II. The Reason Model of Precedent

I begin by outlining the reason model of precedential constraint.Footnote ⁵ I will then relate it to the alternative “result” model.Footnote ⁶ My presentation of both models largely follows Horty’s, with some small stylistic differences.Footnote ⁷ Readers who are familiar with both models may wish to proceed directly to section 3.

B. The Reason Model of Precedential Constraint

On the reason model of precedential constraint, subsequent courts are constrained to maintain consistency with a priority order over reasons that is revealed by the ratio formulated in past decisions. First, a priority order over reasons is derived from a single case $ c=\left\langle X,R,s\right\rangle $ by taking any reason at least as strong as the ratio $ R $ of the case to be stronger than any reason for the opposing outcome $ \overline{s} $ that was satisfied on the facts of the case.

Definition 1 (Priority Order Derived from a Case). Where $ c=\left\langle X,R,s\right\rangle $ is a case, a priority order $ {<}_c $ is derived from $ c $ by taking $ U{<}_cV $ just in case $ U\subseteq {X}^{\overline{s}} $ and $ R\subseteq {V}^s $ .

This account of a derived priority order for a case $ c $ is then lifted to a priority order for a case base $ \varGamma $ by supposing that one reason has priority over another according to a case base when the corresponding priority can be derived from some case in $ \varGamma $ .

Definition 2 (Priority Order Derived from a Case Base). Let $ \varGamma $ be any case base. A priority order $ {<}_{\varGamma } $ is derived from $ \varGamma $ by taking $ U{<}_{\varGamma }V $ just in case $ U{<}_cV $ for some case $ c $ in $ \varGamma $ .

The priority order is then used to define a notion of consistency for a case base—a case base will be inconsistent just in case its derived priority order contains conflicting priorities.

Definition 3 (Consistency of a Case Base). Let $ \varGamma $ be a case base and $ {<}_{\varGamma } $ be its derived priority ordering. $ \varGamma $ is inconsistent just in case there are two reasons $ U $ and $ V $ such that $ U{<}_{\varGamma }V $ and $ V{<}_{\varGamma }U $ . $ \varGamma $ is consistent just in case it is not inconsistent.

From this account of consistency, we can then offer an account of when it will be permissible to decide for a given outcome $ s $ by formulating the ratio $ {R}^s $ in a new fact scenario $ X $ , against the background of a consistent case base $ \varGamma $ . Informally, the decision will be permissible just in case the addition of the new case representing the decision to $ \varGamma $ does not introduce any inconsistency into the case base.

Definition 4 (Permissible Decisions and Required Outcomes). On the reason model of precedential constraint, where $ X $ is a new fact scenario and $ \varGamma $ is a consistent case base, it is permissible for a court to reach a decision in $ X $ with outcome $ s $ based on the ratio $ {R}^s $ just in case the case base $ \varGamma \cup \left\{\left\langle X,{R}^s,s\right\rangle \right\} $ is also consistent. A decision in $ X $ with outcome $ s $ is required just in case there is no reason $ {R}^{\overline{s}} $ satisfied in $ X $ such that it would be permissible to reach a decision with outcome $ \overline{s} $ on the basis of $ {R}^{\overline{s}} $ .

Imagine that a court is being asked to consider whether an airline discriminated against a passenger by prohibiting them from taking their assistance animal—a dog—onboard a flight. They consider a case base $ {\varGamma}_1 $ that contains the sole precedent $ {c}_1 $ , in which it was held that it was not discriminatory to prohibit a passenger from taking his animal on board the flight, because the animal—in that case the passenger’s pet crocodile—presented a safety or hygiene risk. We will suppose that the passenger was the plaintiff, so that the outcome $ \pi $ is associated with the conclusion that the prohibition was discriminatory, and that the airline was the defendant, so that outcome $ \delta $ is associated with the conclusion that the prohibition was not discriminatory. The factor $ {f}_1^{\delta } $ represents the fact that the animal presented a safety or hygiene risk (favoring the conclusion $ \delta $ that it was not discriminatory to prohibit it), and $ {f}_1^{\pi } $ represents the fact that the animal was either on a leash or in a cage (favoring the conclusion that the prohibition was legally discriminatory). The precedent considered by the court is represented in the case $ {c}_1=\left\langle {X}_1,{R}_{(1)}^{\delta },\delta \right\rangle $ , where, in the fact scenario $ {X}_1=\left\{{f}_1^{\pi },{f}_1^{\delta}\right\} $ , the court formulated its decision in terms of the reason $ {R}_{(1)}^{\delta }=\left\{{f}_1^{\delta}\right\} $ , which supported a finding for outcome $ \delta $ . In other words, the court reached the conclusion that the prohibition was not discriminatory on the basis that the animal presented a safety or hygiene risk to passengers. The priority order derived from this decision is $ \left\{{f}_1^{\pi}\right\}{<}_{c_1}\left\{{f}_1^{\delta}\right\} $ , which tells us that the court afforded greater weight or priority to the reason to decide for the defendant due to the safety or hygiene risk over the reason to decide for the plaintiff because the animal was kept on a leash or in a cage.

We now suppose the facts of the new dispute involving a dog are characterized by the set $ {X}_2=\left\{{f}_1^{\pi },{f}_2^{\pi },{f}_1^{\delta}\right\} $ , so that the new case involves an additional factor favoring the conclusion that the prohibition was discriminatory— $ {f}_2^{\pi } $ —which represents the fact that the dog involved was the passenger’s “assistance animal.” On the reason model, it would be impermissible for the court in the new case to rule that the prohibition was discriminatory on the basis of the reason $ {R}_{(1)}^{\pi }=\left\{{f}_1^{\pi}\right\} $ , since that reason was included in the factor set $ {X}_1 $ that characterised the precedent case $ {c}_1 $ and was afforded less weight than the reason $ \left\{{f}_1^{\delta}\right\} $ , which is also satisfied in the new fact scenario $ {X}_2 $ . It would, however, be permissible to rule that the prohibition was discriminatory because the animal was the passenger’s assistance animal–that is, on the basis of the reason $ {R}_{(2)}^{\pi }=\left\{{f}_2^{\pi}\right\} $ , since this reason was not satisfied in the precedent case. Let $ {c}_2 $ represent a case in which the court, faced with the same fact scenario $ {X}_2 $ rules for the plaintiff by formulating its decision in terms of the reason $ {R}_{(2)}^{\pi }=\left\{{f}_2^{\pi}\right\} $ . The case base $ {\varGamma}_1\cup \left\{{c}_2\right\} $ is consistent. Since $ {R}_{(1)}^{\pi }{<}_{\varGamma_1}{R}_{(1)}^{\delta } $ and $ {R}_{(1)}^{\delta }{<}_{c_2}{R}_{(2)}^{\pi } $ , the addition of $ {c}_2 $ to $ {\varGamma}_1 $ does not introduce any inconsistency into the case base.

This account of the structure of a case and its constraining effect on future decisions can also be represented visually.Footnote ⁸ In Figure 1, the enclosed rectangular plane represents the entire universe of possible factors $ F $ , with the diagonal line dividing the universe of factors into the sets $ {F}^s $ and $ {F}^{\overline{s}} $ of all factors favoring outcomes $ s $ and $ \overline{s} $ , respectively. As in a standard Venn diagram, closed curves on the plane represent sets of factors. A case that involves both plaintiff and defendant favoring factors can be represented as a closed curve across the dividing line, so that the set of all factors that characterize the case $ X $ also divides into a set $ {X}^s $ of factors supporting the outcome $ s $ and a set $ {X}^{\overline{s}} $ of factors supporting $ \overline{s} $ . In Figure 1, the ellipse drawn with a continuous line represents the set of factors $ X $ that characterized the precedent case $ c=\left\langle X,R,s\right\rangle $ , in which a judgment with outcome $ s $ was given by the court on the basis of the reason $ R $ . By stipulation, the ratio $ R $ is a subset of the set of factors $ {X}^s $ supporting the court’s judgment in favor of the outcome $ s $ . It too is represented by a closed ellipse drawn with a continuous line.

Figure 1. The precedential effect of case $ c=\left\langle X,R,s\right\rangle $ in new fact scenario $ Y $ is illustrated, with the set of possible distinguishing factors shaded.

The ellipse enclosed by the dashed line represents a new fact scenario $ Y $ . The precedent case $ c $ is applicable in $ Y $ , since the ratio $ R $ is a subset of the set of factors characterizing the new scenario. For any reason $ R^{\prime } $ that is included in the set $ {X}^{\overline{s}} $ (i.e., such that $ {R}^{\mathrm{\prime}}\subseteq {X}^{\overline{s}} $ ), we know that $ {R}^{\mathrm{\prime}}{<}_cR $ can be derived from case $ c $ . In the new fact scenario $ Y $ , consistency with the precedent case is ensured by either applying the precedent to the new fact scenario (which will broaden the precedent’s effect) or by relying on a set comprised of at least one factor favoring the outcome $ \overline{s} $ present in $ Y $ that was not present in the facts $ X $ of the precedent case. In Figure 1, the set of all such “possible distinguishing factors” is shaded. A permissible decision with outcome $ \delta $ must cite as a reason a set of factors that intersects with the shaded set. (The court only needs to include one of these factors.) The court cannot base a decision for $ \delta $ on a reason $ {R}^{\overline{s}} $ that includes only factors present in $ {X}^{\overline{s}} $ , since that would introduce an inconsistency with the priority order $ {<}_c $ derived from $ c $ .

The account of the constraining effect of precedent offered by the reason model is compatible with the idea that the doctrine of precedent imposes a “disjunctive” obligation to follow or distinguish on subsequent courts.Footnote ⁹ If the set of possible distinguishing factors is empty, then courts are required to follow the reasoning of the previous case. If the set of possible distinguishing factors is non-empty, then consistency with the case base requires the court to either apply the precedent or base their judgment on a reason that contains at least one of these possible distinguishing factors. Note, as well, that rationes formulated in terms of larger set of facts will have less constraining effect. The larger the set of factors that are contained in the ratio, the more difficult it is for the same ratio to be satisfied in a new fact scenario. For instance, if the court in $ c=\left\langle X,R,s\right\rangle $ had formulated its ratio in terms of the reason $ {X}^s $ —the set of all factors present in $ X $ that favored $ s $ —then $ c $ would not be applicable in the new fact scenario $ Y $ , because some of the factors in $ X $ are not in $ Y $ . A court faced with the new fact scenario would not even need to distinguish $ c $ .

III. Reinterpreting Precedent: A Challenge for the Reason Model

The reason model as presented in the previous section raises a variety of interpretive problems. To begin with, the model is “flat,” in the sense that it postulates a direct relationship between a series of factors and an outcome. In reality, legal reasoning often proceeds through intermediate steps.Footnote ¹³ The models ignore the effects that judicial hierarchy can have on the reception of a precedent by a subsequent court.Footnote ¹⁴ To simplify presentation and discussion, I have also omitted any extension of the model to accommodate “dimensional” reason—reasoning with ordered sets of legally relevant values rather than simple factors.Footnote ¹⁵ Most relevant to the current discussion is the problem that the models do not address the complex process by which courts interpret (and reinterpret) precedents. For a given case $ c=\left\langle X,{R}^s,s\right\rangle $ , both the fact scenario $ X $ and the ratio $ {R}^s $ are exogenously fixed—the models do not straightforwardly accommodate any scope for interpretive disputes about either the facts that characterized the precedent case or the interpretation of its ratio.Footnote ¹⁶

Rather than proposing a model that accommodates the broadest possible variety of interpretive disputes, my focus here will be on a particular interpretive practice, which I will call restrictive reinterpretation, and which is identified by replacement of the clear ratio of a precedent with a reason that includes a larger set of factors and thus has less constraining effect. (It is possible for courts to reinterpret a precedent in a manner that extends its ratio. But this form of reinterpretation is less problematic for the reason model.Footnote ¹⁷)

The adoption of a restrictive reinterpretation of a precedent case is most often motivated by the desire to avoid the unwanted constraints of precedent. Consider the precedent $ {c}_2=\left\langle {X}_2,{R}_{(2)}^{\pi },\pi \right\rangle $ , interpreted as a case in which the court held that the prohibition of a passenger’s dog was discriminatory on the basis of the reason $ {R}_{(2)}^{\pi } $ , which cited the fact that the dog in question was the plaintiff passenger’s assistance animal. A subsequent court could note that the fact scenario in $ {c}_2 $ , $ {X}_2=\{{f}_1^{\pi },{f}_2^{\pi },{f}_1^{\delta}\} $ , also included the factor $ {f}_1^{\pi } $ , representing the fact that the animal was kept on a leash or in a cage. In order to avoid a situation where they are constrained to hold that it is discriminatory to prohibit a passenger from being accompanied by a dog that is unleashed, the court could decide to reinterpret the precedent in $ {c}_2 $ so that it only applies in cases where the animal is both the passenger’s assistance animal and kept on a leash or in a cage. In other words, they may seek to reinterpret the case as one that involved the ratio $ {R}_{\left(1,2\right)}^{\pi }=\left\{{f}_1^{\pi },{f}_2^{\pi}\right\} $ , involving both plaintiff-favoring factors present in $ {X}_2 $ .Footnote ¹⁸

The practice that I am describing as restrictive reinterpretation needs to be distinguished, at least for the time being, from scenarios in which there are genuine interpretive disputes about the scope of a precedent. In his influential work on precedent in US law, Richard Re describes a practice, which he calls “narrowing,” whereby courts “interpret a precedent more narrowly than it is best read.”Footnote ¹⁹ Crucially, Re argues that the permissibility of narrowing a precedent depends on the precedent being ambiguous, in the sense that it is open to multiple reasonable interpretations, even if there is a “best” reading.Footnote ²⁰ In recent work, Rigoni has sketched an extension of the reason model that accommodates the possibility of multiple reasonable interpretations and thus accommodates narrowing.Footnote ²¹ As I understand it, the practice that Re and Rigoni describe as narrowing and the practice that I am describing as restrictive reinterpretation are extremely closely related. In reality, the distinction between interpretation of an ambiguous precedent and reinterpretation of a clear precedent will often be hard to draw. For the purposes of the present discussion, I will set the possibility of ambiguous precedents aside. Within the constraints of the reason model, the practice that I am describing as restrictive reinterpretation occurs where courts replace the clear formulation of the ratio offered by the precedent court with another reason that has less constraining effect.

Several accounts of precedent directly assimilate restrictive reinterpretations into their account of distinguishing.Footnote ²² Authors who have focused on the English doctrine of precedent, in particular, are tolerant of the practice. In their treatise on precedent in English law, Cross and Harris note the susceptibility of precedents to be reinterpreted in light of both their facts and surrounding case law, endorsing Raz’s view that both distinguishing and reinterpretation involve the use of a “residual power to restrict the scope” of a precedent that is conferred on courts.Footnote ²³ Grant Lamond, a proponent of a different version of the reason model of precedent, writes that the “second common way” of distinguishing occurs where “some fact in the earlier case that was not included in the ratio is relied upon by a later court.”Footnote ²⁴ Lamond notes that this form of distinguishing can also be understood as a decision to “reinterpret” the ratio, adding that for “most purposes it does not matter which is the proper analysis of what the court has done.”Footnote ²⁵ Setting aside questions relating to the permissibility or coherence of restrictive reinterpretations, descriptively it seems hard to deny that it is a feature of legal discourse, at least at the appellate level. To provide just one example: prior to its abolition, the doctrine of common employment, which made employees rather than their employers liable for injury caused to fellow employees, was avoided by the House of Lords by reinterpreting earlier precedents as more narrowly confined to cases in which the two employees were engaged in the same enterprise or activity.Footnote ²⁶

One strategy for accommodating restrictive reinterpretation involves interpreting the absence of a factor that favored the outcome in a precedent case as equivalent to the presence of a negated factor favoring the opposing outcome. Consider a court deliberating in the fact scenario $ {X}_3=\{{f}_2^{\pi },{f}_1^{\delta}\} $ against the background of a case base that includes the precedent $ {c}_2=\left\langle {X}_2,{R}_{(2)}^{\pi },\pi \right\rangle $ . Suppose the court wishes to avoid deciding for the plaintiff (and thus that the prohibition was discriminatory) because the animal in this instance was unleashed. Recall that the salient difference between the current fact scenario $ {X}_3 $ and the fact scenario $ {X}_2 $ of the precedent is that $ {X}_2 $ contained the plaintiff-favoring factor $ {f}_1^{\pi } $ —that the animal in question was kept on a leash or in a cage—where the current fact scenario does not. The difference between the current fact scenario and $ {X}_2 $ could be recast in terms of the presence of a new defendant-favoring factor $ {f}_3^{\delta }=\neg {f}_1^{\pi} $ , equivalent to the negation of $ {f}_1^{\pi } $ , representing the fact that the animal was not kept on a leash or in a cage. The precedent $ {c}_2 $ could then be straightforwardly distinguished by invoking this factor. This framing of the example in question seems innocuous and even attractive: the fact that an animal was not kept on a leash or in cage does appear to be a reason favoring the defendant.

While recasting restrictive reinterpretation as a form of distinguishing involving negative factors might allow us to salvage the reason model, and certainly would pose few problems in this simple example, I think it is worth exploring alternative models of the practice.Footnote ²⁷ To begin with, I think courts do occasionally use interpretive language to justify revising the clear rationes of precedents. As Rigoni has noted in a related context, there is value in exploring whether formal models of precedential constraint can accommodate these interpretive practices on their own terms.Footnote ²⁸ Additionally, I do not think that recasting restrictive reinterpretation as distinguishing based on negative factors provides a universally sound strategy for accommodating cases of restrictive reinterpretation within the reason model. The assumptions needed to rely on such a strategy in every case are strong. The introduction of negated factors raises difficult questions about how we should reason in cases where we are lacking information. Supposing a court lacks any information about whether an animal was kept on a leash, should the court then conclude that the animal was not kept on a leash? To conclude that the absence of information supporting a given factor entails the presence of a negated factor, we must make the “closed-world” assumption.Footnote ²⁹ There are good arguments that the closed world assumption needs to be evaluated on a factor-by-factor basis, and thus is not universally sound.Footnote ³⁰ Moreover, there are likely to be cases in which the absence of a factor favoring one side does not in and of itself favor the other side of the dispute. In a trade secrets case, for instance, while the fact that the defendant bribed one of the plaintiff’s employees is a factor supporting the plaintiff, the absence of any bribery does not count in the defendant’s favor. Negated factors can be neutral as to the outcome they support.Footnote ³¹ These complications suggest that it is worth taking the practice of restrictive reinterpretation at face value and exploring the extent to which it can be accommodated within the reason model.

The version of the reason model of precedent presented in section 2 does not offer an account of the judicial practice of reinterpreting precedent. In fact, the idea that subsequent judicial decisions involve reinterpretation of past cases is more familiar as one of the central claims made by proponents of a rival model of precedential constraint. On the approach favored by those who advocate an account of the constraining effect of precedent as akin to the effect of a statutory rules (what Horty calls the ‘rule model’ of precedent), distinguishing a precedent involves a kind of modification—or reinterpretation—of the rule adopted in the previous decision.Footnote ³² On this approach, distinguishing is just a particular form of reinterpretation of a binding precedent, and even a covert form of overruling. This can be contrasted with the account of distinguishing offered by the reason model. Unlike the rule model, which treats all cases of distinguishing as a form of modification—or reinterpretation—of the rule cited as the ratio in the previous decision, the reason model views permission to distinguish as arising when there is a reason for the opposing outcome that falls outside of the constraints of the previous decision (because it was not satisfied on the facts of that decision).

At first glance, then, the practice of reinterpreting precedent seems more compatible with the rule model than the reason model; it can be more readily assimilated with its proponents’ understanding of judicial practice. Since theorists who accept the rule model already regard distinguishing as an act of modifying an existing precedential rule, they have little trouble accommodating the idea that courts can narrow precedent.Footnote ³³ Proponents of the reason model therefore have good reasons for being wary of accepting the practice of restrictively reinterpreting precedent as a legitimate form of distinguishing. Horty refers to the practice of adopting restrictive reinterpretations as “negative distinguishing,” and argues that it is not, in fact, a legitimate form of distinguishing, since it amounts to a form of implied overruling of precedent.Footnote ³⁴

Viewed in another light, the practice of reinterpretation can also be used to challenge the reason model’s reliance on courts’ formulations of their reasons for decision in constraining subsequent deliberation. Early legal realists like Karl Llewellyn noted the tendency of courts to reinterpret precedents to avoid unwanted outcomes:

The skepticism expressed by Llewellyn is reflected, in a considerably more moderate form, in the view that courts’ verbal formulations of the ratio do not provide any independent constraint, and that subsequent courts are constrained only by the “material facts” of the case and the outcome favored by the court.Footnote ³⁶ This approach to determining the constraining effect of a precedent was introduced as an interpretation of the result model in section 2. Since precedents can always be reinterpreted restrictively, and in extreme cases confined to their facts, their susceptibility to reinterpretation seems to suggest that it is only the material facts of the case that have any constraining effect. In terms of the reason model, this would suggest that rather than assigning any constraining effect to the court’s formulations of the ratio, the precedent should be associated with the reason represented by the entire set of factors that favored the outcome of the precedent case. In section 4 I will demonstrate that if no constraints are placed on the practice of restrictive reinterpretation, the practice does in fact lead to the collapse of the reason model into the result model.

IV. Restrictive Reinterpretations on the Reason Model: Extension and Collapse

In the standard form presented in section 2, the reason model does not accommodate restrictive reinterpretation. But it is not difficult to extend the reason model to accommodate the practice. Unfortunately, as I demonstrate below, extension of the reason model to accommodate restrictive reinterpretation leads to its collapse into the result model.

It will be useful, for the purposes of the present discussion, to focus only on restrictive reinterpretations of precedent—reinterpretations that reduce the constraining effect of the ratio of the precedent case. Reinterpretation restricts the constraining effect of a precedent when it replaces the previous ratio with a new ratio that includes a larger set of factors. More formally, we can define the restrictive reinterpretation $ {\varGamma}^{c/{R}^{\prime }} $ of a case base as the result of replacing some case $ c=\left\langle X,R,s\right\rangle $ in the original case base $ \varGamma $ with the case $ c^{\prime } $ that contains the restricted ratio $ R^{\prime } $ , where $ R\subseteq R^{\prime } $ .

A restrictive reinterpretation $ {\varGamma}^{c/{R}^{\mathrm{\prime}}} $ of a consistent case base $ \varGamma $ will be permissible just in case the set that results from replacing $ R $ with $ R^{\prime } $ in $ c $ is also a case base.

Note that the procedure in Definition 7 does not involve altering the fact scenario or outcome of the original case. Definition 7 also ignores the effects of judicial hierarchies on the permissibility of reinterpretation—I assume, for the purposes of discussion, that the present court is permitted to reinterpret any precedent in the case base.Footnote ³⁷

To illustrate the effect of a restrictive reinterpretation of a case base, consider the example offered in section 3 of a court wishing to reinterpret the precedent $ {c}_2=\left\langle {X}_2,{R}_{(2)}^{\pi },\pi \right\rangle $ , in which the court held in favor of the plaintiff (and thus that the prohibition was discriminatory) in fact scenario $ {X}_2=\left\{{f}_1^{\pi },{f}_2^{\pi },{f}_1^{\delta}\right\} $ , on the basis of the reason $ {R}_{(2)}^{\pi }=\left\{{f}_2^{\pi}\right\} $ , which represents the reason to find for the plaintiff because the animal that they wished to bring with them was an assistance animal. Suppose that the court wishes to reinterpret $ {c}_2 $ so that the judgment was instead based on the reason $ {R}_{\left(1,2\right)}^{\pi }=\left\{{f}_1^{\pi },{f}_2^{\pi}\right\} $ , representing the reason to find for the plaintiff on the basis that the animal involved was kept on a leash or in a cage and the passenger’s assistance animal. Letting $ {\varGamma}_2=\left\{{c}_2\right\} $ be a case base containing only $ {c}_2 $ , this restrictive reinterpretation can be represented in terms of the case base $ {\varGamma}_2^{{\mathrm{c}}_2/{R}_{\left(1,2\right)}^{\pi }} $ containing the sole case $ {c}_2^{\prime }=\left\langle {X}_2,{R}_{\left(1,2\right)}^{\pi },\pi \right\rangle $ .

The requirement that the reinterpretation be restrictive, so the ratio $ R $ is replaced by a larger set of factors $ {R}^{\prime } $ of which $ R $ is a subset, rules out reinterpretations that replace a ratio favoring outcome $ s $ with a reason favoring the opposing outcome $ \overline{s} $ . Recall that $ {\varGamma}_1 $ contains only the case $ {c}_1=\left\langle {X}_1,{R}_{(1)}^{\delta },\delta \right\rangle $ , in which the court ruled for the defendant (and thus that an animal could not accompany the passenger) on the basis of the reason $ {R}_{(1)}^{\delta } $ —that the animal presented a safety and hygiene risk to other passengers. The effects of the requirement that the reinterpretation be restrictive can be illustrated by considering the reinterpretation $ {\varGamma}_1^{{\mathrm{c}}_1/{R}_{(1)}^{\pi }} $ of $ {\varGamma}_1 $ that purports to reinterpret $ {c}_1 $ as a case decided on the basis of the reason $ {R}_{(1)}^{\pi } $ (the reason to decide for the plaintiff because the animal was on a leash or in a cage). This does not meet the requirements imposed by Definition 7, since $ {R}_{(1)}^{\pi } $ is not a subset of $ {R}_{(1)}^{\delta } $ .

The requirement that the restrictive reinterpretation be permissible rules out any attempt to replace a ratio $ R $ with a reason $ R^{\prime } $ that was not satisfied in the fact scenario of the case in which $ R $ was formulated. This can be illustrated by considering the impermissible reinterpretation $ {\varGamma}_1^{c_1/{R}_{\left(1,2\right)}^{\delta }} $ of $ {\varGamma}_1 $ , which reinterprets the precedent $ {c}_1=\left\langle {X}_1,{R}_{(1)}^{\delta },\delta \right\rangle $ as decided on the basis of the reason $ {R}_{\left(1,2\right)}^{\delta } $ , which includes both the factor that the animal in question was a hygiene or safety risk and another factor $ {f}_2^{\delta } $ . We can suppose that $ {f}_2^{\delta } $ represents the fact that the passenger failed to give advance notice to the airline that the animal would be accompanying them. The problem with such a reinterpretation is that the fact scenario $ {X}_1=\left\{{f}_1^{\pi },{f}_1^{\delta}\right\} $ does not satisfy $ {R}_{\left(1,2\right)}^{\delta } $ , because the passenger had not failed to give advance notice of the animal. The resultant set $ {\varGamma}_1^{c_1/{R}_{\left(1,2\right)}^{\delta }}=\left\{{c}_1^{\prime}\right\} $ contains the sole structure $ {c}_1^{\prime }=\left\langle {X}_1,{R}_{\left(1,2\right)}^{\delta },\delta \right\rangle $ . But $ {c}_1^{\prime } $ is not a case and $ {\varGamma}_1^{c_1/{R}_{\left(1,2\right)}^{\delta }} $ is not a case base, since $ {R}_{\left(1,2\right)}^{\delta } $ is not satisfied in $ {X}_1 $ .

Because of the constraints imposed on restrictive reinterpretations, their effect on the original case base is relatively conservative. Although they can reduce the constraining effect of the ratio of past cases, they cannot change the outcome. The constraints also prevent a restrictive reinterpretation from introducing inconsistency into a case base. Supposing a case base $ \varGamma $ is consistent, if a restrictive reinterpretation $ {\varGamma}^{c/{R}^{\prime }} $ of $ \varGamma $ is permissible in the sense identified in Definition 8, then $ {\varGamma}^{c/{R}^{\prime }} $ will also be consistent.

(All observations are restated and proved in the Appendix).

The relative conservativeness of restrictive reinterpretation can also be demonstrated in terms of the relationship between reinterpreted bodies of case law and the result model of precedent introduced in section 2. As we have already noted, restrictive reinterpretations are constrained by the facts of the cases that are being reinterpreted. More generally, we can see that if a decision with outcome $ s $ is required in fact scenario $ X $ against the background of some case base $ \varGamma $ on the material facts model, then for any permissible reinterpretation $ {\varGamma}^{c/{R}^{\prime }} $ a decision with outcome $ s $ will be required in $ X $ against the background of $ {\varGamma}^{c/{R}^{\prime }} $ on the standard factor-based account.

This result is unsurprising given the relationship between restrictive reinterpretations and the fact scenario of the reinterpreted case. The most extreme form of restrictive reinterpretation of a precedent case is a decision to confine the case to its facts—the precedent is associated, not with any more broadly applicable reason, but with a judgment that is made only on the balance of those reasons before the court.Footnote ³⁸ For any case $ \left\langle X,R,s\right\rangle $ , the set $ {X}^s $ of all factors included in $ X $ that favors outcome $ s $ represents the least-constraining ratio of that case—that is, the ratio that is formulated in terms of the largest set of factors supporting $ s $ satisfied by the fact scenario $ X $ . Any attempt to expand the ratio beyond the material facts of the case will be incoherent. To the extent that the material facts of a precedent case require a given conclusion, they constrain any restrictive reinterpretation.

On the other hand, if it is permitted, the practice of restrictive reinterpretation undoubtedly dilutes the constraining effect of precedent. The process of restrictive reinterpretation captured in Definition 7 is capable of iteration—any court could, supposing they were permitted to do so, reinterpret a series of cases in the case base to avoid each of their constraining effects. At its limit, such a process of iteration results in the collapse of the reason model into the result model, at least in terms of formal constraint. Since the ratio of every case in the case base could be replaced by the largest set of factors favoring its outcome satisfied on its material facts, it is only the material facts of the case that have any constraining effect.

To grasp this problem it is useful to consider a variation on the example formalized here in terms of the reinterpretation $ {\varGamma}_2^{c_2/{R}_{\left(1,2\right)}^{\pi }} $ , which represents a reinterpretation of the precedent $ {c}_2=\left\langle {X}_2,{R}_2^{\pi },\pi \right\rangle, $ in which the court framed its decision in terms of a reason $ {R}_{(2)}^{\pi } $ to decide for the plaintiff because the animal in question was an assistance animal in terms of the stronger reason $ {R}_{\left(1,2\right)}^{\pi } $ that also relies on the fact that the animal in question was kept on a leash or in a cage. A court deciding a case in the fact scenario $ {X}_3=\left\{{f}_2^{\pi },{f}_1^{\delta}\right\} $ could use the restrictive interpretation $ {\varGamma}_2^{{\mathrm{c}}_2/{R}_{\left(1,2\right)}^{\pi }} $ to avoid being required to find the plaintiff on the basis of $ {c}_2 $ . We now suppose that a court is deciding in the same factual scenario, $ {X}_3=\left\{{f}_2^{\pi },{f}_1^{\delta}\right\} $ , this time against the background of a case base $ {\varGamma}_3=\left\{{c}_2,{c}_3\right\} $ that contains both $ {c}_2 $ and the additional case $ {c}_3=\left\langle {X}_4,{R}_{(2)}^{\pi },\pi \right\rangle $ . Suppose that in $ {c}_3 $ the court found for the plaintiff (the passenger) in a fact scenario $ {X}_4=\left\{{f}_2^{\pi },{f}_3^{\pi },{f}_1^{\delta}\right\} $ , which included the additional plaintiff-favoring factor $ {f}_3^{\pi } $ , representing the fact that the passenger had purchased a separate ticket for the animal, again on the basis of the reason $ {R}_{(2)}^{\pi} $ (that the animal was an assistance animal). On the reason model, against the background of the case base $ {\varGamma}_3 $ , the court would once again be constrained in the fact scenario $ {X}_3 $ to decide for the plaintiff on the basis of the reason $ {R}_{(2)}^{\pi } $ . This outcome is forced by both cases in the case base; the priority orders derived from both $ {c}_2 $ and $ {c}_3 $ include the information that $ \left\{{f}_1^{\delta}\right\} $ is lower in priority than $ \left\{{f}_2^{\pi}\right\} $ .

Supposing that its power to reinterpret precedent is unconstrained, however, a court could offer successive restrictive reinterpretations of both $ {c}_2 $ and $ {c}_3 $ to avoid their constraining effects. First, the court could adopt the familiar restrictive reinterpretation $ {\varGamma}_3^{c_2/{R}_{(1,2)}^{\pi }} $ that replaces $ {c}_2 $ with the case $ {c}_2^{\mathrm{\prime}}=\langle {X}_2,{R}_{(1,2)}^{\pi },\pi \rangle $ . Let $ {\varGamma}_{3^{\mathrm{\prime}}}=\{{c}_2^{\mathrm{\prime}},{c}_3\} $ be the case base that results from this first reinterpretation. The court could then adopt the further reinterpretation $ {\varGamma}_{3^{\mathrm{\prime}}}^{{\mathrm{c}}_3/{R}_{(2,3)}^{\pi }} $ that reinterprets $ {c}_3 $ as decided on the basis of $ {R}_{\left(2,3\right)}^{\pi }=\left\{{f}_2^{\pi },{f}_3^{\pi}\right\} $ , replacing $ {c}_3 $ with $ {c}_3^{\prime }=\left\langle {X}_4,{R}_{\left(2,3\right)}^{\pi },\pi\;\right\rangle $ . Against the background of the case base $ \left\{{c}_2^{\prime },{c}_3^{\prime}\right\} $ that results from these successive reinterpretations, the court will no longer be constrained to find for the plaintiff in $ {X}_3 $ , since neither $ {R}_{\left(1,2\right)}^{\pi } $ (the ratio of the reinterpreted) $ {c}_2 $ nor $ {R}_{\left(2,3\right)}^{\pi} $ (the ratio of the reinterpreted) $ {c}_3 $ is satisfied in $ {X}_3 $ . It is also easy to see that if both restrictive reinterpretations are adopted, then the constraints imposed by the resultant case base $ \left\{{c}_2^{\prime },{c}_3^{\prime}\right\} $ on the reason model are no different from the constraints imposed by the original case base $ {\varGamma}_3 $ on the result model. Both rationes have been replaced with the largest set of factors favoring the outcome $ \pi $ satisfied on the facts of their respective cases.

At its limit, this process of iterated reinterpretation of a case base will result in the replacement of every case $ \left\langle X,R,s\right\rangle $ in the case base with a new case $ \left\langle X,{X}^s,s\right\rangle $ , in which the original ratio $ R $ is replaced with the set $ {X}^s $ of all factors favouring the outcome $ s $ of the case. Each case, in other words, could be reinterpreted as confined to its facts. This process of iterative reinterpretation would lead to the collapse of the reason model into the result model, as the following observation confirms.

As I demonstrate in the Appendix, Observation 3 follows directly from Horty’s observations about the relationship between the reason model and the result model.Footnote ³⁹ What this result tells us is that if each case in the case base is reinterpreted as restrictively as possible, so that its ratio is understood in terms of the strongest possible reason (that is, for the largest set of factors) favoring the outcome satisfied on the facts of the case, then the formal constraints imposed by the reason model are identical to the formal constraints imposed by the result model of precedent. Once again, this result is unsurprising when viewed in light of the formal relationship between the reason and result models that Horty demonstrates. Nonetheless, I suspect that proponents of the reason model are likely to be disturbed by the collapse. I return to considering its implications in section 6.

V. Restrictive Reinterpretation as Constrained Overruling

As I observed in section 3, the practice of reinterpreting precedent appears to threaten the viability of the reason model on two opposing fronts. Against the rule model, a proponent of the reason model must offer an account of courts’ use of restrictive reinterpretations that preserves a distinction between this practice and distinguishing precedent. Against the result model, on the other hand, the proponent of the reason model needs to defend the attribution of constraining effect to courts’ formulations of their reasons for judgment. Having presented both of these challenges, in the following sections I wish to offer a qualified defense of the rule model. I will begin by examining the relationship between restrictive reinterpretation and distinguishing. I will follow Horty in arguing that restrictive reinterpretation is a form of overruling and therefore should not be understood as a form of distinguishing.Footnote ⁴⁰ I suggest, however, that we should view restrictive reinterpretation as a form of constrained overruling: unlike more straightforward instances of overruling, restrictive reinterpretation is at least constrained by the facts of the precedent case. As a result of this constraint, restrictive reinterpretation is less destructive than overruling.

At least superficially, distinguishing and reinterpretation are very similar. Both practices rely on the later court’s ability to draw a distinction between the facts of the precedent case and the new case before the court. Nonetheless, I think that proponents of the reason model can and should resist accounts that assimilate distinguishing and reinterpretation into a single form of rule modification. The extended framework developed in section 4 allows proponents of the reason model to offer a semantic characterization of the difference between distinguishing and reinterpreting. This characterization, moreover, supports the view that two practices are rooted in different sources of constraint on the authority of the precedent court.

The meaningfulness of the distinction between reinterpreting and distinguishing can be illustrated by their differing effect on the priority information that can be derived from a particular case base. Distinguishing, as Horty notes, has a uniformly constructive effect on the priority information that can be derived from a case base.Footnote ⁴¹ Recall the example featured in section 2, in which a court ruled for the plaintiff on the basis that the animal they wished to bring with them was their assistance animal, against the background of a case base $ {\varGamma}_1 $ that contained the single decision $ {c}_1 $ . In $ {c}_1=\left\langle {X}_1,{R}_1^{\delta },\delta \right\rangle $ the court ruled in favor of the defendant, on the basis of reason $ {R}_1^{\delta }=\left\{{f}_1^{\delta}\right\} $ , in the fact scenario $ {X}_1=\left\{{f}_1^{\pi },{f}_1^{\delta}\right\} $ . The complete priority order derived from the case base $ {\varGamma}_1 $ containing only $ {c}_1 $ is therefore $ \{{f}_1^{\pi}\}{<}_{\Gamma_1}\{{f}_1^{\delta}\} $ . The court’s new decision is represented by the case $ {c}_2=\left\langle {X}_2,{R}_{(2)}^{\pi },\pi \right\rangle $ , where $ {X}_2=\left\{{f}_1^{\pi },{f}_2^{\pi },{f}_1^{\delta}\right\} $ . In other words, the decision in $ {c}_2 $ distinguished the decision in $ {c}_1 $ from the problem represented by the fact scenario $ {X}_2 $ on the basis of the reason $ {R}_{(2)}^{\pi } $ (representing the reason to decide for the plaintiff because the animal was an assistance animal). The decision in $ {c}_2 $ is associated with a derived priority order that includes $ \left\{{f}_1^{\delta}\right\}{<}_{c_2}\left\{{f}_2^{\pi}\right\} $ and $ \left\{{f}_1^{\delta}\right\}{<}_{c_2}\left\{{f}_1^{\pi },{f}_2^{\pi}\right\} $ , providing information that cannot be derived from $ {c}_1 $ . Where $ {\varGamma}_4={\varGamma}_1\cup \left\{{c}_2\right\} $ is the case base that results from adding $ {c}_2 $ to $ {\varGamma}_1 $ , the priority order $ {<}_{\Gamma_4} $ derived from $ {\varGamma}_4 $ is an extension of the priority order $ {<}_{\varGamma_1} $ derived from $ {\varGamma}_1 $ . It can be shown, in other words, that $ {<}_{\varGamma_4} $ contains priority information that $ {<}_{\varGamma_1} $ does not.

On the other hand, as Horty also notes, restrictive reinterpretations have a destructive effect, since they subtract from the priority information that can be derived from a particular case base.Footnote ⁴² The effect of a restrictive reinterpretation is always to replace the ratio of some set of precedents in the original case base with a larger set of factors. This reduces the priority information that can be derived from any reinterpreted case. Consider, by way of example, the difference between the original case base $ {\varGamma}_2=\{{c}_2\} $ and its reinterpretation $ {\varGamma}_2^{{\mathrm{c}}_2/{R}_{(1,2)}^{\pi }}=\{{c}_2^{\mathrm{\prime}}\} $ , replacing the ratio $ {R}_{(2)}^{\pi } $ with $ {R}_{\left(1,2\right)}^{\pi } $ in $ {c}_2 $ . The original case base $ {\varGamma}_2 $ contains the sole case $ {c}_2=\left\langle {X}_2,{R}_{(2)}^{\pi },\pi \right\rangle $ . According to its derived priority order, we have both that $ \{{f}_1^{\delta}\}{<}_{\varGamma_2}\{{f}_2^{\pi}\} $ and $ \left\{{f}_1^{\delta}\right\}{<}_{\Gamma_3}\left\{{f}_1^{\pi },{f}_2^{\pi}\right\} $ . On the other hand, the reinterpretation $ {\varGamma}_2^{{\mathrm{c}}_2/{R}_{(1,2)}^{\pi }} $ contains the sole case $ {c}_2^{\prime }=\left\langle {X}_2,{R}_{\left(1,2\right)}^{\pi },\delta \right\rangle $ , with the resultant derived priority order $ {<}_{c_2^{\prime }} $ containing only the information that $ \left\{{f}_1^{\delta}\right\}{<}_{c_2^{\prime }}\left\{{f}_1^{\pi },{f}_2^{\pi}\right\} $ . The reinterpretation of $ {\varGamma}_2 $ replacing $ {R}_{(1)}^{\pi } $ with $ {R}_{\left(1,2\right)}^{\pi } $ in $ {c}_2 $ results in removal of priority information.

More formally, it can be demonstrated that for any permissible restrictive reinterpretation $ {\varGamma}^{c/R} $ , the priority order $ {<}_{\varGamma } $ derived from the original case base $ \varGamma $ is an extension of the priority order $ {<}_{\varGamma^{c/R}} $ , in the sense that for any two reasons $ U $ and $ V $ , if $ U{<}_{\varGamma^{c/R}}V $ then $ U{<}_{\varGamma }V $ . The priority order derived from $ \varGamma $ contains at least the same information as the priority order derived from $ {\varGamma}^{c/R} $ . As the previous example shows, however, it may be possible to derive priority information from $ \varGamma $ that cannot be derived from $ {\varGamma}^{c/R} $ .

To understand the significance of this observation, consider the difference between the case base $ {\varGamma}_4=\left\{{c}_1,{c}_2\right\} $ , which is the result of adding $ {c}_2 $ (distinguishing $ {c}_1 $ ) to the case base $ {\varGamma}_1=\{{c}_1\} $ , and the reinterpreted case base $ {\varGamma}_2^{{\mathrm{c}}_2/{R}_{(1,2)}^{\pi }} $ . Recall that the addition of $ {c}_2 $ means that $ {\varGamma}_4 $ contains additional priority information—namely that $ \{{f}_1^{\delta}\}{<}_{\varGamma_4}\{{f}_2^{\pi}\} $ . Although it allows courts to avoid the constraining effect of other precedents, distinguishing increases the constraining effect of the case base as a whole. Suppose a court is faced with the fact scenario $ {X}_3=\left\{{f}_2^{\pi },{f}_1^{\delta}\;\right\} $ , which includes only the defendant-favoring factor that the animal presented a safety or hygiene risk and the plaintiff-favoring factor that the animal was the passenger’s assistance animal, against the background of the case base $ {\varGamma}_4 $ . In this fact scenario, due to the inclusion of $ {c}_2 $ in the case base, the court would be constrained to decide for the plaintiff (the passenger) on the basis of ratio $ {R}_{(2)}^{\pi }=\left\{{f}_2^{\pi}\right\}. $ Compare this with a situation in which the court decides in the same fact scenario, $ {X}_3, $ against the background of the case base $ {\varGamma}_2^{{\mathrm{c}}_2/{R}_{\left(1,2\right)}^{\pi }} $ that reinterprets $ {c}_2 $ as decided on the basis that the animal in question was both on a leash or in a cage and the passenger’s assistance animal. The effect of this reinterpretation is that the case base has less constraining effect. Because the reinterpreted case base no longer supports the conclusion that $ \left\{{f}_1^{\delta}\right\} $ has lower priority than $ \left\{{f}_2^{\pi}\right\} $ , a court would no longer be required to decide for the plaintiff in $ {X}_3. $

I think that the differing effects of reinterpreting and distinguishing on the underlying priority order reflect the fact that the practices are grounded in different sources of constraint on the authority of precedent court. A court’s permission to distinguish is derived directly from the fact-sensitivity of the original court’s authority. The precedent court is only vested with the power to make authoritative determinations on the facts before it.Footnote ⁴³ A precedent’s authority cannot expand beyond the facts on which it was decided. Distinguishing is thus by its very nature constructive—it involves the consideration of reasons that were not considered by the precedent court. On the other hand, because restrictive reinterpretation is destructive, rather than constructive, it requires direct derogation from the authority of the precedent court.

Although little seems to have been written about the practice of restrictive reinterpretation, commentators do seem to have noted that it has this derogative effect. Re’s treatment of the practice of “narrowing” precedent in US law argues that it is legitimate only where the precedent is ambiguous. Without such ambiguity, the adoption of a restrictive reinterpretation of a precedent case is treated as partial overruling and, therefore, requires special justification.Footnote ⁴⁴ Writing about use of precedent in English law, Cross and Harris argue that empowering later courts to develop their own interpretations of precedent is an important countermeasure to the power given to precedent-setting courts to offer rationes for their decisions.Footnote ⁴⁵ Where the permissibility of distinguishing is a reflection of the limits of the precedent court’s authority, the power to reinterpret, if it is to be granted at all, must be understood as a power to alter the effect of the precedent court’s judgment. To the extent that the practice of reinterpretation can be justified, it must be justified as an express conferral of power on the later court to derogate from the authority of the precedent court.

It would nonetheless be a mistake to view restrictive reinterpretation as indistinguishable from simple overruling. Restrictive reinterpretation is constrained, at least, by the facts of the cases that are being reinterpreted and their outcomes. As I noted in section 4, the constraints placed on the permissibility of restrictive reinterpretation rule out any attempt to replace the ratio of a case with a reason that was not satisfied on the fact scenario of the case in which the ratio was formulated, or with a reason that favored the opposing outcome. In section 4, I demonstrated that these constraints ensure that restrictive reinterpretation cannot introduce inconsistency into a reinterpreted case base, nor can it allow a court to reach a result that would be impermissible on the result model of precedent. These constraints also distinguish restrictive reinterpretation from more typical instances of overruling.

In a typical case of overruling, the entire precedent case is removed from the case base “root and branch.”Footnote ⁴⁶ This means that the overruled case is no longer a source of any priority information whatsoever. When a case $ c=\left\langle X,R,s\right\rangle $ is removed from a case base $ \varGamma $ , its derived priority order $ {<}_c $ is likewise removed entirely from the case base’s derived priority order $ {<}_{\varGamma } $ . Where a restrictive reinterpretation of the case $ c=\left\langle X,R,s\right\rangle $ is adopted, however, then the revised case base retains, at minimum, the priority information provided by reinterpreting its ratio $ R $ in terms of $ {X}^s $ , the largest possible set of factors favoring the case’s outcome. The restrictive reinterpretation $ {\varGamma}_2^{{\mathrm{c}}_2/{R}_{\left(1,2\right)}^{\pi }} $ , for instance, still contains the case $ {c}_2^{\prime }=\left\langle {X}_2,{R}_{\left(1,2\right)}^{\pi },\pi \right\rangle $ . The resultant derived priority order $ {<}_{c_2^{\prime }} $ therefore contains the information that $ \{{f}_1^{\delta}\}{<}_{c_2^{\mathrm{\prime}}}\{{f}_1^{\pi },{f}_2^{\pi}\} $ . Had $ {c}_2 $ been subject to wholesale overruling, on the other hand, the case would have been removed entirely from the case base, along with any priority information derived from it.

VI. The Reason Model and the Role of Rationes

If the practice of adopting restrictive reinterpretations is treated as permissible, then this appears to vindicate the claim that it is the material facts of the case and its outcome, rather than the precedent court’s formulation of its reasons for judgment, that have any constraining effect. As I demonstrated in section 4, in its most extreme form the process of restrictive reinterpretation could be repeated for every case in the case base, resulting in a collapse of the reason model into the result model. Some proponents of the reason model will take this to be an unacceptable consequence of extending the model to allow for restrictive reinterpretation. As Horty notes, the orthodoxy among legal theorists seems to be that “the reasons offered by a court to justify its decisions must be taken seriously as an element influencing precedential constraint.”Footnote ⁴⁷ For this reason, proponents of the reason model have generally been at great pains to avoid any collapse of the reason model into the result model.Footnote ⁴⁸

The collapse result could be avoided entirely if we followed Horty in viewing restrictive reinterpretation as nothing other than a form of overruling. Unfortunately, I do not think this is an adequate response. Restrictive reinterpretation is undeniably an aspect of judicial practice, at least in some appellate jurisdictions. As I noted in section 5, moreover, the practice is subject to important constraints that distinguish it from simple overruling. In this section, I will offer a more qualified response to the reason model’s collapse into the result model.Footnote ⁴⁹ Elsewhere, Horty and Broughton have demonstrated that other plausible extensions of the reason model result in similar forms of collapse into the result model.Footnote ⁵⁰ I think these results, alongside the result demonstrated in this paper, should lead to broader reflection on the constraining effect of precedent courts’ formulations of their reasons for judgment.

This section will explore the possibility that these formulations, which are in any case are rarely easy to identify, are not a source of formal constraint. This approach to differentiating the two models is not entirely unprecedented. Although he later revised his view, Horty initially suggested that the result model and reason model could be differentiated by placing additional constraints on the acceptability of the rationales formulated by courts to justify their decisions.Footnote ⁵¹ In presenting his version of the reason model, Lamond places little emphasis on the court’s formulation of its reasons. What we typically call the “ratio” of a judgment is “rarely expressly stated by the precedent court; and even if it is stated, it is the substance of the decision, not the formulation of the ruling, that is ultimately binding.”Footnote ⁵² Lamond suggests that, rather than being immediately obvious, the ratio is typically constructed from the precedent decision by subsequent courts.Footnote ⁵³ Building on Lamond’s observations, I will argue here that even where courts do offer a clear formulation of their reasons for judgment, that formulation is only a source of informal constraint on subsequent courts in their interpretation of the substance of the decision. Formally, interpretation of rationes is constrained by the facts treated as material in the precedent case. Informally, the interpretation of the ratio of a precedent case is constrained by its reception in related case law. The interplay between these two forms of constraint can raise the salience of a particular interpretation of the precedent, making its effects harder to avoid through reinterpretation.

At the outset, it is important to note that the collapse result rests on the assumption that the practice of restrictive reinterpretation is permissible and unrestricted. Given the destructive effects of restrictive reinterpretations demonstrated in section 5, however, there are good reasons to limit the permissibility of the practice to jurisdictions that are also empowered to overrule precedents. The extended model developed in section 4 may only be appropriate for modelling judicial reasoning at the highest appellate levels, where courts appear to be given relatively broad power to reinterpret both their own decisions and the decisions of lower courts. The collapse of the reason model into the weaker result model is less concerning in the context of modelling precedential constraint in these jurisdictions than it would be if the practice of restrictive reinterpretation were a commonplace of precedential reasoning at every level of the judicial hierarchy.

It is also important to note that although there is a limit at which reinterpretation of every case would mean that the result model and the reason model coincide in their constraining effects, this does not indicate that the reason model is redundant. As Horty suggests, the reason model allows us to develop a more charitable interpretation of the material facts account, according to which the true constraining effect of the precedent $ c=\left\langle X,R,s\right\rangle $ is associated with an implicit decision on the basis of the set $ {X}^s $ of all applicable factors favoring its outcome.Footnote ⁵⁴ More importantly, without the additional information provided by the reason model, the practice of reinterpretation is simply unintelligible—there is no formulation of the ratio for a subsequent court to reinterpret. Suppose a court is faced with the precedent $ \left\langle X,R,s\right\rangle $ in a fact scenario $ Y $ that satisfies the ratio $ R $ formulated by the precedent court, but where the set of all factors favoring outcome $ s $ in the new fact scenario $ {Y}^s $ is smaller than the set of all factors $ {X}^s $ that favored the same outcome in the precedent case. On the result model, the precedent case is not a source of any constraint whatsoever. The reason model, on the other hand, is able to explain why a court in the new fact scenario is at least constrained to offer a reinterpretation of the precedent: as it was formulated by the precedent court, the ratio $ R $ is satisfied in the new fact scenario $ Y. $

Finally, and relatedly, the fact restrictive reinterpretation is permissible in a given jurisdiction does not necessarily indicate that subsequent courts are completely unconstrained in the reinterpretations they can offer. In section 5, I noted that the process of reinterpretation is formally constrained by the material facts of the precedent case. In addition, we can use the model of reinterpretation developed in section 4 to explore the idea that reinterpretation of an individual case is subject to further constraints imposed by surrounding case law.

As I have already noted, in its standard form the reason model takes the ratio to be exogenously fixed. Most presentations of the reason model thus proceed on the discursive assumption that the ratio of a case can be clearly identified from the language used by the precedent court. But on a different view of common law reasoning, it is not the ratio as it is formulated by the precedent court that is a source of formal constraint. In English law, it is still relatively rare for courts to expressly state the ratio of the precedent, and a settled ratio for a precedent often emerges from a process of construction in subsequent cases.Footnote ⁵⁵ Even supposing the rationes of precedents are subject to restrictive reinterpretation, it seems plausible that the reinterpretations adopted might be constrained in various ways by the subsequent reception of the precedent in surrounding case law. Developed bodies of case law will entrench certain interpretations of a precedent’s ratio and make it harder for the court to entertain arguments for restricting the ratio of a past case. The tendency of the rationes of cases to be revealed through repeated application is a recognizable feature of common law reasoning.Footnote ⁵⁶

Consider the example involving the case base $ {\varGamma}_3=\{{c}_2,{c}_3\} $ that was introduced in section 4, where $ {c}_2=\left\langle {X}_2,{R}_{(2)}^{\pi },\pi \right\rangle $ and $ {c}_3=\left\langle {X}_4,{R}_{(2)}^{\pi },\pi \right\rangle $ are both cases in which the court justified its decision for the plaintiff on the basis of the reason $ {R}_{(2)}^{\pi } $ . As we demonstrated in section 4, although a court faced with the new fact scenario $ {X}_3=\left\{{f}_2^{\pi },{f}_1^{\delta}\right\} $ against the background of $ {\varGamma}_3 $ would ordinarily be required to find for the plaintiff on the basis of $ {R}_{(2)}^{\pi } $ , they could avoid this requirement by selectively reinterpreting both cases in the case base. First, the court could adopt restrictive reinterpretation $ {\varGamma}_3^{c_2/{R}_{(1,2)}^{\pi }} $ that replaces $ {c}_2 $ with the case $ {c}_2^{\prime }=\left\langle {X}_2,{R}_{\left(1,2\right)}^{\pi },\pi\;\right\rangle $ . The court could then reinterpret $ {c}_3 $ as decided on the basis of $ {R}_{\left(2,3\right)}^{\pi }=\left\{{f}_2^{\pi },{f}_3^{\pi}\right\} $ , replacing $ {c}_3 $ with $ {c}_3^{\prime }=\left\langle {X}_4,{R}_{\left(2,3\right)}^{\pi },\pi\;\right\rangle $ . In presenting the collapse result in section 4, I proceeded on the assumption that courts were completely unconstrained in their ability to reinterpret cases in this way. This assumption seems to me to be unwarranted. Suppose, for example, that the court’s judgment in $ {c}_3 $ involved an explicit application of the ratio formulated by the court in $ {c}_2 $ . This would indicate that the court in $ {c}_3 $ agreed with the earlier court’s characterization of their decision as justified solely by the reason to decide for the plaintiff because the animal in question was their assistance animal. In these circumstances, the idea of interpreting $ {c}_2 $ and $ {c}_3 $ as decided on the basis of different reasons would be, if not formally impermissible, then improper. Alternatively, we might suppose that although $ {c}_2 $ and $ {c}_3 $ were independent decisions, both cases are “related” in the sense that they addressed similar legal problems. The more frequently a ratio has been applied in related factual scenarios, the less appropriate it will be to reinterpret one case without reinterpreting the other in the same way.

The account of restrictive reinterpretation developed in section 4 can be modified to capture the constraining effect that surrounding case law seems to have on reinterpretation. Unfortunately, the reason model does not allow us to track when two cases have any sort of doctrinal relationship. It only allows us to identify when two cases share the same ratio. In the preceding example, it seems quite plausible that $ {c}_2 $ and $ {c}_3 $ are related to one another in some doctrinal sense, so that the two decisions can be seen as offering mutual support to one another. Both seem to give the reason $ {R}_{(2)}^{\pi } $ greater salience or importance than other plaintiff-favoring reasons satisfied in the same cases. The two cases share the same defendant-favoring factors, share a common ratio, and differ only in terms of the plaintiff-favoring factors $ {f}_1^{\pi } $ , which is in $ {c}_2 $ but not $ {c}_3 $ , and $ {f}_3^{\pi }, $ which is in $ {c}_3 $ but not $ {c}_2 $ . But what if $ {c}_3 $ featured entirely different defendant-favoring factors? It seems implausible that in that scenario the reinterpretation of $ {c}_2 $ would have any implications at all for the interpretation of $ {c}_3 $ , or vice versa. The two cases would be responding to substantially different legal problems.

In response to this difficulty, an account of the constraining effect of surrounding case law might be developed in any number of ways. One possibility would be to extend our case structures to include information about whether other cases were followed or distinguished in the decision. Another possibility is to develop a metric of similarity between cases that allows us to determine when two cases are sufficiently similar to one another that they cannot be interpreted or reinterpreted in isolation. Here, largely for purposes of illustration, I will explore a very simple version of the latter approach, which stipulates that any two cases are “related” for the purposes of reinterpretation if they share the same ratio favoring outcome $ s $ as well as sharing factors favoring the opposing outcome $ \overline{s} $ . (I use the function $ Fact{s}^s $ to identify a set of factors favoring the outcome $ s $ satisfied in the fact scenario of a case. So, for the case $ {c}_1=\left\langle {X}_1,{R}_{(1)}^{\delta },\delta \right\rangle $ where the fact scenario $ {X}_1=\left\{{f}_1^{\pi },{f}_1^{\delta}\right\} $ , $ Fact{s}^{\pi}\left({c}_1\right)=\left\{{f}_1^{\pi}\right\} $ .)

What this definition tells us is that, for our purposes, for some case $ c $ we can find a set of related cases by identifying any case that has the same ratio as $ c $ and shares with it at least one factor that supported the opposing outcome. In the case base $ {\varGamma}_3 $ , for instance, the decision $ {c}_3 $ will be related to $ {c}_2 $ , since both cases involve the same plaintiff-favoring ratio $ {R}_{(2)}^{\pi } $ , and share the common defendant-favoring factor $ {f}_1^{\delta } $ . This account of what makes two fact scenarios sufficiently similar to be related is quite weak—they need only to satisfy the same ratio and share at least one opposing factor. I use it only for the purposes of illustration. The metric of similarity could be strengthened in several ways.

The account of restrictive reinterpretation offered in section 4 can then be expanded, so that—rather than restricting the ratio of a single case—any restrictive reinterpretation of a case must also be performed on all related cases in a case base. (It is easy to see that given Definition 9, the relation of being related to $ c $ by its ratio and facts is reflexive: any case c is related to itself. Thus, for any case $ c\in \varGamma $ the set of cases $ Relate{d}_{\varGamma }(c) $ is guaranteed to contain at least $ c $ .)

• • For each case $ {c}^{\prime }=\left\langle Y,R,s\right\rangle $ in the set of cases $ Relate{d}_{\varGamma }(c), $ replace c’ in $ \varGamma $ with the case $ {c}^{\prime \prime }=\left\langle Y,{R}^{\prime },s\right\rangle $ .

The permissibility of a restrictive reinterpretation will then be defined as it was in Definition 8, in terms of the requirement that the set resulting from restrictive reinterpretation of related cases also be a case base.

The purpose of the approach to restrictive reinterpretation of related cases outlined in Definitions 9 and 10 is to reflect the tendency for more developed bodies of case-law to entrench certain interpretations of a ratio and make it harder for the court to entertain arguments for its reinterpretation. Once again, consider the case base $ {\varGamma}_3=\{{c}_2,{c}_3\} $ . Clearly, $ {c}_2 $ and $ {c}_3 $ are related by the ratio $ {R}_{(2)}^{\pi } $ and their common defendant-favoring factor, $ {f}_1^{\delta }. $ In light of the inclusion of $ {c}_3 $ in $ {\varGamma}_3 $ , the reinterpretation $ {\varGamma}_3^{c_2^{\ast }/{\mathrm{R}}_{(1,2)}^{\pi }} $ that replaces the ratio $ {R}_{(2)}^{\pi } $ of $ {c}_2 $ with the stronger $ {R}_{\left(1,2\right)}^{\pi } $ in all related cases is not permissible. Given the restrictions placed on restrictive reinterpretations of related cases by Definitions 9 and 10, the reinterpretation $ {\varGamma}_3^{c_2^{\ast }/{\mathrm{R}}_{(1,2)}^{\pi }} $ would also reinterpret the ratio of $ {c}_3 $ , leading to inclusion of a structure $ \left\langle {X}_4,{R}_{\left(1,2\right)}^{\pi },\pi \right\rangle $ in the resultant set that does not conform to the requirements placed on a case (because $ {R}_{\left(1,2\right)}^{\pi } $ is not satisfied in $ {X}_4 $ ).

The constraints that repeated application of a ratio in related cases places on restrictive reinterpretations can also be illustrated using a similar form of visual representation to that which was deployed in section 2. In Figure 2, the ratio $ R $ has been applied in two related cases: $ \left\langle X,R,s\right\rangle $ and $ \left\langle Y,R,s\right\rangle $ . The shaded area represents the space of alternative permissible restrictive reinterpretations of either case restricting $ R $ (noting that the restriction of a ratio is reflected by an expansion of the set of factors). One such restriction, represented by the set of factors $ R^{\prime } $ , has been included in the diagram. In Figure 3, we assume that the original ratio $ R $ was applied in a further related case $ \left\langle Z,R,s\right\rangle $ , where the court was not constrained to decide for s on the basis of $ R $ but nonetheless opted to do so. The addition of $ \left\langle Z,R,s\right\rangle $ to the case base shrinks the space of permissible reinterpretations. A restrictive reinterpretation replacing $ R $ with $ R^{\prime } $ in all related cases is now impermissible, since its expanded set of factors is not satisfied in the fact scenario $ Z $ .

Figure 2. The reinterpretation of $ \left\langle X,R,s\right\rangle $ and the related case $ \left\langle Y,R,s\right\rangle $ replacing $ R $ with $ {R}^{\prime } $ in both cases is represented. The space of alternative permissible restrictive reinterpretations is shaded.

Figure 3. The addition of the further related case $ \left\langle Z,R,s\right\rangle $ to the casebase shrinks the space of permissible restrictive reinterpretations. It is no longer permissible to replace $ R $ with $ {R}^{\prime } $ .

On this approach, as a case base grows, it is possible—and probably likely—that the scope for permissible restrictive reinterpretations restricting a ratio will shrink. The more a ratio has been applied in related fact scenarios, the harder it will become to satisfy these basic constraints on reinterpretation. The ratio becomes more settled and harder to avoid by relying on restrictive reinterpretation.

There are two ways in which the constraints that surround case law can place on the interpretation of its ratio might be understood. On the first understanding, we could take the further constraints on reinterpretation imposed by Definitions 9 and 10 to reflect formal constraints on the permissibility of reinterpreting a case, so that it is simply impermissible for a court to adopt a restrictive reinterpretation of a case without reinterpreting related cases. There is something to be said for this idea. The more deeply embedded a particular ratio is in the case base—the more frequently it has been applied in diverse fact scenarios by previous courts—the harder it will be to reinterpret. Even supposing that courts have the power to restrictively reinterpret precedent, it seems reasonable to think that power is not entirely unfettered. Interpreting these requirements as strict requirements, however, is inconsistent with the idea that the constraining effect of precedents arises from the requirement to treat precedent cases as correctly decided on their facts. The notion of factual similarity incorporated in Definitions 9 and 10 allows for rationes to have a constraining effect that reaches beyond the facts of the cases in which they have been applied.

On the second way of understanding these constraints, which I am inclined to think is more plausible, they reflect the limits placed by general considerations of coherence and intelligibility on reinterpretation.Footnote ⁵⁷ On this understanding, the further constraints imposed by the addition of related cases to the case base do not affect whether it is legally permissible to adopt a restrictive reinterpretation of a single case without considering surrounding case law, but rather represent general considerations that help to determine whether the adoption of such a reinterpretation is justifiable or socially acceptable. In the case of a court seeking to reinterpret the case base $ {\varGamma}_3=\left\{{c}_2,{c}_3\right\} $ , it seems plausible that a court could, if it wished, adopt a restrictive reinterpretation of $ {c}_2 $ without also reinterpreting $ {c}_3 $ in the same way—this reinterpretation can be adopted without implying that $ {c}_3 $ was wrongly decided on its facts. But once the decision in $ {c}_3 $ is taken to have some important doctrinal relationship to $ {c}_2 $ , the decision to selectively reinterpret $ {c}_2 $ by adopting a reinterpretation that is not available in $ {c}_3 $ will seem opportunistic or unprincipled. It will be permissible, in a formal sense, to reinterpret $ {c}_2 $ without similarly reinterpreting $ {c}_3 $ , but it will be difficult for a court to offer a coherent and satisfying justification for doing so.