Judicial Discretion as a Result of Systemic Indeterminacy

The topic of ‘judicial discretion’ has been at the center of the debate on legal interpretation in the philosophy of law.1 In a general sense, ‘discretion’ here refers to the exercise of a judgment by a decision-maker due to the lack of legal constraints affecting one’s ability to decide a case. The most fundamental question on this topic is ‘do judges have discretion when interpreting the law?’ There are three kinds of answers to this query. One kind of answer states that judges never have discretion.2 Another kind of answer states that judges always have discretion in interpretation.3 The third kind of answer states that judges sometimes have discretion when interpreting the law, and sometimes they do not.4


Introduction
The topic of 'judicial discretion' has been at the center of the debate on legal interpretation in the philosophy of law. 1 In a general sense, 'discretion' here refers to the exercise of a judgment by a decision-maker due to the lack of legal constraints affecting one's ability to decide a case. The most fundamental question on this topic is 'do judges have discretion when interpreting the law?' There are three kinds of answers to this query. One kind of answer states that judges never have discretion. 2 Another kind of answer states that judges always have discretion in interpretation. 3 The third kind of answer states that judges sometimes have discretion when interpreting the law, and sometimes they do not. 4 A common feature of all the aforementioned answers is that judicial discretion presupposes the fact that interpretative problems in law can yield a plurality of lawful outcomes. Since law provides a multiplicity of interpretative results for a given case, judges, it is said, have the discretion to choose how law will be interpreted for its application.
The issue of the plurality of legal outcomes has been studied within a broader issue in the legal theory termed, 'legal indeterminacy'. Legal indeterminacy is the multiplicity of norms that can be ascribed ex ante to a norm formulation. A norm formulation is a positivized legal rule. A norm is the meaning ascribed to a norm not have redundant rules, a plurality of interpretative directives will yield a plurality of interpretative results. On the other hand, due to the non-liquet rule, judges are obligated to choose among the different interpretative results provided by the interpretative code. By building a thesis on legal indeterminacy as the consequence of having a plurality of interpretative directives that necessarily yield different results, I proceed to provide an account of discretion as a necessary feature of legal adjudication.
My argument is stated as follows: a) a typical rational legal system has an interpretative code with more than one interpretative directive; b) a typical rational legal system contains the non-liquet rule; c) necessarily, if a rational legal system has an interpretative code with more than one interpretative directive and it contains the non-liquet rule, then in this system, legal interpretation is discretionary; d) thus, in a typical rational legal system, legal interpretation is discretionary. This paper is structured as follows. In §2, I stipulate the definitions of the central concepts of my paper. In §3, I provide an overview of both the treatment of legal indeterminacy by the scholarship and my account of legal indeterminacy. In §4, I provide reasons for accepting premises a) and b) of my argument. In §5, I provide a reason for accepting premise c) of my argument. In §6, I finish with the conclusions.
Legal interpretation is the ascription of meaning to a text originated by a source of law in accordance with certain interpretative guidelines (e.g., the norm formulation 'T' means 'S'). 7 Judicial discretion is the exercise of judgment performed by a legal decisionmaker in making a choice among a closed list of alternatives, in which there is no prevalence of one alternative over another. 8 This distinction is in the following order: 1) having discretion; and 2) exercising discretion. On the one side, 'having discretion' is the possibility of choice. On the other side, 'exercising discretion' is making a choice. Exercising discretion entails having discretion but not the other way around.
Norm formulations are authoritative legal texts which originated from the sources of law of a given community that have not yet been interpreted. One example is the article 6n°3 of the Consolidated Version of the Treaty on 7. I will use both the expressions, 'legal interpretation' and 'interpretation' interchangeably. 8. I will use both the expressions, 'judicial discretion' and 'discretion' interchangeably.
European Union states, "fundamental rights, as guaranteed by the European Convention for the Protection of Human Rights and Fundamental Freedoms and as they result from the constitutional traditions common to the Member States, shall constitute general principles of the Union's law". 9 Norms are the meaning that the interpreter has ascribed to a norm formulation (i.e., the end result of the process of legal interpretation). 10 For example, article 6n°3 of the Consolidated Version of the Treaty on European Union means that the right to free movement of European citizens within the EU is a general principle of the Union's law.
Legal indeterminacy is the multiplicity of norms that can be ascribed ex ante to a norm formulation.
Interpretative directive is a second-order norm formulation that prescribes how judges ought to interpret a first-order norm formulation. Examples of these interpretative directives include, the plain meaning rule, the history of the legislature rule, and the intention of the legislature rule.
Interpretative code is the set of interpretative directives, techniques of reasoning (analogical reasoning, deductive reasoning, etc.), techniques of argumentation (a fortiori; a pari ratione; a maiori ad minus; a minori ad maius, etc.), and, although they do not address issues of interpretation, meta-rules that solve conflicts between rules (lex posteriori, lex specialis, legal hierarchies), contained in a legal system to guide and constrain the interpretation of first-order norm formulations. 11 A typical legal system is one that shares features commonly ascribed to Western legal systems. These features are, among others, upholding of the rule of law, the interdiction of arbitrariness in adjudication, the justified exercise of coercion by the state, the principle of legal certainty, solution of conflicts, the enforcement and protection of rights, and the supremacy of the constitution. A rational legal system is one that has formal/logical properties, such as the avoidance of antinomies, gaps, and redundancies. 12

Mapping the Scholarship
There are two prevailing accounts of legal indeterminacy within legal philosophy. As I stated in §1, I call these two accounts 'the realist account' of legal

372
Reyes Molina indeterminacy and the 'natural language' account of legal indeterminacy, respectively. Both accounts are not mutually exclusive. Arguably, legal systems are often affected by both kinds of indeterminacy. While the realist account claims that legal indeterminacy is the result of the multiplicity of legal materials, the natural language account claims that legal indeterminacy is the result of the indeterminacy of natural languages. To the extent that the legal materials are formulated in natural language, they are also subject to linguistic indeterminacy. In this section, I will present an overview of these two accounts with the purpose of situating my claim within the scholarly debate on legal indeterminacy.

The 'Realist Account' of Legal Indeterminacy
Before I describe the 'realist account', a comment is in order. I do not claim that there is an intrinsic connection between American legal realism and legal indeterminacy. This connection is, as far as I see, purely historical. It is a matter of fact that this kind of claim about the indeterminacy of law has been held by some authors that are part of the American legal realist movement.
As stated in §1, the realist account holds that law is indeterminate due to the multiplicity of legal materials that yield competing but equally legally valid results. By 'legally valid' I mean that the decision produces legal consequences. 13 As Kress notices, "[i]n adjudication, law is indeterminate to the extent that authoritative legal materials and methods permit multiple outcomes to lawsuits." 14 According to Llewellyn, [o]ne does not progress far into legal life without learning that there is no single right and accurate way of reading one case, or of reading a bunch of cases : : : [i]n the work of a single opinion-day I have observed 26 different, describable ways in which one of our best state courts handled its own prior cases, repeatedly using three to six different ways within a single opinion. What is important is that all 26 ways (plus a dozen others which happened not to be in use that day) are correct. 15 The realist account explains legal indeterminacy by resorting to an unclear set of factors that include, inter alia, the personal traits of judges, value-preferences of the interpreter, legal materials (which includes scholarly work), and methodologies used for interpretation. It is the interplay of the aforementioned factors that yield competing but equally legally valid results. This interplay depends mostly on the personality of the judge. As Frank points out, 13. As the reader may have noticed, the expression 'legally valid' used here is different from the three notions of 'validity' advanced by Kelsen. That is, validity as: 1) existence; 2) binding force; and 3) efficacy. For an analysis of the three Kelsenian notions of validity, see Bulygin, supra note 12 at 81-83. 14. Ken Kress, "Legal Indeterminacy" (1989) 77 Cal L Rev at 283. 15. Llewellyn, supra note 5 at 395.
[t]hose who today complain of any 'judicial legislation' in statutory interpretation are complaining of the intrusion of the judges' personalities. However, : : : legal thinkers, in increasing numbers, have shown that the personal element in statutory construction is unavoidable : : : [t]he creativeness of the judges should always be limited; but, within proper limits, it is a boon not an evil. 16 If, according to the realist account, the judges' personality is 'unavoidable' when interpreting law, then legal outcomes will vary according to the personality of the judge who decides the case. In other words, the content of the law is insufficient to ensure a single legal answer. This point has been stressed by Leiter in his reconstruction of the realist account. For him, problems of indeterminacy are problems that are associated with what he calls, 'the class of legal reasons'. The class of legal reasons is comprised of the sources of law, the interpretative techniques of a given legal system, the legitimate way of deciding over the quaestio facti, and the techniques of reasoning (e.g., analogical reasoning, a fortiori arguments, and deductive reasoning). 17 He claims that the class of legal reasons is indeterminate and insufficient to both cause the judge to reach only one outcome and to justify only one outcome. 18 In other words, for the realist account, what causes and justifies a legal outcome is derived from both legal and extralegal factors. That law is not enough to cause or justify a legal outcome does not mean that we cannot predict the courts' decisions. The realist account leaves open the possibility to predict legal outcomes on the basis of discernible patterns that a specific court/judge will follow when deciding similar cases. As Leiter points out, "[j]udicial decisions : : : fall into discernible patterns (making prediction possible), though the patterns are not those one would expect from the existing legal rules. Rather, the decisions fall into patterns correlated with underlying factual scenarios of the dispute at issue". 19 For example, suppose that two people come to the lawyer's office saying that they were wrongfully terminated from their job. They want to sue their employer, but they are unsure whether the grounds for the lawsuit are strong enough to ensure a favourable result. Following the realist account, the best way to provide satisfactory counseling is knowing which judge could potentially decide the case. Once the judge is identified, then the lawyer will base the legal advice on knowledge about the political allegiances, religious beliefs, and moral commitments of the potential judge. The lawyer's knowledge, in turn, will be based on the historical records of the judge's decisions on similar cases. The task of the lawyer, then, is to identify that judge's decision-patterns in similar cases to the one that requires their services, as well as the different factors that inform said patterns.  Reyes Molina

The 'Natural Language' Account of Legal Indeterminacy
The natural language account holds that certain features of natural languages bring about legal indeterminacy. 20 These features of natural language are ambiguity and vagueness. On the one hand, ambiguity in law arises when a norm formulation can be interpreted in different ways. On the other hand, vagueness in law manifests as hard cases. These are borderline cases in which it is indeterminate if a specific state of affairs can be subsumed in the class of state of affairs that a norm formulation governs. In this subsection, I focus on two versions of the natural language account. I call these versions 'the double-indeterminacy' thesis and 'the open texture' thesis respectively. As I will demonstrate, these two versions are related in that the double indeterminacy version entails the open texture version.
The double-indeterminacy version is held by Guastini. 21 This version claims that law is ambiguous and vague. 22 However, ambiguity and vagueness are to be found in different stages within the process of interpretation. Guastini points out that the term 'interpretation' is ambiguous and it can refer to, at least two kinds of activities that the judge engages in when deciding a case: 1) interpretation in abstracto; and 2) interpretation in concreto. 23 Interpretation in abstracto is the task of ascribing meaning to a legal text (e.g., the legal rule 'R' means 'T'). This kind of interpretation deals with ambiguity. Under the double indeterminacy thesis, ambiguity refers to the multiplicity of norms that can be ascribed to one norm formulation. 24 In other words, prior to the act of interpretation, a plurality of interpretations can be ascribed to a norm formulation, thereby rendering them indeterminate (e.g., the legal rule 'R' can mean 'T 1 ' , 'T 2 ' , 'T 3 ', and so on). 24. This way of understanding ambiguity differs from the way that philosophers have understood it. Although I will not develop the different kinds of ambiguity, suffice it to say that the common understanding of this phenomenon refers to the property enjoyed by signs that bear a plurality of interpretations. However, in a more specific way, ambiguity refers to two or more lexical entries corresponding to the same word that can or cannot be related. Thus, for example, a bank can mean both a financial institution and a riverside. This is not the way that Guastini understands the ambiguity of norm formulations. It is not the case that norm formulations have different unrelated meanings, but that each norm formulation can be reworded in a different, but somewhat related, way in light of interpretative directives used by the interpreter. In this sense, issues of legal interpretation are more similar to the phenomenon of polysemy, in which the different meanings ascribed to a lexeme are related, than to ambiguity.
Interpretation in concreto is the subsumption of a case within the class of cases that an interpreted in abstracto legal provision governs. This kind of interpretation deals with vagueness. Sometimes it is indeterminate if the case at hand belongs to the class of cases governed by the meaning ascribed to the legal rule by the interpreter (e.g., if there is a norm formulation that prohibits the entrance of animals to government buildings, do service animals count as an animal for the purpose of this rule?). The cases that interpretation in concreto covers bear resemblance to the easy cases/hard cases distinction with which legal philosophers are acquainted. The second version of the natural language account is the open texture thesis promoted within the legal scholarship by Hart. 25 The open texture thesis states that all rules have a core of settled meanings and a penumbra zone where its application is indeterminate. 26 Since law is a manifestation of natural language, legal rules will suffer from open texture. According to Hart, [t]here must be a core of settled meaning, but there will be, as well, a penumbra of debatable cases in which words are neither obviously applicable nor obviously ruled out. These cases will each have some features in common with the standard case; they will lack others or be accompanied by features not present in the standard case. 27 He argues that because norm formulations are posited in general terms aimed at regulating particular cases, there will be both cases that will be covered by the plain meaning of the norm formulation, and cases that will fall into its penumbra, zone rendering its application indeterminate. As Hart notes, [a]ll rules involve recognizing or classifying particular cases as instances of general terms, and in the case of everything which we are prepared to call a rule it is possible to distinguish central clear cases, where it certainly applies and others where there are reasons for both asserting and denying that it applies. Nothing can eliminate this duality of a core of certainty and penumbra of doubt when we are engaged in 25. HLA Hart, "Positivism and the Separation of Law and Morals" (1958)  thesis stresses the fact that law qua manifestation of natural language is vague and in borderline cases, law is necessarily indeterminate. Note that the double indeterminacy thesis entails the open texture thesis. In the next subsection, I will provide a general overview of the account that I advance in order to localize it within the scholarly debate.

On the 'Systemic Indeterminacy' Account
In this subsection, I will advance a revised version of the realist account that I call 'systemic indeterminacy'. This account holds that the plurality of legal outcomes is the consequence of a typical rational legal system having an interpretative code with a multiplicity of interpretative directives. Since the source of the plurality of interpretative results is the plurality of interpretative directives, the systemic indeterminacy entails the realist account but not the other way around. This is because the interpretative directives are a subset of what the realist account refers to as 'legal materials'. Since I identified the source of indeterminacy with a specific kind of legal rule (i.e., the interpretative directives), my account is narrower than the realist account. The systemic indeterminacy account places emphasis on structural features of legal systems (e.g., interpretative codes), instead of features of natural languages and/or extra-legal factors, such as the personality of judges, value preferences of the interpreter, legal education, and so on. To the extent that the interpretative code is posited in natural language, my account does not exclude the natural language account.
The systemic indeterminacy thesis stresses the relationship between the number of interpretative directives and the number of interpretative results. By this, I mean that a legal system will have the same number of interpretative results as the number of interpretative directives that are contained in the interpretative code. As will be developed further in §5, because typical rational legal systems do not allow for redundant norms-what I call the non-redundancy clausenecessarily different interpretative directives will provide different interpretative results. Roughly put, the connection between interpretative codes and the nonredundancy feature of typical rational legal systems is the following: since typical 28. Hart, supra note 4 at 123. 29. Schauer, supra note 24 at 200. rational legal systems do not allow for redundant rules necessarily each interpretative directive provides a different interpretative result.
It is not surprising that the fact that interpretive codes would contribute to legal indeterminacy could seem counterintuitive. After all, it is a common assumption that the deficiencies of the semantic content of language call for the need of interpretative directives that allow the deficiencies to be addressed. In other words, typical rational legal systems try to reduce the indeterminacy of natural language that affects law through a set of second-order rules known as interpretative directives/canons of interpretation. However, the interpretative code cannot complete eliminate indeterminacy. Interpretative codes cannot eliminate indeterminacy because the interpretative directives are also affected by the indeterminacy of natural language. As Hart points out, "[c]anons of 'interpretation' cannot eliminate, though they can diminish, these uncertainties; for these canons are themselves general rules for the use of language, and make use of general terms which themselves require interpretation". 30 Hart's claim relies on the natural language account of legal indeterminacy. He stresses the fact that the canons of interpretation are affected by the open texture of the words and are thus susceptible to vagueness themselves. Therefore, in cases in which the canons of interpretation are in themselves vague, they will fail to reduce indeterminacy of law.
The systemic indeterminacy account does not rely on the open texture of the interpretative code to hold that the canons of interpretation contribute to legal indeterminacy. Under my account, the plurality of legal outcomes is caused by a multiplicity of interpretative directives, independent of their semantic content. 31 In §1, I stated that there is a relationship between legal indeterminacy and discretion. In the next section, I will present an overview of this relationship in the three accounts of legal indeterminacy that I have laid down.

On the Relationship between Legal Indeterminacy and Judicial Discretion
The existence of a relationship between legal indeterminacy and discretion is a point that all three accounts of legal indeterminacy share. In all these accounts, discretion can be formulated as a choice made by the interpreter. Both the content of the choice and its constraints will vary depending on the account that is analyzed. In this subsection, I will provide an overview of judicial discretion and its relation to legal indeterminacy. The purpose for this is twofold: 1) to define this relationship; and 2) to say something about the scope of judicial discretion.
The realist account holds that it is possible to interpret the law in multiple different ways due to the multiplicity of legal materials. This account

378
Reyes Molina acknowledges the fact that judges and lawyers can ascribe and justify competing legal results by using the legal materials available to them. Which legal material will be used and how that legal material will be used will depend on a choice based on criteria that escape the law. In other words, when determining the use of the legal materials, judges exercise discretion. As Leiter points out, "[For realists,] lawyers and judges have this interpretive latitude often enough to inject a considerable degree of indeterminacy into law". 32 Recall the distinction of having discretion compared to exercising discretion. The realist account explains having discretion (i.e., the potential of choosing) based on the multiplicity of legal materials. The exercise of discretion, (i.e., the choice itself) is explained by a complex set of non-legal factors.
The two versions of the natural language account treat discretion in a similar way. The double indeterminacy thesis holds that discretion is exercised in both interpretation in abstracto and interpretation in concreto. When judges interpret in abstracto, they need to choose among the different norms that can be ascribed to the norm formulation, thus putting an end to the ambiguity affecting the uninterpreted norm formulation. When judges interpret in concreto, discretion is exercised in borderline cases (i.e., when it is indeterminate whether the specific case can be subsumed in the general class of cases governed by the norm).
The open texture account claims that in cases that fall into the penumbra zone of a legal rule, judges have the discretion to settle the question if a certain case falls into the core meaning of the norm formulation they want to apply. As Hart states, [w]hen the unenvisaged case does arise, we confront the issues at stake and can then settle the question by choosing between the competing interests in the way which best satisfies us. In doing so we shall have rendered more determinate our initial aim, and shall incidentally have settled a question as to the meaning, for the purposes of this rule, of a general word. 33 From the standpoint of the distinction between having and exercising discretion, according to the natural language approach, judges have discretion because of ambiguity and vagueness. Moreover, judges exercise discretion when deciding issues of ambiguity and vagueness.
The systemic indeterminacy account holds that discretion is the result of having an interpretative code with more than one interpretative directive and the nonliquet rule. 34 The relationship between the non-liquet rule and discretion could be stated as follows: although there are different interpretative results to decide a case, judges are legally obliged to choose the interpretative directive on which they will justify the interpretation of the norm formulation that will govern the case. Since the non-liquet rule prescribes the obligation to decide tout court, this decision refers to a choice among the interpretative directives. Moreover, this 32. Leiter, supra note 16 at 73. 33. Hart, supra note 4 at 129. 34. I develop this in depth in §4 and §5. decision must be taken even if the interpretative code does not contain legal rules that guide the use and/or establish hierarchies among the interpretative directives.
The systemic indeterminacy account claims that judges have discretion because of the multiplicity of interpretative directives contained in the interpretative code. It also claims that judges exercise discretion because of the nonliquet rule. Thus, if judges have discretion, they are legally obligated to exercise it. Whilst it is unclear if having discretion entails its exercise in both the realist account and the natural language account, it is clear that having discretion entails exercising such discretion in the systemic indeterminacy account.
The notion of discretion used in all the accounts of legal indeterminacy analyzed in this subsection can be reconstructed as a choice. The content of the choice will slightly vary from one version to another. In both the realist account and the systemic indeterminacy account, the choice is about the legal materials to be used and interpreted. However, in the natural language account, both versions seem to share that the content of the choice ultimately has to do with settling the 'question of the meaning' of a norm formulation that is either ambiguous and/ or vague.
A point of divergence regarding discretion is that it is possible to identify the issue regarding the scope of discretion in these accounts. Discretion in legal interpretation will be present in borderline cases. This is true for both Hart's 'open texture' account and Guastini's account of the interpretation in concreto. Thus, for these accounts, discretion is a marginal phenomenon in legal interpretation because the majority of cases will fall within the core meaning of the norm formulation. This is not the case for the realist account, the systemic indeterminacy account, and Guastini's account of the interpretation in abstracto. For all these accounts, discretion is present in all litigated cases.
The realist account, the systematic account, and the natural language account help to identify at which stage within the process of adjudication judges will have discretion. In addition, the combination of these different accounts of indeterminacy show that this phenomenon is present in all litigated cases.
In the next section, I will develop the main argument of this paper. This argument consists of three premises and the conclusion. In §4 I provide reasons to accept premise a) and b). Premise c) will be developed in §5.

Main Argument
Recall the main argument of this paper: a) a typical rational legal system has an interpretative code with more than one interpretative directive; b) a typical rational legal system contains the non-liquet rule; c) necessarily, if a rational legal system has an interpretative code with more than one interpretative directive, and it contains the non-liquet rule, then in this system legal interpretation is discretionary; d) thus, in a typical rational legal system legal interpretation is discretionary.

380
Reyes Molina In this section, I will provide reasons to accept premise a) and b). Premise a) is supported by a two-part reason. Premise b) is supported by three independent reasons.

On premise a): Interpretative Codes with Multiple Interpretative Directives
In §4.1.1 and §4.1.2, I provide reasons for the plausibility of the claim that in typical rational legal systems, interpretative codes contain more than one interpretative directive. Premise a) is supported by a two-part reason.

Part 1 Reason for Premise a): Providing Legal Justification for the Interpretation of a Norm Formulation
In §2, it was stated that there are some features of Western legal cultures that are contained in typical legal systems. Two of these features are the upholding of the rule of law and the justified exercise of coercion. In general, these two features are aimed at controlling and limiting the exercise of state power. This aim is fulfilled in different ways. One way is the legal obligation that state authorities have to provide legal justification for their decisions. One method of providing legal justification for their decisions is by having legal grounds for the judge's interpretation of norm formulations. Interpretative codes with more than one interpretative directive provide legal grounds for the interpretation of norm formulations. Conversely, if judges do not interpret the norm formulation based on one of the interpretative directives then that interpretation lacks legal grounds.
Notice that interpreters do not need to use all of the interpretative directives when interpreting a norm formulation. It is sufficient for a norm to be legally valid if the interpreters grounded their interpretation of the norm formulation in only one interpretative directive. Following Spaak, one factor that is relevant to the interpretation of statutes: the literal meaning of the statutory text, consistency and coherence in the legal system, legislative intent, and statutory purpose, respectively. These interpretative arguments [as he calls them] are legal, not moral or political, in the sense that their common starting-point is the law itself. They are also fundamental in the sense that almost any controversy about the content of the major premise of the legal syllogism can be framed in terms of them. 35 A legal system without an interpretative code will have at minimum, a deficit in legal reasons on which the norm ascribed to the norm formulation in which it could be grounded. This is not to say that there would be no reasons at all to justify the interpretative decision; however, it does mean that these reasons will not be legal in origin.
Notice that the reason presented in this subsection argues for accepting the claim that typical legal systems will have interpretative codes. It does not say anything about the fact that these interpretative codes will have a multiplicity of interpretative directives. This second feature (i.e., a multiplicity of interpretative directives) will be addressed in the next subsection.

Part 2 Reason for Premise a): Preventing Arbitrariness and Limit Vagueness and Ambiguity
The second part of the reason for premise a) consists of two parts. The first part holds that interpretative codes with more than one interpretative directive prevent arbitrariness in legal interpretation. Recall that in §2, interdiction of arbitrariness is one of the features of typical legal systems. To both avoid uncertainty and reduce indeterminacy in legal decisions, interpretation of norm formulations is guided by the interpretative codes. Conversely, the lack of interpretative codes will amount to, at minimum, discretion, and at maximum, arbitrariness when interpreting the law. The interpretative codes will provide a set of legal reasons by which the interpreter legally ought to support a decision about which norm will be ascribed to the norm formulation. Since the interpretative directives are finite in number, judges will be constrained regarding the number of reasons upon which they can base their decision. Thus, a finite number of interpretative directives will prevent arbitrariness in adjudication. By providing a method of interpretation, the interpretative codes will prevent arbitrariness in interpretation albeit the matter of discretion is still debatable. A lack of interpretative codes will allow for unknown or extremely unclear methods by which interpreters apply the law.
The second part holds that interpretative codes with more than one interpretative directive limit the vagueness and the ambiguity of norm formulations. In order to reduce the effects of the natural language indeterminacy in legal adjudication, the interpretative codes tend to contain a multiplicity of interpretative directives. Of course, a multiplicity of interpretative directives is not a requirement sine qua non for having an interpretative code. It is possible to imagine a legal system with only one interpretative directive (e.g., the plain meaning of the text rule). Such a legal system would find problems, for instance, whenever a norm formulation that contains vague formulas must be applied or whenever the meaning of a word has changed in a given society, and so on. It seems that in those types of cases, the interpretative code would not be able to achieve its goal of reducing indeterminacy.

On Premise b): The Non-Liquet Rule
The term non-liquet can be translated as 'it is not clear' and it refers to the situation in which a competent court fails or abstains to decide a case either because there is no law that regulates the matter or the existent law is unclear. 36 In law, there is a general prohibition to non-liquet. This prohibition can be expressed as a rule that prohibits judges from not deciding cases. I call this rule, the non-liquet rule. This rule stems from a general principle in law that is known as the principle of completeness. In general, the principle of completeness states that legal systems ought to provide a solution to every case.
The principle of completeness can be subjected to two readings. I call these two readings the strong and the weak reading, since the weak reading implies the strong reading. The strong reading claims that the principle of completeness binds both the legislative and the judiciary power. In other words, both the legislative and the judiciary ought to provide solutions to every case. The way that the legislators comply with the strong reading is by introducing the so-called 'closure' rules. A closure rule is "a rule that would qualify deontically all actions not already qualified by the system in question". 37 An example of a closure rule in criminal law is the nullum poena sine lege rule, which can be formulated roughly as follows: it is prohibited to punish an action that is not prohibited by law. An example of a closure rule in public law is the following: everything that is not allowed is forbidden. These examples illustrate the fact that all actions that have not been deontically qualified by the law in a specific way fall under these closure rules and there are thus, deontically qualified in a general way. In other words, closure rules provide necessary guidance to the judge for determining the solution to cases that encompass actions that are not specifically deontically qualified by the law.
The weak reading claims that the principle of completeness prescribes a duty to the judges to decide every case, even when there are no norm formulations that govern those cases. Thus, the weak reading's claim only binds the judiciary power. For example, article 4 of the French Code civil prescribes that "[a] judge who refuses to give judgment on the pretext of legislation being silent, obscure or insufficient, may be prosecuted for being guilty of a denial of justice". 38 Another example of a non-liquet rule can be found in article 76, paragraph 2 of the Chilean Constitution that prescribes that "[a]fter intervention is requested in legal form and for matters of their competence, [the courts] cannot excuse themselves from exercising their authority, not even in the absence of a law to resolve the dispute or issue submitted to their decision". 39 One of the defining features, if not the defining feature of the courts, is the power to decide disputes. This power is granted by competence norms that serve as a ground for the legal consequences that such decisions bear. However, courts 36. For a historical approach of the non-liquet rule, see Alfredo Mordechai Rabello, "Non Liquet: From Modern Law to Roman Law" (1974) 9 Israel L Rev 63. 37. Alchourrón & Bulygin, supra note 11 at 135. 38. Original text: "Le juge qui refusera de juger, sous prétexte du silence, de l'obscurité ou de l'insuffisance de la loi, pourra être poursuivi comme coupable de déni de justice". 39. Original text: "Reclamada su intervención en forma legal y en negocios de su competencia, no podrán excusarse de ejercer su autoridad, ni aun por falta de ley que resuelva la contienda o asunto sometidos a su decisión". not only have the power to decide but also the legal obligation to do so. This obligation, according to the weak reading of the principle of completeness, is imposed by the non-liquet rule. In other words, judges are bound to decide a legal matter even if there is a lack of norm formulations that govern it. Although it is possible to imagine a legal system without the non-liquet rule, such legal systems will allow for the possibility of Déni de Justice and thus fail to both solve conflicts and protect rights.
In §4.2.1, §4.2.2, and §4.2.3, I provide three reasons to support premise b). Although these reasons are independent of each other, they are all related to the role that legal systems play within a given community.

Reason 1 for Premise b): Law is Supposed to Solve Conflicts
There is a widely shared idea that law is a tool for solving conflicts among members of a community. This idea is popular in private law and procedural law literature. The non-liquet rule is in line with the conflict-solving view. Since it is debatable whether the non-liquet rule is a conceptual requirement of a legal system, one could argue against the idea that legal systems have this type of rule. Thus, someone could argue that it is possible to imagine a legal system without the non-liquet rule in which courts do not decide cases because they think that someone else should decide them. This can be termed, 'jurisdiction objection'. This objection could go as follows: it is possible that courts would consider not having the competence to decide such matters because it falls in the jurisdiction of a different institution. In other words, the court does not have the competence to adjudicate a case. This is something that happens, for instance, in constitutional litigation. Constitutional courts can decide that the matter should be resolved by the legislative power and not by constitutional courts.
Before addressing the 'jurisdiction objection', it is necessary to distinguish two kinds of legal decisions. There are two types of decisions in law: 1) substantive; and 2) adjectival. Substantive decisions are those that provide a solution to a legal case based on the merits of the case (e.g., the decision of awarding compensation for pecuniary loss). Adjectival decisions are those that solve or put an end to procedural matters (e.g., decisions about the legal competence of the court to decide a case).
Having the substantive/adjectival distinction in mind, one can answer the 'jurisdiction objection' along the following lines. This objection is based on a narrow conception of 'legal decision'. This narrow conception assumes that law can only provide substantive answers. Either the action is granted or denied, and thus, it leaves out issues of procedural nature that do not deal with the litigated matter. However, it is possible to argue for a broad conception of 'legal decision' that encompasses the decisions that do not provide a legal solution to the case. The broad conception of legal decision states that the obligation to decide does not ensure specific kinds of decisions. By this, I mean that a decision can be either a substantive or adjectival decision. The latter type would not resolve the matter per se but it would put an end to the case regarding the participation of the specific court that was required to solve the matter at hand. Thus, a decision over issues of legal competence would satisfy the obligation imposed by the non-liquet rule. This does not mean that the case does not have a solution or it would not be solved, but rather that the legal system has a different instance or institution with the competence to do so. For the broad conception of the non-liquet rule, the duty to decide will be fulfilled whenever the legal systems provide an answer, either substantive or adjective, to a case.

Reason 2 for Premise b): No Decision Would Make the Legal System Inapt for Enforcement
Besides solving conflicts, the law, at least how it has been conceived after the Second World War, is a method of ensuring and enforcing the protection of rights. 40 From the standpoint of enforceability of legal rights, the rule of nonliquet provides a safeguard for the enforcement of rights in situations in which there are no norm formulations that can govern the case. The rule of non-liquet would be the legal system's answer to inaction in the face of legal gaps or borderline cases.
A legal system without a non-liquet rule would be ineffective in its task of protecting rights whenever borderline cases or legal gaps arise and the judges decide to remain silent on the matter by not providing a legal solution. Furthermore, law enforcement in a legal system without the non-liquet rule will be dependent on the whim of the judge when deciding both clear and borderline cases. 41

Reason 3 for Premise b): Legal Certainty Requires a Decision
The non-liquet rule functions as a safeguard of the Rule of Law to the extent that it prevents the Déni de Justice and contributes to legal certainty. 42 It does so by assuring that an answer, whatever this answer is, will be given by the legal system whenever a legally relevant conflict arises.
Legal systems ensure that providing answers to cases prevents the undesirable consequence of having conflicts extended over indefinite periods of time. Conversely, the lack of a response from the legal system makes both the goals of ensuring and protecting rights and solving conflicts among people unattainable for cases that judges decide to remain silent on.
A legal system without a non-liquet rule will depend on the judge's whim to decide a case, thus, making legal certainty impossible.
40. On law as a tool for protection of rights see for example, Mauro Barberis, Ética para Juristas (Trotta, 2008); Luigi Ferrajoli, Derecho y Razón: Teoría del garantismo penal (Trotta, 1995). 41. As the reader may notice, I distinguish the notion of 'discretion' from 'whim'. A 'whim' is a sudden desire that lacks justification whilst 'discretion' is a choice among a set of alternatives, in which there is no prevalence that can be justified of one alternative over another.

Conclusion on the Plausibility of the Conditions a) and b)
Typical rational legal systems with interpretative codes, a plurality of interpretative directives, and non-liquet rules seem to accommodate our intuitions about how the law and courts work in a typical society. As I have argued, premises a) and b) are plausible. In the next section, I will provide a reason for accepting premise c). This reason consists of three parts. I call the first part the non-redundancy clause. I call the second part the interpretative negative spaces scenario. I call the third part the obligation to decide.

On Premise c)
Premise c) states that if a rational legal system has an interpretative code with more than one interpretative directive, and it contains the non-liquet rule, then in this system legal interpretation is necessarily discretionary.
In this section, I will provide one reason to support this premise. This reason consists of three parts. The first part focuses on the consequences of having a legal system that is rational. The second part focuses on the lack of rules that govern the use of the interpretative codes. The third part focuses on the interplay between the legal obligation to decide all cases, prescribed by the non-liquet rule, and the lack of legal rules that regulate the use of the multiplicity of interpretative directives contained in the interpretative code.
As I will advance further in the next subsections, premise a) allows for judges to have discretion, and premise b) allows for the legal obligation of judges to exercise discretion.

Part One of the Reason for Accepting Premise c): The Non-Redundancy Clause
The non-redundancy clause states that rational legal systems can have neither redundant norm formulations nor norms. Following Ross, "[r]edundancy occurs when a norm lays down a legal effect which in the same factual conditions is authorized by another norm". 43 If a rational legal system has an interpretative code with more than one interpretative directive, then each interpretative directive will necessarily lead to a different interpretative result. This is because if two or more interpretative directives lead to the same result, then they are redundant and all but one needs to be eliminated from the system. In other words, a plurality of interpretative directives amounts to a plurality of norms ascribable ex ante to a norm formulation, thus making typical rational legal systems indeterminate. For example, suppose that in order to determine the evidential threshold prescribed by the legal system, judges need to interpret the 'beyond a reasonable doubt' standard of proof. They will 43. Ross, supra note 11 at 132. turn to the interpretative code that contains three interpretative directives: the 'plain meaning of the text' rule, the 'intention of the legislator' rule, and the 'history of the legislation' rule. Due to the non-redundancy feature, the interpretative codes cannot provide the same interpretative result; thus, the judge will have three different norms (i.e., meanings) that could be ex ante ascribed to the formula 'beyond a reasonable doubt'. In other words, the 'plain meaning of the text' rule necessarily provides a different interpretative result than the 'intention of the legislator' rule and the 'history of the legislation' rule. Otherwise, these three rules are redundant and all but one ought to be eliminated from the system.
Due to the non-redundancy clause, interpretative codes with more than one interpretative directive will necessarily multiply norms that can, ex ante, be ascribed to norm formulations. Thus, the interpreter can ascribe as many norms as there are interpretative directives to a norm formulation. Instead of reducing indeterminacy, interpretative codes will contribute to it by allowing for a multiplicity of ex ante ascribable norms to norm formulations. Thus, a norm formulation can bear, at least the same number of norms as the number of interpretative directives contained in the interpretative code of a legal system. From the multiplicity of interpretative directives, it follows that it is impossible to determine ex ante which norm will be chosen by the interpreter. Therefore, the non-redundancy clause supports the claim that the multiplicity of interpretative tools contributes to the multiplicity of norms, thus, making law indeterminate.
This multiplicity of ascribable norms allows for judges to have discretion when interpreting the law. That is, the judge will need to choose which norm from all the interpretative options will be ascribed to the norm formulation applicable to a case. However, this is not the entire story for the discretion of judges. Further argumentation will be provided in §5.2.
The non-redundancy clause supports the claim that the multiplicity of interpretative tools contributes to the multiplicity of norms, thus making law indeterminate. In other words, a norm formulation can bear, at least, the same number of norms as the number of interpretative directives contained in the interpretative code of a legal system. From the multiplicity of interpretative directives, it follows that it is impossible to determine ex ante which norm will be chosen by the interpreter. Now, notice that this argument does not need to rely on a specific theory of meaning nor does it need to deny the existence of a right answer in legal interpretation. It could be the case that among the multiplicity of norms, there is one that can be deemed as such. However, from the perspective of the legal system, all norms equally legally valid. Those who support the 'right answer' thesis will need to rely on extra-legal criteria to identify it. This argument does not imply that from an empirical point of view, different interpretative directives cannot yield the same result. By result, I am referring to the decision of a case. In contrast, it is possible that several interpretative directives lead to the same result and still have a rational legal system. 44 Notice that the use of only one interpretative 44. Alchourrón & Bulygin, supra note 11 at 64. directive is a sufficient condition for a norm to be legally valid. Thus, having the same interpretative result in practice is inconsequential for the validity of the legal decision and at best it could be considered a rhetorical tool to support why a particular norm should be preferred above others.

Part Two of Reason for Accepting Premise c): The Existence of Interpretative Negative Spaces
There are two kinds of interpretative codes. One kind of interpretative code is the one that is regulated by legal meta-rules. Thus, it will contain legal meta-rules that will guide and constrain the use of the interpretative directives. These meta-rules, as far as I can see, can be of two kinds: syntactic meta-rules and semantic meta-rules. The first group of rules prescribes hierarchies among the interpretative directives. Thus, they guide the interpreter in the use of the interpretative code. For example, a syntactic meta-rule can be formulated as follows: "When the wording of the law is clear the judge ought to interpret the law according to the plain meaning of the text. If the wording of the law is unclear the interpreter ought to interpret the law according to the intention of the legislator. If interpreting the law according to the intention of the legislator produces a result that infringes fundamental rights then the interpreter ought to interpret the law according to the historical context in which the legislation was enacted". This rule prescribes an order that the judge is legally obliged to follow when interpreting the law. The second group of rules prescribes how we are to interpret the interpretative directives. Semantic meta-rules are meta-meta rules that help to reduce the natural language-based indeterminacy of the interpretative code. For example, a semantic meta-rule can be formulated as follows: "When applying the plain meaning of the text rule, the judge ought to interpret the formula 'plain meaning of the text' as the lexicographical meaning of the words of the law".
The second kind of interpretative code is one that is not regulated by metarules. Since there are no legal meta-rules that guide and constrain the use of the interpretative directives, the interpreter will have discretion when using the interpretative code.
The lack of legal-meta rules that govern the use of the interpretative code is what I call 'interpretative negative space'. Roughly put, in art, negative space is the area between and around an object in an image. A negative space plays at least three functions. First, it says something about the positive space (i.e., objects, identifying their shape and limits). Second, it represents the absence of a positive space. Third, it says something about the limits or boundaries of the said absence. 45 I think there is an analogous case to be made in legal reasoning. I use the aesthetic concept of 'negative space' to illustrate the lack of legal meta-rules that regulate the use of the interpretative rules contained in the interpretative code (i.e., the interpretative negative space). The interpretative negative space fulfils three functions in legal reasoning. First, it says something about the legal rules that are used in legal reasoning (i.e., the positive spaces). Thus, it identifies the interpretative code and the types of interpretative directives that it contains as the rules used by judges in legal interpretation. Second, it represents the absence of rules that regulate the use of interpretative directives (i.e., the absence of semantic and/or syntactic meta-rules). Third, it defines the boundaries of the absence of rules. These boundaries refer to the scope of the decision that judges need to make regarding the use of the interpretative code. That is, the decision about both establishing a hierarchy among interpretative directives and the way that the interpretative directives are to be interpreted when applied. Since there is a lack of meta-rules, this decision is discretionary. In other words, interpretative negative spaces define the scope of discretion in legal interpretation. As I will develop further, the scope of discretion in legal interpretation extends to the decision of which interpretative directive the judge will use and how the selected interpretative directive will be interpreted. Thus, the advanced analogy stresses the fact that there are negative spaces between interpretative directives.
Due to the lack of legal meta-rules that regulate the interpretative code, all of the different norms that can be ascribed to a norm formulation are in equal standing before the interpreter. Therefore, since there are negative spaces when interpreting the law, judges do not have legal guidance to choose which interpretative directive they will use to ground their interpretation of a norm formulation. Due to the existence of interpretative negative spaces and the multiplicity of interpretative directives, judges have discretion when interpreting law.
There are two conceptually possible scenarios in which interpretative negative spaces can be found and these scenarios are mutually exclusive. One scenario is the case in which the interpretative code lacks legal meta-rules that governs how this code is to be used. Another scenario is that the interpretative codes have meta-rules, either semantic or syntactic, that governs the use of interpretative codes. It is conceptually the case that one of these two scenarios will arise. These two scenarios will be explored in §5.2.1 and §5.2.2.

In the Event of No Legal Meta-Rules
It is possible that the interpretative code contains neither syntactic nor semantic legal meta-rules. That is, there is no hierarchy among the interpretative directives nor methods for their interpretation.
The point that I will be stressing in this section is the importance of the existence of discretion regarding the lack of syntactic legal meta-rules. From the absence of legal hierarchies, it follows that all interpretive directives are in an 'equal stance' before the interpreter. As Spaak notes, "there is no agreement among lawyers on how to rank the interpretative [directives]". 46 Thus, the decision about which interpretative directive will be used, and thus which norm will be ascribed to the norm formulation, is left to choice. Lack of meta-rules allows for the use of competing norms to decide the case at hand. This is what makes the law. That is, it is not possible ex ante to determine what norm will decide the case because it is not possible to predict which interpretative directive the interpreter will use insofar as there are no hierarchies among them. Notice that the lack of syntactic legal meta-rules does not entail the absence of rules for the interpreter. It just means that if the interpreter follows meta-rules to guide a decision about how to use the interpretative code, then these meta-rules are not legal in origin but something else.
Due to the multiplicity of interpretative directives and the lack of legal metarules, interpreters will necessarily find instances of discretion when interpreting norm formulations (i.e., instances where they will need to exercise judgment). These instances of discretion necessarily arise whenever the use of the interpretative code is needed. In other words, if a legal system has an interpretative code with more than one interpretative directive and there are no meta-rules that regulate its use by establishing hierarchies among them, then legal interpretation necessarily entails discretion in the form of the choice of the interpreter among a closed set of options (i.e., the interpretative directives).
In the scenario for which there are no legal meta-rules, judges will necessarily have discretion when interpreting the law.

In the Event of Legal Meta-Rules
It is possible to think about a legal system that contains an interpretative code with both syntactic and semantic meta-rules of legal origin. However, even if this is the case, negative spaces will necessarily arise when interpreting these legal meta-rules. Due to the indeterminacy of natural languages, cases of ambiguity and vagueness will eventually affect those legal meta-rules. Thus, the interpretation of legal meta-rules is left to the interpreter's discretion. In §5.2.2.1 and §5.2.2.2, I further develop the claim that negative spaces are necessary albeit the existence of legal meta-rules; thus, judges will necessarily have discretion in interpreting the law.

In the Event of Syntactic Legal Meta-Rules
What if interpretative codes do contain syntactic meta-rules? Although legal systems do not usually expressly contain these kinds of rules, it is possible to imagine a legal system that does. If the legal meta-rules are posited as norm formulations, they will require interpretation in order to be used. It is possible 46. Spaak, supra note 10 at 57.

390
Reyes Molina as a result to foresee a situation of infinite regress in which meta-meta rules will be required to interpret meta-rules and so on. Arguably, the key issue about the interpretation of legal meta-rules has to do with the decision of whether some interpretative directive is suitable for the interpretation of norm formulations in a given case. This is of central importance to the extent that a decision about the suitability of a given interpretative directive will have a direct impact on the norm that will decide the case in the end. 47 The decision regarding the meaning of 'suitable' will be left to the discretion of the judge. Instances of discretion will move from the choosing of interpretative directives to the interpretation of the syntactic meta-rules. In other words, interpretative negative space will arise when interpreting syntactic meta-rules, and thus judges will-since they must decide-necessarily have discretion in interpreting the law. Once again, an absence of syntactic meta-rules does not follow from this claim. Interpreters will probably follow certain rules or guidelines to establish hierarchies among interpretative directives, but if that is the case, these rules cannot be legal in origin.

In the Event of Semantic Legal Meta-Rules
What if interpretative codes do form semantic legal meta-rules?
It is possible to imagine an interpretative code that does contain meta-meta rules of interpretation. One can claim that the origin of these interpretative directives can be judge-made or the product of the legislator. However, whatever its origin, these meta-meta rules require interpretation in order to be used to interpret the norm formulation that will be applied to decide a given case.
Suppose that we have chosen the intention of the legislator rule for our code. Now we need to apply it to norm formulation "X" that is used to solve a case. The first thing that the interpreter must do is ascribe a meaning to the interpretative directive. What do we mean by 'the intention of the legislator'? The interpretation of the first-order norm formulation depends on the answer to this question.
The problem with this, prima facie, seems rather obvious: we do not have meta-meta-interpretative legal rules (i.e., semantic meta-meta rules), and proposing the addition of more rules to the system would amount to an infinite regress. 48 It follows that when interpreting interpretative directives, legal operators will eventually find interpretative negative spaces. Thus, the decision about how we are to interpret these meta-rules is discretionary. We are in the same situation as in §5.2.2.1: discretion is moved one level up. That is, indeterminacy will affect the semantic-meta rule, making it impossible to know ex ante how this rule is to 47. As the reader might have noticed, the issue described above has to do with the indeterminacy of natural language. This further supports the main point of this essay that law is affected by at least two kinds of conceptual indeterminacy: 1) the indeterminacy of natural language; and 2) the 'systemic' indeterminacy accounted for in this paper. 48. For an example of adding rules to the system in order to deal with issues of indeterminacy, see Sebastián Reyes Molina, "On Legal Interpretation and Second-order Proof Rules" (2017) Analisi e Diritto at 179. be interpreted. Since how we are to understand the interpretative guideline will depend on how the semantic meta-rule will be interpreted, the indeterminacy of the latter will affect the interpretation of the former. As Waldron notes, [s]ince words do not apply themselves, since it is we who apply them to cases, of course we may need further rules for their application, and of course we will eventually run out of analytic meaning-rules before a precise application is determined. There cannot be a rule to tell us how to apply every rule: sooner or later one simply makes a judgment. 49 Following Waldron, at some point the interpreter needs to make a judgment. In other words, interpretative negative space will arise when interpreting semantic meta-rules, and thus judges will necessarily have discretion when interpreting the law.

Part Three of Reason for Accepting Premise c): The Obligation to Decide
From §5.1 and §5.2, it is possible to conclude that there is a multiplicity of interpretative results and an absence of legal rules that guide and constrain the interpreter as to how to choose one among the many interpretative results. However, one additional step is required in order to make the case that discretion is a necessary feature in legal interpretation. This further step has to do with prescribing a legal obligation that judges have to decide a case. As I have covered in §4.2, a typical rational legal system contains the non-liquet rule. This rule prescribes the obligation to decide all cases. This obligation is applicable even when there are no legal rules that govern the case at hand. Its scope encompasses both the final decision of the case, as well as all the other decisions that judges need to make throughout the legal process. One of these decisions is the one regarding which interpretative tool the norm that will decide the case will be grounded. As stated in §4, whenever the interpretation of a norm formulation is required, the interpretative result ought to be grounded in an interpretative directive. Although there are interpretative negative spaces regarding how to use the interpretative code, the non-liquet rule demands a decision that will settle this specific matter regardless. Due to the existence of these interpretative negative spaces, the decision about the interpretative directive necessarily amounts to a choice over which interpretative tool will be selected. In other words, discretion is a necessary feature in legal interpretation.
Due to the obligation to decide, judges must necessarily exercise discretion in legal interpretation.

Identifying Discretion in Legal Decision-Making
The existence of interpretative negative spaces allows for a moderate account of discretion in legal interpretation. By moderate account, I mean that interpreters will exercise their discretion, in the form of judgment, whenever they choose both the interpretative directive and its interpretation. In other words, discretion can be reduced to a choice about which interpretive directive will ground the norm that will be applied in a given case.
Two comments are in order. First, discretion does not pervade the entire process of legal decision-making, this is not what I am claiming. It is only limited to the decision about the interpretative directives. Second, discretion is not merely the result of certain features of natural language which would figure prominently with a 'plain meaning' rule.
Although interpretative codes are thought to help reduce ambiguity and vagueness, they would explain the presence of a plurality of norms, thus contributing to the indeterminacy of law. Notice, however, that this plurality of different norms is not the result of certain features of natural languages that affect the law. Although ambiguity and vagueness play a role in legal interpretation, the picture of legal indeterminacy is complete when it is understood in combination with the notion of a rational legal system. Since legal systems cannot have redundant legal rules, each interpretative code will yield a different norm. Thus, legal interpretation necessarily involves a choice. This choice is about which norm the judge will apply to decide the case. In other words, discretion is a necessary feature in legal interpretation.

Conclusion
Unlike the prevailing account of legal indeterminacy laid down in §3, I do not think that indeterminacy and discretion in law are explained mostly by certain features of natural language or by extra-legal factors that might operate in legal adjudication. In this paper, I have advanced a thesis that bases legal indeterminacy and judicial discretion on features within the structure of typical rational legal systems. My argument states that under certain premises, there are specific features of legal systems that allow for a conceptual kind of both legal indeterminacy and discretion in legal adjudication.
In §4, I defended the premises a) and b) of my arguments. These premises are sound and accommodate our intuitions about how typical rational legal systems are structured. These conditions are plausible in the sense that they will commonly be found in existing typical legal systems.
In §5, I provided a three-part reason that supports premise c). The first part, the non-redundancy clause, states that rational legal systems do not have redundant norms. This means that in the case of legal interpretation, interpretative codes with multiple interpretative directives will necessarily yield multiple norms that an interpreter can ascribe to a norm formulation. Each norm necessarily provides a different solution for a given case, otherwise they are redundant and all but one ought to be eliminated from the legal system. The second part, the existence of interpretative negative spaces, stresses the fact that interpretative directives are typically not ranked by other meta-rules from a legal origin. It follows then that all of the interpretative directives are on an equal footing and each necessarily provide a different norm for a specific norm formulation. These interpretative negative spaces will necessarily arise and they can be found in two scenarios. The first scenario refers to an interpretative code that lacks legal meta-rules for its use. The second scenario refers to the existence of legal meta-rules that governs how to use the interpretative code. As I have shown in both scenarios, judges will necessarily have discretion when interpreting the law. The third part, the obligation to decide, establishes that interpreters need to make a decision on how to use the interpretative code despite the existence of interpretative negative spaces. The judges' situation is as follows: there are many interpretative directives, each one providing necessarily a different result; judges need to choose one in order to ground an interpretation of the norm formulation; there are no meta-rules to help them decide which directive they will use and they are bound to do so regardless of the existence of interpretative negative spaces due to the non-liquet rule. Thus, judicial discretion is the result of the combination of an interpretative code with a multiplicity of interpretative directives, the non-redundancy clause, the existence of interpretative spaces, and the non-liquet rule. This is the core of premise c). In other words, judges necessarily exercise discretion. Notice, however, that it does not follow that there are no rules at all, but if the interpreter uses rules to guide a decision, such rules are not legal in origin but something else.
The decision about how to use the interpretative code amounts to a choice. A judgment made by the interpreter decides two things. First, which interpretative directive one will use. Second, how the interpretative directive will be interpreted. Discretion is limited to certain instances where interpretative negative spaces can be found. These instances have a direct impact on the way in which the norm formulation is interpreted. Notice that these negative spaces are conceptual in nature, thus yielding discretion as a necessary element in legal interpretation.
Some prima facie consequences of this approach are as follows: first, as stated throughout the text, law is necessarily indeterminate. What is indeterminate is which interpretative directive the judge will use. Bear in mind that each interpretative directive provides a different interpretative result. It follows that it is not possible ex ante to determine which norm will be ascribed to the norm formulation. This means that it is indeterminate which norm will be applied to the case. Second, legal adjudication is necessarily discretionary, since the interpreter is under an obligation to select one among many interpretative directives that are on equal grounds without any rule that offers guidance. Thus, the decision about which interpretative directive will be used is necessarily discretionary. Third, neither indeterminacy nor discretion is solely the product of law understood as a manifestation of natural language. They are also a product of how 394 Reyes Molina a legal system is designed. That is, which type of formal properties we ascribe to them and how we regulate the use of interpretative codes. Fourth, I do not claim that indeterminacy is radical and we are in an 'anything goes' type of situation.
There is a plurality of norms that can be ascribed to a single norm formulation, but this plurality is a finite number. That is, there are as many norms for any single norm formulation as the number of interpretative directives contained in an interpretative code. Fifth, I do not claim that discretion pervades the entire adjudication process or that the interpreter can make whatever judgment they desire. Discretion, in my account, is limited only to the choice among the interpretative directives contained in the interpretative code. Lastly, two practical concerns are raised by this paper. First, if a given society deems discretion to be problematic, the fact that it is a necessary element of adjudication then prevents the solving of this problem through more legal rules. The multiplication of legal rules will inevitably lead to an infinite regress and just 'move' the problem of indeterminacy and discretion to a higher level. Second, if discretion is a necessary feature of legal interpretation, then studying extralegal factors (i.e., biases, political allegiances, moral convictions, the overall set of stimuli, etc.) could be an advisable strategy to reduce indeterminacy in law.