Keywords

1 Regulative and Constitutive Rules

The idea of a rule is traditionally associated with the guidance of behaviour. Rules prescribe behaviour, they can be followed, obeyed and disobeyed, and after the behaviour has taken place, rules can be used to evaluate behaviour as correct or incorrect. This association between rules and the guidance of behaviour is reflected in philosophical discussions about rule followingFootnote 1, and in jurisprudential accounts of the nature of law. According to Aquinas, law is a rational ordering of thingsFootnote 2; according to KelsenFootnote 3, law consists of norms, where norms are defined in terms of ‘ought’, and according to HartFootnote 4, primary rules are behaviour guiding rules. Moreover, recent discussions emphasise the behaviour guiding role of law.Footnote 5

And yet some rules seem not to guide behaviour at all, or only in the marginal sense that they allow the evaluation of behaviour as correct (in agreement with the rule) or incorrect. Examples of rules which seem not to aim at guidance at all are the rules that confer competences, or define the institutions of, for example, the European Union. Examples of rules which do not strictly guide behaviour but which can nevertheless be used to evaluate behaviour as correct or incorrect are the rules of language and of mathematics, but also the definitions of terms used in legislation, and the rules that specify which procedure should be followed in order to reach a particular result.

Rules of games form an in between category in the sense that they only specify what should be done in order to play the game correctly. They do guide behaviour, but only for those who want to play the game. Some would see the ‘rules’ of moralityFootnote 6 and legal rulesFootnote 7 in a similar way: they only specify what somebody should do who wants to comply with morality or with the law.

The phenomenon that not all rules can easily be said to guide behaviour has found philosophical recognition in the distinction between regulative and constitutive rules. This distinction has gained most of its popularity from the work of Searle.Footnote 8 According to Searle regulative rules regulate antecedently or independently existing forms of behaviour.Footnote 9 He mentions rules of etiquette as an example. Constitutive rules, on the contrary, would create or define new forms of behaviour. In a later work Searle discusses constitutive rules as a means to impose status.Footnote 10 An example would be the status of money which is imposed on pieces of paper. Constitutive rules are then assumed to have the form ‘X counts as Y under circumstances C’; they are what will later be called ‘counts as-rules’.

My aim in this paper goes far beyond Searle’s however. I will argue that all rules are constitutive and that strictly speaking no rule guides behaviour. Norms may be defined as rules that prescribe or prohibit behaviour. My purpose is to separate these two aspects of norms into the rule-aspect of a norm and the normative aspect of a norm. The rule-aspect as such has no normative aspect, while the normative aspect can exist separate from the norm. It is this normative aspect, what I will call a deontic or normative fact, which guides behaviour and which can be followed or violated.

To this purpose I will argue that:

  1. 1.

    Searle’s account of constitutive rules is too limited, and that there are more kinds of constitutive rules than only counts as-rules;

  2. 2.

    regulative rules can very well be interpreted as a kind of constitutive rules, and that therefore Searle’s opposition between regulative rules and constitutive ones is a bogus one.

  3. 3.

    constitutive rules can be interpreted as a kind of constraints on what counts as possible.

The ultimate purpose of the argument is to disconnect the notion of a rule from that of normativity and to emphasise the role of rules as tools by means of which human beings impose ontological structure upon the world.Footnote 11 In my argument I will mostly use legal examples, but I do not think that this bias damages the strength of the overall argument.

2 Three Kinds of Constitutive Rules

The first step in my argument that regulative rules are constitutive is to distinguish between three different kinds of constitutive rules. It is not my intention to argue that these three kinds are the only constitutive rules, but their existence suffices to show that there is no need to assume the existence of regulative rules which are not at the same time constitutive rules. The three kinds of constitutive rules that will be distinguished are dynamic, fact-to-fact, and counts-as rules.

2.1 Dynamic Rules

Dynamic rules create new facts, or modify or take away existing facts, as the consequence of an event. Examples of events to which a dynamic rule attaches consequences are that:

  • John promised Richard to give him € 100, which leads to an obligation for John to pay Richard € 100;

  • Eloise was appointed as chair of the French Parliament, which has as consequence that Eloise has become chair of the French Parliament;

  • a creditor informs his debtor that the latter will not have to repay the money before next year, which has as consequence that the debtor is not under an obligation to repay the money before next year;

  • the legislator derogates a law, which has as consequence that the law does not exist anymore.

Dynamic rules may be conditional, in which case their consequences are only attached to the events under certain conditions. An example is the rule that if it is dark, the occurrence of a car accident obligates the drivers to place a light on the road next to the cars.

In law, some dynamic rules attach the presence of an obligation to the occurrence of a particular kind of event. The obvious examples are the rules of contract and tort law.

2.2 Fact to Fact-Rules

Fact to fact-rules attach a fact to the presence of some other fact. They make that facts of one type (almostFootnote 12) always go together with facts of some other type. Fact to fact-rules are different from dynamic rules because the relation between the connected facts does not involve the lapse of time. An example is the rule that attaches the fact that P is competent to alienate O to the fact that P owns O. For example, if Smith owns Blackacre, she is competent to transfer her property right in this real estate to Jones. This is a relation between two facts—being an owner and having the competence to transfer—which does not involve any change that occurs in time.

Fact to fact-rules may be conditional too. An example is the rule that the mayor of a city is competent to evoke the state of emergency in case of emergencies. This rule attaches the fact that some person has a competence to the fact that this person is the mayor under the condition that there is a state of emergency.

Fact to fact-rules often attach a duty to the presence of some status, such as being a house-owner, being the public prosecutor, or being the mayor of a city. Such rules are also duty-imposing rules.

2.3 Counts as-Rules

Counts as-rules have the following 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. Often, however, the ‘individuals’ are events. For instance, under suitable circumstances, causing a car accident counts as committing a tort, or offering money to another person counts as attempting to bribe an official.

Usually counts-as rules are conditional, meaning that individuals of type 1 only count as individuals of type 2 if certain conditions are satisfied.Footnote 13 An example from Dutch law 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.Footnote 14

2.4 Constitutive Rules and the Direction of Fit

All three kinds of constitutive rules, dynamic, fact to fact-, and counts as-rules, affect the facts in the world. A dynamic rules generates new facts, modifies existing ones, or takes existing facts away as the result of some event. Fact to fact-rules make that facts of particular kinds go together with other facts in a timeless fashion. This makes that the first-mentioned facts are attached by the fact to fact-rules to the last-mentioned facts. Counts as-rules, finally, make that some kinds of ‘things’, often events, are also ‘things’ of another kind.

Rules have a lot in common with descriptive sentences: they have a propositional content and they can in some sense ‘correspond’ with facts. For example, the rule that criminals are liable to be punished ‘corresponds’ to the fact that criminals are liable to be punished, just as the descriptive sentence ‘Criminals are liable to be punished’ corresponds to this fact. However, in this correspondence lies also a major difference with descriptive sentences. Descriptive sentences are ‘successful’ in the sense of ‘true’ if they match the facts. They have the ‘word-to-world direction of fit.Footnote 15 Constitutive rules are successful in the sense of ‘valid’, if the facts match the rule. With this match I do not mean that the rule is obeyed, but that the content of the rule is reflected in the world.Footnote 16 For instance, the rule that thieves are punishable is reflected in the world if (because of this rule) thieves are punishable. Valid constitutive rules—which I take to be the same as existing constitutive rules—impose themselves on the world. They have the world-to- word direction of fit because they constrain the world in the sense that not all combinations of facts are possible. As a consequence, these rules bring about facts, and in this sense they are constitutive.Footnote 17

3 Regulative Rules

Having argued that there are more constitutive rules than counts as-rules only, the next argument step will lead to the conclusion that regulative rules can well be seen as a special kind of constitutive rules. To make this step, an intermediate conclusion is required, namely that there can be deontic facts.

3.1 Deontic Facts

Sentences that tell one what to do often take the form of declarations. For example, the sentence ‘You ought to go to the supermarket’ looks very much like the sentence ‘You will go to the supermarket’. Is this similarity misleading, covering up a kind of order as a description? Or does the former sentence, if it is true, stand for a fact? Because the description contains a ‘deontic operator’, such as ought, should, must, is to … etc. the kind of fact at stake would then be a ‘deontic fact’.Footnote 18 I will argue for the view that these sentences express deontic facts indeed.

The following argument that there can be deontic facts is unavoidably too brief to convince readers who firmly believe in the gap between is and ought. It makes presuppositions about the role of the mind in structuring the world which I defended elsewhere.Footnote 19 Here my argument will essentially be limited to clearing away two possible misunderstandings.

One misunderstanding is that a sentence such as ‘You ought to go to the supermarket’ is ‘really’ a kind of order or exhortation, rather than the description of a fact, because it is, or can be, used to make somebody do something. Underlying this misunderstanding is the—often implicit—assumption that the world is inert and that (beliefs about) facts in the world cannot guide behaviour unless accompanied by a motivating factor such as a desire. This would be the reason why the fact that somebody ought to do something, or—probably better—that somebody is aware that he ought to do something, guides his behaviour.Footnote 20 Because facts by themselves, without accompanying desire, could not guide behaviour, it could not be a fact that somebody ought to do something.

This first misunderstanding has been attacked by GeachFootnote 21 and SearleFootnote 22 basically because the speech act which can be performed with a sentence does not determine the meaning of the sentence. The sentence ‘You ought to do A’ means the same when it stands on its own as when it is used the conditional sentence ‘If you ought to do A, I will eat my hat’. So if the sentence expresses a fact in its second use, it also expresses a fact in its first use.

The other misunderstanding is that a ‘real’ fact cannot depend on what humans think, believe, project, accept or recognise. On the assumption that standards for goodness and for what should be done are mind-dependent, the misunderstanding becomes that ‘real’ facts cannot depend on standards. This misunderstanding is essentially that of applying an ontological realist stance to all domains, including those in which this is less suitable. One such domain is that of social reality, because social reality depends to a large extent on what people accept or recognise about it.Footnote 23 If there are some domains of facts (e.g. social reality) for which an ontological realist stance is not appropriate, there is no decisive reason why other domains could not be mind-dependent too. This opens up the possibility to recognise the existence facts that depend on standards, including facts that depend for their existence on rules. In law, the existence of such facts is assumed in a routinely fashion, for instance when it is said that (it is a fact that) somebody is the owner of a good. If facts can depend on rules, there is no good reason why there cannot be deontic facts.

Because normative judgments can very well, and often are, expressed by means of declarative sentences, apparently there are deontic facts which are expressed by these sentences. The two objections do not suffice to take the force of this appearance away.

3.2 Rules, Facts and Descriptions

A regulative rule is a rule that guides behaviour, either by prescribing or prohibiting particular kinds of behaviour or by indicating what should be done, or how something should be done, without imposing a duty or an obligation to do so.

Examples would be the rules:

  • It is forbidden to torture sentient beings.

  • Car drivers must drive on the right hand side of the road.

  • If the king is in chess, the threat should immediately be removed.

However, these sentences can not only be interpreted as rule formulations, but also as descriptive sentences which, if they are true, express deontic facts. Moreover, if these sentences are true, the most likely explanation is that this is so because the rules with the same formulation brought about the facts which made the sentences true. In schema (Fig. 2.1):

Fig. 2.1
figure 1

Rules, facts and descriptions

Full size image

Let us take a closer look at this mechanism according to which rules lead to facts which can be described by sentences that are identical to the rule formulations. To show that this has everything to do with the relation between constitutive rules and facts, and nothing in particular with regulative rules as such, I will start with a competence conferring rule. An example would be the fact to fact-rule that owners of real estate are competent to mortgage this real estate. Then, if Smith owns Blackacre, the rule makes that Smith is competent to mortgage it. The rule applies only to individual cases and makes that in all these cases the owners of real estate have the competence to mortgage it. Indirectly, via these individual cases, the rule also makes that (all) owners of real estate are competent to mortgage this real estate. And this general fact is truly described by the sentence ‘Owners of real estate are competent to mortgage this real estate’.

The same line of reasoning can—almost literally—be applied to rules that lead to deontic facts. An example would be the fact to fact-rule ‘Car drivers should drive on the right’. Let us assume that this is an existing (valid) rule. Then, if Schmidt drives a car, the rule makes that Schmidt should drive on the right. The rule is applicable to all persons who drive cars and imposes on all these persons the duty to drive on the right. The rule applies to all individual drivers and makes that they should drive on the right. As a result, car drivers should drive on the right, and this general fact is truly described by the sentence ‘Car drivers should drive on the right’.

A similar line or reasoning can be used in the case of dynamic obligation imposing rules. An example would be the dynamic rule that those who promise somebody else to do something are from then on under an obligation towards this other person to do what was promised. If this is a valid rule, then, if Smith promised Jones to pay her € 500, the rule makes that Smith is under an obligation towards Jones to pay her € 500. The rule is applicable to all individual persons who made a promise to do something and imposes on all these individuals the obligation to do what they promised. That all these individuals are under this obligation is truly described by the sentence ‘Those who promise somebody else to do something are from then on under an obligation towards this other person to do what was promised’.

3.3 Regulative Rules Are also Constitutive Rules

I have discussed two examples of rules that constitute deontic facts, primarily duties or obligations for individual persons, but in a derived sense also for categories of persons. These two examples represent two kinds of regulative rules, namely fact to fact-rules which impose duties on categories of persons, and dynamic rules which impose obligations on persons as the result of events. There are still other categories of regulative rules, for instance the rule of etiquette, which do not really prescribe behaviour, but which still tell agents what to do.Footnote 24 Another example is the rule that if the king is in chess, the threat to the king should be removed immediately. These regulative rules are also constitutive in the sense that they make that agents should do something. Such should-facts exist in the same manner as duties and obligations, and if the latter two categories of facts are constituted by rules, so is the former. Therefore these non-mandatory regulative rules are no counter-examples against the general thesis that regulative rules are also constitutive rules. I submit here the thesis that all regulative rules are also constitutive rules, and that they only differ from other constitutive rules in that they constitute deontic facts, rather than other kinds of facts. This is not a sufficient reason to make a special category out of them.

4 Rules as Soft Constraints on Possible Worlds

If, as argued above, regulative rules are a subcategory of constitutive rules, one may wonder whether all rules are constitutive ones. A first reason to assume that this is indeed the case is that as yet no other kinds of rules than regulative and constitutive ones have been identified. If all these rules are constitutive, then at least it seems that all rules are constitutive. However, it is also possible to give a positive account of the nature of rules which leads to the conclusion that all rules are constitutive. According to this account, rules are a kind of ‘constraints’ on possible worlds, and more in particular a special kind of constraints, namely ‘soft’ ones.Footnote 25

Possible worlds play an important role in model-theoretic semantics, an important part of modern logic. Therefore it is important to understand what these possible worlds are. The best technical approximation of logically possible worlds is the interpretation function that explains how truth values are assigned to compound sentences on the basis of the truth values of atomic sentences and to atomic sentences on the basis of individuals being elements of sets.Footnote 26 The nature of possible worlds is seldom discussed in an informal manner, however.Footnote 27 It may therefore be useful to say a little more about possible worlds, and to explain the role of constraints in defining which worlds count as possible ones.

4.1 States of Affairs, Possible Worlds and Constraints

For the purpose of this article, the terms ‘possible world’ and ‘state of affairs’ will be defined as follows:

A state of affairs is what is expressed by a declarative sentence. Notice that it is not necessary that the sentence is true. States of affairs are merely ‘potential facts’. True sentences express facts, a subset of the states of affairs.

A possible world is a complete set of compatible states of affairs. Both completeness and compatibility are defined below.

Logic cannot determine which states of affairs are compatible, because logic presupposes the notion of (logical) compatibility rather than defining it. Which states of affairs are deemed compatible is not something that is ‘objectively’ givenFootnote 28 but depends on the constraints that are imposed on the world. Constraints determine which states of affairs can go together in a possible word. It is possible that it rains and the sun is shining at the same time, but not that John is both a thief and not a thief.

Compatibility of states of affairs is by definition relative to a set of one or more constraints. The states of affairs that John is a thief and that he is not a thief are incompatible because of the constraints that a state of affairs cannot both exist and not. Another constraint is that the compound state of affairs that John is both a thief and a minor can only exist if both the states of affairs that John is a thief and that he is a minor exist. Such constraints are usually called logical constraints.

Besides logical constraints there are also other constraints. Physical constraints prevent a piece of metal being heated without expanding. Conceptual or semantic constraints make it impossible that something is both a square and a circle.

What is possible depends on the constraints that are taken into account. This brings us back to the notion of a possible world. A possible world is a set of states of affairs that are compatible relative to some set of constraints C, in the sense that the facts of that world satisfy the constraints in C. For instance, a logically possible world does not contain both the fact that the capital of Belgium is Brussels and the fact that the capital of Belgium is not Brussels.

Since a set of constraints will usually not determine all the states of affairs in a possible world, every set of constraints defines a set of worlds that are possible relative to this set. For instance the set of logically possible worlds may contain one world in which Brussels is the capital of Belgium, and another possible world in which Belgium does not even exist. But, relative to a plausible set of constraints, there is no possible world that contains both facts.

If the traditional constraints of propositional logic are taken into account, a possible world cannot contain both states of affairs ‘Snow is white’ and ‘Snow is not white’. If physical constraints are taken into account, no possible world will contain the states of affairs that a piece of metal is heated without expanding. But if only proposition-logical and physical constraints are taken into account, some possible world may contain a square which is at the same time a circle. To rule out that latter possibility, also some conceptual or semantic constraints need to be taken into account

Different sets of constraints may lead to different sets of possible worlds, and there is not one single set of possible worlds. So there is the set of worlds that are both logically and conceptually possible, and this set may include a world that is physically impossible. Another set includes the worlds that are physically and conceptually possible, but not necessarily logically possible. And so on …

There are not only constraints on the states of affairs that can exist together simultaneously. Many natural laws, for instance, operate in time and make that some things must happen after something else happened. These constraints confine which possible worlds can follow after a particular possible world in time.

Not any set of states of affairs that satisfies a set of constraints c is a possible world relative to c; the set must also be complete. Intuitively completeness means that a possible world determines for every sentence whether it is true or false.Footnote 29 This idea can be implemented by demanding that it is not possible to add any state of affairs to the possible world without violating a constraint on that world. For instance, if a possible world contains the states of affairs that John is a criminal and that the legal rule that criminals are punishable is valid, then it is possible (and—if the world is to be legally possible—even required) to add the state of affairs that John is punishable, but not to add the state of affairs that John is not punishable.

4.2 Contraints and Directions of Fit

The formulation of a constraint is very similar to a declarative sentence. For instance, one physical constraint might be that all pieces of metal that are heated expand, and a mathematical constraint would be that there is exactly one line parallel to some line L, which is through a point P which does not lie on L. However, there is a difference in the direction of fit between declarative sentences and constraints. Declarative sentences, even if they deal with what is possible, are true or false depending on the states of affairs that exist in some particular possible world, or—if they aim to describe a regularity—in the set of all worlds that are possible relative to some set of constraints.Footnote 30

Constraints, even though they can be formulated in language, are not linguistic utterances, let alone declarative sentences. The existence of ‘hard’ constraints, about which we are presently talking, does not depend on the facts in a possible world, or the facts in all possible worlds. It is rather the other way round: a world only counts as possible if it satisfies the constraint, if the facts in this world are as the constraint says they are. The constraint in that sense ‘imposes itself’ on the world.Footnote 31

4.3 Constraints and Conditionals

Because of their nature, constraints support conditional and even counterfactual sentences. An example of a conditional sentence would be: ‘If John has committed theft, he is punishable’. This sentence tells us what is the case in the hypothetical situation that John has committed theft, without also informing us whether this hypothetical situation is actually the case. Metaphorically speaking, this sentence tells us what is the case in a possible world in which John has committed theft. Or rather, it tells us that John is punishable in all possible worlds in which John has committed theft, because it does not mention any other conditions for the punishability of John. The conditional sentence ‘If John has committed theft, he is punishable’ might be reformulated in possible world terminology as ‘In all possible worlds in which John has committed theft, he is punishable’. These possible worlds may include the actual world, but the sentence does not inform us whether this the case.

A counterfactual sentence such as ‘Even if Jane would have thrown a brick at the window, the window would still not have broken’ is very much like a conditional sentence, informing us what is the case in all possible worlds in which Jane has thrown a brick at the window. But it also tells us that Jane did not throw a brick at the window in reality, that is in the actual world.

Constraints on possible worlds are the reason why conditional sentences, including counterfactuals, are true or false. If all possible worlds would be constrained in such a way that if they contained the state of affairs that Jane threw a brick to the window, they must also contain the state of affairs that the window did not breakFootnote 32, then the counterfactual sentence ‘Even if Jane would have thrown a brick at the window, the window would still not have broken’ would for that reason be true. If all possible worlds would be constrained in such a way that if they contain the state of affairs that John has committed theft, they must also contain the state of affairs that John is punishable, then for this reason the conditional sentence ‘If John has committed theft, he is punishable’ is true.

Constraints also make that some facts are necessarily the case, or impossible. The constraint that if something is a square, it must have four straight angles makes that it is necessarily (in all possible worlds) the case that all squares have at least three straight angles.Footnote 33 The constraint that pieces of metal expand if heated makes it impossible that a heated piece of metal does not expand.Footnote 34 Analogously, a time constraint makes it impossible that a train arrives before it departed.

4.4 Rules as Soft Constraints

The reader may have noticed that the example about John who committed theft was based on a legal rule, and that this illustrates that rules can be seen as constraints. And, indeed, rules have a lot in common with more traditional constraints such as the logical and physical ones. They make that some things are necessarily the case (e.g. that thieves are punishable), and they support conditionals (If Jane bought this car, she owns it now) and counterfactuals (If the witness would not have lied, he would not have been punished for perjury).

However, many rules can be created or derogated, and in that sense they differ from the more traditional constraints which somehow seem outside the scope of human manipulation. As a consequence, there can be some logically and physically possible worlds in which a particular rule exists, and other similarly possible worlds in which the same rule does not exist.

In the world in which a rule exists, the rule functions as a constraint next to the other constraints to which this world is subjected. So if w1 is a world that is logically and physically possible and this world contains the rule that thieves are punishable, then in this world thieves are punishable, not merely as a contingent matter of fact (e.g. because all thieves happen to own a gun, which is punishable), but necessarily, because being a thief makes one punishable. Moreover, in this possible world the counterfactual state of affairs ‘If John had been a thief, he would have been punishable’ exists, and it exists because the rule is valid (exists). However, if the same rule does not exist in world w2, which is also logically and physically possible, thieves in w2 would not be necessarily punishable.

Apparently, rules are like other constraints, because they impose themselves upon the world and because they make that some facts necessarily obtain and that counterfactual states of affairs exist. But they differ from other constraints in that their existence is contingent. They can be created, and then the world is subjected to new constraints. They can be derogated, with as effect that existing constraints disappear again. For this reason, rules will be categorised as ‘soft constraints’, opposed to the hard constraints which do not depend for their existence on human decisions.

If rules are constraints on possible worlds, this explains their constitutive nature. The facts that are constituted by the application of a rule are the facts that must be present if the world is to count as a possible world. It also explains why all rules are constitutive, on the assumption that all rules function as constraints on possible worlds.

5 Logic for Rules

In this final section the focus will be on some logical implications of the idea that rules are soft constraints on possible worlds.

5.1 No Derivation of Rules

Rules have formulations which make them look like declarative sentences, but—amongst others—the fact that rules have the world-to-word direction of fit, while declarative sentences have the word-to-world direction of fit makes that rules are not declarative sentences. Moreover, there is much to be said for the view that rules are, from a logical point of view, individuals. Rules can be the topic of declarative sentences which are not part of a meta-language. Examples are statements that inform us how long a rule exists, when it was created and by whom, how popular it is, and—for logical purposes important (see the next subsection)—whether it exists (is valid).

Therefore rules cannot be the conclusions of arguments in the traditional sense, and it is not possible to derive rules from (declarative sentences and) other rules. Notice, by the way, that this has nothing to do with the supposedly ‘normative’ nature of rules, and not even with the fact that rules are constraints on possible worlds. However, the fact that rules are from a logical point of view individuals both makes it impossible to use rules as premises or conclusions in arguments, and makes it possible to treat them as soft constraints on possible worlds.

5.2 The Logic of Rule Application

Arguably a logic of rule application is possible. Take for instance the following rule applying argument:

  • Thieves ought to be punished

  • John is a thief

  • Therefore: John ought to be punished.

There is nothing wrong with this argument, if the first premise is interpreted as a declarative sentence about all thieves. But what if the first premise is interpreted as a rule? Then the argument is not deductively valid. Not only because rule-applying arguments tend to be defeasible (which they are), but first and foremost because rules are logical individuals and can for that reason not function as premises or conclusions of deductively valid arguments. (Only sentences with truth values can.)

The problem that rules are logical individuals and can therefore not occur in deductively valid arguments can to some extent be overcome by replacing the rule formulation as premise by the statement that the rule with this formulation exists (is valid). The argument then becomes:

Valid ([Thieves ought to be punished]).

John is a thief

Therefore: John ought to be punished.

This argument is not deductively valid either, but that problem can be worked around by a logic that has as an axiom (and allows as a well-formed sentence):

Valid ([Rule-formulation]) → ⃞ Rule-formulation.Footnote 35

The use of this axiom is justified on the assumption that rules function as constraints on possible worlds. If a rule is such a constraint, the combination of states of affairs as presented by the conditions and the conclusion of the rule exists in all possible worlds. This means that the sentence that describes this combination is true in all possible worlds, or—to state it differently—necessarily true.

However, this sentence is not true in all possible worlds, because a rule in the actual world does not affect possible worlds in which the rule does not exist. Here the difference between hard and soft constraints plays a role, because hard constraints affect all worlds which are possible relative to these hard constraints (e.g. all worlds that are logically and physically possible).

The axiom ‘Valid([Rule-formulation]) → ⃞ Rule-formulation’ for a logic of rules can deal with this soft nature of the constraint if the ⃞ -operator is interpreted as expressing that in all possible worlds in which the rule [Rule-formulation] exists, the sentence ‘Rule-formulation’ is true.Footnote 36

On a more general level it should be noted that the above axiom nicely indicates the status of rules as soft constraints. The effects of hard constraints can be represented as sentences which are true in all worlds that are possible relative to these constraints, and in that sense necessarily true. The sentence that describes the effects of a rule is only necessarily true on the condition that the rule exists.

Conclusion

In this paper the thesis was argued that regulative rules are a subcategory of constitutive rules. In order to reach that conclusion, it was first shown that there are more constitutive rules than counts as-rules only. In particular fact to fact-rules and dynamic rules turned out to be important kinds of constitutive rules too. The second step was to argue that there can be ‘deontic facts’, facts that specify what should be done, and which can therefore take the behaviour guiding function that is often ascribed to rules. The third step was to show that mandatory and should-rules are both constitutive rules. On the assumption that all regulative rules are like mandatory rules or should-rules that conclusion leads to the further conclusion that regulative rules are a subcategory of constitutive rules.

If regulative rules are a kind of constitutive rules, this makes the latter category of rules even more important. Therefore the question needs to be addressed how constitutive rules relate to facts and to the sentences that describe these facts. That is the topic of the second part of paper, which discusses how constitutive rules can fruitfully be seen as ‘soft’ constraints on possible worlds. To this purpose the ideas of possible worlds and constraints on them were discussed. Building upon that discussion, constitutive rules are posited as ‘soft constraints’ in between the constraints that define sets of possible worlds and the declarative sentences that are contingently true. Moreover, from the result that rules are soft constraints on possible worlds and are therefore from a logical point of view individuals, the conclusion was drawn that rules themselves cannot occur as premises in deductively valid arguments because they are not sentences with truth values. A second conclusion was that statements about the validity of rules can function as a proxy, in particular if the axiom ‘Valid ([Rule-formulation]) → ⃞ Rule-formulation’ is added to the logic.