explain why political conservatives
generally believe more (than liberals) in free will (Everett et al., 2017). If those with more
punitive personalities believe more in free will in general and perhaps display larger motivated
free will belief tendencies, this could have important implications for jury selection, criminal
justice reform, and possibly the empirical conclusions drawn by scientists of different personality
types.
Conclusion
The present research consistently replicated motivated free will beliefs and ruled out
potential alternate explanations for these effects. Thus we can say with greater confidence that
desires to punish lead people to increase their broad beliefs in the human capacity for free action.
These findings provide additional support for Nietzsche’s insight over 100 years ago: Free will
beliefs may reflect, at least in part, desires to create and uphold human moral responsibility.
MOTIVATED FREE WILL BELIEFS
References
Appelbaum, P. S., Scurich, N., & Raad, R. (2015). Effects of behavioral genetic evidence on
perceptions of criminal responsibility and appropriate punishment. Psychology, Public
Policy, and Law, 21, 134.
Alicke, M. D. (1992) Culpable causation. Journal of Personality and Social Psychology, 63,
368-378.
Alicke, M. D. (2000). Culpable control and the psychology of blame. Psychological Bulletin,
126, 556574.
Aristotle. (1985). Nicomachean ethics. (T. Irwin, Trans.). Indianapolis, IN: Hackett.
Boyer, P., & Bergstrom, B. (2008). Evolutionary perspectives on religion. Annual Review of
Anthropology, 37, 111-130.
Cinyabuguma, M., Page, T., & Putterman, L. (2006). Can second-order punishment deter
perverse punishment?. Experimental Economics, 9, 265-279.
Clark, C. J., Baumeister, R. F., & Ditto, P. H. (2017). Making punishment palatable: Belief in
free will alleviates punitive distress. Consciousness and Cognition, 51, 193-211.
Clark, C. J., Bauman, C. W., Kamble, S. V., & Knowles, E. D. (2017). Intentional sin and
accidental virtue? Cultural differences in moral systems influence perceived
intentionality. Social Psychological and Personality Science, 8, 74-82.
Clark, C. J., Liu, B. S., Winegard, B. M., & Ditto, P. H. (2019). Tribalism is human
nature. Current Directions in Psychological Science, 28, 587-592.
Clark, C. J., Luguri, J. B., Ditto, P. H., Knobe, J., Shariff, A. F., & Baumeister, R. F. (2014).
Free to punish: A motivated account of free will belief. Journal of Personality and Social
Psychology, 106, 501-513.
MOTIVATED FREE WILL BELIEFS
Clark, C. J., Shniderman, A., Luguri, J. B., Baumeister, R. F., & Ditto, P. H. (2018). Are morally
good actions ever free?. Consciousness and Cognition, 63, 161-182.
Clark, C. J., Winegard, B. M., & Baumeister, R. F. (2019). Forget the Folk: Moral Responsibility
Preservation Motives and Other Conditions for Compatibilism. Frontiers in
Psychology, 10, 215.
Clark, C. J., & Winegard, B. M. (2019). An evolutionary perspective on free will belief. Science
Trends.
Clark, C. J., & Winegard, B. M. (2020). Tribalism in war and peace: The nature and evolution of
ideological epistemology and its significance for modern social science. Psychological
Inquiry, 31, 1-22.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence
Hillsdale, NJ: Lawrence Erlbaum.
Confer, J. A., & Chopik, W. J. (2019). Behavioral explanations reduce retributive punishment
but not reward: The mediating role of conscious will. Consciousness and Cognition, 75,
102808.
Cushman, F., Gray, K., Gaffey, A., & Mendes, W. B. (2012). Simulating murder: the aversion to
harmful action. Emotion, 12, 2-7.
Cushman, F., Knobe, J., & Sinnott-Armstrong, W. (2008). Moral appraisals affect
doing/allowing judgments. Cognition, 108, 281-289.
De Brigard, F., Mandelbaum, E., & Ripley, D. (2009). Responsibility and the brain
sciences. Ethical Theory and Moral Practice, 12, 511.
MOTIVATED FREE WILL BELIEFS
Ditto, P. H., Clark, C. J., Liu, B., Wojcik, S. P., Chen, E. E., Grady, R. H., Celniker, J. & Zinger,
J. F. (2019). Partisan bias and it’s discontents. Perspectives on Psychological Science, 14,
304-316.
Ditto, P. H., Liu, B., Clark, C. J., Wojcik, S. P., Chen, E. E., Grady, R. H., Celniker, J. & Zinger,
J. F. (2019). At least bias is bipartisan: A meta-analytic comparison of partisan bias in
liberals and conservatives. Perspectives on Psychological Science, 14, 273-291.
Ditto, P. H. & Lopez. (1992). Motivated Skepticism: Use of different decision criteria for
preferred and nonpreferred conclusions. Journal of Personality and Social Psychology,
63, 568-584.
Everett, J. A., Clark, C., Meindl, P., Luguri, J., Earp, B. D., Graham, J., ... & Shariff, A. (2020).
Political differences in free will belief are associated with differences in
moralization. Journal of Personality and Social Psychology.
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using
G*Power 3.1: Tests for correlation and regression analyses. Behavior Research
Methods, 41, 1149-1160.
Fehr, E. & Gächter, S. (2005). Altruistic punishment in humans. Nature, 415, 137140.
Feltz, A., Cokely, E. T., & Nadelhoffer, T. (2009). Natural compatibilism versus natural
incompatibilism: Back to the drawing board. Mind & Language, 24, 1-23.
Feltz, A., & Cova, F. (2014). Moral responsibility and free will: A meta-analysis. Consciousness
and Cognition, 30, 234-246.
Gildersleeve, K., Haselton, M. G., & Fales, M. R. (2014). Do women’s mate preferences change
across the ovulatory cycle? A meta-analytic review. Psychological Bulletin, 140, 1205.
MOTIVATED FREE WILL BELIEFS
Goh, J. X., Hall, J. A., & Rosenthal, R. (2016). Mini metaanalysis of your own studies: Some
arguments on why and a primer on how. Social and Personality Psychology
Compass, 10, 535-549.
Greene, J. D., & Cohen, J. D. (2004). For the law, neuroscience changes nothing and everything.
Philosophical Transactions of the Royal Society of London. Series B: Biological
Sciences, 359, 17751785. doi:10.1098/rstb.2004.1546
Greene, J., & Haidt, J. (2002). How (and where) does moral judgment work?. Trends in
Cognitive Sciences, 6, 517-523.
Haggard, P., Mele, A., O’Connor, T., & Vohs, K. (2015). Free will lexicon. In A. R. Mele (Ed.),
Surrounding free will (pp. 319- 326). New York, NY: Oxford University Press.
Haider-Markel, D. P., & Joslyn, M. R. (2008). Beliefs about the origins of homosexuality and
support for gay rights: An empirical test of attribution theory. Public Opinion
Quarterly, 72, 291-310.
Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach to moral
judgment. Psychological Review, 108, 814834.
Haidt, J., Bjorklund, F., & Murphy, S. (2000). Moral dumbfounding: When intuition finds no
reason. Unpublished manuscript, University of Virginia, 191-221.
Haselton, M. G., & Nettle, D. (2006). The paranoid optimist: An integrative evolutionary model
of cognitive biases. Personality and social psychology Review, 10, 47-66.
Henrich, J., McElreath, R., Barr, A., Ensminger, J., Barrett, C., Bolyanatz, A., Cardenas, J.
C., Gurven, M., Gwako, E., Henrich, N., Lesorogol, C., Marlowe, F., Tracer, D., & Ziker,
J. (2006). Costly punishment across human societies. Science, 312, 1767-1770.
Kant, I. (1953). Critique of pure reason. (N. K. Smith, Trans.). London: Macmillan.
MOTIVATED FREE WILL BELIEFS
Knobe, J. (2003). Intentional action and side effects in ordinary language.” Analysis, 63, 190-
194.
Knobe, J. (2014). Free will and the scientific vision. In E. Machery & E. O’Neill (eds.), Current
controversies in experimental philosophy (pp. 6985). New York: Routledge
Knobe, J. & Fraser, B. (2008). Causal judgment and moral judgment: Two experiments. In W.
Sinnott-Armstrong (Ed.), Moral psychology (Vol. 2, pp. 441-448). Cambridge, MA: MIT
Press.
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108, 480498.
Lerner, J. S., Goldberg, J. H., & Tetlock, P. E. (1998). Sober second thought: The effects of
accountability, anger, and authoritarianism on attributions of responsibility. Personality
and Social Psychology Bulletin, 24, 563-574.
Leslie, A. M, Knobe, J., & Cohen, A. (2006). Acting intentionally and the side-effect effect:
Theory of mind and moral judgment. Psychological Science, 17, 421-427.
Liu, B. S. & Ditto, P. H. (2012). What dilemma? Moral evaluation shapes factual belief. Social
Psychological and Personality Science, 00, 1-8.
List, C., Caruso, G., & Clark, C. J. (2020). Free will: Real or illusionA debate. The
Philosopher, 108.
Mandelbaum, E., & Ripley, D. (2012). Explaining the abstract/concrete paradoxes in moral
psychology: the NBAR hypothesis. Review of Philosophy and Psychology, 3, 351-368.
Mercier, H. & Sperber, D. (2011). Why do humans reason? Arguments for an argumentative
theory. Behavioral and Brain Sciences, 34, 57-111.
MOTIVATED FREE WILL BELIEFS
Monroe, A., & Ysidron, D. W. (2019). Not so motivated after all? Methodological and
theoretical challenges to a morally-motivated belief in free will. Journal of Personality
and Social Psychology.
Nahmias, E., Morris, S., Nadelhoffer, T., & Turner, J. (2005). Surveying freedom: Folk
intuitions about free will and moral responsibility. Philosophical Psychology, 18, 561-
584.
Nichols, S. (2007). The rise of compatibilism: A case study in the quantitative history of
philosophy. Midwest Studies in Philosophy, 31, 260-270.
Nichols, S. & Knobe, J. (2007). Moral responsibility and determinism: The cognitive science of
folk intuitions. Noûs, 41, 663685.
Nietzsche, F. (1954). Twilight of the idols. (W. Kaufmann, Trans.). New York: Penguin Books.
(Original work published 1889).
Paolacci, G., & Chandler, J. (2014). Inside the Turk: Understanding Mechanical Turk as a
participant pool. Current Directions in Psychological Science, 23, 184-188.
Paulhus, D. L., & Carey, J. (2011). The FAD-Plus: Measuring lay beliefs regarding free will and
related constructs. Journal of Personality Assessment, 93, 96104.
Rakos, R. F., Laurene, K. R., Skala, S., & Slane, S. (2008). Belief in free will: Measurement and
conceptualization innovations. Behavior and Social Issues, 17, 20-39.
Reeder, G. D., & Spores, J. M. (1983). The attribution of morality. Journal of Personality and
Social Psychology, 44, 736-745.
Sarkissian, H., Chatterjee, A., De Brigard, F., Knobe, J., Nichols, S. and Sirker, S. (2010), Is
belief in free will a cultural universal?. Mind & Language, 25, 346358.
MOTIVATED FREE WILL BELIEFS
Shariff, A. F., Greene, J. D., Karremans, J. C., Luguri, J. B., Clark, C. J., Schooler, J. W., ... &
Vohs, K. D. (2014). Free will and punishment: A mechanistic view of human nature
reduces retribution. Psychological Science, 25, 1563-1570.
Sperber, D., & Baumard, N. (2012). Moral reputation: An evolutionary and cognitive
perspective. Mind & Language, 27, 495-518.
Stroessner, S. J. & Green, C. W. (1990). Effect of belief in free will or determinism on attitudes
toward punishment and locus of control. Journal of Social Psychology, 130, 789-799.
Tabb, K., Lebowitz, M. S., & Appelbaum, P. S. (2019). Behavioral genetics and attributions of
moral responsibility. Behavior Genetics, 49, 128-135.
Tetlock, P. E. (2002). Social functionalist frameworks for judgment and choice: intuitive
politicians, theologians, and prosecutors. Psychological Review, 109, 451-471.
Tetlock, P. E. (2020). Gauging the Politicization of Research Programs. Psychological
Inquiry, 31, 86-87.
Valdesolo, P., & DeSteno, D. (2006). Manipulations of emotional context shape moral
judgment. Psychological Science, 17, 476-477.
Von Hippel, W., & Trivers, R. (2011). The evolution and psychology of self-
deception. Behavioral and Brain Sciences, 34, 1-16.
Vonasch, A. J., Clark, C. J., Lau, S., Vohs, K. D., & Baumeister, R. F. (2017). Ordinary people
associate addiction with loss of free will. Addictive Behaviors Reports, 5, 56-66.
Vonasch, A. J., Reynolds, T., Winegard, B. M., & Baumeister, R. F. (2018). Death before
dishonor: Incurring costs to protect moral reputation. Social Psychological and
Personality Science, 9, 604-613.
MOTIVATED FREE WILL BELIEFS
Walster, E. 1966. Assignment of responsibility for an accident. Journal of Personality and Social
Psychology, 3, 73-79.
Wegner, D.M. (2002). The illusion of conscious will. Cambridge, MA: MIT Press.
Wegner, D. M. (2003). The mind’s best trick: How we experience conscious will. Trends in
Cognitive Sciences, 7, 65-69.
Winegard, B. M., Clark, C. J., Hasty, C. R., & Baumeister, R. F. (2018). Equalitarianism: A
source of liberal bias. Unpublished manuscript.
Winegard, B. M., & Clark, C. J. (2020). Without Contraries is no Progression. Psychological
Inquiry, 31, 94-101.
Wood, W., Kressel, L., Joshi, P. D., & Louie, B. (2014). Meta-analysis of menstrual cycle effects
on women’s mate preferences. Emotion Review, 6, 229-249.
Yen, C. L., & Cheng, C. P. (2013). Researcher effects on mortality salience research: A meta-
analytic moderator analysis. Death studies, 37(7), 636-652.
... Hence, one specific goal of the present study was to test the prediction that manipulating information about the effects of punishment on reoffending would causally affect the severity of punishment recommendations. Second, Clark et al. (2014, p. 504) suggested, based on a mediation analysis, that the desire to punish causally influences belief in free will (for disputes, see Monroe & Ysidron (2021); but see also Clark et al. (2021)). The hypothesis that the desire to punish causally affects belief in free will (or moral responsibility) challenges the viability of non-skeptical reform: even if people could recommend less severe punishment when presented with information that is conceptually unrelated to skepticism, that might still causally make them less confident in free will and moral responsibility, which would amount to skepticism entering through the backdoor. ...
... 504-505) hypothesis. However, as I will further discuss in what follows, the studies previously mentioned (Clark et al., 2021;Monroe & Ysidron, 2021), which were published after the present study was run, reveal that the present sample size was insufficient to properly address this issue. Condition B was designed to check whether either punishment or treatment would be preferred when both were available and thought to have similar effects on recidivism; it was also less artificial because actual interventions to prevent criminal behavior are unlikely to be as effective as suggested in condition A. ...
... In addition to the observation that general and specific beliefs about free will tend to be stronger after exposure to immoral behavior, Clark et al. also argue that the explanation of the observation resides in the stronger desire to punish that arises when one is exposed to immoral behavior. This issue is further addressed in subsequent sections of this paper (see alsoClark et al., 2021;Monroe & Ysidron, 2021). ...
Article
Derk Pereboom and Gregg Caruso argue that humans are never morally responsible for their actions and take that thesis as a starting point for a project whose ultimate goal is the reform of responsibility practices, which include expressions of praise, blame, and the institution of legal punishment. This paper shares the skeptical concern that current responsibility practices can be suboptimal and in need of change, but argues that a non-skeptical pursuit of those changes is viable and more promising. The main lines of the argument are developed by assessing the prospects of implementing one of the changes favored within the skeptical project (namely, the reduction of punishment severity) in light of how human moral psychology works. An original vignette experiment (N = 180; participants from Facebook groups related to Brazilian universities) asked participants to recommend sentences for a fictitious criminal after considering alternatives to regular punishment that varied in their effectiveness to prevent reoffending. The results suggest that people can become less punitive even if they continue to believe in moral responsibility and free will. The paper further argues that reforming responsibility practices is more likely to occur without the endorsement of skepticism.
... Most relevant to Nietzsche's assertion, research has found that people attribute more free will to morally bad actors than to comparable morally neutral (e.g., Clark et al., 2014) and morally good actors. And people report stronger beliefs in free will and are more critical of science that threatens the existence of free will when they are experiencing desires to punish (Clark et al., 2021). For example, in one study, students in a social psychology course received an email from their professor shortly after a midterm exam informing them that one of their classmates had cheated, or in the control condition, students were simply told there 7 would be an activity in the next class. ...
... Thus no situational features, including counternormativity, varied between the actions except that people are more likely to care about harm to ingroup members than harm to outgroup members (e.g., Aboud, 2003;Tajfel, 1982) and wish to excuse the harmful behavior of ingroup members more than the harmful behavior of outgroup members (e.g., Brandt et al., 2014;Claasen & Ensley, 2016;Jordan et al., 2014). In other studies, it was found that people attributed more free will to harmful actions that were judged as less counternormative than paired neutral actions (Clark et al, 2021). And in others, a reverse pattern was detected for one's own actions such that people attributed less free will to their own harmful behavior than to their own harmless behavior (Vonasch et al., 2017). ...
... Motivated attributions of responsibility findings likely receive more attention in the literature than the effect sizes warrant because they seem somehow surprising, irrational, or normatively mistaken. For example, a meta-analysis of motivated free will attribution effects found a small to medium effect size, r =.25, and found evidence that such effects may not be particularly long-lasting (Clark et al., 2021). One can easily imagine cases in which a person might have extremely strong desires to blame or punish but find a causal agent completely free of responsibility (e.g., if a person killed one's sister in a car accident because they had a seizure while driving). ...
Chapter
Full-text available
The Blame Efficiency Hypothesis applies insights from evolutionary psychology to resolve the apparent conflict between rationalist and intuitionist perspectives on moral judgment. First, people reserve moral condemnation for actors and actions that are likely to be deterred by moral condemnation. This includes intended and controllable actions (consistent with rationalist perspectives that highlight how considerations of intentionality and causal control influence blame judgments) because blame cannot deter intentions that never existed in the first place, nor can it alter unalterable actions. And this includes actors who care about their moral reputations (i.e., cognitively competent and mentally healthy humans), because moral condemnation cannot deter those with no regard for their moral reputations. Second, moral judgment is subject to error management principles. Failing to condemn a morally responsible harmdoer signals exploitability, and so it would be more costly to erroneously not blame a harmdoer who could have been deterred by blame than to erroneously blame a harmdoer who could not have been deterred by blame (called The Don Corleone Principle). In contrast, no obvious cost asymmetry exists for overestimations or underestimations of moral responsibility for helpful or neutral actions. Therefore, in ambiguous cases, people should err on the side of assuming harmdoers are morally responsible (consistent with intuitionist perspectives which highlight how punitive desires influence judgments of intentionality and moral responsibility). Blaming efficiently means that people attribute moral responsibility to those who appear capable of controlling their behavior in future situations and to those who appear controllable—likely to change their behavior in response to harsh moral judgment, with a small bias toward ascribing responsibility in ambiguous cases of harm to minimize the costly error of under-blaming.
... Likewise, a debate around the correct explanation of why people increase ascriptions of free will to abnormally behaving agents has emerged. While some have argued that people's desire for punishment underpins their belief in the free will of a norm-deviating agent (Clark et al., 2014;Clark, Winegard, & Shariff, 2021), others have made the case that dispositional inferences about the agent's character make up this effect (Monroe & Ysidron, 2021). ...
... The basic idea that unites some of these proposals is that norm-violating actions or negative consequences usually trigger a negative evaluative response, and that the respective judgment-e.g., attributions of causality, free will or intentional action-is increased in order to justify this evaluation (Alicke, 2000). Accordingly, all of these heterogeneous norm effects are merely a by-product of a more general moral cognitive process, and may be subsumed under some kind of unified account according to which participants are "motivated" to judge norm-deviating behavior (Clark et al., 2021). ...
Preprint
Full-text available
Norm violations have been demonstrated to impact a wide range of seemingly non-normative judgments. Among other things, when agents' actions violate prescriptive norms they tend to be seen as having done those actions more freely, as having acted more intentionally, as being more of a cause of subsequent outcomes, and even as being less happy. The explanation of this effect continues to be debated, with some researchers appealing to features of actions that violate norms, and other researchers emphasising the importance of agents' mental states when acting. Here, we report the results of two large-scale experiments that replicate and extend twelve of the studies that originally demonstrated the pervasive impact of norm violations. In each case, we build on the pre-existing experimental paradigms to additionally manipulate whether the agents knew that they were violating a norm while holding fixed the action done. We find evidence for a pervasive impact of ignorance: the impact of norm violations on non-normative judgments depends largely on the agent knowing that they were violating a norm when acting. Moreover, we find evidence that the reduction in the impact of normality is underpinned by people's counterfactual reasoning: people are less likely to consider an alternative to the agent’s action if the agent is ignorant. We situate our findings in the wider debate around the role of normality in people's reasoning.
Article
Within the criminal justice system, one of the most prominent justifications for legal punishment is retributivism. The retributive justification of legal punishment maintains that wrongdoers are morally responsible for their actions and deserve to be punished in proportion to their wrongdoing. This book argues against retributivism and develops a viable alternative that is both ethically defensible and practical. Introducing six distinct reasons for rejecting retributivism, Gregg D. Caruso contends that it is unclear that agents possess the kind of free will and moral responsibility needed to justify this view of punishment. While a number of alternatives to retributivism exist - including consequentialist deterrence, educational, and communicative theories - they have ethical problems of their own. Moving beyond existing theories, Caruso presents a new non-retributive approach called the public health-quarantine model. In stark contrast to retributivism, the public health-quarantine model provides a more human, holistic, and effective approach to dealing with criminal behavior.
Article
Norm violations have been demonstrated to impact a wide range of seemingly non-normative judgments. Among other things, when agents' actions violate prescriptive norms they tend to be seen as having done those actions more freely, as having acted more intentionally, as being more of a cause of subsequent outcomes, and even as being less happy. The explanation of this effect continue to be debated, with some researchers appealing to features of actions that violate norms, and other researcher emphasizing the importance of agents' mental states when acting. Here, we report the results of two large-scale experiments that replicate and extend twelve of the studies that originally demonstrated the pervasive impact of norm violations. In each case, we build on the pre-existing experimental paradigms to additionally manipulate whether the agents knew that they were violating a norm while holding fixed the action done. We find evidence for a pervasive impact of ignorance: the impact of norm violations on non-normative judgments depends largely on the agent knowing that they were violating a norm when acting. Moreover, we find evidence that the reduction in the impact of normality is underpinned by people's counterfactual reasoning: people are less likely to consider an alternative to the agent's action if the agent is ignorant. We situate our findings in the wider debate around the role or normality in people's reasoning.
Article
Previous research suggests that free will beliefs and moral responsibility beliefs are strongly linked, yet ultimately distinct. Unfortunately, the most common measure of free will beliefs, the free will subscale (FWS) of the Free Will and Determinism Plus, seems to confound free will beliefs and moral responsibility beliefs. Thus, the present research (1,700 participants across two studies) details the development of a 2-factor FWS – the FWS-II – that divides the FWS into a free will subscale and a moral responsibility subscale. The FWS-II showed good fit compared to standard fit thresholds and superior fit compared to the original FWS. The moral responsibility subscale was moderately correlated with general punitive attitudes and specific punitive assignments, even when controlling for the free will subscale. Conversely, the free will subscale was moderately correlated with conservativism and religiosity, even when controlling for the moral responsibility subscale. These results provide evidence that the FWS is better suited – psychometrically, theoretically, and practically – as a 2-factor measure of free will beliefs and moral responsibility beliefs than as a 1-factor measure of free will beliefs.
Article
Good self-control is a crucial factor in the distribution of life outcomes, ranging from success at school and work, to good mental and physical health, and to satisfying romantic relationships. While in the last decades psychologists have learned much about this all-important trait, both social theory and politics have not caught up. Many academics and policymakers still seem to believe that everybody has unlimited capacity for self-control and that maintaining discipline is purely a matter of volition. This book shows that such beliefs are fundamentally mistaken. It presents the state-of-the-art in research on self-control, explains why this trait has been largely overlooked, and sets out the profound implications of this psychological research for moral responsibility, distributive justice and public policy. It shows that the growing emphasis in politics on 'personal responsibility' is deeply problematic, and outlines alternatives more in accord with human psychology.
Article
We are more likely to judge agents as morally culpable after we learn they acted freely rather than under duress or coercion. Interestingly, the reverse is also true: Individuals are more likely to be judged to have acted freely after we learn that they committed a moral violation. Researchers have argued that morality affects judgments of force by making the alternative actions the agent could have done instead appear comparatively normal, which then increases the perceived availability of relevant alternative actions. Across five studies, we test the novel predictions of this account. We find that the degree to which participants view possible alternative actions as normal strongly predicts their perceptions that an agent acted freely. This pattern holds both for perceptions of the prescriptive normality of the alternatives (whether the actions are good) and descriptive normality of the alternatives (whether the actions are unusual). We also find that manipulating the prudential value of alternative actions or the degree to which alternatives adhere to social norms, has a similar effect to manipulating whether the actions or their alternatives violate moral norms. This pattern persists even when what is actually done is held constant, and these effects are explained by changes in the perceived normality of the alternatives. Together, these results suggest that across contexts, participants' force judgments depend not on the morality of the actual action taken, but on the normality of possible alternatives. More broadly, our results build on prior work that suggests a unifying role of normality and counterfactuals across many areas of high-level human cognition.