Two Concepts of Constitutive Rules
Jaap Hage
University of Maastricht
Abstract
In this article, it is argued that rules have two main functions, the practicedefining function and the constraining (fact-to-fact) function. These two functions
are compatible. In their function as constraints, some rules are also indirectly regulative. In both of their functions, rules differ from the summaries (rules of
thumb) that Rawls discussed and opposed to the constitutive (fact-to-fact) rules
which make that some decisions are the right ones.
In his work, first on the philosophy of language and later on social ontology,
Searle focused on one kind of constitutive rules: counts-as rules, which are constitutive in the sense that they attach new facts to the existence of “old” ones. In doing so, Searle created the scientific interest in constitutive rules which they deserve. However, because of his narrow focus on counts-as rules, Searle also created the impression that counts-as rules are all there is to constitutive rules. This
impression is wrong, if only because it overlooks dynamic rules.
Keywords: constitutive rules, counts-as rules, duties, fact-to-fact rules, obligations,
practice-defining rules, regulative rules.
1. Introduction
Through the ground-breaking work of Searle, the notion of a constitutive rule
received the attention it deserves. However, this very notion was used by Searle
in two different senses. The one sense is that of a rule which partly defines a social practice by being part of it. I will call such constitutive rules “practicedefining rules”. Constitutive rules in the other sense are rules which attach new
facts to old ones. I will call such constitutive rules “fact-to-fact rules”.1 Perhaps
because Searle did not distinguish the two notions clearly, he overlooked that
counts-as rules, on which he focuses, are only one kind of fact-to-fact rules.
In this paper I will elucidate the distinction between the two kinds of constitutive rules, elaborate on fact-to-fact rules, and distinguish in that connection
between static rules and dynamic rules. Moreover, I will also argue that there
are two kinds of regulative rules—duty-imposing rules and rules that lead to obIn other work (Hage 2015b, 2016 and 2018) I use the term “fact-to-fact rules” for a subcategory of what is called “fact-to-fact rules” in this article.
1
ligations—and that both kinds are also constitutive and only in an indirect sense
regulative.
2. Rules that Define Social Practices
In Speech Acts, Searle distinguishes between regulative and constitutive rules by
claiming that regulative rules regulate antecedently or independently existing
forms of behaviour, while constitutive rules do not merely regulate but also define or create new forms of behaviour. An example of regulative rules would be
the rules of etiquette, while the rules of football or chess would be examples of
constitutive rules (Searle 1969: 33). Later, in The Construction of Social Reality,
Searle repeats this characterization of constitutive rules, but the emphasis has
shifted to the discussion of counts-as rules as a kind of constitutive rules (Searle
1985: 27-29 and 43-51).
Although Searle does not mention Rawls in Speech Acts, the idea of constitutive rules as rules that regulate a practice which itself depends on the rules that
regulate it, seems to be inspired by Rawls’ paper Two concepts of rules (Rawls
1955). In that paper Rawls distinguishes between the summary and the practice
concept of rules.2 Rawls’ analysis of what he called the practice concept of rules,
is very similar to Searle’s notion of constitutive rules. Since Rawls’ analysis is
more detailed than what Searle wrote, I will try to elucidate this concept of constitutive rules through a brief discussion of Rawls’ paper.
2.1 Two Levels of Justification
Rawls’ central concern in his paper is the distinction between justification of a
practice and justification of a particular action falling under a practice. More in
particular Rawls tries to defend the utilitarian theory of punishment against the
criticism that it would allow punishment of innocent persons. 3 The defence goes
as follows. Utilitarianism contains a justification of the practice of punishing
people who committed crimes, and this justification is consequentialist. The
practice contains rules which specify under which circumstances a criminal can
be punished. The justification of the punishment of a concrete person should be
based on these rules. Let us assume that the practice of punishment is justified
on utilitarian grounds and that it does not contain a rule that permits the punishment of innocent persons. In that case such punishment would not be justified, even though in this particular case punishment might be justified on utilitarian grounds. Attempts to justify punishment by pointing out the desirable
consequences in a concrete case would be misguided, since (rule-)utilitarianism
is not meant as a theory to justify the outcomes of concrete cases, but only as a
theory for the justification of rules and the social practices to which they belong.
This connection to Rawls’ paper was also emphasized by Żełaniec (2013: 27-28). At the
same place Żełaniec also mentions several other precursors who contributed to the theory
of constitutive rules, whether or not under that name.
3
I ignore here the complication that sanctions might not be called “punishment” (or even
“sanctions”) if inflicted upon innocent persons.
2
Two Concepts of Constitutive Rules
3
2.2 Two Concepts of Rules
Following up on his distinction between two levels of justification, Rawls distinguishes two concepts of rules. The first concept, which Rawls calls the “summary concept”, takes rules to be rules of thumb which make it easier to determine what ought to be done according to some independent standard. This independent standard might, for instance, be the utilitarian one. Let us assume for
the sake of argument that practically every case of punishing a thief would increase utility. This means that, according to the independent standard, almost
all thieves should be punished. For courts that must take decisions about individual cases of punishment it is much easier to apply the rule that thieves should
be punished, than to compute the costs and benefits of punishment in terms of
utility for every case of theft that is brought before them. Epistemic efficiency
suggests that judges apply the rule, rather than compute utility for individual
cases. Given our assumption that the rule is merely an epistemic tool (a “rule of
thumb”), the punishment of a concrete suspect is justified by utilitarian considerations. However, the judgment of the court is epistemically justified by the application of the rule.4
The second concept of rules, which Rawls calls the “practice concept”,
takes rules to be the ultimate justification of the decisions based upon them. For
example, a thief can be punished because there is a rule to this effect, and not
because punishment would maximize utility. The use of this rule, as part of the
social practice of criminal law, may be justified on utilitarian grounds, but the
justification of this general use should not be confused with the justification of a
concrete instance of punishment.
To elucidate the distinction between the two concepts of rules, Rawls lists a
number of characteristics of rules on the summary concept and on the practice
concept, and opposes them. For my present purposes the relation between the
rules and the cases to which they are applied is particularly interesting. According to Rawls, on the summary concept, the decisions made on particular cases
are logically prior to rules. He means that it is determined whether a person can
be punished independently of the rule on punishment. This determination results, for instance, from application of the utilitarian standard. Moreover, the
rule of thumb is correct or incorrect depending on whether its outcomes match
the outcomes of the utilitarian standard. On the practice concept of rules, however, the rules are logically prior to the proper decisions in particular cases, because these decisions depend on the rules that must be applied. This means that
the correctness of the rules cannot be tested against the results they produce in a
concrete case.
When I wrote that in the practice concept of rules, the rules are logically
prior to the proper decisions in particular cases, I (intentionally) misrepresented
Rawls a bit. Rawls, actually, wrote that “the rules of practices are logically prior
to particular cases” (Rawls 1999: 36). By that formulation he meant to say that
some kinds of behaviour are only possible because there are rules regulating this
behaviour. An example would be the rule of article 310 of the Dutch Criminal
Code, which, amongst other things, defines which kinds of behaviour count as
theft. The punishment of thieves (those who committed theft) is only possible
This use of the expression “epistemically justified” stems from the present author, not
from Rawls.
4
thanks to this rule which defines what counts as theft. This kind of logical priority does not follow from the distinction between the summary concept of rules
and the practice concept. That some rules constitute, rather than predict, the
outcomes of cases does not mean that these rules also define particular kinds of
events. Therefore I reformulated Rawls’ characterization of the logical priority
of rules on the practice concept to a version in which the logical priority does
follow. However, the fact that the logical priority of the kinds that Rawls had in
mind does not follow from the distinction between the two concepts of rules,
does not mean that this kind of logical priority is uninteresting. On the contrary,
it is this kind of logical priority which Searle had most likely in mind when he
used it to found the distinction between regulative and constitutive rules.
2.3 The Ontological Priority of Constitutive Rules
We must distinguish between the priority of rules over concrete decisions for
which Rawls made a case and which is based on the distinction between the
summary concept and the practice concept of rules and the priority which Searle
uses to argue for the distinction between regulative and constitutive rules. I will
call the former priority “logical” and the latter priority “ontological”. The ontological priority of constitutive rules boils down to this: these constitutive rules
are necessary for the existence of the phenomena with which these rules deal.
For instance, rules which make thieves punishable can only make sense if they
also define theft. I use on purpose the rather vague verb “to deal”, because we
will see that there are several reasons why it cannot well be said that constitutive
rules also guide the phenomena which they constitute.5
Let us first look how this ontological priority functions in the classical example of playing chess.6 Some forms of behaviour, such as moving sculptured
pieces of wood over a board with a pattern of light and dark squares, count as
playing chess partly because of the existence of the rules of chess. These rules
make playing chess possible and are in that sense constitutive for chess and
therefore they are called constitutive rules. However, these rules do not prescribe
behaviour and if “regulating” is interpreted as determining what is and what is
not allowed, the constitutive rules of chess do not regulate the game. They determine what counts as a valid move in chess, they determine what the players
should strive for in order to play the game seriously, they define certain situations such as check-mate and stale-mate, and they determine when a game is finished and who the winner is. The rules of chess “deal with” playing chess, but
they are not prescriptive or permissive.7 If Searle distinguishes regulative rules
5
To anticipate the argument that will follow: one reason is that constitutive rules typically do not prescribe behaviour at all (this subsection), not even in a wide sense, while the
other reason is that if rules ‘prescribe’ behaviour in a wide sense, strictly speaking they do
not prescribe at all, but merely constitute obligations or duties (section 5.1).
6
The example is classical, not only because it was used by Searle (1969: 33-34), but also
because it was used to more or less the same purpose by Alf Ross (1959: 11-17). Ross was
in turn inspired by Kelsen (1934/1992: 10-11), who did not use the chess example, but
wrote about norms as schemes of interpretation, a kind of constitutive rules avant la lettre.
7
So I disagree with Żełaniec when he writes (Żełaniec 2013: 103) that, once it has been
established that a particular item is a bishop, ‘we must put that item under the obligation
to move diagonally only’ (italics in the original). Żełaniec raises this point to argue that
the rule that a bishop can only move diagonally is not constitutive for what counts as a
Two Concepts of Constitutive Rules
5
from constitutive ones only by means of the practice which they deal with, he
seems to overlook that regulative rules are by definition deontic (prescriptive or
permissive), while constitutive rules very often are not. Therefore it is less than
felicitous to say that constitutive rules “regulate” the practice which they constitute.
It might be objected that the rules of chess also regulate the game by prohibiting some moves (“you are not allowed to castle when your king is in check”)
and allowing others (“on its first move a pawn is allowed to go two fields forward”). This objection does not cut ice, though. Characteristic of a prescription
is that violation is normally possible. In case of the rules of chess, violation leads
to invalid moves which have no influence on the progress of the game. These
rules, which define the game, cannot be violated because a violation would require that an illegal move be made in the game, while “moves” that violate the
rules do not count as moves in the game.8 Therefore it is not possible to violate
the rules that define how pieces move in a game of chess, which is a sure sign
that these rules are not prescriptive.
Criminal law is a social practice which is in part constituted by mandatory
(prescriptive) rules. This practice consists of cooperating organizations (police,
courts, prisons) the functioning of which is governed by rules. Moreover, the
rules that belong to the practice define kinds of behaviour (e.g. theft), prohibit
these kinds of behaviour (the mandatory rules), and empower courts to impose
sanctions in case these prohibitions were violated. As this example illustrates,
rules are only part of the practice, where the practice does not only consist of
rules, but also of organizations. Moreover, the rules which belong to the practice, and in that way partially constitute the practice, fulfil rather different roles.
They include counts-as rules, such as the rule that defines what counts as theft9,
but also mandatory rules, such as the rule that prohibits theft, and powerconferring rules10, such as the rule that makes it possible to sanction criminals.
All these kinds of rules—counts-as, mandatory, and power-conferring—
contribute to the existence of the practice and are in that sense constitutive. So it
is not only counts-as rules which are constitutive rules in the sense of practicedefining rules.
It is worthwhile to consider how counts-as rules, mandatory rules and power-conferring rules operate together to partly constitute the social practice of
criminal law. If we take the prohibition of theft, we see a mandatory rule which
regulates behaviour which both does and does not exist prior to the existence of
the prohibition. Would it be possible to have thefts without a rule that prohibits
bishop in a particular instance of the game. Perhaps Żełaniec is right on that score, but
this rule is constitutive for what counts as a valid move for a bishop, and does not impose
any obligations on bishops in chess, or on chess players.
8
If the pieces are made of wood, the only thing that moves in case of an illegal move, is
the piece of wood, and perhaps also the chess-piece (the piece of wood in the meaning it
has in chess). The game did not move (progress), however.
9
I do not want to suggest that theft cannot exist without such a counts-as rule. However,
counts-as rules are essential for the existence of theft in the sense of criminal law.
10
I only use the expression ‘power-conferring rule’ here because it is established. In
Foundations and Building Blocks of Law (Hage 2018: 200) I argue that, strictly speaking,
there are no power-conferring rules, but that there are competence-conferring rules,
which are not the same thing under a different name.
theft? It seems so. It is true that theft in the sense of criminal law cannot exist
without a rule defining what counts as theft, but this counts-as rule does not
prohibit theft. For the existence of theft in the sense of the criminal law, the existence of a counts-as rule is required, but not the existence of a mandatory rule.
To this extent, the rule that prohibits theft does not constitute the behaviour that
it regulates. However, this rule makes it possible to violate the criminal law, and
such a violation is required for the competence of a court to sanction the criminal.11 Although it is possible to commit theft even if theft is not prohibited, and
the prohibition can therefore be said to regulate antecedently existing behaviour,
the counts-as rule that defines theft only makes sense in the context of a social
practice which also contains the prohibition and the power-conferring rules. It is
only possible to commit theft in the sense of criminal law if the social practice of
criminal law exists, and all the different kinds of rules that go into criminal law
are conditions for the existence of this practice, and therefore indirectly also for
the existence of concrete cases of theft. In this sense, the prohibition of theft is
ontologically prior to concrete cases of theft.
2.4 Conclusion on Practice-Defining Rules
Starting from Searle’s distinction between regulative and constitutive rules and
Rawls’ distinction between the summary and the practice concept of rules, we
found that social practices depend for their existence on rules. These rules are
ontologically prior to facts that exist in the context of these social practices, and
are in this sense constitutive for both the practices and the facts that exist in their
context. I call these constitutive rules practice-defining rules and emphasize that
practice-defining rules may be of many different kinds, including counts-as rules,
mandatory rules and power-conferring rules.
In subsection 3.3 we will encounter fact-to-fact rules, which are constitutive
in a different sense, because they function as constraints on possible worlds.
However, first I will provide a setting for these rules by going into some detail
concerning directions of fit, and constraints on possible worlds (subsections 3.1
and 3.2).
3. Rules as Constraints on Possible Worlds12
Often the notion of a rule is connected to the guidance of behaviour: rules indicate what we should do. Typical examples are the rule that impose the duty on
home-owners to clean away the snow on the pavement in front of their houses
and the rule that prohibits building in a zone that the municipality reserved for
environmental purposes. However, there are also rules whose primary function
does not seem to be to guide behaviour. Examples would be the rule that gives
the municipality council the power to make a parking regulation for the city and
the rule that makes persons against which a serious suspicion exists that they
committed a crime into criminal suspects in the sense of the Code for Criminal
Procedure.
11
For the sake of argument I ignore the complication that the crime needs to be proven if
the competence is to arise.
12
The argument in this section is a re-working of the argument of Hage 2015a. See also
Hage 2018: 57-70.
Two Concepts of Constitutive Rules
7
If it is not the primary function of all rules to guide behaviour, the question
is whether there is a function that all rules share. This question can be answered
affirmatively: all rules have in common that they attach the presence of facts to
the presence of other facts. A proper understanding of this common characteristic requires that we pay some attention to what will be called the “world-to-word
direction of fit” of rules.
3.1 Directions of Fit
Perhaps the best way to introduce the distinction between directions of fit is by
means of an example of Anscombe's (Anscombe 1976: 56). Suppose that Elisabeth makes a shopping list, which she uses in the supermarket to put items in
her trolley. A detective follows her and makes a list of everything that Elisabeth
puts into her trolley. After Elisabeth and the detective are finished, the list of the
detective will be the same as Elisabeth’s shopping list. However, the lists had
different functions. If Elisabeth used the list correctly, she placed exactly those
items into her trolley that are indicated on the list. Her behaviour is to be
adapted to what is on her list. In the case of the detective it is just the other way
round; the list should reflect Elisabeth’s shopping behaviour. The two different
functions of the lists with regard to Elisabeth’s behaviour represent the two different directions of fit that we are looking for.
The two items involved in Anscombe’s example are a linguistic one—the
list of items—and the world—the collection of all facts. The directions-of-fit distinction can also be applied to other items than linguistic ones, but let us focus
on the linguistic case first. The relation between language and the world goes in
two directions. If the linguistic entities, in particular descriptive sentences, are to
be adapted to the world, as when the detective writes down which groceries are
in the trolley, the fashionable expression is “word-to-world direction of fit”. If
the world is to be adapted to the linguistic entities, as when Elisabeth puts those
items in her trolley that are mentioned in her shopping list, the fashionable expression is “world-to-word direction of fit” (Searle 1979: 3-4).
For the world-to-word direction of fit we must distinguish between three
kinds. For all three kinds it holds that somehow the facts in the world are
adapted, in order to “fit” what is expressed by the words. One case is when the
words function as a directive, as when James shouts “Carol, stop!” when he fears
that his young daughter Carol will cross the busy street. This order aims at making Carol stop, and if the order is successful in the sense of “efficacious”, Carol
will stop and the facts in the world fit the content of the order. In this case the
relation between the utterance of the order (the performance of the speech act)
and the facts in the world is causal. I will therefore write about the “causal
world-to-word direction of fit”.
A second form of the world-to-word direction of fit manifests itself in constitutive speech acts. Constitutive speech acts are speech acts performed with the
intention to bring about a particular change through the operation of a rule or
convention. They differ from directives which operate by means of a causal, rather than a rule-based, connection. Examples of constitutive speech acts are the
baptism of a ship (“I hereby baptize you the Princess Victoria”), making a promise (“I promise you to be back before 8 o’clock this evening”), granting a power
(“You can consider every promise made by Michal on my behalf as a promise
made by me”), and the issuing of a command (as distinguished from an order).
In all these cases the facts in the world have come to fit the content of the speech
act, but in contrast to the operation of directives, the result is brought about by a
rule or convention, and not through a causal connection. In connection with
constitutive speech acts we will speak of the “constitutive world-to-word direction of fit”.
The result of a successful command, such as “I hereby forbid you to cross
the street” directed by James to his daughter Carol, is that a duty enters into existence.13 In this case it is the duty for Carol not to cross the street. If such a
command is successful, the facts in the world come to match the content of the
speech act and Carol has from that moment on the duty not to cross the street.
In this case the relation between the speech act and the facts in the world is constitutive by nature; the performance of the speech act constitutes the duty. This
is a case of the constitutive world-to-word direction of fit, and I mention it explicitly to emphasize the difference between orders, conceived as a kind of directives, and commands, conceived as constitutive speech acts. The terminological
distinction between orders and commands is stipulative: that is how I will use
these words here. However, the difference between directives, based on a causal
connection, and the creation of duties by means of constitutive speech acts and
based on the operation of rules, does not depend on this terminological convention.
The third kind of world-to-word direction of fit concerns the effects of rules.
Take for example the conceptual rule (the meaning postulate) that the word
“rectangle” denotes quadrilaterals with four orthogonal angles. Given this rule,
if something is a rectangle, it must be a quadrilateral with four orthogonal angles. This “must” depends on the conceptual rule that defines the relation between being a rectangle and being a quadrilateral. The facts in the world adapt
themselves to the rule—the quadrilateral now also is a rectangle—and that is
what is meant by the world-to-word direction of fit of rules, or—as we will see in
section 3.3—more in general the world-to-word direction of fit of constraints.
The idea that rules have a special form of the world-to-word direction of fit
deserves an elaborate explanation, but that requires more background on possible worlds and necessity. That is the topic we will therefore turn to now.
3.2 Possible Worlds
We are all familiar with the distinction between what the facts actually are and
what the facts might have been. It happens to be snowing, but the sun might just
as well have been shining. In Syria there is a war, but there might have been
peace.
Logicians use possible worlds-terminology to deal with this distinction between what the facts actually are and what they might have been. 14 They say, for
instance, that in the actual world it is snowing, but that in some other possible
world the sun is shining. Intuitively, a possible world is an exhaustive set of facts
which makes some descriptive sentences true and others false. The actual world
is one of the many worlds that are possible and in the actual world it is snowing.
13
It is not that Carol does not cross the street. That would be a causal effect of Carol’s
belief that she has the duty not to cross the street.
14
To keep the exposition relatively simple, the following accounts of possible worlds semantics and of propositional logic are not very precise.
Two Concepts of Constitutive Rules
9
However, in some other possible world, the sun is shining. That is another way
of saying that although actually it is snowing, the sun might have been shining.
What is necessary is the case in all possible worlds, while what is impossible is
not the case in any possible world. What is contingent is the case in some, but
not in all possible worlds. In all possible worlds, nine is the square of three.
Moreover, in all possible worlds, if Luis is either in Madrid or in London, and
he is not in London, then he is in Madrid. However, in some possible worlds,
Luis is in Madrid, while in some other possible worlds he is in London.
3.3 Constraints on Possible Worlds
The intuitive idea about the connection between possible worlds and necessity is
quite simple, but there turns out to be a complication: what is the difference between a possible world and an impossible one? Rather than trying to answer this
question in abstract, it is better to study an example in which the idea of possible
worlds has turned out to be fruitful. This example is propositional logic.
As its name indicates, propositional logic deals with propositions, the
meanings of descriptive sentences. Some propositions are basic or atomic; they
do not exhibit any internal structure. Other propositions are compound or molecular. They contain a logical operator and—most of the times—more than one
basic proposition. Examples, with the logical operators italicized, are “Jane
loves Mary and Mary loves Jane”, “If Siobhan talks to Joachim, then Joachim
listens”, and “It does not rain”.
An important difference between elementary and compound propositions is
that the truth-values (true or false) of the former are completely independent
from one another, while the truth-values of the compound propositions are
completely determined by the truth-values of the elementary propositions that
are part of them. To be more concrete, the truth-values of “Jane loves Mary”
and of “Mary loves Jane” are, logically speaking, independent from each other,
while the truth-value of “Jane loves Mary and Mary loves Jane” completely depends on the truth-values of the former two propositions. The latter proposition
is true if and only if the two former propositions are both true.
If two worlds agree in their truth-values for “Jane loves Mary” and “Mary
loves Jane”, and if they do not agree in their truth-values for “Jane loves Mary
and Mary loves Jane”, at least one of these worlds cannot be logically possible.
Logically possible worlds are constrained in such a way that the truth-values of
compound propositions in these worlds are determined in a particular way—
depending on the form of the compound sentences and the logical operators that
occur in them—by the truth-values of the elementary propositions that constitute these compound propositions.
Constraints on logically possible worlds can be characterized in terms of
constraints on the facts that are elements of these possible worlds. In a logically
possible world some compound states of affairs must obtain if some elementary
states of affairs obtain, and the other way round. For example, the compound
state of affairs “Jane loves Mary and Mary loves Jane” must obtain in a logically
possible world that contains the elementary states of affairs “Jane loves Mary”
and “Mary loves Jane”. More in general, worlds that are possible according to
propositional logic satisfy the constraints of propositional logic. A world that
does not satisfy these constraints does not count as logically possible.
Constraints on possible worlds make that some things are necessarily the
case, while other things cannot be the case. Given the constraints of propositional logic, it must be the case that either it rains or it does not rain. Given
these same constraints it cannot be the case that both it rains and it does not
rain. However, these constraints make it possible that both it rains and it is
snowing.15 Constraints on possible worlds do not only support necessity and
possibility judgements, but also conditional judgements (“conditionals”). For an
arbitrary possible world we cannot know whether it snows, but we do know that
if it snows, then the state of affairs “It snows or the sun shines” obtains. The reason why constraints support modal judgements—judgements about what is
(im)possible and what is possible—and conditionals, is that constraints have the
world-to-word direction of fit. They determine which worlds are possible and in
that way impose themselves on possible worlds. A world in which it rains can
only be logically possible if either it rains or the sun shines. Possible worlds are
adapted to the constraints to which they are subjected. Of course this adaptation
is not something that happens in time. It is a conceptual matter; the world only
counts as logically possible if it satisfies the constraints of logic.
The constraints of propositional logic are useful to illustrate the operation
of constraints on possible worlds, because they are well-defined and—at least to
the readers who know propositional logic—quite familiar. However, other constraints than the logical ones are also important. Physical constraints, for example. A world in which sound travels faster in vacuum than light is not physically
possible, at least not according to our present knowledge of physics. The same
holds for a world in which metals do not expand when heated, or a world in
which the total amount of energy fluctuates. The constraints of physics limit the
facts that can physically go together.
Just as logical constraints underlie logical necessity judgments, physical
constraints underlie physical necessity (and possibility) judgments. We have already seen some examples of judgements that particular facts cannot go together
(impossibility judgements). Examples of physical necessity judgements are that a
piece of metal necessarily expands when heated, that light must travel in a right
line through vacuum if not subject to gravitational forces, and that an unsupported body with gravitational mass must drop when in the neighbourhood of
the Earth. Physical constraints also support physical conditionals and counterfactual judgements: “If this piece of metal were heated, it would expand” and
“Had this stone not been supported, it would have dropped”.
What holds for physical constraints also holds for conceptual constraints
(meaning postulates; semantic conventions). The convention that defines bachelors as unmarried males underlies the judgment that necessarily all bachelors are
unmarried, while the convention that defines what a book is underlies the judgment that it is possible to read books. Semantic conventions also support conditionals such as “If this is a skate board, it is a vehicle in the sense of the Traffic
Act”, and counterfactuals such as “If this vehicle had a motor, it would have
counted as a car”.
15
Perhaps snow and sunshine cannot go together but that would not be a matter of logic,
but of what is physically possible. Some things are logically possible, but physically impossible.
Two Concepts of Constitutive Rules
11
The last two examples concern semantic rules that could also have been legal rules. In general, conceptual constraints depend for their existence on their
being adopted by a language-using community. They illustrate that constraints
can depend for their existence on being adopted by human beings.
3.3 Rules as Constraints on Possible Worlds
Why are rules a kind of constraints? Because they behave like other constraints.
In a world in which a rule exists, the rule imposes itself on the facts of that
world with the world-to-word direction of fit that other constraints also have. So
if some possible world contains the rule that thieves are punishable, then in that
world thieves are punishable. In that world, it is not merely a contingent matter
of fact that thieves are punishable, but a necessary one, because being a thief
makes one punishable. Moreover, the rule also supports conditional and counterfactual judgements: if Jane had been a thief, she would have been punishable.
Rules have a lot in common with more traditional constraints such as the
logical and physical ones, but they also have a characteristic that is not shared
by all other constraints: rules only apply locally. The laws of one country are for
example different from the laws of another country. The necessity of rule-based
judgements is therefore merely local necessity: in some African countries practicing homosexuals count as criminals, while that is not the case (anymore) in
European countries. This is different for logical and physical laws, which seem
to have a universal scope of application.16 The scope of rules is not only limited
in space, but also in time. Many rules can be created or repealed and in that
sense they differ from more traditional constraints which somehow seem outside
the scope of human manipulation. When the rule that thieves are punishable is
introduced, suddenly all thieves become punishable. And when the rule is repealed again, the possibility to punish thieves disappears with the rule.
As a consequence of these differences, there can be some logically and
physically possible worlds in which a particular rule exists, and other possible
worlds in which the same rule does not exist. In a sense it might be said that logical and physical constraints create necessities that are themselves necessary,
while rules create contingent necessities. For this reason, rules will be categorized as “soft constraints”, as opposed to the hard constraints that seemingly do
not depend for their existence on human decision making or social practices. 17
3.4 Factual and Descriptive Counterparts of Rules
16
This difference should not be overestimated, however. The geometrical law that the
three corners of a triangle add up to 180 degrees only holds for relatively small triangles
and (which may be the same issue) for triangles in a flat plane. See also the discussion of
the scope of physical laws in Toulmin 1953: 69 and 78.
17
‘Seemingly’, because it is not clear in which respect physical laws are more than mere
regularities, with their aspect of necessity added through social practice. Similarly, with
the proliferation of logical systems it becomes increasingly clear that logical necessity
may be more mind-dependent than used to be assumed. This is an important theme,
closely related to the nature of constraints and the way they differ from mere regularities,
but it deserves a more extensive treatment than it can be given here.
The rule that thieves are punishable makes it impossible that thieves are not
punishable.18 Or, to state the same thing affirmatively, the rule necessitates that
thieves are punishable. If some rule—or, more in general, a constraint—exists,
this means that some general descriptive sentence must be true.
Such sentences are open generalizations which deal with potentially infinitely many items. The open generalizations that describe the effects of rules
typically have the same formulation as the rule the effects of which they describe, and they are true because that rule exists. If a rule has the formulation
“Thieves are liable to be incarcerated for a maximum of five years”, the effect of
this rule can be described by saying that thieves are liable to be incarcerated for
maximum five years. The sentences that describe the consequences of a rule describe facts that will be called the “factual counterpart” of the rule.
The sentences themselves may be called the “descriptive counterparts” of
rules. Where rules impose themselves on the world by way of their world-toword direction of fit, but are not true or false, the descriptive counterparts of
rules are descriptive sentences with the word-to-world direction of fit, which are
true or false, usually depending on the existence of the rules of which they are
the counterpart.
Since the facts which make these sentences true obtain because of rule, descriptive counterparts of existing rules must be true. The facts which are expressed by these sentences and which exist because of rules can be designated as
the factual counterparts of these rules.
The differences between rules, their factual counterparts and their descriptive counterparts, may be confusing, but the three can be kept apart if one realizes that rules have the world-to-word direction of fit (they constrain the facts in
the world), that descriptive counterparts have the word-to-world direction of fit
(they describe the facts in the world), and that facts are just elements of the
world and have no direction of fit.
4. Fact-to-Fact Rules
All rules have the world-to-word direction of fit: they impose themselves on the
worlds (places and times) in which they exist by attaching facts to other facts.
That is why they are also called “fact-to-fact rules”. The connected facts may be
simultaneous, or they may succeed each other in time.
4.1 Dynamic Rules
The latter is the case with dynamic rules: they create new facts, or modify or
take away existing facts as a consequence of the occurrence of an event. One example is that Jane promised Richard to give him €100, with the consequence
that from the moment of the promise on Jane has the moral obligation to give
Richard €100. Another example is that Eloise was appointed chair of the French
Parliament, with the consequence that from the starting point of the chair’s new
term on, Eloise will be the chair of the French Parliament.
These examples also illustrate that the world-to-word direction of fit of constitutive speech acts is based on the world-to-word direction of fit of dynamic
18
For our present purposes we will not take into consideration the possibility of exceptions to rules.
Two Concepts of Constitutive Rules
13
rules. Because of the operation of the dynamic rule that making a promise creates an obligation, promising is a constitutive speech act with the world-to-word
direction of fit that it creates an obligation. The distinction between the worldto-word direction of fit of constitutive speech acts and the world-to-word direction of fit of constraints and more in particular dynamic rules may be useful for
expository purposes, but they are two sides of the same coin.
4.2 Static Rules
Whereas dynamic rules govern the succession of facts in time, static rules govern the co-existence of facts. They make (constitute) that one kind of fact goes
together with some other kind of fact, where the latter fact depends (supervenes)
on the former. The relation between the kinds of facts is timeless, in the negative
sense that the one kind of fact is not the occurrence of an event after which the
second kind of fact comes into existence.
Legal examples of static rules are that:
1. the owner of a good is permitted to use this good;
2. home-owners must (have the duty to) keep the pavement in front of their
houses clean;
3. the Mayor of a municipality has the power to issue emergency regulations
for that municipality;
4. the King of the Belgians is the commander in chief of the Belgian army.
4.3 Counts-as Rules
One kind of fact-to-fact rules has become particularly well-known: the counts-as
rules. Counts-as rules have the structure: Individuals of type 1 count as individuals of type 2. These “individuals” may be human beings, as in the rule that the
parents of a minor count as the minor’s legal representatives, or the rule that the
king of the Belgians is the commander in chief of the Belgian army. Often, however, the “individuals” that count as another kind of individual are events. For
instance, under particular circumstances, causing a car accident counts as committing a tort, or offering money to another person counts—given “suitable” circumstances—as attempting to bribe an official. Frequently counts-as rules are
conditional, meaning that individuals of type 1 only count as individuals of type
2 if certain conditions are satisfied. An example from Dutch law (art. 3:84 of the
Civil Code) would be the rule that the delivery of a good counts as the transfer
of that good if the person who made the delivery was competent to transfer and
if there was a valid title for the transfer.
4.4 Practice-Defining and Fact-to-Fact Rules
All rules are fact-to-fact rules, and if there are practice-defining rules, they must
be fact-to-fact rules at the same time. When we look at examples of practicedefining rules, for examples the rules that constitute the social practice of criminal law, we find confirmation of this claim. There are counts-as rules which define the crimes, there are duty-imposing rules, which forbid acts that would constitute a crime, and there are power-conferring rules, granting courts the power
to impose sanctions upon convicted criminals. These rules, in combination with
each other and with organizations which apply them, together constitute the
practice of criminal law, and they are all fact-to-fact rules at the same time. So
there is no opposition between practice-defining rules and fact-to-fact rules:
sometimes fact-to-fact rules (partly) define a social practice and then they are also practice-defining rules. There is no opposition, but only a contrast between
different functions which rules can fulfil, and there is no reason why a rule cannot fulfil both functions at the same time.
5. Regulative Rules
5.1 Duties and Obligations
The English language does not make a strict distinction between duties and obligations, but in the sphere of normativity there is an important distinction and
the words “obligation” and “duty” are quite suitable to mark this distinction. 19
Obligations are relations between typically two concrete persons, where the one
person, the debtor, is obligated to do something, or to refrain from doing something, while the other person, the creditor, has a claim on the debtor that he perform his obligation. Obligations are the result of an event that took place, typically a promise, a contract, or a tort, to which the obligation was attached by a
dynamic rule. The content of the obligation is partially defined by this event.
Examples of obligations are the obligation that Jane undertook towards Richard
by promising him that she would pay him €100, and the obligation of Carol towards Nina to compensate the damage that Carol caused to Nina by bumping
into Nina’s car.
Duties, on the contrary, do not involve other persons than the duty-holder,
although the content of the duty may mention a person. For example, Heinrich
may have the duty not to hurt his dog Bonzo and also the duty not to hurt his
brother Franz when they have a quarrel. Neither Bonzo nor Franz has a claim
against Heinrich not to be hurt, although Franz may receive a claim against his
brother for damages if his brother breaches his duty not to hurt Franz. While obligations are attached to events by a dynamic rule, duties are typically 20 attached
by a static rule to the possession of a certain status. For example, a static rule
attaches the duty to clean the pavement in front of a house to the status of being
the home-owner, while another static rule attaches the duty to turn on the car
lights when it gets dark to the status of being a car-driver. The most general duties are attached to the status of being human, or to the status of a legal subject. 21
We find that a person can in at least two ways become obligated to do
something: via a dynamic rule and via a static rule. In the first case the dynamic
rule creates an obligation; in the second case the static rule creates a duty. Both
the dynamic rule and the static rule are in themselves not deontic: they do not
obligate but they create, respectively an obligation and a duty. It may seem that
rules that create obligations or duties are nevertheless normative because of their
normative impact. To some extent this is indeed the case, but they are not so
very different from other dynamic and static rules. A dynamic rule which creates
an obligation is not very different from a dynamic rule which makes somebody
The suitability has to do with the use of the word ‘obligation’ in the civil law tradition
(Zimmerman 1996: 1-10), which matches the use of the word that will be proposed here.
20
Typically, but not always: a duty may be the result of a command.
21
Notice that the status to which a static rule attaches duties (or other consequences)
may, but need not, itself be the result of a fact-to-fact rule.
19
Two Concepts of Constitutive Rules
15
into the chair of the French parliament. The only difference between them is the
kind of fact which they attach to an event. An obligation-creating rule makes
that some obligation exists, while the rule which makes somebody into the parliament chair makes that somebody is the parliament chair. A static rule which
attaches the duty to clean the pavement to the status of home-owner is not very
different from the rule that attaches the power to dispose of the house to this
same status.
The impact of this observation that regulative rules, that is obligationcreating and duty-imposing rules, are not very different from dynamic, respectively static rules which do not have normative consequences can be seen clearly
in the following reinterpretation of Searle’s “derivation” of Ought from Is.
5.2 Searle’s “Derivation” of Ought from Is
It is often assumed that Is and Ought represent different ontological spheres, and
that therefore it is not possible to derive ought-conclusions from merely ispremises. That would give regulative rules a special ontological status, because
they would provide the deontic content for duties and obligations. In a famous
article Searle attempted to show that this alleged gap between Is and Ought can
be bridged through the speech act of promising. This would undermine the special role of regulative rules, because they would not be necessary anymore for
the existence of obligations. Searle’s conclusion was correct, be it that his argument can be improved. The improved version which is presented in this section
provides us with additional insight into the nature of fact-to-fact rules and is for
that reason included here.22
Searle produced his famous argument by means of which he tried to show
that it is possible to deduce Ought from Is alone in an article from 1964 (Searle
1964). In this argument an actual derivation of an ought-conclusion from ispremises takes the central role. The derivation went as follows:
1. Jones uttered the words “I hereby promise to pay you, Smith, five dollars”.
2. Jones promised to pay Smith five dollars.
3. Jones placed himself under (undertook) an obligation to pay Smith five
dollars.
4. Jones is under an obligation to pay Smith five dollars.
5. Jones ought to pay Smith five dollars.
Searle argued that the relation between a statement in this list and its successor is either an entailment or at least not contingent, and moreover that the
relation could, where needed, be made into an entailment by the addition of a
premise which was neither an evaluative statement, nor a moral principle, nor
anything of the sort.
Searle’s argument is less than convincing, but nevertheless correct in its underlying idea. It is less than convincing because of the presupposition needed to
get from (2) to (5), which is presumably something like:
For any x and any A, if x promised to do A, then x ought to do A.
22
The argument was presented earlier, in a more elaborate form and in a different setting,
in Hage 2011 and Hage 2013. See also Hage 2018: 97-102.
The problem is that this presupposition is a sentence which expresses,
among other things, also an ought. Searle tries to tackle this by the claim that
the sentences which make the transition between the sentences of the argument
deductively valid are analytic (literally: “tautologies”), but even if Searle is correct in this claim, the issue remains that analytic ought-sentences are still oughtsentences. Therefore, Searle did not succeed in deriving Ought from Is alone. 23
Nevertheless, Searle’s argument that it is possible to derive Ought from Is
has an underlying idea which is correct. This underlying idea is that some events
lead to new obligations and therefore new oughts, and that the presence of these
events itself does not depend on pre-existing obligations or oughts. In the following, the correctness of this idea is illustrated by using contracts as example, instead of promises.
5.3 Contracts and their Consequences
Contracts are a means by which legal subjects can change the legal positions of
themselves or other persons. Contracts can only operate in a setting of rules,
which define, amongst other things, how contracts can be made, which persons
are competent to and have the capacity to contract and which legal consequences are connected to the successful creation of a contract. These latter rules are
the dynamic rules which play such a crucial in Searle’s argument.
The only point of making promises is to undertake obligations. Although it
is possible to undertake obligations by means of contracts as well, contracts can
be also used for other purposes that cannot be achieved by promises. It is for instance possible to appoint by means of a contract an arbiter who is empowered
to decide over conflicts that might arise in connection with the execution of (the
rest of) the contract. In a contract, the parties create or repeal facts, to the extent
that they are empowered by dynamic rules to do so. Although contracts do not
necessarily lead to obligations, they often do. And when contracts lead to obligations, they almost always also lead to the oughts that follow from these obligations.24 In this way, contracts seem to bridge the gap between Is and Ought.
It may be objected that contracts only seemingly do so because the obligation to do what was contracted is based on the regulative rule that contracts
ought to be complied with. The contract itself would on this view be nothing
more than a specification for a concrete situation of what the general obligation
to perform contracts, laid down in the regulative rule, implies. However, it is
questionable whether the rule that contracts ought to be complied with really
exists. The point of contracts is more general than merely that contracts facilitate the intentional creation of obligations. Their point is that the facts established by means of the contract hold between the contract parties.
Underlying contracts is not the rule that contracts ought to be performed,
but the dynamic rule that the facts which a contract aims to bring about, actually
come into existence. The rule that what parties agreed to holds between the parties, does itself not impose any obligations. If obligations result from most contracts, this is because by means of most contracts the contract parties create obli23
See Hare 1964.
It is possible that somebody has the obligation to do something, for instance because he
contracted to do so, while at the same he legally ought to refrain from doing so. More on
the relation between obligations and what ought to be done in Hage 2018: 148-52.
24
Two Concepts of Constitutive Rules
17
gations between themselves. Notice the emphasis on “create”. Before the contract was concluded, the obligations were not there yet; they are the result of the
contract. The presumed rule that one ought to obey one’s contracts is superfluous. If the contract does not create obligations, but aims for instance at cancelling existing obligations, there is nothing to obey. If the contract does create obligations, the rule that one ought to comply with these obligations would effectively be that one ought to do what one is under an obligation to do. That would
be an almost analytical rule. So, apart from this analytical rule according to
which an obligation typically leads to an ought, there is no role for the rule that
contracts ought to be obeyed. The obligation to do what one contracted to do
does not, therefore, derive from such a rule. This obligation is in a concrete case
created by means of the contract and it is a new obligation that did not yet exist
before the contract, not even in the more abstract form of an obligation to comply with one’s contracts.25
Searle’s derivation of Ought from Is rests on two steps. One step is the logical representation of constitution. Given a setting of rules, some events lead to
the coming into existence of new facts, and these facts may include the existence
of obligations. This mechanism is the main step in Searle’s argument, and the
interesting thing about it is that it has almost nothing to do with oughts or obligations. It is a mechanism by means of which all kinds of facts in social reality
are generated by means of rules. That some of these facts involve the existence
of obligations is almost a coincidence. The second step is the “derivation” of an
ought from the existence of an obligation. This is also an important step, but it is
not important for the point that Searle—perhaps unknowingly—made, namely
that through the application of rules—in his example dynamic rules—new facts
can come into existence, and that these new facts may very well involve the existence of obligations.
The relevance of this point can hardly be overestimated, because it illustrates the distinction between two kind of normativity. On the one hand there is
the kind of normativity that is also called “deontic”. It has to do with what
agents have the duty or the obligation (not) to do. On the other hand there is the
normativity of facts that are the result of rule-application, rule-based facts, which
are sometimes also called “institutional facts”.26 It is tempting to treat these two
kinds of normativity as one and the same thing, a temptation which is only
strengthened if one assumes that all oughts must be rule-based. However, it is
possible to create new facts by means of contracts and other juridical acts and
these new facts can, but certainly need not, involve obligations. The conventional world-to-word direction of fit of constraints has nothing to do with duties,
obligations and what agents ought to do. These latter entities have at best—that
is: if they are complied with—a causal world-to-word fit.
5.4 Conclusion on Regulative Rules
25
Notice, by the way, that this “obligation” to comply with one’s contracts cannot even
be an obligation. It holds in general, is not based on a specific event, and has no creditor.
In other words, if it would have existed, it would have been a duty.
26
See Anscombe 1976, MacCormick 1986, and Searle 2010. It has always escaped me
why these facts should be called “institutional”, since only in a limited number of cases,
social institutions are involved in the existence of rule-based facts.
Regulative rules create either duties or obligations. They create them, which
means that before the duty, respectively the obligation, was created, there was
nothing mandatory. Not even the regulative rule itself was mandatory, because
it merely creates, constitutes new facts. Rules that create duties typically mention the duties which they create, and as a consequence it may seem that they involve the duties themselves. For example the rule that home-owners have the
duty to clean the snow in front of their houses mentions the duty to clean the
snow. However, if there are no home-owners, there is no duty, not even if the
rule exists. Therefore the mere existence of a duty-imposing rule does not involve the existence of a duty.
In the case of a rule that creates an obligation this is even more pronounced. Some rules that lead to an obligation at least mention the obligation.
This is the case with the rule that creates the obligation to pay damages in case
of a tort. However, the rule that the facts to which contract parties agreed hold
between the parties, does not even mention an obligation. If this rule leads to an
obligation, this is because the contract parties agreed on the creation of an obligation.
Rules that impose duties or create obligations—in short: regulative rules—
are merely constitutive, and if they lead to an ought, a new ought, it is only the
created ought that is regulative. The rule itself is merely constitutive. This means
that all rules are constitutive and that strictly speaking no rule is regulative. This
need not stop us from calling duty-imposing rules and some27 obligation-creating
rules regulative, as long as we are aware that they are only indirectly regulative,
namely through the duties and obligations that they constitute.
6. Final Conclusion
In this article I have argued that rules have two main functions, the practicedefining one and the constraining (fact-to-fact) function. These two functions are
compatible. In their function as constraints, some rules are also indirectly
regulative. In both of their functions, rules differ from the summaries (rules of
thumb) that Rawls discussed and opposed to the constitutive (fact-to-fact) rules
which make that some decisions are the right ones.
In his work, first on the philosophy of language and later on social ontology, Searle focused on one kind of constitutive rules: counts-as rules, which are
constitutive in the sense that they attach new facts to the existence of “old”
ones. In doing so, Searle created the scientific interest in constitutive rules which
they deserve. However, because of his narrow focus on counts-as rules, Searle
also created the impression that counts-as rules are all there is to constitutive
rules. This impression is wrong, if only because it overlooks dynamic rules. A
possible consequence is that Searle underestimated the relevance of his derivation of Ought from Is in his early paper.
References
27
Only some, because only some dynamic rules have the creation of obligations as a legal
consequence. Other dynamic rules have more abstractly defined consequences, which
may involve obligations.
Two Concepts of Constitutive Rules
19
Anscombe, G.E.M. 1976, Intention, 2nd ed., Oxford: Basil Blackwell.
Hage, J.C. 2005, Studies in Legal Logic, Dordrecht: Springer.
Hage, J.C. 2007, De wondere wereld van het recht [The Wondrous World of the Law],
Inaugural Address, University of Maastricht.
Hage, J.C. 2011, “Legal Transactions and the Legal Ought”, in Stelmach, J. and
Brożek, B. (eds.), The Normativity of Law, Kraków: Copernicus Press, 167-90.
Hage, J.C. 2013, “Juridical Acts and the Gap between Is and Ought”, Netherlands
Journal of Legal Philosophy, 42, 50-66.
Hage, J.C. 2015a, “Separating Rules from Normativity”, in Araszkiewicz, M., Paweł
Banaś, Gizbert-Studnicki, T. and Pleszka, K. (eds.), Problems of Normativity, Rules
and Rule-Following, Cham: Springer, 13-30.
Hage, J.C. 2015b, “The (Onto)logical Structure of Law: A Conceptual Toolkit for
legislators”, in Araszkiewicz, M. and Pleszka, K. (eds.), Logic in the Theory and
Practice of Legislation, Cham: Springer, 3-48.
Hage, J.C. 2016, “Facts, Values and Norms”, in Taekema, S., van Klink, B. and
Been, W. (eds.), Facts and Norms in Law, Cheltenham: Edward Elgar Publishing,
23-52.
Hage, J.C. 2018, Foundations and Building Blocks of Law, Den Haag: Eleven International Publishing.
Hare, R.M. 1964, “The Promising Game”, Revue Internationale de Philosophie, 70,
1964, 398-412. Also in Foot, P. (ed.), Theories of Ethics, Oxford: Oxford University Press, 1967, 115-27.
Kelsen, H. 1992, Introduction to the Problems of Legal Theory, transl. of the first edition
of the Reine Rechtslehre (1934), Litschewsky Paulson, B. and Paulson, S. (eds.),
Oxford: Clarendon Press.
MacCormick, N. 1986, “Law as Institutional Fact”, in MacCormick, N. and Weinberger, O. (eds.), An Institutional Theory of Law, Dordrecht: Reidel, 49-76.
Rawls, J. 1955, “Two Concepts of Rules”, Philosophical Review, 64, 3-32. Also in his
Collected Papers, Cambridge, MA: Harvard University Press, 1999, 20-46.
Rawls, J. 1999, Collected Papers, Freeman, S. (ed.), Cambridge, MA: Harvard University Press.
Ross, A. 1959, On Law and Justice, Berkeley: University of California Press.
Searle, J.R. 1964, “How to Derive ‘Ought’ from ‘Is’”, Philosophical Review, 73, 43-58.
Also in. Foot, P. (ed.), Theories of Ethics, Oxford: Oxford University Press, 1967,
101-14.
Searle, J.R. 1969, Speech Acts, Cambridge: Cambridge University Press.
Searle, J.R. 1979, Expression and Meaning. Studies in the Theory of Speech Acts, Cambridge: Cambridge University Press.
Searle, J.R. 1995, The Construction of Social Reality, New York: The Free Press.
Searle, J.R. 2010, Making the Social World, Oxford: Oxford University Press.
Toulmin, S. 1953, The Philosophy of Science. An Introduction, New York: Harper Row.
Żełaniec, W. 2013, Create to Rule. Studies on Constitutive Rules, Milano: Edizioni
Universitarie di Lettere Economia Diritto.
Zimmerman, R. 1996, The Law of Obligations. Roman Foundations of the Civilian Tradition, Oxford: Oxford University Press.