Content uploaded by Ivar R. Hannikainen
All content in this area was uploaded by Ivar R. Hannikainen on Apr 13, 2020
Content may be subject to copyright.
Judgment and Decision Making, Vol. xx, No. x, Month 20xx, pp. XX–XX
An experimental guide to vehicles in the park
Noel Struchiner∗Ivar R. Hannikainen†Guilherme da F. C. F. de Almeida‡
Prescriptive rules guide human behavior across various domains of community life, including law, morality, and etiquette.
What, speciﬁcally, are rules in the eyes of their subjects, i.e., those who are expected to abide by them? Over the last sixty
years, theorists in the philosophy of law have oﬀered a useful framework with which to consider this question. Some, following
H. L. A. Hart, argue that a rule’s text at least sometimes suﬃces to determine whether the rule itself covers a case. Others,
in the spirit of Lon Fuller, believe that there is no way to understand a rule without invoking its purpose — the benevolent
ends which it is meant to advance. In this paper we ask whether people associate rules with their textual formulation or their
underlying purpose. We ﬁnd that both text and purpose guide people’s reasoning about the scope of a rule. Overall, a rule’s
text more strongly contributed to rule infraction decisions than did its purpose. The balance of these considerations, however,
varied across experimental conditions: In conditions favoring a spontaneous judgment, rule interpretation was aﬀected by
moral purposes, whereas analytic conditions resulted in a greater adherence to textual interpretations. In sum, our ﬁndings
suggest that the philosophical debate between textualism and purposivism partly reﬂects two broader approaches to normative
reasoning that vary within and across individuals.
Keywords: experimental jurisprudence, the concept of law, rules, legal psychology, Hart, Fuller.
In 1958, H. L. A. Hart and Lon Fuller held a heated exchange
about the nature of rules in the Harvard Law Review. Hart
maintained that a rule’s text was enough to solve most legal
cases, while Fuller insisted that it was never suﬃcient to de-
termine its scope. According to Fuller, the purposes behind
rules — their morally laudable ends — were always at play
when judging whether they had been obeyed or violated.
Their debate was brought to life through competing lines of
reasoning about a rule forbidding vehicles in a public park.
Hart ﬁrst proposed the example in an attempt to show that,
no matter how indeterminate legal language might be in cer-
tain cases, rules retain a core of undisputed meaning and this
meaning is capable of settling legal disputes: We may debate
whether “bicycles, roller skates, [and] toy automobiles” are
We would like to thank Alice M. de Almeida and Fiery Cushman for
their invaluable assistance at diﬀerent stages of this research. We are also
indebted to all other members of the NERDS research group and to Fábio
Shecaira and Joshua Knobe for comments on earlier drafts. Finally, we
would like to thank both the National Council for Scientiﬁc and Technolog-
ical Development (CNPq) and of the Carlos Chagas Filho Research Support
Foundation (FAPERJ) for funding this research.
Copyright: © 2020. The authors license this article under the terms of
the Creative Commons Attribution 3.0 License.
∗Corresponding author. PUC-Rio. Email: firstname.lastname@example.org. ORCID
†PUC-Rio & Universidad de Granada. Email: email@example.com. ORCID
‡FGV Direito Rio & PUC-Rio. Email: firstname.lastname@example.org.
vehicles according to the park’s rule, but regular cars are
clearly prohibited. We do not need to exercise any moral
reasoning in order to apply the no-vehicles rule to cars.
Fuller took issue with Hart’s understanding of the no-
vehicles rule. In Fuller’s view, even the most paradigmatic
violations of a rule demand a consideration of the purpose
behind the rule’s existence. Fishing for both his readers’ and
Hart’s intuitions, Fuller asked:
What would Professor Hart say if some local pa-
triots wanted to mount on a pedestal in the park a
truck used in World War II, while other citizens,
regarding the proposed memorial as an eyesore,
support their stand by the “no vehicle” rule? Does
this truck, in perfect working order, fall within the
core or the penumbra (Fuller, 1958, p. 663)?
For Fuller himself, this was an easy case: The intuitively
obvious response is that the memorial truck is allowed in
the park. To see that regular cars are prohibited and the
memorial truck is allowed, we need to think beyond the
rule’s text, resorting to its purpose. According to Fuller, a
rule’s purpose is the moral goal it pursues. For example, it
would be reasonable to infer that the purpose of a prohibition
on vehicles in the park is to keep park-goers safe.1Since a
regular car driving through the park would pose a risk to
visitors’ safety, it falls within the scope of the rule. The
1Fuller is not aiming towards the intentions of the rule-maker. For him,
the task of ﬁnding a rule’s purpose is necessarily normative. See Fuller
(1969, pp. 228–229).
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 2
memorial truck poses no comparable risk, and hence falls
Although the debate may seem inconsequential, the “no
vehicles in the park” rule has become a centerpiece of legal
theory2— so pervasive in fact that, on April 1st, 2012,
Lawrence Solum posted the following on his Legal Theory
Frederick Schauer (University of Virginia School
of Law) has posted No More Vehicles in the Park
on SSRN. Here is the abstract:
In prior work, I have examined the memorable
controversy about the ﬁctional legal rule prohibit-
ing vehicles in the park, which ﬁrst appeared in
the 1958 debate between Lon Fuller and H.L.A.
Hart. That essay focused on the original version
of the thought experiment as presented by Hart. In
this essay, I examine a series of classic variations
found in the work of other theorists, including “am-
bulance in the park,” “tricycle in the park,” “mo-
torized wheelchair in the park,” “radio-controlled
toy car in the park,” “tank memorial in the park,”
and “silent hovercraft in the park.” Drawing on
Daniel Dennett’s critique of thought experiments
as intuition pumps, this essay shows that many (if
not all) of these variations are simply incapable of
generating valuable insights about legal rules, le-
gal interpretation, and the nature of legal language.
I conclude by suggesting that the proliferation of
vehicle-in-the-park thought experiments be termi-
This, of course, was an April Fools’ joke inspired by
Schauer’s (2008) own take on the hackneyed thought exper-
iment.4But the joke is telling. For decades, legal philoso-
phers have obsessed over the numerous variants of this
thought experiment that helped to immortalize the dispute
between Hart and Fuller. Why did this particular thought
experiment become so fundamental to legal philosophy?
One of the reasons is its implications for a deeper ju-
risprudential disagreement about the nature of law. Legal
positivists, including Hart, hold that law and morality are
2A brief and certainly non-exhaustive list of works that discuss the rule
includes Schlag, 1999; Bix, 1991; Soames, 2012; Schauer, 2008; Fish,
2006; Hurd, 2015; Marmor, 2005, chapter 7; Shecaira, 2015; Slocum,
2015, chapter 5; Tobia, forthcoming. Attesting to the lasting relevance of
the exchange about the no-vehicles-in-the-park rule, Peter Cane edited a col-
lection of essays in 2010 titled “The Hart-Fuller Debate in the Twenty-First
Century”, stating in the preface that the essays “demonstrate that this debate
between two of the twentieth century’s greatest legal theorists continues to
present a rich, and by no means exhausted, seam of jurisprudential ideas
waiting to be mined in the years to come” (Cane, 2010, p. vi).
4Schauer may have thought he was putting the debate to rest with his
extremely thorough piece entitled “A critical guide to vehicles in the park”
(2008). He was wrong about this.
conceptually distinct and that one can ascertain what the
law is without resorting to moral criteria (Gardner, 2001).
In Schauer’s words, “if law is to be understood as not nec-
essarily incorporating moral criteria for legal validity, then
there must exist some possible rules in some possible le-
gal systems that can be identiﬁed as legal without resort to
moral criteria” (Schauer, 2008, p. 1113).5Since the textual
interpretation of a rule, according to its ordinary meaning,
does not in principle demand moral scrutiny, this so-called
textualist view of rules is compatible with positivism.6
Advocates of natural law such as Fuller, on the other hand,
take law and morality to be conceptually intertwined, so that
the identiﬁcation of law is necessarily a matter of moral eval-
uation. According to Schauer, “in oﬀering [his] example,
Fuller meant to insist that it was never possible to determine
whether a rule applied without understanding the purpose
that the rule was supposed to serve” (2008, p. 1111). If
we always — in all conceptually possible legal systems —
need to engage in purpose-driven moral reasoning in order
to determine what the law requires, then there is a neces-
sary conceptual connection between law and morality. So,
Fuller’s purposivism is a natural law position. In this way,
the Hart-Fuller debate, though focused on legal interpre-
tation speciﬁcally, bears upon a more fundamental debate
concerning the nature of law.
Aside from these divergent conceptual implications, the
Hart-Fuller debate also turns centrally on a particular empir-
Fuller is arguing not only that his purpose-focused
approach is a necessary feature of law properly
so called, but also that it is an accurate descrip-
tion of what most judges and other legal agents
would actually do in most common law jurisdic-
tions. On this point Hart might well be read as
being agnostic, but there is still a tone in Hart
of believing that Fuller not only overestimates the
role of purpose in understanding the concept of
law, but may well also be overestimating the role
of purpose and underestimating the role of plain
language in explaining the behavior of lawyers and
judges (Schauer, 2008, p. 1130).
5Note that Schauer, in line with the conceptual nature of Hartian posi-
tivism, speaks only of “possible” rules and legal systems. Although we take
the conceptual argument to be correct, we are not only interested in mak-
ing purely hypothetical points about law. Hart himself stated several times
that his interests were in general and descriptive jurisprudence. We share
Hart’s aims. Throughout the paper, we are therefore more concerned with
a descriptive version of legal positivism, that claims not only that people
can possibly hold non-moral criteria for the identiﬁcation of legal rules, but
that they actually do so.
6Note that other interpretative methods might also be compatible with
this insight. Some authors defend intentionalism — the method of looking
for the meaning actually intended by the rule-maker — along positivist
lines, such as Alexander & Sherwin (2008).
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 3
The way lawyers and judges actually think and use the con-
cept of a rule is, according to this characterization of the
debate, decisive. If legal professionals generally decide that
legal rules cover certain cases without resorting to purposive
or moral considerations, positivism may better capture legal
cognition. If, instead, the ordinary concept of a rule gener-
ally begets a moral appraisal, natural law theory may oﬀer a
more accurate account.
But what are the reasons one could have to prefer textual-
ism or purposivism? To understand this normative question,
it pays to examine one interesting feature of rules. Legal
rules typically regulate action-types (e.g., driving under the
inﬂuence), but these action-types acquire value insofar as
they probabilistically inﬂuence one or more of the author-
ity’s objectives (e.g., protecting people’s lives). As a result,
action proscriptions in the law are in general both imper-
fectly sensitive (since some drivers may ingest alcohol yet
pose no special risk to road safety) and imperfectly speciﬁc
(since other drivers pose a risk to road safety even without
ingesting alcohol). These imperfections lead us to ask why
we should proscribe types of actions, if these proscriptions
are mere heuristics and the ultimate goal of law is to promote
desirable outcomes. One possible answer is that it is easier,
and less subject to controversy, to apply a literal rule than to
consider every consequence and assess its probability. An-
other answer is that simple rules sometimes lead to better
consequences (on average) when the attempt to decide on
the basis of consequences is subject to bias or large errors
(Gigerenzer & Todd, 1999; Gigerenzer & Engel, 2006). The
same kind of reasoning applies to rules outside of the legal
domain, such as moral norms.
Those two advantages are highlighted by the literature on
normative models of decision-making in law. Schauer (1991,
1988) points out that formalism — the commitment to apply
the rule’s text even when this seems to be wrong vis-à-vis an
all things considered reasoning — yields faster and cheaper
decisions than the alternative of trying to ﬁnd the optimal
moral solution to every legal case (a position he calls partic-
ularism). He also points out that formalist judges will often
be better on aggregate than their particularistic counterparts,
because particularists will err more often, given the com-
plexity of moral decision-making and the relative simplicity
of following rules’ authoritative linguistic formulations.
Furthermore, when interpreted textually, rules that estab-
lish linguistically determinate action-types tend to better pro-
mote predictability and coordination among their subjects.
Legislators in heterogeneous societies where values are dis-
puted frequently do not agree with regards to outcomes but
they will eventually vote on an agreed text. A norm presented
in clear wording which requires certain actions and excludes
others will at least allow subjects with diﬀerent moral out-
looks to know what is expected of them, thus allowing them
to plan their lives accordingly. Even among groups com-
posed exclusively of well intentioned people who hope to do
what is good or right, clear-cut proscriptions of actions are
necessary for coordination and predictability, if these people
diverge regarding the right and the good. Given the indeli-
ble predicament of profound moral disagreement in plural
societies, some legal philosophers have advocated, contra
Fuller, that law be taken as a set of determinate rules to be
understood according to their text:
Thus, if people were gods — morally omniscient,
but not angels, morality would be an adequate
guide to behavior and posited norms would be
unnecessary. If, however, people were angels but
not gods, then posited norms in the form of de-
terminate rules would be necessary to implement
morality. Law as determinate rules is a solution to
a cognitive, not a motivational, problem (Alexan-
der & Sherwin, 2001, 219).
The mock abstract Solum attributed to Schauer culminated
with a suggestion that “the proliferation of vehicle-in-the-
park thought experiments be terminated”. We would not go
that far, but we deﬁnitely sympathize with the urgent need
for a change of focus. Hart and Fuller can be viewed as
advancing competing empirical predictions about the way
legal agents understand, interpret, and apply rules. As such,
we may be able to arbitrate between their contrasting views
through empirical research. In order to shed light on legal
agents’ concept of a rule, thought experiments won’t do. We
need to conduct actual experiments.
Furthermore, there are important reasons to examine naive
subjects’ intuitions as well. After all, the law is directed at
all of us, and citizens’ intuitions about rules are important in
order to understand how they conceive the rules’ demands.
As one prominent legal philosopher of law recently put it:
In one way or another the law plays a role in the
practical reasoning of everyone in society, and in
reasonably well functioning societies, law works
as an internal guide to (nearly) everyone in society,
and not just to appellate judges. It is to say that a
general jurisprudential theory would be radically
incomplete and seriously misleading, if it failed
to give some account of the place of law in the
practical reasoning of oﬃcials, lawyers, and lay
citizens alike (Postema, 1991, pp. 799–800).
Following in the footsteps of recent work on experimen-
tal jurisprudence (Donelson & Hannikainen, 2020; Kneer
& Machery, 2019; Macleod, 2019; Sommers, forthcom-
ing; Tobia, 2018, forthcoming), we employ the methods
of experimental philosophy to reveal how people ordinarily
understand and apply the concept of a rule. This ﬁlls an im-
portant gap, since current experimental investigations into
the tension between text and purpose (Turri & Blouw, 2015;
Turri, 2019; Garcia, Chen & Gordon, 2014) miss some of
the most important conceptual features of rules in law and in
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 4
life (Schauer, 1991). We take the ordinary concept to be the
concept that non-philosophers, understood as both the folk
and legal professionals, employ in their use of language.
We tested both legal and non-legal rules because the con-
cern of the Hart-Fuller debate — the tension between text and
purpose in the making of a rule — obviously extrapolates the
ﬁeld of legal theory. We also used non-legal rules because it
seems to be a common practice among legal philosophers to
assume that discussing non-legal prescriptive rules can help
illuminate legal rules themselves. In this sense, sport rules
(Hart, 1994), household rules (Twining & Miers, 2010), par-
enting rules (Schauer, 1991) and many others can be found
in discussions within jurisprudence. As one book clearly
All of us are confronted with rules every day of
our lives. Most of us make, interpret and apply
them, as well as rely on, submit to, avoid, evade
and grouse about them; parents, umpires, teach-
ers, members of committees, business- people,
accountants, trade unionists, administrators, logi-
cians and moralists are among those who through
experience may develop some proﬁciency in han-
dling rules. Lawyers and law students are spe-
cialists in rule-handling, but they do not have a
monopoly of the art. A central theme of this book
is that most of the basic skills of rule-handling
are of very wide application and are not con-
ﬁned to law. There are certain speciﬁc techniques
which have traditionally been viewed as ‘legal’,
such as using a law library and handling cases and
statutes. But these share the same foundations as
rule-handling in general: they are only special in
the sense that there are some additional considera-
tions which apply to them and are either not found
at all or are given less emphasis in other contexts
(Twining & Miers, 2010, xiv).
We investigated people’s intuitions about a series of putative
rule infractions through correlational (Study 1) and exper-
imental (Studies 2, 3 and 4) methods. Our ﬁndings reveal
that people spontaneously consider both a rule’s text and
its purpose when determining whether a particular incident
constitutes an infraction. Yet experimental manipulations
of textual compliance yielded stronger eﬀects than did ma-
nipulations of purposive violation — a pattern mirrored in
participants’ subjective assessments. Finally, the weight of
moral considerations upon judgments of rule infraction var-
ied across experimental conditions. In spontaneous condi-
tions, judgments of rule infraction depended more on the
blameworthiness of the agent — but analytic conditions an-
nulled this eﬀect.
2 Study 1
272 volunteers (mean age = 24.6, 169 women, 227 reported
no legal training) were recruited through snowball sampling
on social media and completed the survey.
In a correlational design, we asked participants to consider
a rule and report whether an agent committed an infraction,
and whether they had violated the text and/or the purpose of
Participants ﬁrst read an adaptation of Schauer’s (1991) “no
dogs in the restaurant” rule. The introduction described a
previous incident involving a customer’s dog jumping, run-
ning and barking in the restaurant. This incident led the
owner to “ban dogs from the restaurant” (the rule’s text) in
order to “avoid behaviors that cause nuisances to the restau-
rant’s customers” (the rule’s purpose). Thus, both the rule’s
text and its purpose (the normative goals the creator of the
rule aimed to achieve) were explicitly stated in the introduc-
tion to the study.
Respondents then viewed a random subset of four sce-
narios from a battery of eight (see7), describing a putative
infraction of the no-dogs rule. The scenarios introduced vari-
ation along four related dimensions: being a dog, looking
like a dog, behaving like a dog, and annoying other cus-
tomers. In some scenarios, the client brought a misbehaved
dog (“. . . a dog that runs, jumps, barks and eats food from
the ﬂoor”), in others, a well-behaved dog (“. . . a purse con-
taining what seems to be a teddy bear. Actually, it’s her dog,
who doesn’t bark and barely moves, being easily mistaken
for a toy”). Other scenarios included something that looks
like a dog, but doesn’t behave like one (“. . . a taxidermied
dog”), and something that acts like a dog, but doesn’t look
like one (“. . . a dog in an extremely realistic pig costume”).
After each scenario, participants were asked “Did he/she
break the rule?”. Responses were recorded as either (1)
“Yes” or (0) “No”. Alongside their rule infraction decisions,
participants were asked to assess four features of the case at
hand — namely, whether the animal/object:
(i) “was a dog” (identity),
(ii) “looked like a dog” (appearance),
(iii) “behaved like a dog” (behavior), and
(iv) “bothered other customers” (purpose).
Each assessment was recorded on a seven-point scale,
ranging from 1: “Clearly not” to (7) “Clearly yes”. Of
7Available at: Supplementary Materials
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 5
the ﬁrst three assessments, the ﬁrst (i.e., identity) reﬂects
the most direct interpretation of the rule’s text. Thus, we
predicted that identity judgments would reveal a stronger
association with rule infraction decisions than either appear-
ance or behavior judgments. The fourth assessment asked
directly about the rule’s stated purpose.
The introduction to the rule stipulated that the rule’s text
is “no dogs allowed in the restaurant” (text) in order not
to “bother other customers” (purpose). By asking whether
there is a dog in the restaurant, and whether it bothered other
customers, we sought to capture commensurate estimates of
the eﬀect of textual and purposive considerations in infrac-
We entered rule infraction judgments as the dependent mea-
sure in a mixed-eﬀects, logistic model with participant and
scenario as crossed random eﬀects. We evaluate the ﬁxed
eﬀect of each assessment in a series of simple linear regres-
sions, allowing random slopes by participant and scenario.
Both identity and purpose judgments revealed robust eﬀects
(identity: OR = 1.57, z = 7.03; purpose: OR = 1.30, z = 3.70;
ps < .001). In turn, judgments of appearance and behavior
revealed weaker associations with infraction decisions (ap-
pearance: OR = 0.91, z = −1.21, p = .23; behavior: OR =
1.14, z = 2.1, p = .034).
Querying our causal assumptions helps to further inter-
pret this pattern of results. While ‘being a dog’ is a cause of
‘looking like a dog’ and ‘behaving like a dog’, the opposite
is not true. Thus, a ‘backdoor path’ connects behavior to
infraction decisions (i.e., Behavior ←Identity →Infraction;
see Pearl, 2009) — confounding the bivariate analysis above.
In a multiple regression analysis, participants’ infraction de-
cisions were predicted by assessments of behavior (OR =
1.34, if anything higher than without the covariate, z = 2.67,
p = .007), after accounting for the eﬀect of assessments of
identity (OR = 1.67, z = 5.10, p < .001).
A similar line of reasoning may suggest that the eﬀect
of purpose too is confounded (i.e., Purpose ←Behav-
ior/Identity →Infraction): Speciﬁcally, one could think that
‘being a dog’ and ‘behaving like a dog’ both cause ‘bother-
ing other customers’ and, ex hypothesi, rule infractions. And
yet, in a multiple regression analysis, participants’ purpose
assessments predicted their infraction decisions (OR = 1.51,
z = 3.53, p < .001), even when conditioning on both con-
founds (identity: OR = 2.00, again higher than when alone,
z = 5.27, p < .001; behavior: OR = 1.42, z = 2.39, p = .017).
Finally, the above analyses do not distinguish subject vari-
ation from case variation. Looking at the point-biserial cor-
relation between purpose assessments and infraction deci-
sions revealed positive coeﬃcients (rs) ranging from .02 to
.21 (.80 < ps < .015) for all but one scenario8— suggesting
that the eﬀect of purpose assessment is not due solely to vari-
ation across cases (see Figure 1). Still, ﬁrst and foremost,
the Hart-Fuller debate concerns our intuitions regarding dif-
ferent cases, i.e., the intuitions we feel when considering
the car entering the park versus the memorial truck, and not
diﬀerences between individuals when perceiving the same
case. Thus, the primary concern in Study 1 was to examine
the features of cases that are seen as involving a rule infrac-
tion (Figure 1). For instance, an overwhelming majority of
participants judged the misbehaved dogs (97% without and
96% with a pig costume) to be in violation of the rule, while
only a slim minority (7%) judged a goldﬁsh to violate the
no-dogs rule (see Table 1). Finally, cases where textual and
purposive assessments supported opposing verdicts yielded
substantial division: For instance, the guide dog was judged
to be in violation of the rule by 48% of participants.
‘Being a dog’ (identity) and ‘bothering customers’ (purpose)
were the strongest predictors of participants’ decisions about
whether the no-dogs rule had been violated. Meanwhile,
‘acting like a dog’ and especially ‘looking like a dog’ yielded
notably weaker eﬀects. Thus, in the context of a non-legal
rule, lay decisions regarding a series of putative infractions
appeared to reﬂect the very concerns that dominate the ju-
risprudential debate between Hart and Fuller: the rule’s text
and its purpose.
The results of Study 1 also provide tentative support for a
textualist approach to interpretation. Participants were more
likely to report that the rule was broken by well-behaved dogs
than by bothersome non-dogs9— despite having stipulated
both the text and the purpose of the rule. This pattern of
results echoes a broad theme in moral psychology: People
exhibit a strong adherence to simple action proscriptions,
adopting this policy even in contexts in which an alterna-
tive action plan would yield the superior outcome (Baron &
Spranca, 1997; Bartels, 2008).
However, the conclusions of Study 1 rest on a single rule.
In addition, Study 1 relied on participants’ subjective as-
sessment of whether the rule’s text and purpose had been
8The exception was the dog in pig costume scenario, where the correla-
tion was -.07, p= .41
9We did not control for the seriousness of the textual and purposive
violations presented. So, it might be the case that the observed eﬀect
is due to the presence of more serious textual than purposive violations
in the stimuli. Although we did not introduce a manipulation check to
verify whether that was the case, we feel that the diﬀerences in seriousness
between textual and purposive violations in our stimuli were not great.
Another objection that might be raised about comparing the size of textual
and purposive violations is that those might be incommensurable. In other
words, it might be the case that there is no way to translate the degree of
textual violation to the same scale used to measure the degree of purpose
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 6
Table 1: Displays summary statistics for each of the putative infractions in Study 1.
Scenario Rule Identity Appearance Behavior Purpose
Prop. (%) M SD M SD M SD M SD
Dog in pig costume 98 5.80 1.50 2.68 1.48 5.65 1.20 4.72 1.31
Misbehaved dog 96 6.31 1.00 6.30 0.93 6.30 0.71 4.94 1.25
Quiet dog 90 6.10 1.49 4.67 1.99 3.45 1.73 2.58 1.15
Guide dog 48 6.45 0.92 6.55 0.76 4.82 1.94 2.52 1.14
Taxidermied dog 44 4.99 2.10 5.97 1.28 1.50 0.81 3.09 1.51
Robot dog 27 2.34 1.59 5.74 1.17 5.87 1.12 4.62 1.40
Cat 23 1.71 1.31 1.99 1.28 2.62 1.42 3.84 1.03
Goldﬁsh 7 1.40 0.76 1.35 0.64 1.42 0.74 2.18 1.19
Not at all
Was it a dog?
Not at all
Did it bother other costumers?
Figure 1: Eﬀects of identity and purpose assessments on infraction decisions. Case means are overlaid.
violated. We address these limitations in Study 2 by survey-
ing a broader set of rules, while experimentally manipulating
whether the text and purpose were violated.
3 Study 2
One of the deﬁning features of rule-based decision-making
is the possibility that text and purpose will diverge. Imagine
that John, in hopes of keeping his apartment clean, estab-
lishes a rule according to which no one is allowed to enter
his home with shoes on. To make his decision clear, he
hangs a sign on the front door saying “no shoes allowed in
Suppose that a friend walks barefoot in the mud outside
John’s apartment. She would not, according to the rule’s text,
be barred from entering John’s apartment — though doing
so would most certainly dirty the ﬂoor. Schauer (1991) calls
cases like this, where a rule’s text fails to cover actions that
violate the rule’s purpose, cases of underinclusion.
Now, imagine that a second friend bought a brand new pair
of shoes. According to the rule’s text, this friend would not
be allowed to try her new shoes on inside John’s apartment
— even though doing so would not dirty the apartment in
any way. Schauer calls cases like this, where a rule’s text
proscribes behavior that does not violate the rule’s purpose,
Thus, in two diﬀerent ways, appraisals of the text of a
rule give rise to verdicts that fall short of the rule’s guiding
spirit. In addition to Schauer’s categories, we can also deﬁne
core cases as those in which the rule is violated on both
grounds of text and purpose (for instance, dirty boots), and
oﬀ-topic cases where neither text nor purpose prohibit the
action (someone enters John’s apartment barefoot and with
In Study 2, we devised scenarios of all four types: core,
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 7
overinclusion, underinclusion, and oﬀ-topic cases. What
would Hart and Fuller predict about each? They would agree
that core cases are rule violations while oﬀ-topic cases are
not. Their theories, however, make competing predictions
about overinclusion and underinclusion cases.
Hart’s textualism would predict that people are more will-
ing to take overinclusion cases to be rule violations than
underinclusion cases. What’s more, by agreeing that cases
of overinclusion constitute rule violations nonetheless, peo-
ple would be expressing that, at least sometimes, rules cover
circumstances in which they yield morally undesirable out-
comes. By the same token, if people denied that underinclu-
sion cases violate the rule at hand, they would be implying
that purposive considerations do not suﬃce to determine
whether a rule has been violated. This pattern of results
would be congenial to a positivist like Hart.10
Fuller would disagree. He held that law makes sense only
as a purposive enterprise. According to this view, people are
much more concerned with advancing the goals that inspired
a rule than with abiding by their speciﬁc textual formulation.
As such, the Fullerian prediction would be the exact opposite
of the Hartian prediction: namely, that underinclusion cases
should be viewed as rule violations more often than are
200 volunteers (mean age = 28.6, 116 women, 101 reported
no legal training) were recruited through sponsored posts
and snowball sampling using social media and completed
In a 2 (text: abide vs. violate) × 2 (purpose: abide vs.
violate) Latin square design, participants read about four
scenarios: No Shoes in the House, No Vehicles in the Park,
No Sleeping at the Station11, and No Cellphones in Class
(see Supplementary Materials).
After each scenario, participants were asked whether the
agent violated the rule, and responses were recorded as ei-
10This does not commit the positivist to the idea that the rule should be
followed, i.e., that the person who broke the rule should be punished. One
could note that an overinclusion case is an instance of rule violation and still
maintain that the oﬀending party should not be punished for moral reasons.
This feature is often referred to as the normative inertia of legal positivism
11The cases under this rule were inspired by Fuller’s (1958, p. 664)
discussion of text and purpose under the “No Sleeping in the Train Station”
rule. We have used his original proposal as a canvas, introducing changes
in some scenarios to emphasize the over or underinclusiveness of each
ther (1) “Yes” or (0) “No”. Alongside their rule infraction
decisions, participants were asked to make two assessments
about the case at hand — namely, whether:
(i) the rule’s text was violated (e.g., “Did Jane wear shoes
in the house?”), and
(ii) the rule’s underlying purpose was violated (e.g., “Did
Jane dirty the house?”).
Both assessments were made on seven-point scales, rang-
ing from 1: “Clearly not” to (7) “Clearly yes”.
Manipulation checks revealed that violations of the rule’s
text were rated higher on the textual assessments, B = 3.64, t
= 24.11, p < .001. Correspondingly, violations of the rule’s
purpose were rated higher on the purposive assessment, B =
1.55, t = 10.48, p < .001.
Looking at a wider variety of legal and non-legal rules, we
replicated the eﬀect in Study 1. Experimental manipulations
of both text (B = 3.76, z = 11.69, p < .001) and purpose (B =
2.04, z = 7.56, p < .001) violations increased the probability
of perceived infraction. Critically, in a pairwise comparison
(OR = 5.60, z = 6.73, p <.001), cases of overinclusion (ˆ
.62) were more likely to be considered rule violations than
were cases of underinclusion (ˆ
p = .22), as depicted in Figure
2.12 The same pattern of results holds when we restrict the
analysis to those with legal training.13
Both textual and purposive violations were treated as rule
infractions. However, extending the results of Study 1, over-
inclusion cases (violations of the text that do not infringe
upon the rule’s purpose) were more likely to be seen as in-
fractions than were underinclusion cases (violations of the
purpose that are not captured by the rule’s text).14 Thus,
once again, a rule’s text appeared to play a predominant role
in infraction decisions.
Rules may be thought of as proscribing actions that hinder
the rule’s purpose in most conditions. The results of Study
2 add nuance to this picture: Infraction decisions were de-
termined primarily by the action (i.e., bringing a vehicle to
12But see the limitations mentioned on footnote nº 9.
13The 98 respondents who reported some legal training (38 law school
graduates and 60 law students) were more likely to report that the rule was
broken when text (B = 3.54, z = 7.97, p < .001) and purpose (B = 2.13, z
= 5.53, p < .001) were violated. In a pairwise comparison (OR = 4.10, z
= 4.04, p < .001) cases of overinclusion) (ˆ
p = .58) were more likely to be
considered rule violations than cases of underinclusion ( ˆ
p = .25).
14Ironically, the only scenario where this pattern of results did not obtain
was the no-vehiclesin the park scenar io. Most respondents probably thought
that our toy car was a toy and not a car at all and therefore not a case of
overinclusion. This interpretation is consistent with the manipulation check:
in aggregate, participants did not consider that the rule’s text was broken in
this scenario (M = 2.53, t-test against the midpoint (4), t(54) = −4.59, p <
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 8
Was the rule violated?
Figure 2: Predicted probability of a rule violation by scenario
the park, wearing shoes in the house), even when the result-
ing outcome was known (e.g., whether the park-goers were
endangered, or the apartment was dirtied). This result sug-
gests that abiding by the rule’s text is encoded as a source
of intrinsic, and not merely instrumental, value (Blair, 1995;
4 Study 3
Studies 1 and 2 focused on cases that we might describe as
clear-cut: Each potential infraction unambiguously violated
the rule’s text and/or its purpose (or else, the action clearly
violated neither). However, the cases that stoke jurispruden-
tial interest often involve some degree of uncertainty: i.e.,
either it is unclear whether the putative infractor’s action is
described by the rule’s text, or it may be unclear whether the
action violates the rule’s purpose.
For Study 3, we developed a new battery of 24 borderline
cases. For instance:
One day, in a high proﬁle case, a 21 year old
young woman got into a deadly traﬃc accident.
The accident happened because the young woman
was using her smartphone in one hand to text her
friends while driving. Congress, recognizing the
graveness of the situation and with the goal of
avoiding this type of accident, passed a law with
the following textual formulation: “it is forbidden
to send text messages while driving”.
This time, however, the particular action did not clearly
violate the text of the rule:
Felipe uses the voice-to-text functionality of his
smartphone to text his friends. While doing so,
Felipe suﬀers a serious accident with another ve-
hicle, severely injuring the occupants of both cars.
There is some ambiguity in whether Felipe complied with the
rule’s text. On the one hand, he was sending text messages
by using his phone (which processed his voice into text).
On the other hand, neither his hands nor his eyesight were
diverted from driving.
Analogously, other cases were unclear as to whether or not
the protagonist’s behavior violated the rule’s stated purpose:
e.g., a case where someone who was harassed for being
agnostic in a Catholic country uses an ecumenical chapel
created to protect religious minorities to read the biography
of an atheist mathematician.
Study 3 introduced a further change to the way partici-
pants’ decisions were elicited. In our ﬁrst two studies, we
asked participants to simultaneously consider the rule’s text
and its purpose — while they decided whether the rule had
been violated. This feature of our protocol could have primed
participants to consider these factors in their infraction deci-
sions to a greater extent than they spontaneously would have.
To reveal the balance of textual and purposive considerations
in spontaneous circumstances, in Study 3 participants made
rule infraction judgments in isolation. Participants in other
conditions judged whether the text was violated, or whether
the violator was morally blameworthy.
Finally, instead of asking whether the purpose had been
violated ("Did Felipe’s behavior cause an accident?"), we
asked whether putative infractors were morally blamewor-
thy for the outcomes they brought about ("Is Felipe morally
blameworthy for using the voice-to-text functionality of his
smartphone to communicate with his friends while he was
driving?"). Insofar as the purposes in our scenarios were uni-
vocally good, ascriptions of blame to the infractor should be
closely linked to assessments of whether the rule’s purpose
had been violated.
175 volunteers (mean age = 27.3, 89 women, 40 reported
no legal training) were recruited through sponsored posts
and snowball sampling on social media and completed the
In a 3 (Question: text, moral, rule) × 1 between-subjects
design, each participant viewed a random subset of four
cases drawn from a total set of 24 cases.
Each case described an incident giving rise to a rule (e.g.,
the accident involving a young woman texting), followed by
a subsequent target incident (e.g., Felipe’s texting). For each
case, participants were asked to make a single judgment,
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 9
which varied by condition: in the Rule condition, after each
case, participants read a statement that the agent violated
the rule (e.g., “Felipe broke the law passed by Congress”).
In turn, participants in the Text condition read a statement
that the agent violated the text of the rule (e.g., “Felipe
sent a text message while driving”). Finally, participants
in the Moral condition read a statement that the agent’s be-
havior was morally blameworthy (e.g., “Felipe should be
morally chastised for using the voice-to-text functionality of
his smartphone to communicate with his friends while he
was driving”). In each condition, we asked participants to
report whether they agreed or disagreed with the statement
on a seven-point scale, ranging from 1: “Strongly disagree”
to 7: “Strongly agree”.
To understand whether participants in the Rule condition
were spontaneously incorporating assessments of the law’s
text and/or the behavior’s immorality, we averaged textual,
moral and rule judgments by scenario (see Table 2). In a
by-scenario analysis, rule judgments correlated with both
textual (r= .55, 95% CI [.19, .78], p = .005) and moral
blameworthiness (r = .61, 95% CI [.27, .81], p = .002) judg-
ments — as displayed in Figure 3. Meanwhile, textual and
blameworthiness judgments were themselves uncorrelated,
r= .28, 95% CI [−.14, .61], p = .19. This pattern of results
held true when looking at the partial correlation coeﬃcients
(text: partial r = .50, 95% CI [.11, .76], p = .015; moral:
partial r = .57, 95% CI [.20, .79], p = .005). Again, results
were robust as to legal training.15
In Study 3, textual compliance and moral blameworthiness
each predicted judgments about rule infraction in a separate
group of participants. Unlike Studies 1 and 2, in which tex-
tual compliance was more important than abidance with the
purpose, in Study 3, the eﬀects of compliance and blame-
worthiness were comparable in magnitude.
These results coalesce well with Fuller’s position. Inas-
much as this study was concerned with borderline cases,
Hart could be just as willing to acknowledge the prominence
of purposes as Fuller. After all, his claim was not that the
“no vehicles in the park” rule’s text was always clear (as we
have seen, he conceded that it was not, at least as applied
15132 out of the 175 participants reported legal training. 64 were law
school graduates, while 68 were law students. When we analyze only
responses of this subset, both results remain signiﬁcant, with signiﬁcant
correlations between rule judgments and text (r = .54, 95% CI [.18, .78], p =
0.006) and moral (r= .68, 95% CI [.39, .85], p < .001) assessments, and no
correlation between text and moral judgments (r= .31, 95% CI [−.01, .63],
p = .14). The partial cor relations are also still signiﬁcant for text (partial r
= .48, 95% CI [.08, .74], p = .02) and purpose (partial r = .65, 95% CI [.32,
.84], p < .001).
to “bicycles, roller skates, [and] toy automobiles”), but only
that it clearly covered certain cases (such as regular cars).
Therefore, if this characteristic of Study 3’s design was the
reason for the increased inﬂuence of moral evaluation, Hart
can still claim to be fully vindicated by the data. If, how-
ever, the second change introduced in Study 3 — asking only
one question per participant — proves decisive, Fuller might
claim that moral purposes, under some conditions, are cer-
tainly more empirically relevant than Studies 1 and 2 made
them out to be.
5 Study 4
Study 3 revealed stronger eﬀects of moral-purposive con-
cerns than were observed in Studies 1 and 2. Two method-
ological shifts could account for these divergent results:
First, in Study 3 we asked participants to consider borderline
cases rather than clear cases; second, participants in Studies
1 and 2 were asked to simultaneously determine whether the
text and the purpose of the rule had been violated.
We speculated that the supplementary questions regard-
ing moral blameworthiness and textual compliance could
have modulated participants’ spontaneous concept of a rule
violation. Previous work investigated the eﬀects that ques-
tions regarding part of a concept exert over answers to a
more general question regarding the same concept. Specif-
ically, when asked to rate their marital or dating life before
assessing their overall life quality, participants answered dif-
ferently depending on the conversational setting introduced
by researchers (Schwarz, Strack & Mai, 1991). If the ques-
tions were introduced outside of a joint evaluation context
(i.e., if both questions were presented sequentially through
no unifying context), participants showed an assimilation
eﬀect whereby there was a higher correlation between both
answers than in a control condition. If, however, both ques-
tions were introduced by a shared prompt that made it salient
that they pertained to one single conversational context, par-
ticipants showed a contrast eﬀect, whereby the correlation
between questions was (non signiﬁcantly) lower than in the
control condition. The authors theorized that this is due to
the Gricean maxim of non-redundancy: since participants
have already expressed that they have a good (bad) roman-
tic life, the remaining general question must be probing for
something else that is unrelated to the ﬁrst question.16
Study 4 asked whether the diﬀerences in the eﬀect size of
purpose stem from contrast eﬀects when textual and moral
16Assimilation could also explain the results of Studies 1 and 2: in
both studies, participants were asked about text and purpose (parts) before
assessing whether or not the rule was broken (whole). Thus, the detected
eﬀects of text and purpose could be less a feature of the concept of rule
and more a feature of how people answer questions in surveys. The results
of Study 3 ruled out these strong concerns about assimilation — since
participants appeared to spontaneously incorporate the same considerations
(i.e., textual compliance and moral blameworthiness) in separate evaluation.
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 10
Table 2: Rule violation judgments (y-axis) by textual (left) and moral (right) assessments. Each observation represents a
Rule infraction Text violation Moral blame
Scenario M SD M SD M SD
Id required 3.58 2.11 2.00 0.76 3.22 2.39
60 km/h speed limit 7.00 0.00 6.92 0.29 6.50 0.58
No bags in the library 1.33 0.50 1.45 0.93 1.27 0.47
No objects at the fair 1.82 1.83 5.79 1.89 1.00 0.00
Pet friendly nightclub 3.00 2.27 4.67 2.27 4.00 2.10
No bags in the library (v2) 6.6 0.55 6.00 1.96 6.42 0.51
No writing on library books 3.71 2.81 4.75 1.91 2.42 1.73
No smoking in the restaurant 3.88 2.42 4.27 2.43 5.25 1.04
No sports in the playground 4.00 2.00 3.80 2.20 1.69 1.38
Both hands on the wheel 2.60 2.12 3.29 1.89 2.38 2.00
No children at the spa 3.20 2.49 1.50 0.53 4.56 2.13
No kids in the nightclub 5.33 2.00 6.33 1.32 4.5 1.95
No texting while driving 3.11 2.15 4.22 2.33 5.38 2.13
No drinking and driving 3.90 2.47 2.67 1.88 6.71 0.49
No phones in the bank 5.33 1.97 4.80 2.20 3.00 2.31
Religious minorities 3.90 2.51 2.30 1.64 2.6 1.78
No littering 4.67 2.88 2.90 2.28 2.11 1.17
Zero tolerance policy 2.83 2.14 2.36 1.91 2.00 1.41
Insider trading 6.0 1.32 5.50 2.27 3.71 2.06
Insulin gun 1.3 0.48 2.00 1.79 1.40 0.52
Driving license 5.2 2.04 3.00 1.91 4.64 2.01
Max earnings for aid eligibility 6.00 0.82 4.31 2.18 3.44 1.90
No eating in class 3.87 2.42 6.22 1.09 1.80 1.55
No hats in the bank 1.88 2.10 3.90 2.23 1.70 0.95
assessments are made simultaneously. Prompting partici-
pants to take into account the degree of textual compliance
and purpose violation may have aﬀected the balance of these
considerations in participants’ infraction decisions — rel-
ative to spontaneous judgments made in isolation. If the
rule concept is composed of both text and purpose, con-
trast eﬀects are predicted by the maxim of non-redundancy:
inquiring about one of the sub-elements in the same conver-
sational context as the general rule-breaking question should
make participants more likely to answer the latter based on
the second, latent sub-element. To test the hypothesis of a
contrast eﬀect, in Study 4 we manipulated whether partici-
pants were asked to decide rule judgments either in isolation
or while simultaneously considering its sub-elements, i.e.,
textual compliance and/or moral blameworthiness.
364 people (mean age = 30, 184 women, 125 reported no
legal training) were recruited through sponsored posts and
snowball sampling on social media and completed the survey.
In a 2 (Case-type: overinclusion vs. underinclusion) × 2
(Semantic-prompt: present vs. absent) × 2 (Moral-prompt:
present vs. absent) between-subjects design, participants
viewed a battery of four clear-cut cases adapted from Studies
1 and 2.
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 11
r = 0.55 [0.19, 0.78], p = 0.005
...violated the rule's text?
...violated the rule?
...is morally blameworthy?
Figure 3: Rule violation judgments (y-axis) by textual (left) and moral (right) assessments. Each observation represents a
As in previous studies, participants learned about a rule, its
literal wording and its underlying purpose, and were asked
to determine whether a case of either under or overinclu-
sion was in violation of that prior rule. We also manipulated
whether participants were asked to make textual and/or moral
judgments simultaneously. When the semantic prompt was
present, participants were asked to determine whether the
act violated the text of the rule. When the moral prompt
was present, participants were asked to determine whether
the agent’s behavior was morally blameworthy. Semantic
and moral prompts were orthogonal factors, such that par-
ticipants could see either, both or neither (Table 3).
In every condition, participants read a statement that the
agent violated the rule (e.g., “Tim broke the No Shoes
rule”)17, and reported whether they agreed or disagreed
with the statement on a seven-point scale, ranging from 1:
“Strongly disagree” to 7: “Strongly agree”. Participants
in the Semantic Prompt Only condition read a statement
that the agent violated the text of the rule (e.g., “Tim wore
shoes in the house”) immediately above or below the rule-
violation statement in a counterbalanced order across par-
ticipants. Similarly, participants in the Moral Prompt Only
condition read a statement that the agent violated the rule’s
moral purpose (e.g., “Tim is morally blameworthy for walk-
ing in the house with dirty feet”). Lastly, participants in the
Both Prompts condition saw both the semantic and moral
17The “no shoes” rule was not one of the scenarios presented in Study 4.
We have kept the same example used to explain over and underinclusion in
the text for ease of exposure. We used a mix of legal and non-legal cases
that can be found in the Supplementary Materials.
prompts, together with the rule-violation statement, in an
order randomized across participants.
If moral-purposive concerns play a greater role in the
spontaneous concept of a rule, participants should treat un-
derinclusion as prohibited in the No Prompts condition —
but treat overinclusion as prohibited in the Both Prompts
condition. The inclusion of the Semantic Prompt Only and
Moral Prompt Only conditions enables us to infer whether
either prompt drives the hypothesized shift away from a spon-
taneous (i.e., moralized) concept of rules.
In a 3-way ANOVA, we entered Case-Type (overinclusion,
vs. underinclusion), Semantic Prompt (present, vs. absent),
and Moral Prompt (present, vs. absent) in the ﬁxed eﬀects
portion of the model (see Table 4). To account for the non-
independence of observations, we include crossed random
eﬀects of participant and scenario.
We replicated the eﬀect of case-type documented in Study
2, F(1, 262) = 7.52, p = .007. The eﬀect of case-type was
qualiﬁed by a two-way interaction between case-type and
moral prompt, F(1, 262) = 9.99, p = .002. Meanwhile, no
corresponding interaction emerged between case-type and
textual prompt, F(1, 262) = 1.07, p = .30. The marginal
means by condition are displayed by Figure 4.
Inspection of the signiﬁcant two-way interaction helps to
interpret the overall pattern of results: In the absence of the
moral prompt (i.e., No Prompts & Semantic Prompt Only
conditions), participants did not distinguish overinclusion
(M = 3.46, 95% CI [3.08, 3.94]) from underinclusion (M =
3.54, 95% CI [3.15, 3.93]) cases, B = −0.08, t = −0.29, p =
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 12
Table 3: Experimental conditions in Study 4.
(No prompt) (Semantic prompt only)
Absent Tim broke the No Shoes rule. Tim wore shoes in the house.
Tim broke the No Shoes rule.
(Moral prompt) (Both prompts)
Present Tim is morally blameworthy for. . . Tim wore shoes in the house.
Tim broke the No Shoes rule. Tim is morally blameworthy for. . .
Tim broke the No Shoes rule.
Table 4: Type-2 ANOVA Table with Satterthwaite approximation for degrees of freedom.
Semantic Prompt 2.36 1 262 0.45 0.5
Moral Prompt 39.17 1 262 7.43 0.007
Case Type 39.62 1 262 7.52 0.007
Semantic Prompt*Moral Prompt 0.87 1 262 0.17 0.68
Semantic Prompt*Case Type 5.62 1 262 1.07 0.3
Moral Prompt*Case Type 52.65 1 262 9.99 0.002
Semantic Prompt*Moral Prompt*Case Type 8.32 1 262 1.58 0.21
.99. Meanwhile, in its presence (i.e., Both Prompts & Moral
Prompt Only), participants treated cases of overinclusion
(M = 4.58, 95% CI [4.19, 4.97]) as rule violations more
often than cases of underinclusion (M = 3.45, 95% CI [3.10,
3.81]), B = 1.13, t = 4.24, p < .001 — as in Studies 1 and 2.
Again, the overall pattern of results remains the same when
controlling for legal training.18
Thus, the textualist distinction between case types ap-
peared to arise from a contrast eﬀect — i.e., requesting
that people simultaneously assess the blameworthiness of
the infraction at hand. The inclusion of the moral prompt
appeared to lead participants to interpret the infraction deci-
sion as distinct from the question of moral blameworthiness
— resulting in the tendency to view cases of overinclusion,
but not underinclusion, as violations of the rule at hand.
Finally, we garner further evidence for this interpreta-
tion by analyzing participants’ subjective assessments. We
look at the association between textual compliance and rule
infraction judgments, and ask whether this association is
moderated by the presence of the moral prompt. Specif-
18235 out of the 364 respondent’s to Study 4 were legally trained. 87
graduated law school, while 87 were law students. The main ANOVA
taking into account only lawyers reveals the same overall patterns, although
only the main eﬀect of case type remained signiﬁcant, F(1, 162) = 4.8, p =
0.03. The interaction between moral prompt and case type, although in the
predicted direction, drops below signiﬁcance F(1, 162) = 1.92, p = 0.17.
ically, in a two-way ANCOVA, controlling for Case Type,
we enter Moral Prompt,Textual Assessment, and the Moral
Prompt×Textual Assessment interaction. (Since only partic-
ipants in the Semantic Prompt Only and Both Prompts condi-
tions made assessments of textual compliance, this analysis
draws on data from two of four groups.)
If a contrast eﬀect is present, adding the moral prompt
ought to strengthen the relationship between textual assess-
ments and rule infraction decisions. Indeed, we observed
a two-way interaction between Moral Prompt and Textual
Assessment, F(1, 438) = 10.93, p = .001. Simple slopes
analyses revealed that the eﬀect of textual assessments on
infraction decisions was larger when accompanied by the
moral prompt (B = 0.70, 95% CI [0.59, 0.81], vs. unaccom-
panied: B = 0.45, 95% CI [0.33, 0.58]; t = −3.27, p = .001),
as shown in Figure 5.
For completeness’ sake, we mirror the above analysis with
moral blameworthiness as the continuous moderator: Did
the presence of the semantic prompt weaken the eﬀect of
perceived moral blameworthiness on infraction decisions?
This ANCOVA revealed no two-way interaction between
Text Prompt and Moral Assessment, F(1, 321) = 0.88, p
= .35. The eﬀect of moral blameworthiness on infraction
decisions was comparable whether accompanied by the se-
mantic prompt or not (B = 0.41, 95% CI [0.26, 0.57] vs.
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 13
Moral Prompt Only Both Prompts
No Prompts Semantic Prompt Only
1 4 7 1 4 7
Was the rule violated?
Figure 4: Rule violation judgments (marginal mean, and
95% conﬁdence interval) in cases of under- and over-
inclusion, for each of the four experimental conditions. We
observe a distinction between cases of under- and over-
inclusion when the moral prompt is present (bottom panel),
but not absent (top panel).
unaccompanied: B = 0.50, 95% CI [0.35, 0.64]; t = −0.93,
p = .35; see Figure 5).
Study 4 yielded evidence of a contrast eﬀect (Schwarz, Strack
& Mai, 1991). Speciﬁcally, making a separate yet simultane-
ous moral judgment of the agent’s conduct strengthened the
association between textual compliance and infraction judg-
ments. In other words, in the absence of a moral prompt,
participants spontaneously assigned greater weight to the
moral-purposive dimension of rules (and as a result treated
cases of underinclusion as violations to a comparable extent).
Meanwhile, the presence of the moral prompt led participants
to distinguish morally blameworthy agents from rule viola-
tors. In sum, when people interpret rules spontaneously, they
tend to ’moralize’ the concept of rule — an eﬀect that can
be weakened by demanding that participants simultaneously
reason about the morality of the putative infractor’s con-
duct, which leads to a stronger textualist approach to rules.
This pattern of results suggests that the discrepancy between
Studies 1 and 2 on one hand, and Study 3 on the other, is
driven to some extent by a part-whole contrast eﬀect of the
prompt about the agent’s moral blameworthiness.
One question that might arise is that the part-whole hy-
pothesis should predict contrast eﬀects not only with the
inclusion of the moral probe, but also (and in the opposite
direction) with the inclusion of the semantic probe. We did
not detect such an eﬀect. A possible explanation is that, al-
though composed of both textual and moral elements (a dual
concept account of rules), the primacy of the textual compo-
nent in the folk concept of rule is such that it overrides the
maxim of non-redundancy. Instead, one hypothesis is that
we should expect assimilation to be the outcome dictated by
conversational pragmatics in such circumstances.19
An alternative explanation for the eﬀects is to do away
completely with the dual concept account and maintain that
the folk concept of rule is only textual, but that moral con-
cerns often interfere with our ability to apply that concept
correctly. Under this alternative account, the reason why the
introduction of the moral prompt increases the correlation
between textual and rule-violation judgments is that par-
ticipants have been given an opportunity to vent their moral
views in a way independent of their rule-violation judgments,
which leaves them free to correctly apply the latter concept
(see Turri & Blouw, 2015; Turri, 2019). This explanation is
consistent with our results and further research should test
6 General Discussion
Social life is characterized by the ubiquitous presence of
prescriptive rules of all kinds — e.g., legal, etiquette, moral,
political, and so on. From the simple “no shoes in the
house” rule to the highly institutional governmental rules
prohibiting insider trading, what all of them have in com-
mon is that they exist in the hope of exerting pressure in the
world, by guiding judgments and channeling behaviors. As
long as we are dealing with non-ultimate linguistic prescrip-
tions — i.e., those rules that aim at achieving background
purposes and are communicated through language — there
remains the question characteristic of the Hart-Fuller debate:
Is a rule its textual formulation understood according to its
ordinary meaning, or is a rule ﬁrst and foremost the back-
ground moral purposes it aims at achieving? From their
armchairs, philosophers have advanced many diﬀerent com-
peting theories about the nature of rules. However, what
do non-philosophers — those who are the target of rules in
their daily lives and operate with and under rules — actually
believe them to be?
19Imagine that your meal is composed of a large bowl of soup and three
small pieces of bread. Your meal is composed of two parts: the soup and
the bread, albeit the soup is a much stronger component than the bread. If
someone asks you about the bread and then about the meal, conversational
pragmatics seems to dictate (via the maxim of non-redundancy) a contrast
eﬀect. However, if someone asks you about the soup and then about the
meal, they seem to dictate an assimilation eﬀect. Imagine the following con-
versations: (a) “How was the soup?”, “Great!”, “And the meal?”, “Oh, the
meal was only OK”; (b) “How was the soup?”; “Great!”, “And the meal?”,
“It was awesome!”. Conversation (a) seems to cry out for explanation, while
conversation (b) sounds natural.
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 14
Semantic Prompt Only Both Prompts
1 4 7 1 4 7
Moral Prompt Only Both Prompts
1 4 7 1 4 7
Figure 5: Rule violation judgments in overinclusion (red) and underinclusion (blue) cases, by textual (top) and moral (bottom)
assessments. Left-side panels display responses from the single prompt conditions (i.e., semantic prompt only and moral
prompt only), while right-side panels display responses from the Both Prompts condition. The introduction of a moral prompt
strengthened the association between textual assessments and rule violation judgments. In contrast, the introduction of a
semantic prompt did not moderate the association between moral assessments and rule violation judgments.
In this paper we embrace the experimental turn in order to
answer the age-old question about the concept of rule, trying
to capture its ordinary meaning and seeing which philosoph-
ical theory would be vindicated by the results. We ﬁnd that,
for both legal experts and non-experts, both text and purpose
are relevant components of rules (legal or otherwise). Even
though both components are relevant, our evidence points to
a predominance of text over purpose in people’s understand-
ing of rules.
The main reason for the primacy of text has been noted
by psychologists, rule consequentialists in normative ethics,
as well as by some legal scholars working on the nature of
rules. As discussed in the introduction, rules that are applied
according to their text (assuming that their text is reasonably
determinate) are more prone to achieving certain objectives:
predictability, certainty, coordination, decision-making eﬃ-
ciency, and avoidance of moral errors due to complex moral
reasoning. Some research suggests that many times we are
better at pursuing outcomes indirectly, through clearly ex-
pressed rules that are followed according to their text, than
trying to get at these goals directly (Gigerenzer & Todd,
1999; Gigerenzer & Engel, 2006; Schauer, 1991). This
could explain why people have taken the stance of acknowl-
edging text as a major component of the concept of rule
through and through our set of experiments.
However, text is not the only relevant feature of rules. To
a lesser, but signiﬁcant extent, participants also feel that pur-
poses matter when making judgments about rule violations.
There are two ways to conceptualize this ﬁnding. Previous
work (Turri & Blouw, 2015; Turri, 2019) dealt with a similar
phenomena by assuming that the concept of rule is entirely
determined by the rule’s text, but that our ability to correctly
apply this concept might be aﬀected by our moral commit-
ments. So, when someone blamelessly violates a rule’s text,
we feel the temptation to excuse them and respond that there
was no rule violation not because they did not violate the rule,
but because we are conﬂating two diﬀerent speech acts: one
purely descriptive that states the fact that they violated the
rule and another one normative that states that they should
be punished. This account is compatible with our results.
It might be the case that the folk concept of rule is entirely
textual, but that our capacity to correctly apply this concept
is mitigated by our normative commitment not to blame peo-
ple who did nothing wrong (and, on the other hand, to blame
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 15
people who did something wrong despite complying with
A diﬀerent way to make sense of our results is to hypoth-
esize that the concept of rule is composed of at least two
elements: one textual and descriptive, the other purposive
and normative. Under this account, it is not the case that
we have a capacity to apply a descriptive concept that is bi-
ased by the confounding demands of morality. In eﬀect, the
concept of rule itself — like many others (see Knobe, 2010)
— would host these diﬀerent components. Both interpreta-
tions ﬁt the data well. Future research should design novel
experiments to tease apart these diﬀerent theories about the
concept of rule.
6.1 Implications for legal theory
If the strong view attributed to Fuller by Schauer accord-
ing to which it is “never possible to determine whether a
rule applied without understanding the purpose that the rule
was supposed to serve” (2008, p. 1111) hinges on people’s
actual beliefs and attitudes, then it is clearly at odds with
the present evidence. Oftentimes people are willing to say
that a rule was violated even by behaviors that do not hinder
the rule’s explicit purpose at all. There do seem to exist
“(. . . ) some possible rules in some possible legal systems
that can be identiﬁed as legal without resort to moral cri-
teria” (Schauer, 2008, p. 1113), a point in favor of Hart’s
Hart’s views are also only partially vindicated. Even
though people drift towards a more textualist view on
joint-evaluation, they remain susceptible to moral concerns
through and through. The fact that moral and purposive
concerns are just as important in determining people’s spon-
taneous grasp of rules might pose a challenge for Hartian
positivism to the extent that it is concerned with accurately
describing the attitudes that laypeople, who presumably en-
gage with law intuitively, have towards legal rules. If this is
the case, positivist theories may not always be as accurate as
natural law in explaining the way people ordinarily conceive,
or reason about, the law.
Rather, our data suggest that the concept of law reﬂects
both a preferential concern for the legal text as claimed by
positivists, and a default sensibility towards the purpose of
law, as argued by Fuller. In many circumstances, people’s
understanding of whether a rule is broken is spontaneously
informed by both concerns. Yet, we also found evidence
that, upon concerted consideration, people were more likely
to distinguish morality and law, resulting in a positivist un-
derstanding of the concept of rule.
Our results trace the psychological fault lines of the Hart-
Fuller debate. It is possible that the philosophical styles
of Hart and Fuller could explain their diﬀerent views about
the concept of rule and the perennial character of their de-
bate. Hart was the analytic philosopher par excellence and
the methods of analytic philosophy emphasize the decom-
position of concepts. Therefore, one could speculate that
Hart and other analytic philosophers, who succeeded him
and held similar views, are prone to decompose multifaceted
concepts into their constitutive sub-elements, thus privileg-
ing text. On the other hand, as noted by Schauer, “Fuller’s
philosophical forays were far clumsier” (2008, 1132). This
more holistic, and less analytic, approach to legal reasoning
may have contributed to Fuller’s predilection for purposes.
If contextual factors play this critical role in rule infrac-
tion judgments, maybe we should move away from general
jurisprudence (which makes claims about essential features
of law in all possible legal systems and worlds) to particular
jurisprudence (which makes claims about localized legal sys-
tems). If people in diﬀerent cultures and contexts conceive
rules in diﬀerent ways, carving out the interplay between text
and morality in distinct manners, then their very concept of
law may be diﬀerent. Big general claims about the em-
pirical, even if not conceptual20, descriptive correctness of
positivism or natural law should be wary of the experimental
ﬁndings of particular jurisprudence.
Finally, law is a socially constructed concept (Hart, 1994;
Schauer, 2005; Searle, 1995). As such, people may be able
to promote their preferred concept by shifting the conditions
under which others engage with rules. Aware of our results,
those who are optimistic about a morally infused concept of
law may wish to encourage the spontaneous point of view,
while those who are not stand to beneﬁt from fostering a
more analytic perspective on rules.
6.2 Limitations and future work
The experimental research we conducted has several lim-
itations that should be taken into account in discussing its
implications for general jurisprudence. The ﬁrst of such lim-
itations is the extent to which our results generalize across
languages, cultures and jurisdictions. After all, we con-
ducted our studies in Portuguese, and a vast majority of our
subjects were Brazilian. Perhaps, our particular pattern of
results reﬂects a quirk of Brazilian legal culture, or of certain
kinds of legal systems (i.e., civil law systems), and only fu-
ture cross-cultural studies will tell if they are representative
of legal reasoning more broadly.
Another limitation is that, inspired by Schauer, we set out
to experimentally examine rules not only in law, but also in
life. As a result, several of our examples dealt with non-
legal rules, such as the no-dogs, and the no-shoes rules.
Positivists may very well say that this reduces the import of
our results to their theory. After all, exercising one speciﬁc
and particularly stringent form of authority is the distinctive
20But as long as conceptual claims hinge on people’sintuitions or people’s
actual linguistic behavior (as suggested by Hart, 1994), then conceptual
positions also depend on actual experimental ﬁndings concerning these
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 16
way law works. Other normative systems need not be so
stringent and can live with looser, more moralized concepts
of rule. In fact, post-hoc analysis of Study 3 data, classifying
our rules as legal or non-legal showed that people might in
fact think diﬀerently about legal rules.21 In any event, future
research should test in a controlled setting whether or not
judgments about legal and non-legal rules diﬀer.
Another valid objection deals with the relationship be-
tween a rule’s purpose and overall morality. In the ﬁrst two
experiments, we asked whether each case impinged on the
purposes of the relevant rule (i.e., was the animal/object a
nuisance to the clients of the restaurant?). In Studies 3 and
4, in contrast, we asked whether or not rule-breaking agents
were morally blameworthy. These two questions are not nec-
essarily the same. Imagine a group of lawyers who think that
a rule’s purpose has precedence over its text, but that overall
morality should have no inﬂuence over the law. They think,
for instance, that the truck-turned-monument proposed by
Fuller should be allowed in the park because the park’s rule’s
purpose is to avoid accidents and a stationary truck poses no
such risk. If they are committed to this speciﬁc underlying
purpose, but not to overall morality, they should object to
the passage of an ambulance through the park: even though
letting the ambulance through might be the right decision all
things considered, it certainly increases the risk of an acci-
dent inside the park. By design, we surveyed cases where
(to our eyes) both speciﬁc purposive reasoning and general
moral reasoning recommended the same result — and sought
to collect data on both tasks. On the other hand, we are left
with a diﬀuse understanding of how natural law considera-
tions play into processes of legal cognition. In future work,
we hope to distinguish whether the ordinary concept of rule
spontaneously incorporates either (1) concerns with doing
the best thing all-things-considered (De Freitas et al., 2017),
or (2) a preoccupation with the speciﬁc underlying purposes
(2a) ascribed to or (2b) intended by the rule (Rose & Nichols,
2019), or else (3) some special moral domain unique to legal
systems in general (the Fullerian view) or in particular (see
Finally, diﬀerent agents might engage with legal rules
under very diﬀerent circumstances. Think of the rules pro-
hibiting dogs in restaurants. The restaurant’s maître d’ has
to decide whether a man carrying a cat will be allowed in the
restaurant even before he gets a table. If the cat later makes a
21In the case of legal rules, the correlation between text and rule violation
judgments was very high (r = .75, 95% CI [.27, .93], p < .05), while the
correlation between purpose and rule violation judgments was much lower
(r = .43, 95% CI [−.23,.82], p = .19). In contrast, judgments about non-legal
rule violation correlated more strongly with purpose violation (r = .74, 95%
CI [.32, .92], p < .005) than with text violation judgments (r = .51, 95%
CI [−.06, .83], p = .08). The small number of observations and the fact
that legal and non-legal cases were not matched, however, casts doubt on
whether those results are of any signiﬁcance, especially when we take into
account the results of Study 4, where only one of the three rules used was
mess, the restaurant owner or even a judge, if the case some-
how ends in court, will have to decide whether the maître d’
should have allowed the customer with the cat according to
the rule. People in diﬀerent roles may have a diﬀerent take
on what makes a rule while occupying these diﬀerent posi-
tions. This calls for a more nuanced “perspectival theory of
law” (Sinnott-Armstrong, 1999).
In line with this, ongoing work (Struchiner, Almeida &
Hannikainen, in preparation) shows that legal judgments
about rule violation are subject to the abstract/concrete para-
dox. Maybe fact-ﬁnding judges and law enforcement agents
hold the purpose to be more important insofar as they are
exposed to the case’s concrete facts. Appellate judges, for
instance, might be more inclined to adopt a textualist view of
rules, in part as a result of the abstraction with which legal
cases are brought before them.
Alternatively, judges and law enforcement agents might
tend more closely to the textual features of rules insofar as
they are accountable to higher authorities. It seems intu-
itively easier to justify strict adherence to a law’s text than
to defend one’s purposive interpretation of a law and its
relation to the concrete case at hand. If this is the case,
then we should expect that decision-makers may exercise the
freedom to moralize legal rules when making deﬁnitive or
autonomous rulings — i.e., that will not be subject to further
oversight — but stick closely to a rule’s text when issuing
provisional or highly scrutinized decisions.
After 60 years of speculation, the Hart-Fuller debate needed
to leave the armchair. In a series of four studies, we have
tried to draw the psychological fault lines of the philosophi-
cal divide. We have found that people’s judgments about rule
violation are inﬂuenced by both textual and moral-purposive
considerations. At a broad level, this dueling aspect of the
concept of rule leans Hartian, as text is often suﬃcient to de-
termine whether a rule was broken. However, this tendency
appears to be weaker when participants reasoned sponta-
neously about a series of putative infractions. In these con-
texts, participants show a greater comparative concern for
broader moral purposes beyond the text.
These ﬁndings represent an important step toward under-
standing people’s intuitions regarding rules, legal and other-
wise. Much remains to be done both conceptually and ex-
perimentally: First and foremost, subsequent research should
survey people from diverse legal and cultural backgrounds to
understand whether our results generalize. Moreover, more
data need to be collected on legal and non-legal rules to
determine whether or not legal status matters with regards
to the inﬂuence of text and purpose. We also need to tease
apart purpose and morality to understand what drives the nor-
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 17
mative component of a rule. Finally, diﬀerent roles might
privilege diﬀerent aspects of the concept of rule.
Alexander, L., & Sherwin, E. (2001). The Rule of Rules:
Morality, Rules, and the Dilemmas of Law. Durham:
Duke University Press.
Alexander, L., & Sherwin, E. (2008). Demystifying Legal
Reasoning. Cambridge: Cambridge University Press.
Baron, J., & Spranca, M. (1997). Protected values. Organi-
zational behavior and human decision processes, 70(1),
Bartels, D. M. (2008). Principled moral sentiment and the
ﬂexibility of moral judgment and decision making. Cog-
nition, 108(2), 381–417.
Bear, A., & Knobe, J. (2016). Normality: part descriptive,
part prescriptive. Cognition, 167, 25–37.
Bix, B. (1991). H. L. A. Hart and the “open texture” of
language. Law and Philosophy, 10(1), 51–72.
Blair, R. J. R. (1995). A cognitive developmental approach to
morality: Investigating the psychopath. Cognition, 57(1),
Cane, P. (Org.) (2010). The Hart-Fuller debate in the
twenty-ﬁrst century. Oxford: Hart Publishing.
Cushman, F. (2013). Action, outcome, and value: A dual-
system framework for morality. Personality and social
psychology review, 17(3), 273–292.
De Freitas, J., Tobia, K. P., Newman, G. E., & Knobe, J.
(2017). Normative judgments and individual essence.
Cognitive Science, 41, 382–402.
Donelson, R., & Hannikainen, I. (2020). Fuller and the
folk: the inner morality of law revisited. In Lombrozo, T.,
Knobe, J., & Nichols, S. Oxford Studies in Experimental
Philosophy Volume 3. Oxford: Oxford University Press,
Fish, S. (2005). There is no textualist position. San Diego
Law Review, 42, 629–650.
Fuller, L. (1958). Positivism and ﬁdelity to law: a reply to
professor Hart. Harvard Law Review, 71(4), 630–672.
Garcia, S.M., Chen, P., & Gordon, M., (2014). The letter
versus the spirit of the law: A lay perspective on culpabil-
ity. Judgment and Decision Making, 9(5), 479–490.
Gardner, J. (2001). Legal positivism: 5 1/2 myths. The
American Journal of Jurisprudence, 46(1), 199–227.
Gigerenzer, G., & Todd, P. M. (1999). Fast and frugal heuris-
tics: The adaptive toolbox. In Gigerenzer, G., Todd, P.
M., & the ABC Research Group (Eds.) Simple heuristics
that make us smart, pp. 3–34.. Oxford: Oxford University
Gigerenzer, G., & Engel, C. (2006). Law and heuristics:
An interdisciplinary venture. In Gigerenzer, G., & Engel,
C. Heuristics and the law. Cambridge: The MIT Press,
Hart, H. L. A. (1994). The Concept of Law. 2nd ed. Oxford:
Hart, H. L. A. (1983). Essays in Jurisprudence and Philos-
ophy. Oxford: Claredon Press.
Hurd, H. (2015). Interpretation without intentions. In
Pavlakos, G.; Rodriguez-Blanco, V. (Eds.) Reasons and
Intentions in Law and Practical Agency. Cambridge:
Cambridge University Press, 52–71.
Kneer, M., & Machery, E. (2019). No luck for moral luck.
Cognition, 182, 331–348.
Knobe, J. (2010). Person as scientist, person as moralist.
Behavioral and Brain Sciences, 33(4), 315–329.
Macleod, J. (2019). Ordinary causation: a study in exper-
imental statutory interpretation. Indiana Law Journal,
Marmor, A. (2005). Interpretation and Legal Theory. Ox-
ford: Hart Publishing.
Pearl, J. (2009). Causality: models, reasoning and inference.
Cambridge: Cambridge University Press.
Postema, G. J. (1991). Positisim, I presume?... Comments
on Schauer’s ‘Rules and the Rule of Law’. Harvard Jour-
nal of Law & Public Policy, 14(3), 797–823.
Rose, D., & Nichols, S. (2019). Teleological essentialism.
Cognitive Science, 43(4), e12725.
Schauer, F. (1991). Playing By The Rules: A Philosophical
Examination of Rule-Based Decision-Making in Law and
in Life. Oxford: Oxford University Press.
Schauer, F. (2008). A critical guide to vehicles in the park.
New York University Law Review, 83(4), 1109–1134.
Schauer, F. (2005). The social construction of the concept
of law: A reply to Julie Dickson. Oxford Journal of Legal
Studies, 25(3), 493–501.
Schlag, P. (1999). No vehicles in the park. Seattle University
Law Review, 23(2), 381–390.
Searle, J. R. (1995). The construction of social reality. New
York: The Free Press.
Schwarz, N., Strack, F., & Mai, H. P. (1991). Assimilation
and contrast eﬀects in part-whole question sequences: A
conversational logic analysis. Public Opinion Quarterly,
Shecaira, F. P. (2015). Sources of law are not legal norms.
Ratio Juris, 28(1), 15–30.
Sinnott-Armstrong, W. (1999). A perspectival theory of law.
Australasian Journal of Legal Philosophy, 24, 27–55.
Slocum, B. G. (2015). Ordinary meaning: a theory of
the most fundamental principle of legal interpretation.
Chicago: The University of Chicago Press.
Struchiner, N., Almeida, G. F. C. F. & Hannikainen, I.
(In preparation). Legal decision-making and the ab-
Soames, S. (2012). Vagueness and the Law. In: Marmor, A.
(Ed.) The Routledge Companion to Philosophy of Law.
Judgment and Decision Making, Vol. xx, No. x, Month 20xx An experimental guide to vehicles in the park 18
New York: Routledge, 95–108.
Sommers, R. (forthcoming). Commonsense consent. 129
Yale Law Journal. Available at: https://papers.ssrn.com/
Sunstein, C. R. (2005). Moral heuristics. Behavioral and
brain sciences, 28(4), 531–541.
Tobia, K. P. (2018). How people judge what is reasonable.
Alabama Law Review, 70(2), 293–359.
Tobia, K. P. (forthcoming). Testing ordinary meaning: an
experimental assessment of what dictionary deﬁnitions
and linguistic usage data tell legal interpreters. 133 Har-
vard Law Review. Available at: https://papers.ssrn.com/
Turri, J., & Blouw, P. (2015). Excuse validation: a study in
rule-breaking. Philosophical Studies, 172(3), 615–634.
Turri, J. (2019). Excuse validation: a cross-cultural study.
Cognitive Science, 43(8), 1–14.
Twinning, W., & Miers, D. (2010). How to Do Things
with Rules: a primer of interpretation. Fifth edition.
Cambridge: Cambridge University Press.
Waluchow, W. J. (2007). A common law theory of judicial re-
view: the living tree. Cambridge: Cambridge University