The Camel's Nose Is in the Tent: Rules, Theories, and Slippery Slopes

Mario J. Rizzo* and Douglas Glen Whitman**
51 UCLA L. Rev. 539

Slippery slopes have been the topic of a spate of recent literature.  In this Article, the authors provide a general theory for understanding and evaluating slippery slope arguments and their associated slippery slope events.  The central feature of the theory is a structure of discussion within which all arguments take place.  The structure is multilayered, consisting of decisions, rules, theories, and research programs.  Each layer influences and shapes the layer beneath:  Rules influence decisions, theories influence the choice of rules, and research programs influence the choice of theories.  In this structure, slippery slope arguments take the form of meta-arguments, as they purport to predict the future development of arguments in the structure of discussion.  Evaluating such arguments requires knowledge of the specific content of the structure of discussion itself.  The Article then presents four viable types of slippery slope arguments; draws attention to four different factors that, other things equal, tend to increase the likelihood of slippery slopes; and explores a variety of strategies for coping with slippery slopes.

I.  Definitions and Clarifications
    A.  Essential Characteristics of the Slippery Slope Argument
    B.  Slippery Slope Arguments and Slippery Slope Events
II.  The Structure of Discussion and Argument:  The Microfoundations of Slippery Slopes
    A.  Rules
    B.  Theories
    C.  Research Programs
    D.  Arguments
    E.  Example:  Make-Up Exam Meshugas
    F.  Arguments About Arguments
III.  A Microprocess:  Sorites With Precedent
IV.  Macroprocesses
    A.  Altered Economic Incentives
    B.  Separately Validated Propositions
    C.  Humean Beneficence
V.  Responding to the Objections
VI.  Factors Affecting the Likelihood of Slippery Slopes
VII.  Coping With Slippery Slopes
    A.  Accepting the Trade-Off
    B.  Stipulating an Arbitrary Stopping Rule
    C.  Appealing to a Higher Standard
    D.  Adopting an Open-Ended or Standard-Mediated Rule
    E.  Altering the Scope and Power of Precedent
    F.  Establishing Presumptions, Burdens of Persuasion, or Levels of Scrutiny
    G.  Creating Supermajority Requirements and Constitutional Constraints

"The question of questions for the politician should ever be - "What type of social structure am I tending to produce?' But this is a question he never entertains."
- Herbert Spencer, The Coming Slavery

"If you accept a seemingly appropriate argument now, you will be more likely to accept an inappropriate argument later. And if you accept such an inappropriate argument, you will be more likely to make a bad decision or perform a dangerous act." This is a common, general form of the so-called slippery slope argument. More specific slippery slope arguments occur in public policy (instituting a price ceiling on milk will lead to price controls on the sale of cowsn1, in law (forbidding the Nazis to march in Skokie will lead to the forbidding of valuable speech that hurts the feelings of religious or ethnic groups n2), in ethics (acceptance of the abortion of a month-old fetus will lead  [*541]  to acceptance of the abortion of third-trimester fetuses or even to infanticiden3), and, indeed, in almost every arena where decisions must be made.

Slippery slope arguments have been used by thinkers from across the political spectrum. These arguments can be found in the writings of twentieth and twenty-first century civil libertarians, opponents of euthanasia, opponents of some frontier medical procedures including genetic engineering and cloning, and advocates and opponents of abortion rights. A shared characteristic of these arguments is that they are used to oppose some type of change in the status quo. In that sense, and only in that sense, slippery slope arguments are usually employed for "conservative" purposes.

The scholarly literature on slippery slopes and their related arguments is not extensive. Recently, however, there have been some important contributions. Sanford Ikeda has analyzed tendencies toward expansion that inhere in the state's intervention in the economy.n4 Eugene Volokh has produced a wide-ranging study of various possible slippery slope mechanisms in both judicial decisionmaking and legislative action.n5 And Douglas Walton has critically analyzed the various forms of slippery slope arguments used in many types of public debate.n6 None of these authors, however, has attempted to construct a unified framework in which such arguments can be studied and related to actual or potential slippery slope events. This is what this Article attempts to provide.

Sometimes slippery slopes appear to involve only actions: One action leads to another. But in the cases of law, ethics, and public policy, the actions usually require justification. Hence, first and foremost, slippery slopes are slopes of arguments: One practical argument tends to lead to another, which means that one justified action, often a decision, tends to lead to another. When we say that one argument (and its supported action) tends to lead to another, we mean that it makes the occurrence of the subsequent argument more likely, not that it necessarily makes it highly likely or, still less, inevitable.n7 Hence the transition between arguments is not based on strict logical entailment.

 [*542]  The ubiquity of slippery slope arguments should not lead us to believe they are unproblematic or simple in structure. First, there is no single paradigm of a slippery slope argument. Walton has distinguished four types: sorites, precedent, causal, and full (a complex combination of the first three).n8 Our concerns do not exactly parallel these distinctions. In particular, our concerns do not extend to the purely causal argument. This is not to say that the causal mechanisms by which one external event leads to another are irrelevant, but that they must be mediated by arguments. We are interested in those realms of decisionmaking in which justification is the essence. Thus we are concerned with arguments about arguments - also known as meta-arguments. To put it another way, these are arguments about accepting or rejecting arguments for actions. They involve intellectual commitments that, as it were, take on a life of their own.

Second, there is the perplexing question of whether slippery slope arguments invoke some form of irrationality. Can purely rational thought produce a progression from the acceptance of a correct or persuasive argument to the acceptance of a clearly incorrect or unpersuasive argument? Slippery slope arguments appear to be vulnerable to the following three objections from the perspective of rational choice:

1. If the future decision (the "danger case," as Schauer calls it n9) is bad, but the prior decisions are good, why not simply refrain from making the bad decision down the road? The slippery slope argument seems to rob our future selves of the ability to make reasoned decisions. It treats future decisionmakers as automata who cannot resist doing the wrong thing. We call this the automaton objection.
2. If the consequences of a sequence of decisions are undesirable overall, then why are we tempted to defect from the right path now? Suppose, for instance, that undesirable future decisions will somehow flow with high probability from the present decision. Unless we are simply ignorant of the causal chain, the undesirability  [*543]  of the final outcome should be imputed backward to the initial decision,n10 and the initial decision should thus not appear desirable after all. The slippery slope argument seems to violate the assumption, taken for granted by many economists, of rational expectations. This we call the imputation objection.
3. The mere fact that the ultimate decision appears undesirable from today's vantage point does not mean it will appear undesirable tomorrow. After all, if we take that final step when the time arrives, it must look desirable at that time. The slippery slope argument appears to privilege the current over the future point of view, ruling out the possibility that new values will exist at the moment of decision. We call this the presentism objection.n11
Because of these and similar objections, some analysts have concluded that slippery slope arguments are questionable or even fallacious. Nonetheless, slippery slope arguments can constitute a valid form of argumentation. In this Article, we provide a general theory of slippery slope arguments that allows us, among other things, to evaluate their validity and explore strategies for avoiding the events they describe. In Part I, we outline the essential features of a slippery slope argument, and clarify terminology. In Part II, we present a rubric for understanding the structure of discussion in which slippery slope arguments - and all other arguments, for that matter - are made. This structure provides us with a foundation for analyzing particular types of slippery slope arguments. In Parts III and IV, we discuss four different processes that could provide the basis for slippery slopes: the sorites with precedent process, the altered economic incentives process, the separately validated propositions process, and the Humean beneficence process. In Part V, we respond to the three objections to slippery slope arguments presented above, explaining why  [*544]  we think they are not (always) valid. In Part VI, we offer several propositions about factors that make slippery slopes more likely. Finally, in Part VII, we discuss various strategies employed by individuals and systems for dealing with or reducing the likelihood of slippery slopes. We conclude with some general observations about the validity and invalidity of slippery slope arguments.

I. Definitions and Clarifications

A. Essential Characteristics of a Slippery Slope Argument

Although there is no paradigm case of the slippery slope argument, there are characteristic features of all such arguments. The key components of slippery slope arguments are three:

1. An initial, seemingly acceptable argument and decision;
2. A "danger case" - a later argument and decision that are clearly unacceptable;
3. A "process" or "mechanism" n12 by which accepting the initial argument and making the initial decision raise the likelihood of accepting the later argument and making the later decision.
The "processes" invoked as the link between the initial case and the danger case can be quite varied. It is useful for our purposes to distinguish processes that, in principle at least, can be generated by a single individual in isolation from the activities of others - in other words, a Robinson Crusoe process. We call these microprocesses. For example, Robinson Crusoe might be susceptible to a slippery slope from accepting the virtue of relaxation from work to accepting the vice of laziness. (This slope might occur as a result of, for example, Crusoe's commitment to reasoning by analogy from past choices to present ones.) Note that the defining feature of a microprocess is not that it must be generated by a single individual, but that it could be; this point will become clearer later.n13 In contrast, other processes, by their very nature, require the interaction of many  [*545]  individuals to generate the final result. These we call macroprocesses. For example, certain kinds of government regulation may distort the incentives of particular agents, as when retail price controls encourage suppliers to limit production. This may produce unintended consequences for consumers who then try to remedy the situation by voting for controls on the input prices faced by suppliers.n14 In this Article, we examine slippery slope processes of both micro and macro varieties.

B. Slippery Slope Arguments and Slippery Slope Events

Considerable confusion can be forestalled by distinguishing between slippery slope arguments and slippery slope events. A slippery slope argument (SSA) is an argument about how the acceptance of one argument (regarding a decision, act, or policy) may lead to the acceptance of other arguments (regarding other decisions, acts, or policies). It has a hypothetical form: if this, then that - with increased likelihood. A slippery slope event (SSE) refers to the actual manifestation of the events (decisions, acts, or policies) described in the SSA.

It is possible that the persuasiveness of an SSA may preclude the occurrence of an SSE. For example, if (on the basis of some initial argument) decisionmakers are persuaded that allowing first-trimester abortions will lead with high probability to infanticide, they may never accept the initial argument for permitting first-trimester abortions. Thus, acceptance of the SSA may help prevent the more easily observable SSE.

To understand better the distinction between an SSA and an SSE, it is important to recognize that there are two distinct types of ideas in the social sciences: constitutive ideas and speculative ideas.n15 Constitutive ideas are ideas that motivate the actions of individuals. Speculative ideas, on the other hand, are ideas that observers - such as social scientists or policy analysts - have about the actions individuals will take and the results that will follow. For instance, ideas that consumers have about the desirability of goods and services are constitutive ideas, as they affect consumers' buying decisions. The ideas that  [*546]  economists have about the effect of consumers' decisions on market outcomes (such as the prices and quantities of goods sold) are speculative ideas. Although constitutive and speculative ideas are usually distinct, it is possible for a speculative idea to become a constitutive idea. For example, if economists predict that a recession is looming, and consumers believe them, then consumers may respond by altering their buying decisions. Thus, the economists' speculative idea that a recession is coming becomes the consumers' constitutive idea, insofar as it motivates consumers to reduce unnecessary expenditures in the expectation of possible unemployment.

SSAs are typically speculative ideas. They are predictions, made by observers, about how acceptance of some ideas (and resulting actions) can lead to acceptance of other ideas (and resulting actions). But an SSA can become a constitutive idea, if the SSA is accepted by individuals and affects their actions. Indeed, the person who formulates an SSA may do so with the intention of persuading others to change their behavior - that is, with the intention of making it a constitutive idea. For instance, those who argue against voluntary euthanasia, on grounds that it will increase the likelihood of involuntary euthanasia, presumably hope their argument will persuade the public to oppose policies allowing voluntary euthanasia.

In short, an SSA is by nature an idea about other ideas. Like the theories and models used by social scientists, it makes a prediction about the behavior of people who are motivated by their own ideas.

II. The Structure of Discussion and Argument: The Microanalytic Foundations of Slippery Slopes

A distinctive feature of most, perhaps all, SSAs is that they are arguments about arguments. That is, they are meta-arguments. The SSA relies on the notion that the argument (and decision) we take now will, at some time, make people more likely to accept another argument.

Consider the illustrative case in which the Village of Skokie, Illinois, made it a misdemeanor to disseminate material promoting or inciting racial or religious hatred. This included, in the words of the local ordinance, the "public display of markings and clothing of symbolic significance." Accordingly, Skokie tried to stop a Nazi group from demonstrating peacefully, in uniforms and with banners, in front of the village hall. In Collin v. Smith,n16 the Seventh Circuit struck down the ordinance. The court explained that if it were permissible to graft an exception onto the First Amendment for a demonstration  [*547]  that might inflict "psychic trauma" on certain people (such as Holocaust survivors), it might also be permissible to halt any speech that generates "anger," "unrest," or "dispute."n17 At that point nothing would remain of the First Amendment. In our terminology, the court is stating that, if it accepts an argument about the permissibility of the Skokie ordinance, it would also have to accept (or would at least be more likely to accept) future arguments about further speech restrictions. These arguments might allow, perhaps seriatim, restrictions on the kind of speech that generates anger, then unrest, and then simple dispute. Having accepted the initial Skokie argument, the court would find the others "indistinguishable in principle." It would then be led to accept an ultimate argument that, in today's view, is clearly wrong.

In essence, the court makes an SSA that claims that if Argument 1 is accepted, then so will be Argument 2, and then Argument 3, and so on to, say, Argument 10 that would justify some clearly unacceptable outcome. Now, it cannot be the case that Arguments 1 and 10 are identical; otherwise the SSA would be redundant. If Argument 10 were clearly bad, and Arguments 1 and 10 were identical, then Argument 1 would be unacceptable on its face. Therefore, on what basis can the analyst predict that different arguments will be made and accepted? How can he predict what he himself or later decisionmakers will find similar or close to a previous argument? Only, it seems, if he understands the theoretical framework in which the judicial decisionmakers operate.

To understand an SSA, then, it is necessary to think more carefully about the structure in which decisions are made. Our object in this part is to lay out a rubric for thinking about the structure of decisionmaking. We start by offering a discussion of the key concepts within this structure: rules, theories, research programs, and arguments.

A. Rules

A rule is a mapping from a type of factual situation or event to a desirable action. A rule's mapping seeks "to change or channel behavior" relative to what it would be without the rule.n18 In political analysis, the factual situation may be a social problem and the desired action a governmental policy. In law, the situation may be a justiciable dispute and the action a ruling. In ethics, the situation may be a set of moral options and the action a moral decision.

Three clarifications are in order. First, it is important to distinguish our use of the word "rule" from other meanings of the term. There is a difference  [*548]  between rule-conforming and rule-guided behavior.n19 In the case of rule-conforming behavior, agents need not understand that they act or make decisions in accordance with a rule. They simply exhibit a regularity, which the rule describes. Much animal behavior, such as the tendency of birds to migrate, conforms to rules in this way. The laws of physics, such as Newton's Law that bodies tend to move toward each other, are also of this nature. In rule-guided behavior, a rule is prescriptive in nature, as agents use the rule as a reason or justification for their decisions. It is the latter sense of the word "rule" that we employ here.

Second, it is worth noting that rules are often stated along with their rationales, that is, as part of arguments. But a rule in its pure form is simply a mapping. It is also true that rules can map situations to more than one desirable action. But, for simplicity, we shall think of rules as issuing in one fairly specific action.n20

Third, any rule works by reference to a set of characteristics that describe a situation, and this set is necessarily a subset of all those characteristics that might be construed as describing it "fully." Rules are unavoidably abstract, as they omit or abstract from a potentially infinite number of characteristics that could be used. Consider a legal rule that says, "Whenever a car rear-ends another car, the car that came from behind is liable for damages." This rule identifies one characteristic of the situation (which car came from behind) while effectively ignoring an endless number of other characteristics (the color of the cars, the time of the accident, the number of people in each car, whether the occupants were listening to their radios, ad infinitum).n21 Of course, actual rules can be, and generally are, more complex. They may identify a very large number of characteristics. But no matter how many characteristics are identified, an infinite number of other characteristics are ignored. The choice of which characteristics to include, if it is not arbitrary, must be made on the basis of a higher-order conceptual entity; that is, a theory.


B. Theories

A theory is a system of ideas based on general principles designed to organize thought and to explain or justify something. A theory can be positive, normative, or both.n22

The most important role of theory, at least in the present context, is as a source of justifications for rules and decisions. In normative terms, a theory can justify a rule.n23 The normative and positive elements of a theory interact to produce arguments about how rules should be chosen and how decisions should be made. As suggested earlier, a rule itself does not necessarily carry with it any justification; it could be entirely arbitrary in the characteristics it identifies as relevant. But often, rules have their basis in theories about what kinds of characteristics are positively and normatively relevant in a given context.

For instance, the rule "whenever a car rear-ends another car, the car that came from behind is liable for damages" might be justified by a theory that emphasizes the capacity of law to promote accident prevention. Thus, if the driver of the rear car has greater control over whether a rear-end accident takes place (a positive judgment), and it is desirable to minimize the sum of expected accident and accident prevention costs (a normative judgment or standard), then the theory, at least prima facie, justifies the rule.

Theories are closely related to the notions of relevance and similarity. The application of these notions, far from being a matter of direct insight, is theory-laden. What is relevant according to one theory may be irrelevant according to another. For example, an orange is similar to a banana, and a banana is similar to a cigar; therefore, can we say an orange is similar to a cigar? To answer affirmatively would be an example of invalid reasoning because the similarity relations are not the same.n24 The first similarity relation presumably derives from a theory that identifies an object's use or origin in nature as a relevant characteristic, whereas the second similarity relation presumably derives from a theory that identifies an object's shape or length as a relevant characteristic.

One implication of the theory-laden nature of relevance and similarity is that a theory can be either implicit or explicit. Even if someone claims not to have a theory in some context, his statements about similarity and relevance  [*550]  in that context belie his claim. He must have a theory, even if he does not realize what it is. Karl Popper notes:

Generally, similarity, and with it repetition, always presuppose the adoption of a point of view: some similarities or repetitions will strike us if we are interested in one problem, and others if we are interested in another problem. But if similarity and repetition presuppose the adoption of a point of view, or an interest, or an expectation, it is logically necessary that points of view, or interests, or expectations, are logically prior, as well as temporally (or causally or psychologically) prior, to repetition.n25
In short, there cannot be any theory-free identification of "similarity." It can be identified only by use of a (possibly implicit) theory.n26

C. Research Programs

"Research program" is a term we have borrowed from philosopher of science Imre Lakatos, who uses it to refer to a broad set of basic assumptions, premises, and methods shared by a group of scientists working in the same scientific tradition.n27 A research program is sufficiently loose that it can encompass multiple theories held by different scientists, and those theories may contradict each other. The research program places constraints on the types of theories scientists can use without losing credibility in their community of scholars.n28 We use the term "research program" here in much the same way,  [*551]  but without the emphasis on science.n29 In law, the relevant community may include judges, legal scholars, and private actors subject to the law. In ethics, the relevant community may be much broader, including everyone within the same moral community.

A research program is not a theory in the usual sense, as it does not have enough content to provide meaningful predictions or explanations in applied contexts. Instead, it is a meta-theory, a theory about theories. A research program in the law places broad limits on what legal theories can assume, how they can differ, and what sorts of conclusions they can reach. In other words, a research program defines the structure of allowable change and difference among theories, thereby indirectly affecting rules and decisions.

As an example of the sort of guidelines that may characterize a research program, consider Melvin Eisenberg's claim that replicability in legal decisionmaking requires that "the courts employ a consistent methodology across cases."n30 The use of a common methodology, Eisenberg argues, enables "private actors, within limits, to determine before they enter into a transaction the legal rules - including the "new' legal rules - that will govern the transaction if a dispute should arise." n31 Perhaps this conclusion is too strong, because a research program or "methodology" is an incompletely defined structure. It cannot, therefore, provide a great deal of guidance for private actors attempting to predict outcomes in specific disputes. Nonetheless, the use of a common methodology places limitations on how far a legal decisionmaker's approach may differ from the approaches of others in the legal community at large.n32

One example of a legal research program is the economic efficiency approach. The normative premise of this research program is the notion that legal rules should be chosen so as to maximize economic efficiency, understood  [*552]  as social wealth maximization.n33 There are a variety of assumptions inherent in this approach, mostly drawn from the field of economics, including: Agents have relatively stable and well-defined preferences; agents change their behavior in response to legal incentives; wealth maximization is a relevant standard for measuring social welfare; and so forth. But within the economic efficiency approach, there exist differing theories. Richard Epstein, for instance, has emphasized the importance of simple, well-known rules that serve to guide the expectations of litigants in a wide range of cases n34 Richard Posner, on the other hand, has placed more emphasis on the selection of rules that induce wealth-maximizing choices in specific circumstances.n35 Their differing theories have yielded differing conclusions about which rules should be used in specific areas of law; for instance, Posner has generally supported negligence rules in the law of tort, whereas Epstein has leaned toward strict liability rules.n36 More generally, legal efficiency analysts can reach different conclusions because of their differing perspectives on issues such as the magnitude of transaction costs, the relevance of administrative costs, the elasticity of litigants' behavior with respect to expected punishments, or the frequency of efficiency-relevant parameters across cases.

D. Arguments

For our purposes, an argument is a reason or sequence of reasons, usually defeasible, for acting in a particular manner. Often, an argument will take the form of a deductive justification:

If E occurs and has characteristics X, Y, Z, then one should do D (major premise).
 [*553]  Event E has occurred and has characteristics X, Y, Z (minor premise).
Therefore, one should do D.
This argument provides a normative major premise for reaching a conclusion, and a decisionmaker who wishes to apply the argument must verify or support the minor factual premise.

Again, some clarifications are in order. First, we employ the word "argument" in the sense of a justification for taking some action or accepting some proposition. This is distinct from other sorts of arguments, such as the use of empirical data to test a scientific theory. Second, it is important to distinguish a justificatory argument from a causal explanation of some agent's action. Simply because the foregoing syllogism is valid, and even if its premises are true, the agent need not act in accordance with the conclusion. He obviously can ignore his duty to do what he "should" do. Nevertheless, the argument provides a structure for the justification of an action - a reason or set of reasons for acting in a certain way.n37

We reach the definition of argument last, because arguments can take place on different levels of analysis. The structure we have outlined here consists of four levels:

1. Decisions

2. Rules

3. Theories n38

4. Research Programs n39

At each level, there are arguments among the different items at that level, and some of those arguments consist of applying items at the next higher level. At the decisions level, there are arguments made about which decision to make, and some of those arguments involve the application of rules. At the rules level, there are arguments about which rule to select, and some of those arguments involve the application of theories. At the theories level, there are arguments about which theory to employ, and some of those arguments involve the application of research programs.


E. Example: Make-Up Exam Meshugas n40

Suppose a student asks his professor, "Why did you refuse to give me a make-up exam?" The professor responds, "Because there is a rule that, absent a doctor's note, you may not be given a make-up exam. You don't have a doctor's note, so you don't get a make-up exam." This is an argument at the decision level; it is a reason for deciding not to give a make-up exam. It is also the application of a rule.

Now suppose the student responds, "But my great aunt Polly died. One purpose of the rule was to allow exceptions for students who have a good reason for missing the exam. A death in the family is a good reason, so the rule should have an exception for that."n41 This is an argument, but it is no longer at the decision level. Instead, it is at the rules level. The student contends that the rule should be something other than what it is. To bolster his argument, he appeals to the justification for the rule, and in doing so he is applying a theory about how to select rules. He hopes the professor will share his normative judgment (that a death in the family is a good reason to miss an exam) as well as his understanding of one purpose of the make-up rule (to allow exceptions for good reasons). In short, the student hopes the professor will share his theory.

The professor might at this point advance theoretical arguments against the student's position. For instance, he might agree that one purpose of the rule is to make exceptions for good reasons, but respond that the death of a distant family member is not a good enough reason to expand the rule to include the present case. This argument, like the student's, is about the correct application of the theory. That is, the student and professor are having a rules-level discussion, in which appeal is made to a shared theory.

On the other hand, the professor might argue that a purpose of the rule is not merely to allow exceptions, but also to constrain the professor's discretion. In making this argument, the professor posits an alternative theory of the justification for rules, and he thereby raises the discussion to the theories level within a research program.n42

 [*555]  It would not be terribly surprising to hear any of the arguments made so far. But it would certainly be surprising to hear the professor say, "The rationale of the theory underlying the rule is to maximize my personal satisfaction, and I don't want to administer make-up exams." If he did, he would be raising the discussion to the research programs level. A professor with this point of view would stand outside the research program in which most professors, students, and other members of the educational community operate.n43

The different levels at which arguments can be made are not hermetically sealed off from each other. As the example indicates, the different levels of argument can mingle, even within a single conversation. And sometimes, the level at which an argument is made may not be clear. For instance, suppose the student argues, "You should let me take a make-up exam anyway, because my great aunt died." If the professor agrees, has the professor changed the rule (operating at the rules level), or has he simply chosen not to apply the rule in the present case (operating at the decisions level)? After all, arguments may be defeated, so the choice not to apply a rule does not necessarily imply a modification of the rule. The argument for the application of the rule may be simply incorrect. But we contend that, at least in principle, it is possible to distinguish between levels of argumentation.

The structure we've described consisting of four levels (decisions, rules, theories, and research programs) is a simplification. What we've called the theories level may actually consist of multiple layers. Some theories are broader and more abstract, others narrower and more applied. In a discussion about choosing among theories (not merely applying them to choose rules), the arguments need not appeal to the research program, but merely to a higher level of theory.

Our broader point is simply that discussion and argumentation can take place at many different levels of analysis. Those things that are taken more or less for granted at a lower level can become objects of questioning and analysis at higher levels.

F. Arguments About Arguments

A distinguishing feature of an SSA is that it relies on the future acceptance of arguments not yet made or appreciated. But in order to predict, even roughly, the arguments that may be accepted, an SSA must go beyond the level  [*556]  at which the initial argument is constructed. It must go at least one step up in the decisions-rules-theories-research programs structure.

Consider the following hypothetical argument against socialized healthcare: "In a socialized system of healthcare, people are not confronted with the monetary costs of their risky behavior; consequently, a moral hazard problem is likely to result. People will take greater health risks than otherwise. In the aggregate, such behavior will drive up the costs of the system as a whole, fueling demands by taxpayers and legislators to restrain the behaviors that increase costs. Thus they will end up supporting the regulation of lifestyle choices, such as "unsafe' sexual practices, indulgence in dietary fat and sugar, and so forth, on the grounds that some choices cost society more in healthcare expenses than others."

The above argument may or may not be persuasive; that is not the issue here. The issue is the form of the argument. The proponent of the SSA is claiming that the acceptance of an argument (that we should have socialized healthcare) will "cause" people to accept another argument (that lifestyle choices should be regulated) that they would otherwise be less likely to accept. The proponent shifts our attention away from the initial argument itself to the transition between arguments. In this sense, he is making an argument about arguments, or a meta-argument.

The two decisions, whether to socialize healthcare and whether to regulate lifestyles, could be made in isolation. But the proponent draws a connection between them, and the connection is made by a meta-argument. He claims that an affirmative answer to the first question about socialized healthcare will lead to (or make more likely) an affirmative answer to the second regarding lifestyle restriction. On what grounds can this claim be supported? Socialized medicine will result in consequences - such as higher external costs of gluttony - that will make other arguments applicable to the question of lifestyle regulation more persuasive.

What would the proponent need to know to make this claim with reasonable support? He would need to know what arguments are likely to be made in the lifestyle debate, which factors increase or decrease their likelihood of acceptance, and so on. In short, he must have knowledge of the structure of discussion and argument. The arguments people will make in the future depend upon their rules, theories, and research programs, as well as the factual situation.

An SSA relies on a model of how people construct, evaluate, and apply arguments. The validity of the argument depends on the accuracy of the model. Obviously, some models are correct and others not. The validity of the argument against socialized healthcare depends, in large part, on whether taxpayers and legislators do in fact consider the monetary costs of lifestyle  [*557]  choices to the community.n44 In the next few parts, we outline some specific slippery slope processes, or models that sometimes prove accurate. We divide our discussion into consideration of microprocesses, which can in principle be generated by the actions of a single agent, and consideration of macroprocesses, which require the interaction of decisions of more than one agent.

III. A Microprocess: Sorites With Precedent

In this part we consider one kind of process that could, in principle, be the result of the actions of a single individual. Sorites refers to a particular type of logical paradox that occurs in the presence of vague words and phrases. The term "sorites" derives from the Greek soros for "heap," a reference to a classic example of the paradox: "If there is a heap of sand, you can always remove one grain of sand and still have a heap." If this premise is applied repeatedly (a heap of premises), we eventually must conclude that even just one grain of sand is also a heap.n45 Similar reasoning can lead to conclusions such as there are no bald men, pygmies are tall, and so forth.

The root of the paradox is the existence of vague or fuzzy concepts like "heap," "bald," and "tall."n46 While clear cases of heaps and nonheaps do exist, there is a gradient of cases in between that are neither clearly heaps nor clearly nonheaps. The gradient creates the possibility of a chain of reasoning, seemingly valid, that links the ends of the spectrum and effectively erases the distinction between them.

 [*558]  Suppose we begin, following Nicholas Rescher, with two "observable facts": first, that one grain of sand does not make a heap and, second, that a million grains of sand do make a heap.n47 So long as we accept the "seemingly evident general principle" n48 that it is always true that adding only one grain to a nonheap still yields a nonheap, we shall end up contradicting our belief that a million grains of sand make a heap.n49

The sorites paradox is relevant to the present discussion because it can act as a slippery slope process,n50 particularly in systems where precedent plays an important role in the decisionmaking process. If the actual and potential cases in which decisions need to be made are distributed along a spectrum according to some relevant factor, a series of logical steps can link highly disparate cases, leading to the erroneous or undesirable conclusion that unlike cases should be treated alike. To put the problem in mathematical terms, imagine that all cases are arranged on a gradient from zero to one. One is the clearest possible case for taking some action A. Zero is the clearest case for not taking action A. According to whatever theory is used by decisionmakers, two cases are "similar" if the difference between their numbers on the scale is less than 0.1. A case arises with a value of 0.95 on the scale, and this falls within the realm of clear cases for taking action A, so action A is taken. In a subsequent case with value 0.9, the decisionmaker observes that it is similar to the first case, and so he follows precedent by taking action A in the present case as well. Then there arises a case with value 0.85, then 0.8, and so forth. Through a series of decisions based on similarity and precedent, we eventually conclude that case n with a value of 0.1 should also result in action A, even though case n is, or is similar to, a clear case for not taking action A.

 [*559]  It should be noted that the character of the slippery slope is crucially dependent on the initial precedents; it is a path-dependent process. Suppose we started from the other direction, that is, with situations that are clear cases for not taking action A. Then the momentum for the sorites slope would move toward a situation that results in inaction for something that would otherwise have been regarded as a clear case for taking action.n51 From an external point of view, the strict logical error in the numerical example above is that similarity, as defined in the example, is not a transitive relation.n52 If case x is similar to case y, and case y is similar to case z, that does not necessarily mean that case x is similar to case z.n53 While this may be apparent from an external point of view, it may not be apparent to decisionmakers operating within the system. The root of their difficulty is that it is plausible that x is similar to z; indeed it may even be true in some instances,n54 but it is costly to determine this. If the agent's decisions are made by reference to precedent and plausible similarity, then the sorites chain can occur. To recognize and possibly avoid the slippery slope, one must be willing to raise the discussion to a higher level that considers the cumulative effect of many marginal decisions.n55 In other words, the SSA - the argument that draws attention to how a chain of seemingly correct decisions can lead to an undesirable outcome - is a meta-argument.

The meta-argument nature of the sorites SSA may be better appreciated by generalizing the progression illustrated above in the following way:

[alpha]1 is an acceptable argument (to do or decide action A1)
[alpha]2  is close to or similar to [alpha]1
Therefore, [alpha]2 is an acceptable argument
[alpha]10 is close to or similar to [alpha]9
Therefore, [alpha]10 is an acceptable argument (to do A10)
But [alpha]10 (A10) is, in fact, unacceptable
 [*560]  In this generalization of the sorites SSA, the progression is generated by an overall perceived similarity n56 of arguments that is often - but not always - rooted in the vagueness of a central concept. For example, an argument that justifies state subsidization of school lunches may be seen as similar to the argument that justifies state subsidization of education in the first place - perhaps because the education of the mind and the health of the body are empirically related.n57

If the law sometimes manages to resist slippery slopes in the presence of vague terms, it is because the legal profession has adopted various stratagems for resisting them. Such stratagems include the establishment of clear (though arbitrary) rules and the selective use of higher standards. We will delay our discussion of these slope-resisting strategies until later. For now, we will observe that there do not seem to be any foolproof methods of resisting slippery slopes, only methods that have been more or less successful than others.

IV. Macroprocesses

In this part we consider processes that involve, in an essential way, the interaction of more than one decisionmaker.

A. Altered Economic Incentives

Some slippery slopes involve, as an essential feature of the process, changes in the real-world costs and benefits faced by decisionmakers. We refer to these as altered economic incentive processes, or incentive slopes for short. Like all SSAs, the incentives slope argument points out that accepting some Argument 1 will increase the likelihood of accepting some other Argument 2. The crucial difference is that the transition between arguments is eased by some change in incentives resulting from the earlier argument's acceptance.n58

 [*561]  The incentive slopes in which we have the greatest interest are those generated by interventions in complex social systems. Such systems are characterized by two fundamental properties: first, the interaction of individuals' actions and plans; and second, the existence of emergent properties resulting from the interactions of these individual behaviors. In these systems, there is an absence of linearity: The effect of the sum of two or more factors is not necessarily equal to the sum of the individual effects arising from them. Furthermore, agents cannot predict outcomes simply by knowing the initial data, because the path of decisions taken will influence which outcome actually occurs. Even outside of systems - where consequences are foreseeable in principle - they may be unforeseeable in practice, because they follow from changes in incentives that are obscure and not immediately apparent to a decisionmaker subject to, for example, high information costs or bounded rationality.n59

Economists have long emphasized that decisions, especially policy decisions, often have consequences neither intended nor expected by the decisionmaker.n60 Often, the unintended consequences result from changes in incentives wrought by new policies. Simply pointing to the unanticipated and unintended consequences that result from altered incentives, however, does not create an SSA. But when these consequences affect the way future decisionmakers (or the same decisionmaker under different constraints) will form and evaluate arguments in some systematic way, then the foundation for an SSA exists. The sliding takes place not because the arguments made at the various stages are similar (as in the sorites process) but because the stages are causally interlinked, inasmuch as the first step lowers the cost or increases the benefit of taking the next step relative to what it otherwise would have been.n61

More concretely, when unanticipated consequences are caused by a government restriction on individual behavior, an SSE can occur if the initial restriction makes further restrictions more likely. This may happen if the experience of unanticipated consequences is conjoined with an argument that further restrictions are curative or, at least, ameliorative. If this scenario is likely to occur, or is perceived likely, some observers or analysts may construct SSAs on this basis.

 [*562]  Note that even in this "causal" process, theoretical constructs are utilized at every stage of both the event and the argument. Beginning with the argument, the observer recognizes a causal relationship between the government policy and the undesirable and often unanticipated consequences. There are at least four types of theory operative here. First, there is the positive theory that links the initial intervention with its real-world consequences. Second, there is the normative theory that deems the consequences undesirable. Third, there is the further positive theory (typically held by someone other than the observer, such as legislators or voters) that sees the second intervention as ameliorative.n62 Finally, there is the further normative theory (typically held by one or more decisionmakers) that sees amelioration as beneficial, all things considered. With respect to the SSE, however, it is not necessary that the economic agents (legislators and voters) understand the connection between policy and the undesirable outcome. It is sufficient that they find it undesirable and believe that further intervention is the answer (that is, they hold the last three theories above).n63

The socialized healthcare hypothetical, previously mentioned, is a simplified case of the altered economic incentives process. Recall that there is a moral hazard problem resulting from changes, at the margin, in incentives to make risky decisions. Although merely pointing to moral hazard does not create an SSA or describe an SSE, claiming that the moral hazard problem will increase support for regulation of lifestyle choices is to make an SSA or to point to an SSE. The key to the SSA and possible SSE in this case is the notion that voters believe it is possible to lower their tax burden by (further) restrictions on individual autonomy ("No food with a saturated fat content beyond x may be sold!"). This is a belief derived from positive theory. They also believe that it is morally acceptable to so restrict individual autonomy for the purpose of lowering their tax burdens. This is derived from a normative theory, if only an implicit one, held by the voters (though not necessarily shared by the person making the SSA).

This incentive slope produces results that may be unacceptable from the initial point of view of those who decide to implement a program of socialized healthcare. These same agents may put into effect the very regulations that they previously disliked, because of the unpredicted change in their own incentives. Their underlying preferences have not changed but their actions have,  [*563]  because they must now bear costs they previously did not. Whether this new outcome is of such lower utility that agents would choose not to adopt socialized healthcare, if they knew the full consequences in advance, cannot be determined a priori.

B. Separately Validated Propositions

In this process, propositions that have been validated separately result in a conclusion or overall outcome that would not have been validated if considered by itself. To see how this is possible, consider decisionmaking by majority rule. It is a well-known fact that if a majority approves of policy A, and a majority also approves policy B, it does not follow that a majority would also approve the union of A and B. The reason is that the majorities supporting the separate policies may not be the same. If 51 percent support A and 51 percent support B, it is possible that as few as 2 percent support both. This fact may not be relevant for our purposes if policies A and B are totally unrelated, but it takes on special significance if the policies are logically or practically related. If that is so, then separate validation of the two policies could result in an overall "coherent" policy outcome that would not itself be validated and could constitute an SSE.

For example, consider the question of whether a nation should institute a more generous welfare system. Suppose that a majority of the population supports this position. Suppose also that a different but overlapping majority believes that a more restrictive immigration policy should accompany a more generous welfare system, so that natives will not have to support new arrivals. Both policies could be implemented, even though as few as two percent of the population might initially support that outcome. Separate implementation of the two policies leads to an outcome desired only by a small fraction of the public.n64

As with the altered economic incentives process, there are unanticipated consequences in this process. In the former process, decisionmakers fail to predict the results of a change in the economic incentives of the people affected by the policy. In the present process, decisionmakers fail to predict the likely behavior of other decisionmakers. For the second step in the process to occur, it must be the case that at least 51 percent of the population prefers the option "generous welfare with restrictive immigration policy" to the option "generous welfare without restrictive immigration policy." Note that these preferences are consistent with having preferred the status quo over either or both policies;  [*564]  for example, one's first preference might be "neither generous welfare nor restrictive immigration policy." The outcome of the voting process could be "generous welfare with restrictive immigration policy" even if it were the preferred option of as few as 2 percent of the public. These preferences are summarized in Table 1. From the table, it can be seen that 51 percent of the population (Groups A and C) would prefer generous welfare (GW) to the status quo. Given the existence of generous welfare, a different 51 percent (Groups B and C) would impose greater restrictions on immigration (RI), yet 98 percent of the population (Groups A and B) would have preferred the status quo ante over the combined regime.

Table 1
Hypothetical Population Preferences Leading to Separately Validated Propositions

Percentage of Population First Preference Second Preference Third Preference
Group A 49% GW status quo ante GW + RI
Group B 49% status quo ante GW + RI GW
Group C 2% GW + RI GW status quo ante

Unanticipated consequences are involved in this process. The members of Group A, who initially supported generous welfare, did not foresee the voting behavior of Groups B and C. If they had, they might not have supported generous welfare in the first place. This is not, however, the incentives slope process as defined earlier, since it is not a change in the economic incentives of the people affected by the policy that brought about the slope.n65

Readers familiar with the literature on social choice will no doubt recognize the figures above as an instance of Condorcet's paradox,n66 which can occur when voters have preferences that are not "single-peaked."n67 This implies that majority voting on pairs of policies can generate nontransitive "social  [*565]  preferences."n68 Eugene Volokh provides more examples of this nature.n69 We wish to add two observations. First, although it is often suggested that the intransitive nature of the voting process will lead to cycling (policy A is replaced by B, which is replaced by C, which is replaced by A again, ad infinitum),n70 that need not be the case. Often policy reversal involves high costs that effectively prevent a return to the original policy. Thus, it is easier to fall down the slope than to climb back up. Second, the separately validated propositions process does not require non-single peaked preferences like those above, as majority rule is not the only social means of making decisions. In law, the authoritative opinions of one or a few courts can be sufficient to validate a proposition. Distinct legal propositions may be validated through separate precedent-setting decisions, and later cases may reveal the unanticipated consequence of combining them. Unless the courts deciding such cases are willing to break at least one precedent, they may find themselves validating additional propositions even if they seem undesirable.

The issue of fetal personhood n71 is an excellent demonstration of how separately validated propositions have the potential to combine to reach conclusions that are (at least to some observers) unpalatable. The issue is whether, and under what circumstances, fetuses should be treated as legal persons. According to the line of reasoning followed in the U.S. Supreme Court's Roe v. Wade n72 decision, a fetus is not regarded as a legal person with respect to the issue of abortion. But in a number of cases unrelated to abortion, American courts have been willing to treat fetuses as persons, especially in criminal cases involving an intentional or accidental fetal death caused by someone other than the mother, such as an attacker or drunk driver.n73 The apparent conflict between the lines of reasoning has not yet been resolved, but some slippery slope possibilities are apparent. From the perspective of a proponent of abortion rights, the danger is that the recognition of fetal personhood in criminal cases could, in combination with the Fourteenth Amendment's protection of the life, liberty, and property of all persons, eventually undermine Roe v. Wade's protection of abortion rights. From the perspective of abortion opponents, of  [*566]  course, this would be a desirable slope. But there is also the potential for a slope in the other direction: The denial of fetal personhood in abortion cases could, in combination with the traditional definition of murder as the wrongful killing of a person, result in a situation in which the deliberate killing of someone else's unborn child could only be prosecuted as battery, not murder. Thus, both advocates and opponents of abortion rights have reason to fear the effect of consistency in the law.n74

SSAs based on the separately validated propositions process highlight the likelihood that certain arguments, if accepted now, will interact with other arguments to increase the likelihood of accepting different arguments in the future. In the case of more generous welfare policy, opponents might warn potential advocates that accepting the policy would increase the likelihood of persuasive arguments for immigration restrictions they do not support. In the case of fetal personhood, abortion rights advocates might warn against allowing charges of murder against a person who caused the death of another's fetus, since it might reinforce persuasive arguments against Roe v. Wade. As with previous types of SSA, this argument requires a model (implicit or explicit) of how decisionmakers (judges, voters, and legislators) accept arguments. Specifically, the model in the welfare/immigration example states that some percentage of voters are willing to "vote their pocketbooks" and accept arguments for immigration restriction. The model in the fetal personhood  [*567]  example says that judges are inclined to accept, on grounds of precedent or consistency, arguments that follow logically from others already accepted.

To evaluate the models empirically, it is necessary to examine (a) the preference distribution of voters and (b) the power of consistency in the formation of legal doctrine, respectively. Voters may not have preferences like those hypothesized. And although systemic consistency n75 does have influence in the law, it is not an absolute value; courts may rule inconsistently by creating a special doctrine or area containing certain factual or legal presuppositions that are at variance with those in other doctrines or areas. The crux of the matter is how much emphasis the legal system places on consistency among legal doctrines. Ultimately, the persuasiveness of an SSA relying on the described processes depends on the credibility of the underlying models.

There is one more aspect of the separate validation process that bears emphasis. In the structure of discussion and argument, we observed that conclusions at one level are often applied as arguments at the next lower level. Research programs provide arguments in the choice of theories, theories provide arguments in the choice of rules, and rules provide arguments in the making of decisions. In the discussion of the sorites process, we assumed that discussants shared the same theory, as exemplified by identical similarity relations. But here, no such assumption is necessary. Proposition A might result from the influence of theory X, proposition B from the influence of theory Y. Indeed, the process may even require the existence of multiple theories, as it seems unlikely that people would disagree with conclusions that follow from propositions arrived at through the same theory - unless the theory is internally inconsistent or incomplete.n76


C. Humean Beneficence

David Hume argues that private benevolence is a "natural virtue," as distinct from justice, which is an "artificial virtue."n77 Whereas acting on a feeling of benevolence toward specific individuals in difficult situations produces an immediate and direct positive feedback, for most people, acting in accordance with the general rules of justice does not produce positive feedback in every case. The social utility of the rules of justice n78 is based on the convention that if one actor adheres to the rules, so will the other. The utility of justice is thus derived from the "whole plan or scheme" n79 and not from a single application of justice. The rules are acquired primarily through socialization and immersion in the norms of the society. This is not, to Hume, an argument against the fundamental nature of the rules of justice; on the contrary, he contends that the general, inflexible pursuit of justice is indispensable to the general happiness of society.n80 But its artificial character makes it more difficult to act upon than benevolence.n81

Therefore, and of special importance to SSAs, there will sometimes - even often - arise conflicts between justice and beneficence. The principle underlying beneficent action is one that takes note of special circumstances and the particular character of individuals, whereas justice is deliberately blind to such factors. In this sense benevolence is a concrete virtue and justice is an abstract virtue.n82 As a result, a benevolent person focused on particular circumstances will become aware of many seemingly undesirable consequences of specific acts of justice. As Hume argues:

All the laws of nature, which regulate property, as well as all civil laws, are general, and regard alone some essential circumstances of the case,  [*569]  without taking into consideration the characters, situations, and connexions of the person concerned, or any particular consequences which may result from the determination of these laws, in any particular case which offers. They deprive, without scruple, a beneficent man of all his possessions, if acquired by mistake, without a good title; in order to bestow them on a selfish miser, who has already heaped up immense stores of superfluous riches.n83
As a result, decisionmakers will sometimes find themselves torn between the demands of justice, on the one hand, and the demands of pity, compassion, and benevolence on the other.n84

The conflict between justice and beneficence creates the potential for a slippery slope. In a specific case, a judge or other decisionmaker may be tempted to depart from the rules of justice to make a special exception. It may seem undesirable, for instance, to enforce a contract against a well-meaning person who simply failed to think through the consequences of his decision to sign. Or it might seem harsh to extract large liability damages from a poor person who accidentally caused harm to another. Now, the mere act of making an exception does not itself constitute a slippery slope. But if the exception in some way makes future exceptions more likely than they would have been otherwise, then there is the potential for a slippery slope.

But why would one exception increase the likelihood of further exceptions? Consider a simple model of judicial decisionmaking, in which judges weigh their personal preferences about the disposition of cases versus a concern for their reputations. The reputation of a judge is determined primarily by the perception that he abides by precedents set by other judges.n85 The more a judge's decision appears to depart from the pattern established in prior cases, the greater will be the negative impact on the judge's reputation. Now, suppose a judge faces a case that he would prefer to decide in a beneficent manner, but there is a general rule established by prior cases against deciding in that way. Other things equal, he will be more inclined to decide the case beneficently (instead of according to the general rule) when there exist at least some "nearby" cases also decided in that way, because such cases reduce the appearance of  [*570]  renegade behavior. The judge can more plausibly claim that his case follows the pattern of previous cases.

Early on, few or no exceptions may have been made, and so judges who wish to indulge their feelings of benevolence have little support from precedent. Only the most "compassionate" judge, one whose desire to act beneficently is large enough to overcome his desire to safeguard his reputation, would be willing to make an exception. But the few early cases in which exceptions are made establish the basis on which further exceptions can be made later. As more exceptions are made, the margin moves, so that judges who previously had not been willing to make exceptions become more willing to make them. The more exceptions that have been made, the easier it is for further exceptions to be justified as consistent with the body of prior cases, and thus the reputational constraint gradually becomes less binding. The process is comparable to the mathematical description of the sorites slippery slope, in which the movement along a scale from zero to one is made possible by intermediate judgments of similarity. But in the present story, the choices of decisionmakers result not from "blind" application of precedent, but from a weighing of concern for precedent versus a desire to act beneficently in the instant case.

The Humean process, described above, has two welfare consequences. The first is that there is a weaker enforcement of "justice" than any decisionmaker at the outset dares to implement. In terms of initial preferences, most judges would find the later decision unsatisfactory or suboptimal, but some - those with extreme benevolence preferences - would approve of it and would actually be better off. The second is that, at any time during the slippery slope process, most judges believe the system to have more beneficence and less justice than is desirable given their concurrent preferences. This is because acting in a beneficent manner creates a negative externality. There is an immediate positive feedback to the individual judge but a weakening of the security of property and contract (with its attendant social costs) in the system as a whole.

V. Responding to the Objections

We now return to the objections we presented at the beginning of this Article, to explain why they do not always present a problem for SSAs.

The Automaton Objection. The automaton objection is that, if the future decisions in question are "bad," then we can simply choose not to make those decisions when the time comes. Three replies are in order.

First, the decisions we make now can change the incentives we face in the future. We do possess free will (we stipulate), and thus we could in principle refuse to make the "bad" decisions in the future. However, our present decisions can make certain future decisions harder to resist by lowering their perceived  [*571]  costs or increasing their relative benefits. In the socialized healthcare example, we could refuse to engage in lifestyle regulation - but the moral hazard created by socialized costs would give us a stronger incentive to regulate than we would have without socialized costs.n86

Second, even in the absence of changes in direct incentives to action, arguments do not exist in isolation. They exist in the context of a structure of discussion. The acceptance of some arguments can lead, logically or by force of precedent, to the increased likelihood of other arguments also being accepted. Again, a person can in principle refuse to accept an argument, perhaps by resisting its logical relationship or similarity to another, but the point is that the acceptance of certain arguments is eased by the acceptance of others.n87 In making this point, we are asserting that (at least some) people choose what arguments to accept in the same way they decide what clothing to buy, what products to produce, and so on: They weigh the costs and the benefits. Policies that alter costs and benefits do not remove the capacity for choice, but they do push the choices in one direction or another, and that is as true for the acceptance of arguments as it is for any other kind of choice.

Both of these replies are related to the third and most important reply: It is misleading to say that "we" are capable of making correct decisions in the future. The process by which arguments are accepted and decisions made is a social one that derives from the decisions of many individuals.n88 No single decisionmaker can control the evolution of the discussion. The person who makes an SSA does not necessarily claim that the listener himself will be the perpetrator of the future bad decision. Rather, he draws attention to the structure of the discussion that will shape the decisions of many decisionmakers involved in a social process.n89

The Imputation Objection. The imputation objection raised against SSAs is that any bad consequences that flow from the future are imputed backward to the initial decision, and therefore the initial decision should not appear attractive after all. We have two replies.

First, this objection implicitly recognizes the persuasiveness of the SSA. If the current decisionmaker already understands the full consequences of his decision, including its likely impact on future decisions, then his assessment of his  [*572]  current decision should indeed be correct. But the whole point of the SSA is to draw attention to a class of consequences that are typically ignored. Decisionmakers in the real world frequently do not clearly see all the likely results of their decisions. The SSA, like many other forms of argument, tries to emphasize the importance of some set of costs or benefits that the decisionmaker may have failed to consider adequately. Since real-world decisionmakers may exhibit myopic behavior,n90 an SSA could make a real contribution in terms of illuminating distant costs and benefits. In other words, SSAs may be valid precisely because they can become constitutive ideas that encourage the consideration of distant costs.

Second, this objection, like the last one, pays insufficient regard to the social nature of the discussion. The single decisionmaker may not have incentives or interests aligned perfectly with those of the society at large. The Humean beneficence process outlined earlier provides an example of how this can be the case. A single judge may be tempted to make exceptions in specific cases because he gets the personal benefit of performing an act of "compassion." Even if the judge also has a regard for the good of the system as a whole (either directly or through the effects on his reputation), this may not be enough to overcome his other concerns.

The Presentism Objection. The third objection is that a current judgment that some future decision is "bad" may reflect a bias for the present perspective, hence disregarding our future values. Again, we have two responses.

First, the fact that some future decision will seem desirable in light of future circumstances does not imply that the circumstances themselves are desirable. Present decisions often have the capacity to alter the environment in which future decisions will be made. The point of the SSA is not necessarily that the future decisions are bad, in the context of today's point of view, but that we can affect the future context in positive or negative ways by our present  [*573]  decisions. Thus, no judgment against the values in play at some future time is necessarily involved.n91

But second, there is no reason the discussion cannot involve a normative component. If the decisions to be made now have consequences in terms of what values will be held or accepted later, our normative theory need not be indifferent to the outcome. Alternative futures may involve alternative sets of preferences, but that does not mean we have no means of choosing among them. We may have meta-values that are relevant to our choices. Thus it may be rational to avoid the initial decision, or to make that decision but somehow prevent our future selves n92 or future decisionmakers from acting on the then-transformed values. Unfavorable changes in future values are normally dealt with by the prior imposition of constraints.n93 Judges, for example, may try to create precedents or stopping rules that impose constraints on future judges, thereby reducing the likelihood that they will act on the new values.n94

Nevertheless, there is one sense in which we are guilty of presentism. The traditional view of rationality is such that the decisionmaker always attempts to maximize his utility relative to the values and constraints (present and future) that he perceives now at the moment of decision.n95 Thus all his actions are based on that present perception. This is a version of presentism that is perfectly consistent with the theory of rational choice, and thus would not constitute a challenge to slippery slope argumentation from that point of view.


VI. Factors Affecting the Likelihood of Slippery Slopes

Once introduced in an argument, slippery slopes can be difficult to eliminate. This is, in large part, because the slippery slope eliminated at one level of analysis often reemerges at a higher level of analysis. For instance, a commitment to following bright line stopping rules might avoid sorites-style events at the level of decisions, but then the choice of the rule itself may be subject to slippage. This difficulty is exacerbated in the law, where the judges often have the responsibility of both applying rules and choosing them. Any temptation to make exceptions to the rules at the decision level can be recast as a temptation to change the rules at the rules level. This became apparent in the "Make-Up Exam Meshugas" example, where the same position could be cast as a decision-level argument ("Make an exception to the make-up exam rule") or as a rules-level argument ("The make-up exam rule should include an exception for cases like mine").

Still, there are factors that can affect the likelihood and severity of SSEs and hence the persuasiveness of the associated arguments. In this part, we suggest four such factors. We argue that the probability of slipping down a slope is positively related to all of the following, other things equal:

1. The degree of disagreement among decisionmakers in their (lower level) theories.
2. The degree of vagueness in the generally accepted theory.
3. The degree of "empirical vagueness" created by the accepted theory.
4. The degree of looseness of the research program in determining the future development of theories.
We address each of these factors in turn.

The degree of disagreement among decisionmakers in their (lower level) theories. The decisionmakers in a system need not share the same positive or normative theories. Different theories will often lead to different conclusions about how to make decisions and how to select rules. It might seem that disagreement would simply make the system unpredictable or unreliable, but not necessarily more subject to slippage. But slippage can indeed be a problem when multiple theories compete, because multiple theories create a greater potential for problems of nontransitivity in similarity relations. The separately validated propositions process discussed earlier relies on the existence of differing theories that lead to differing judgments about arguments. One argument might be accepted through the efforts of adherents of one theory, a second argument through the efforts of adherents of another theory. The propositions together may encourage the acceptance of yet other arguments that possibly fit neither original theory.

This problem is exacerbated when courts use "analogous reasoning." As the number of acceptable theories becomes greater, relationships of similarity  [*575]  are increased and hence the range of plausible legal doctrines or rules can be extended. For any characteristic a of an established case, there is a greater chance it will be found similar to another characteristic b of a newly arisen case. Furthermore, characteristics may be connected in similarity by groups of jointly incompatible theories: a is similar to b on theory X; b is similar to c on theory Y; and c is similar to d on theory Z, where X, Y, Z are incompatible in whole or part. When courts reason by analogy, the theoretical context for the similarity relation is not always made explicit. As a result, statements of similarity may be made without recognition of their conflicting bases. Hence, "like cases that will be treated alike" may not be truly alike according to a consistent principle or theory. Nevertheless, the rule of a precedent may be expanded n96 beyond the most general intention as manifested in the theory of the original decisionmakers.n97

Furthermore, the existence of multiple theories creates an indeterminacy in the sort of arguments that are viable in a system. Decisionmakers looking for an excuse to decide in a particular way are more likely to find a justification when multiple (and potentially contradictory) justifications exist.n98 If multiple theories have intellectual currency, it is easier to find acceptable reasons to support any given position on a particular case. Thus, for example, the existence of multiple theories creates more room for the Humean beneficence process to operate.

Finally, the existence of multiple theories can lead to the adoption of political, legal, and ethical doctrines that are deliberately vague. For instance, politicians will sometimes pass intentionally vague legislation in order to avoid having to make tough decisions, thereby passing the buck to bureaucratic agencies.n99 Balancing "rules" in the common law, which direct judges to weigh a  [*576]  variety of factors when deciding cases, are arguably a means of finessing the differences among judges' theories. Vague constructs such as the reasonable person may not reflect a consensus among judges about acceptable behavior, but in fact just the opposite: a divergence of opinion about how to identify acceptable behavior. These doctrines can lay the groundwork for sorites-style reasoning, which, as discussed earlier, thrives on the existence of vague words with fuzzy boundaries. Even if vague terms are not deliberately adopted to cover up differences of opinion, they may nonetheless have the same effect. For instance, a precedent for voiding contracts in cases involving "coercion" may turn out to be vague when many different notions of what constitutes coercion exist, even if the judge who first used the coercion standard thought the meaning of coercion to be unambiguous.n100

The degree of vagueness in the accepted theory. Setting aside the existence of multiple theories, it is possible that a single theory can be inherently vague. A theory of ethical behavior, for instance, might rely on the use of terms such as "commitment," "promise," "force," and the like. The meaning of these terms is not self-evident. The theory might provide further definition of these terms, but the definitions themselves may rely on yet other vague terms. Much like "heap," "bald," and "tall," the terms used in political, legal, and ethical discourse may not have clear and obvious boundaries of application.n101 And in the presence of such vague terms, there is again the potential for sorites-style slopes.

 [*577]  The degree of "empirical vagueness" created by the accepted theory. Some theories do a reasonably good job of avoiding conceptual vagueness. The notion of wealth maximization employed in the economic analysis of law, for instance, is a relatively well-defined theory.n102 But the fact that an idea is well defined in theory does not guarantee that it is easily applied in practice. We use the phrase "empirical vagueness" to refer to indeterminacy in the application of a theory, typically created by lack of knowledge on the part of agents and decisionmakers who are expected to apply it.

Consider the question of efficiency (wealth maximization) in the context of tort law. A simple application of the usual economic approach suggests that a rule of negligence plus contributory negligence, with optimally set due-care levels, is the most efficient rule to adopt.n103 The story becomes substantially more complex when it is recognized that the rule adopted must be applied not just to a single case, but to a whole class of cases that will not have identical characteristics. What would be efficient care in one case (considered in isolation) is not necessarily what would be efficient in another. The judgment about what is the efficient rule to apply to the class of cases depends, then, on the distribution of relevant characteristics over both plaintiffs and defendants. Yet this is not information that a single court could reasonably be expected to possess, since (a) each court sees only a subset of all the cases that arise, and (b) the cases that reach the legal system are a biased subset of the class of all the relevant situations that will be affected by the chosen rule.n104

Thus, even if there is broad agreement among decisionmakers about what theory to use, and even if the theory is internally consistent and well defined, the theory may be vague in application. If courts are directed to hold  [*578]  a party liable for actions that are "inefficient," that can be just as vague as telling the court to hold a party liable for actions that are "unreasonable." In practice, decisionmakers will likely have to rely on the precedent set by other courts to decide what is efficient, because they do not have the necessary information to make a direct judgment of efficiency. But following precedent in the context of vague terms is a recipe for the occurrence of sorites-style slopes. The applied boundary between "efficient" and "inefficient" may slide in one direction or the other. The fact that the boundary is sharp in theory does little to prevent the slope, because empirical vagueness creates the problem.

The degree of looseness of the research program in determining the future development of theories. The research program is a broad set of principles that shape the development of theories in a particular area of discussion. Some research programs are relatively tight, putting substantial constraints on the development of theories, while others are relatively loose, allowing more room for divergence among theories consistent with the program. In a loose research program, there is a greater ease of transition from one theory to another, and a greater capacity for individual participants to introduce new theories that are at odds with existing theories. Given the preceding discussion of multiple theories, it follows that a looser research program is more susceptible to the emergence of slippery slopes.n105

VII. Coping With Slippery Slopes

In this part we explore the methods or techniques available, in various decision contexts but especially in the law, to resist or deal with potential slippery slopes. We do not argue that the methods are always used consciously for these purposes but that, at the very least, they respond to the threat of slippery slopes to greater or lesser degrees of success, and that some may have the effect of reducing the likelihood of sliding.


A. Accepting the Trade-Off

Suppose that a decisionmaker has just been exposed to a persuasive SSA. The SSA convinces him that making some desirable decision now will lead to some undesirable decision later. So what should he do? The simplest response is to accept the trade off: The desirable and the undesirable cannot be separated, so they must be accepted or rejected as a package. The good must be weighed against the bad to make a decision. If the bad outweighs the good, then the SSA averts the SSE by preventing the initial decision. If the good outweighs the bad, then the potential SSE becomes an unpleasant but expected consequence of the initial decision.n106

Although accepting the trade-off is one possible response to the SSA, it is not a satisfying one, so decisionmakers are inclined to seek other strategies. Probably the most common strategy is to attempt to create a rule that will prevent the SSE from taking place.

B. Stipulating an Arbitrary Stopping Rule

The decisionmaker may attempt to establish a clear rule, a line between the cases in which future decisionmakers should take a particular action and cases in which they should not. For instance, consider the question of executing murderers who are mentally retarded. Although many people would agree that retarded persons should not be executed for their actions, this question is susceptible to a sorites-style slope because of the vagueness of the concept "retarded." IQ is often considered a summary statistic or proxy for intelligence, although it clearly does not capture everything we mean by intelligence. Nevertheless, IQ is a characteristic located on a continuum, and it is not clear where the line should be drawn to separate those whose IQ is high enough to allow execution from those whose IQ is too low. To resolve this problem, society might adopt a somewhat arbitrary rule saying that a murderer with an IQ of seventy-one or greater may be executed, while all others may not. The decisionmakers in actual cases are directed to decide according to this rule, rather than by analogy to similar cases.n107

 [*580]  This approach could avoid the slippery slope at the level of decisions, but it could reemerge in another form. If the rule itself should ever be called into question, then the very process of rule selection could be susceptible to the same kind of sorites reasoning.n108 If all persons with IQs of seventy or greater may be executed, then why shouldn't the person with an IQ of sixty-nine get the same treatment? In response to this challenge, the rule could be moved by increments in much the same way the decisions were. To a certain extent, entrenched rules in general and an entrenched IQ rule in particular are arbitrary. This is because the rule maker refuses to change them even when they appear to be inconsistent with their underlying justifications. If the rule can be maintained, SSEs may be avoided here, but the very arbitrariness of the rule may weaken the rule maker's resolve to hold firm.

C. Appealing to a Higher Standard

In this approach, the decisionmaker appeals to a higher standard for judgment in cases in which the correct decision is unclear. This approach is  [*581]  most natural when there is already a rule in place, but the rule itself includes a vague term. Consider the well-known example of a rule from H.L.A. Hart: n109 "No vehicles are permitted in the park." The word "vehicle" is inherently vague. It is entirely possible that a myopic analysis of the term "vehicle" - perhaps through a series of analogies - can rationalize the extension of this rule to motorized toy cars or wheelchairs. But if the courts were to make reference to the underlying original rationale (for example, protection of pedestrians from serious traffic accidents) instead of focusing on the meaning of the word "vehicle," the SSA and SSE might be avoided.

Or not. Appealing to a higher standard effectively changes the level of discussion from rules to theories. Whether appealing to a higher standard is an effective means of avoiding slippery slopes depends on the characteristics of the theory itself. In the "no vehicles in the park" rule, it seems likely, though not certain, that substantial agreement will exist about the theoretical rationale for the rule. The clarity of the standard helps to "nail down" the rule. But this need not always be the case. Choosing an appropriate standard involves the selection (perhaps implicit) of a theory, and raising the discussion to the level of theory can actually increase the likelihood of a slippery slope. This may be true for any of the reasons discussed in the last part: Theories may be inherently vague, theories may create empirical vagueness, or there may be disagreement among theories.

As an example of theoretical vagueness, consider the treatment of obscenity in First Amendment law. It is well established that obscene material does not enjoy First Amendment protection, while other material (that does not fall in another unprotected category) does. Uncertainty about the meaning of "obscene" prompted the Supreme Court to adopt a standard, in Miller v. California,n110 under which an allegedly obscene work must not, "taken as a whole, ... have serious literary, artistic, political, or scientific value" to lose its First Amendment protection. In essence, the Court directed lower courts to interpret a rule about obscenity by reference to a higher standard, or theory, that defines obscenity in terms of the social value of the expression. The problem, of course, is that judges and legal scholars who use the term "value" might very well possess different, perhaps radically different, theories of value in literature, art, politics, and science. If judges regularly made reference to the theory enunciated in Miller, we might worry that, little by little, the rule protecting nonobscene works from regulation would be eroded as the social value of the work is  [*582]  increasingly taken into account.n111 If other areas of free speech jurisprudence relied on the alleged social value of individual acts of expression to the same extent as obscenity doctrine, without other buttressing justifications, it is not hard to imagine that courts would be left with an unacceptably weak First Amendment.n112 Freedom of speech becomes more resistant to slippery slopes when we treat it as an "entrenched abstraction" n113 or a generalization that is largely immune to exception-making. The fundamental difficulty with appealing to "social value" as the primary underlying standard for protection of free speech is that social value is an inherently vague term, susceptible to a variety of interpretations.n114

As an example of empirical vagueness, suppose we are interested in the question of when contracts should be voided by the courts. According to one current economic theory, it is desirable for a contract to be voided when it is likely that the transaction did not make both parties better off in expected value than they would have been without the contract. This is an efficiency standard. One possibility is to examine each case individually to determine whether it fulfills the standard. This presupposes that the judge has sufficiently good data to make such an individual determination. In practice, however, the standard is empirically vague, and so courts will probably argue by analogies to clear cases.

Now, there presumably exists a spectrum here, from cases in which the contract clearly made both parties better off, to cases in which one party clearly suffers an ex ante loss because the contract was signed at gunpoint. In between, there are cases in various shades of gray. What if, for instance, one party threatened to withhold sexual favors? Or to inflict mental anguish by reminding him in graphic detail of his abusive childhood? Or to kill a beloved pet that  [*583]  belongs to the threatening party? Or to reveal potentially embarrassing secrets about his personal life? Some of these cases might appear quite similar to the contract signed at gunpoint. It is conceivable, furthermore, that a chain of cases could be found that connects the clear cases for upholding the contract to the clear cases for voiding the contract. As a result, courts following precedent in similar cases might be led to void contracts that should clearly be upheld (if the slope goes in that direction), or to uphold contracts that should clearly be voided (if the slope goes in the other direction).

Therefore, the level of discussion per se is of little significance for sliding. What is significant is whether we have moved to a more or less vague, or to an empirically more or less determinate, level of discussion. In some cases, theories will be particularly susceptible to sliding while rules will be less so; in other cases rules will be relatively susceptible while theories less so.n115

D. Adopting an Open-Ended or Standard-Mediated Rule

This strategy represents a compromise between the previous two approaches. The decisionmaker accepts the first decision, and even some subsequent ones as well, but follows or imposes a rule on other decisionmakers that stops the process short of the danger case. Unlike the strategy of stipulating a somewhat arbitrary stopping rule, the rule here is chosen based on a factor that has a stronger theoretical rationale.

In the contract example above, this would mean specifying the conditions under which contracts may be voided. A rule embodying these conditions would  [*584]  likely refer to factors generally but imperfectly correlated with the likelihood of a value-decreasing contract. One such rule is that "a contract should be voided if it was formed in the presence of duress." A potential difficulty with this rule is the vagueness of the concept "duress."n116 Although there are clear cases of duress and clear cases of no duress, there is also a spectrum of cases in between. Is the infliction of severe mental distress a form of duress, or is some form of physical intimidation necessary?n117 Suppose that the courts (in their capacity as rule makers) take the latter approach, setting physical intimidation as a necessary and sufficient condition for voiding a contract on grounds of duress. This rule might be capable of preventing a slippery slope on the level of decisions, as courts deciding cases would only have to verify the existence of a physical threat.n118 But at the level of rule selection, the rule could be exposed to challenges based on the over-and under-inclusiveness of the rule relative to the underlying standard. Suppose that in case x one party threatened to yank the last hair on a (nearly) bald man's head, and in case y one party threatened to recount the details of the other party's abusive childhood (and evidence shows that this was indeed extremely frightening to him). The stated rule would void the contract in case x but not in case y, even though y appears to present the stronger case for voiding on the basis of a plausible theory of human motivation. Just as in the IQ example given earlier, a questioning of the rule relative to its underlying justification could be used to weaken, change, or move the rule in one direction or another.

As another example, consider the issue of abortion. Suppose the cost of remaining at a postulated status quo of no abortions under any circumstances is high. Suppose also that the cost of accepting the danger argument for infanticide is even higher. Does this mean that the decisionmaker will not take the first step if a persuasive SSA is made that "leads" from abortion to infanticide? Not necessarily. The idea behind the argument is that there is a simple gradient that connects zygote to infant child. But if there are important turning points  [*585]  within that gradient, there may be an effective stopping rule that can serve to differentiate the cases and stop the process. The decisionmaker may focus, for example, on the development of a functioning cerebral cortex as a rule-like criterion. This would permit some early abortions while not permitting infanticide. Whether the stopping rule will hold depends, at least in part, on its being perceived as nonarbitrary. For this to be the case, an argument must be made at the level of theory. Some have claimed that there are good reasons to believe that the development of a functioning cerebral cortex is a non-arbitrary stopping point. The cerebral cortex is responsible for many of the functions or capacities that we usually conceive a human person to have.n119 So a rule that prohibits abortions beyond that stage will be consistent with a theory that privileges the human person rather than merely developing human life or tissue. The persuasiveness of this theory will be imputed downward to the rule and determine its effectiveness as a stopping point.

It seems, then, that neither the use of arbitrary rules nor underlying standards nor quasi arbitrary rules offers an infallible escape route from slippery slopes, because slippery slopes can emerge in the process of rule selection as well as in the process of rule application. Still, it is possible that these approaches can reduce the likelihood of slippery slopes.

E. Altering the Scope and Power of Precedent

The reader may notice that we have taken a somewhat ambivalent position on the desirability of precedent. On the one hand, we have indicated that following precedent could encourage slippery slopes of the sorites variety. On the other, we have implied that the progressive weakening of precedent was responsible for the Humean beneficence slippery slope. The seeming contradiction dissipates once we realize that the role of precedent in retarding a slippery slope depends on the assumed location of the danger case and which decisions are regarded as precedent.

Consider the mathematical version of the sorites story, in which one end of a continuum (the "one" end) is the clearest case for taking action A, and the other (the "zero" end) is the clearest case for not taking action A. The slope, as we described it, involved action A being taken in more and more cases, so that eventually it is taken in some case where it clearly should not be - the danger case. If we suppose that action A makes an exception to some rule, the problem is that when an exception is made in a clear case for doing so, precedent allows and perhaps even requires that the exception be made in similar but less clear  [*586]  cases. If making an exception did not establish a precedent for further exceptions, there would be no problem of a slope in the direction of too many exceptions.n120 To put it differently, what if we supposed that the application of the rule (hence the refusal to do A) was treated as the only relevant precedent? Then the slippery slope, if any, would occur in the opposite direction, resulting in too few exceptions to the rule.

A similar set of observations can be made about the Humean beneficence process. In the early stage of that story, when few or no exceptions had been made to the established rule, precedent played a restraining role: Judges were loath to make exceptions because doing so would deviate too much from the established case law. But in the late stage of the story, when many exceptions had been made, precedent played an enabling role: Judges who wished to make further exceptions could easily find previous decisions to justify their own. If exceptions did not act as precedents, then as in the sorites story, there would have been no slippery slope in the direction of exception-making. The problem, if any, would have been the making of too few exceptions.

Thus, we can see that precedent has both an enabling and a restraining aspect.n121 Whether either aspect is desirable depends crucially on what outcome is identified as the danger case. When the danger case is making excessive exceptions to some rule, then the restraining aspect is SSE-retarding and the enabling aspect is SSE-encouraging. When the danger case is making too few exceptions, then the reverse is true.

Precedent is thus an imperfect attempt to enforce rules adopted for the purpose of avoiding slippery slopes.n122 The rules enforced may be arbitrary or standard-based. As indicated above, such rules can only provide a partial barrier to slippery slopes, in large part because disputes about the application of rules can reemerge as problems of rule selection. The problem is exacerbated by (a) the continual emergence of novel cases for which the application of rules  [*587]  is unclear and (b) the mingling of rule application and rule selection functions in the legal sphere.

As a result, courts may find it very difficult to separate the restraining and enabling aspects of precedent. To do so, they would need to establish a distinction between decisions to be regarded as binding precedents and decisions to be regarded as mere exceptions.n123 How might this be done? Schauer observes that some areas of constitutional law, especially those involving First Amendment prohibitions on the reach of government regulation, incorporate "entrenched" abstractions as a fundamental part of their jurisprudence. n124 The law pertaining to First Amendment freedom of speech has made all manner of activities, such as marches, "speech."n125 Further, it inhibits examination of the empirically vague underlying theories or standards that would determine the social worth of speech relative to its social costs.n126 Once an activity falls under the abstract characterization "free speech," it becomes immune to examination at a deeper level.n127 This approach attempts to enhance the precedential power of decisions that favor freedom of expression, while muting the precedential effect of decisions that do not by characterizing them as narrowly defined exceptions.

With regard to constitutional prohibitions on government behavior, the expansive conception of certain abstractions operates to resist slippery slopes in the direction of excessive restrictions on private behavior. The entrenched abstraction limits the putative danger inherent in overextension of the category, justifiable restrictions on speech, by allowing the possible overextension of another category, desirable acts of expression. This makes perfect sense if the identified danger case is the excessive restriction of speech. But if the Court had identified excessive freedom of expression as the danger case, then far from being SSE-retarding, the entrenched abstraction could be regarded as SSE-encouraging.


F. Establishing Presumptions, Burdens of Persuasion, or Levels of Scrutiny

Another possible means of protecting a rule against erosion is to privilege certain crucial facts by a legal presumption. If a rule refers to some characteristic about a case, but the characteristic is theoretically or empirically vague, then requiring the finder of fact to infer its presence from certain sharper "basic facts" can make sliding less likely.

Consider again the example of executing retarded persons. "IQ of 70 or below" is a simple rule for inferring mental retardation (a vague characteristic) from one's IQ score (a sharper or more observable characteristic). As we observed earlier, this rule could be subject to a slippery slope. But now suppose the rule is treated as a presumption that an IQ of 70 or less indicates mental retardation.n128 Under these circumstances someone (for example, a district attorney trying to obtain an execution) may wish to claim that an IQ of 70 is indistinguishable from an IQ of 71 (not retarded or "normal") and hence the defendant with an IQ of 70 should be subject to capital punishment as well. The presumption throws an obstacle in the way of the argument. At a minimum, the presumption requires the proponent of action to produce or come forth with sufficient evidence that a defendant with an IQ of 70 should be regarded as normal. He may not be able to do it. Furthermore, under the "reformist approach," n129 the presumption will shift the burden of persuasion on this issue to the proponent, requiring him to prove by a preponderance of the evidence that the defendant is of normal intelligence. This, ex hypothesei, he will not be able to do. If all he can say is that there is no reason to differentiate 70 from 71 or from 69, for that matter, then he is saying that there is no better reason to consider 70 retarded as opposed to normal. Hence the preponderance standard cannot be met.

The problem with this approach is very similar to that of the arbitrary rule. Just as any dispute at the rule application level can be recast as a dispute at the rule selection level, any dispute at the presumption-application level can be recast as a dispute at the presumption-selection level. The proponent can ask, why should the presumption apply to IQs of seventy and below, rather than sixty-nine and below? Indeed, a presumption is really just a different sort of rule, possibly a weaker one since it is explicitly defeasible. But paradoxically, the presumption's greater defeasibility is also its virtue. Implicit in the presumption  [*589]  is a recognition of the arbitrary character of the rule, with an allowance for exceptions to be made in cases with sufficient proof. When exceptions are made, they do not constitute changes in the rule itself, nor do they necessarily set precedent for future cases. As a result, a presumption may provide less traction for arguments in favor of shifting the rule. In another example, the rule that sets eighteen as the age of legal majority is actually a presumption, because a person under eighteen can petition for emancipation under unusual circumstances.n130 The existence of a possible exception for very special cases may, oddly enough, add legitimacy to a rule that would otherwise appear excessively arbitrary.

In constitutional law, higher levels of scrutiny will be applied to certain categories of state action. Content-based restrictions on freedom of speech, for instance, are exposed to strict scrutiny, whereas content-neutral restrictions face a lower (intermediate) level of scrutiny.n131 Presumably, the Supreme Court has identified content-based restrictions as more perilous - that is, closer to the danger case. Nonetheless, exceptions are allowed in cases in which the state interest is especially compelling. This approach tips the scales against speech restrictions without prohibiting them entirely. A government that wishes to institute a policy favoring free speech will not be expected to justify its choice, whereas a government wishing to institute a policy restricting the content of speech will be expected to provide substantial justification. This approach probably generates fewer challenges to the rule itself (at the rule-selection level) than would a rule prohibiting all content-restricting policies without exception. By providing a safety valve for the most persuasive exceptions, it protects the rule against direct legal challenge.

It should go almost without saying that the presumption approach has its dangers. If the individual cases allowing exceptions (on grounds that the burden of proof or standard of persuasion has been met) are regarded as enabling precedents, so that similar cases with slightly less support are seen as within their orbit, then the presumption may encourage SSEs rather than inhibit them.

G. Creating Supermajority Requirements and Constitutional Constraints

We have been discussing legislative and judicially created impediments to SSEs. Another approach is to constrain those forms of decisionmaking by supermajority requirements and similar forms of constitutional constraint. These constraints can assure that certain types of change will occur only if there is a sufficiently large amount of support for the change.

 [*590]  Consider the legal voting age. There is nothing special about the age of eighteen that makes it the uniquely correct minimum voting age. The arbitrariness of the rule becomes apparent when, for example, a high school senior whose birthday is November 9th cannot vote in the presidential election, whereas a high school drop-out whose birthday is one week earlier can. But no other voting age, at least within some range, would be any less arbitrary, so this rule may be as good (or bad) as many others. If voting privileges were determined individually - say, in legal proceedings - it is not hard to imagine that the voting age might slide, by increments, in one direction or the other. The imbedding of the voting age in the Constitution assures that this sliding cannot occur. Even if a majority of people agreed that eighteen-and-one-month-olds should not be able to vote, that would not be sufficient to change the rule. Only the passage of a constitutional amendment could achieve that, and amendments are notoriously difficult to pass.

In a sense, the differential barriers to the alteration of different types of law or policy reflect the different layers in the structure of discussion. The debate over what terms should be included in a constitution (say, at a constitutional convention) is likely to occur at the level of theory, as the discussion is explicitly focused on what the basic rules should be.n132 Once the constitution is established, the subsequent discussion takes place largely within the established rules. Some may make arguments against the rules adopted, but those arguments are not generally understood as arguments for and against decisions being made within the current rule structure. For instance, an argument against the presidential veto (because, say, it gives too much power to one man) would not be considered a reason to enforce a bill that was passed by Congress and vetoed by the president. Until the Constitution is changed, the veto remains in place.

This is, of course, an idealized view. In any actual constitution, there exist many vague terms, and so there will inevitably be debate about their meaning. Different theories will exist both about what the terms do mean and what they should mean. There will inevitably be some blurring of the distinction between arguments about rule application and arguments about rule selection. A well known, if controversial, position states that the U.S. Constitution is a "living document" whose content is determined by its interpretation, which changes over time in response to changes in society.n133

 [*591]  It is the possibility for blurring that is the Achilles' heel of the supermajority/constitutional constraints strategy for retarding slippery slopes. We have previously observed that virtually any dispute about rule application can be recast as a dispute about rule selection. We now observe that often the reverse is also true: Disputes about rule selection can be recast as disputes about rule application. If the existing rules are sufficiently vague, skillful advocates can argue that the rules effectively give discretion to the decisionmaker, who can employ whatever normative and positive theories he thinks best.


The key feature that distinguishes SSAs from other forms of argument is that they are arguments about arguments. The proponent of an SSA claims to predict how acceptance of one argument will lead (with increased likelihood) to the acceptance of other arguments not identical to the first. Whether the speaker's prediction is correct depends crucially on the process that he claims will lead from earlier arguments to later arguments. To evaluate such a process, one needs to understand the structure of discussion in which arguments are made and accepted. In this Article, we have attempted to fill this need.

The primary tool of our analysis is the structure of discussion and argument outlined in Part II. This structure characterizes discussion and argument as occurring in a hierarchical fashion. The lowest rung of the hierarchy is the decision to be made. The next rung up is the rules, which are applied (sometimes) in the making of decisions. The next rung is theories, which are applied (sometimes) in the selection of rules. The highest rung is research programs, which are used to constrain the selection of theories. Arguments can take place at any level in the structure: to influence the making of rules, to influence the selection of rules, to influence the debate among theories, and so on.

The maker of an SSA purports (implicitly) to have some knowledge of the actual content of the structure of discussion - that is, the actual rules, theories, and research programs at work in the minds of the participants. To the extent that the speaker's purported knowledge is accurate, his argument may be a good description of the likely development of future arguments in the system. In short, the SSA is valid. If the description is not entirely correct, of course, the argument is less valid (or just plain wrong).

We have described several types of SSAs that we think can, at least under some circumstances, be valid. But in each case, we have emphasized that the evaluation depends crucially on the proponent's model of how people make  [*592]  and adopt arguments and other ideas. Even if readers reject one or more of these types of argument, or the specific examples accompanying them, they will hopefully find the overall structure useful in understanding the nature of slippery slope arguments in general.

Slippery slopes are slippery in more ways than one. Aside from sliding from one argument to another, there can also be sliding from one level of discussion to another. Slopes at the level of decisions can become slopes at the level of rule selection, and vice versa. It is this characteristic, we think, that makes them so difficult to deal with. Nonetheless, there exist a variety of imperfect means for resisting slippery slopes, which we have discussed in Part VII of the Article. There may be yet other means. If slippery slopes can indeed be a legitimate form of argumentation, as we suggest, then finding effective means of coping with them will hopefully become a priority in legal, political, and ethical debates.


* Department of Economics, New York University.
** Department of Economics, California State University, Northridge. We are indebted to the James M. Buchanan Center at George Mason University, the Pierre and Enid Goodrich Foundation, and the W.B. Earhart Foundation for financial support. We are also indebted to the following groups for their comments and suggestions: the participants of the Liberty Fund Colloquium on Slippery Slopes and the Austrian Economics Colloquium at New York University. Many individuals have provided valuable suggestions, especially Peter Boettke, Mark Brady, William Butos, Young Back Choi, Adam Gifford, David Harper, Tsutomu Hashimoto, John Hasnas, Sanford Ikeda, Israel Kirzner, Roger Koppl, Elisabeth Krecke, Chidem Kurdas, Ejan Mackaay, Robert Murphy, Edna Ullmann-Margalit, Joseph Salerno, Eugene Volokh, and Douglas Walton. Responsibility for errors remains ours alone.

n1. For an explanation of how price controls on milk can lead to further controls on the prices of inputs into milk production, see Ludwig von Mises, Middle-of-the-Road Policy Leads to Socialism, in Planning for Freedom and Sixteen Other Essays and Addresses 22-24 (4th ed. 1980). See more generally Ludwig von Mises, Human Action: A Treatise on Economics 762-64 (3d ed. 1966), for a description of the process by which price controls on some goods and services lead to demand for price controls on other goods and services.

n2. See, e.g., Collin v. Smith, 578 F.2d 1197, 1205-06 (7th Cir. 1978).

n3. "Infanticide (killing of newly born children), also called neonaticide, follows abortion like night follows day." Tennessee Right to Life, Human Life Issues, at<uscore>life<uscore>issues/human<uscore>life<uscore>issues<uscore>infanticide.htm. For a more objective analysis of slippery slope arguments in the context of abortion, see, for example, Douglas Walton, Slippery Slope Arguments 45-50 (1992).

n4. Sanford Ikeda, Dynamics of the Mixed Economy: Toward a Theory of Interventionism (1997).

n5. Eugene Volokh, The Mechanisms of the Slippery Slope, 116 Harv. L. Rev. 1026 (2003).

n6. See generally Walton, supra note 3.

n7. "A slippery slope argument claims that permitting the instant case - a case that it concedes to be facially innocuous and that it linguistically distinguishes from the danger case - will nevertheless lead to, or increase the likelihood of, the danger case." Frederick Schauer, Slippery Slopes, 99 Harv. L. Rev. 361, 369 (1985) (emphasis added).

n8. Walton, supra note 3, at 3-7. A sorites argument claims that, with respect to a critical characteristic, it is impossible to say where the dividing line is between the clear presence of the characteristic and its clear absence. A precedential argument is based on the notion that a decision in a particular case commits one to decide the same way in future similar cases. A causal argument claims that an initial event causes further events leading to an ultimate bad outcome, similar to a domino effect. To see the interaction of all three in a full argument, consider a rule that would make the termination of "biologically unworthy" life permissible. This is obviously a vague concept without a clear cut-off point. Initially, it might be applied to cases of infants born without some important part of their brains. If such a decision were to become a precedent, then it could be applied in other, somewhat different, cases. For example, the termination of grossly mentally defective life might lead over a series of events to the termination of the lives of those with grossly deformed facial features.

n9. Schauer, supra note 7, at 365.

n10. Some clarification of the idea of "somehow flowing with high probability" is in order. In order to differentiate this objection from the automaton objection, we do not focus on the (in)voluntariness of the future decisions. Instead, we treat the high probability of undesirable future decisions as emanating from the inability of decisionmakers to find relevant dissimilarities between a future case and the current one. Thus, the imputation is epistemic rather than causal. It is the joint consequence of the decisionmaker's adherence to a principle of universalizability ("treating similar cases in a similar way") and his inability to discern a relevant dissimilarity. "If we judge X to be right, and we can point to no relevant dissimilarities between X and Y, then we cannot judge Y to be wrong." Tom L. Beauchamp & James F. Childress, Principles of Biomedical Ethics 120 (2d ed. 1983).

n11. "Presentism" refers to the stipulation that the individual never acts counter-preferentially. He always acts to maximize the satisfaction of his current preferences, including his current preferences about the future. So at t = 0 the agent wishes mightily to avoid the "danger case," but when the choice arrives at t = 1, with changed preferences, he embraces it (if he has not previously bound himself against it). Presentism excludes the possibility that the agent may avoid certain choices simply because of previous preferences or a commitment to oneself based on them. See Jed Rubenfeld, Freedom and Time: A Theory of Constitutional Government 103-30 (2001).

n12. In this Article, we choose the term "process" over "mechanism." Although mechanism is, in some respects, more precise, it may convey a sense of automaticity or deterministic reaction as in the common use of the word "mechanistic." We wish to avoid that connotation. In what follows, we do not intend to suggest that the processes discussed are completely deterministic.

n13. The results of the microprocesses are usually unexpected. The source of this is the aggregation of individually plausible premises and arguments. These individually plausible statements may not be collectively consistent. The "paradoxical" conclusions reached by a decisionmaker are the results of prior intellectual commitments. See generally Nicholas Rescher, Paradoxes: Their Roots, Range and Resolution 65-70 (2001). Microprocesses can therefore be set in motion by a single mind that is not fully aware of the consequences of the aggregation of its accepted arguments. In this sense, they are generated by individual action or decisions but not by individual design. See generally Friedrich A. Hayek, The Results of Human Action but Not of Human Design, in Studies in Philosophy, Politics and Economics 96 (1967).

n14. In general, macroprocesses are those that stem from the existence of social systems and systems effects. The defining characteristics of systems are interconnections between decisions and emergent outcomes. See generally Robert Jervis, System Effects: Complexity in Political and Social Life 3-25 (1997). In these cases, the aggregation of the actions of many individuals is the essence of the process. The consequences of this aggregation are beyond the anticipation and intention of the actors because of the cost of such knowledge, the effects of bounded rationality, or the lack of intellectual insight and alertness. Thus, in a second sense, these are the results of individuals' actions but not of individuals' designs. See supra note 13.

n15. F.A. Hayek, The Counter-Revolution of Science: Studies on the Abuse of Reason 61-65 (2d ed. 1979).

n16. 578 F.2d 1197 (7th Cir. 1978).

n17. Id. at 1205-06.

n18. Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-Based Decision-Making in Law and in Life 2 (1991).

n19. Edward F. McClennen & Scott Shapiro, Rule-Guided Behavior, in The New Palgrave Dictionary of Economics and the Law 363, 363 (Peter Newman ed., 1998).

n20. Compare, however, Joseph Raz:

The distinction between rules and principles of obligation both in law and outside of it turns on the character of the norm-act prescribed. Rules prescribe relatively specific acts; principles prescribe highly unspecific actions ... . The distinction is ... one of degree, since there is no hard and fast line between acts that are specific and those which are unspecific.
Joseph Raz, Legal Principle and the Limits of Law, 81 Yale L.J. 823, 838 (1972).

n21. "A rule withdraws from the decisionmaker's consideration one or more of the circumstances that would be relevant to [a] decision according to a standard." Isaac Ehrlich & Richard A. Posner, An Economic Analysis of Legal Rulemaking, 3 J. Legal Stud. 257, 258 (1974).

n22. A positive theory explains or predicts an event or state of affairs without reference to the value judgments of the theorist. A normative theory establishes an analytical ideal against which some aspect of the world is evaluated by the theorist or observer.

n23. "Rule-based decision-making ... is a form of decision-making arising within some theory of justification and existing only relative to it." Schauer, supra note 18, at 86.

n24. Walton, supra note 3, at 131-32.

n25. Karl R. Popper, The Logic of Scientific Discovery 421-22 (10th ed. 1980).

n26. This point has not always been recognized, even by distinguished legal scholars. Edward Levi, for example, believed that the "basic pattern of legal reasoning is reasoning by example." Edward Levi, An Introduction to Legal Reasoning 1-2 (1949). The pattern consists of three steps. First, "similarity is seen between cases; next the rule of law inherent in the first case is announced; then the rule of law is made applicable to the second case." Id. (emphasis added). This deceptively simple procedure involves theory at every stage. "Similarity," as we have seen, is dependent on a theoretical construct. Extracting the "inherent" rule of law depends on a theory that correctly identifies a set of factors. To apply a rule requires that we distinguish relevant and irrelevant characteristics of the new case. Furthermore, since a rule is first established in a particular factual context (never exactly repeated), it must change, even slightly, as it is applied. A theory establishes the framework of allowable changes in rules justified by the theory.

n27. Imre Lakatos, Falsification and the Methodology of Scientific Research Programmes, in Criticism and the Growth of Knowledge 91 (Imre Lakatos & Alan Musgrave eds., 1970). For a summary of the components of a research program, see id. at 132-35.

n28. There is an obvious similarity with Thomas Kuhn's idea of a scientific paradigm. See generally Thomas S. Kuhn, The Structure of Scientific Revolutions (2d ed. 1970). We prefer the concept of a research program to that of the paradigm, however, for reasons identified by John Worrall:

Although Kuhn's detailed development of [the paradigm] - especially his emphasis on inarticulable skills, "disciplinary matrices", and the like - can be challenged (and certainly stands in need of clarification), he was surely pointing in the direction of an important and then relatively neglected aspect of mature science. Imre Lakatos, with his notion of research programme complete with "positive heuristic", and Larry Laudan, with his notion of a research tradition, both later underlined this same point in slightly different (and considerably sharper) ways.
John Worrall, Philosophy and the Natural Sciences, in Philosophy 2: Further Through the Subject 203 (A.C. Grayling ed., 1998) (emphasis added).

n29. This usage may appear awkward in the case of law, since it seems that a research program should have something to do with research. Law, in fact, is generated by an intellectual framework with assumptions, premises, and methods. So in this sense, legal decisions are the result of "research," or the development of theories, rules, and arguments within a "research program."

n30. Melvin Aron Eisenberg, The Nature of the Common Law 11 (1988).

n31. Id.

n32. As Gerald Postema notes:

Classical common law jurisprudence resolutely resisted the theoretical pressure to identify law with canonically formulated, discrete rules of law. Law, on this view, is not a set of rules or laws, but a practised framework of practical reasoning, and this practised framework provides a form of social ordering. Its rules and norms can be formulated, perhaps, but no such formulation is conclusively authoritative.
Gerald Postema, Philosophy of the Common Law, in The Oxford Handbook of Jurisprudence and Philosophy of Law 596 (Jules Coleman & Scott Shapiro eds., 2002).

n33. Social wealth maximization may be seen as a generalization of the more specific standard of the minimized sum of expected accident and accident prevention costs mentioned infra Part II.B. Minimization of the latter does not necessarily imply maximization of the former in a model with more than two cost variables.

n34. See generally Richard A. Epstein, Simple Rules for a Complex World (1995).

n35. See generally Richard A. Posner, Economic Analysis of Law (5th ed. 1998).

n36. Over the years the differences between Posner and Epstein have narrowed considerably, coming down to an important matter of emphasis. Posner stresses the multidimensionality of the choice between negligence and strict liability. He believes that there are many efficiency considerations: some more applicable in some contexts than in others, including the levels of care and of activity, and the response to court errors in applying the negligence standard. "Because of these differences between negligence and strict liability, we would not expect the tort system to opt all for one or all for the other. Nor would we expect the balance between the two regimes to be the same at all times." Posner, supra note 35, at 196. Epstein, on the other hand, has concluded that because of high administrative costs in making the judgments required in a negligence system (such as, the costs of untaken precautions and the probabilities of unique accidents), strict liability is generally more efficient than negligence. "But this implicit acquiescence in cost/benefit thinking does not require making cost/benefit analyses as part and parcel of the legal rules. Quite the opposite, achieving the efficient social outcome often requires that the legal rules consciously avoid making any explicit reference to cost/benefit analysis." Epstein, supra note 34, at 97.

n37. Not all arguments are deductive in form. A deductive argument such as that in the text presupposes that there is no problem interpreting an existing normative rule as the generalization in the major premise. It also presupposes that there is sufficient legal or normative warrant for the major premise in the first place. In a context in which there are no explicitly stated rules, one must go beyond the "problem of interpretation" to determine whether there is sufficient warrant in the appropriate sources for the generalization (the "problem of relevance"). Establishing the correct interpretation or finding sufficient warrant for a rule is not arrived at by deductive argument. Other methods, such as analogous reasoning, must be used. See Neil MacCormick, Legal Reasoning and Legal Theory 19-72 (H.L.A. Hart ed., 1978).

n38. Theories can exist at multiple levels; a low-level theory might be a relatively specific normative standard. See the discussion infra Part II.F.

n39. Research programs may include relatively general or abstract normative standards.

n40. "Meshugas" is Yiddish (and New Yorkese) for craziness. For a humorous discussion of the use of Yiddish in court decisions, see Alex Kozinski & Eugene Volokh, Lawsuit, Shmawsuit, 103 Yale L.J. 463 (1993).

n41. Relatedly, in a humorous article, Mike Adams observes that a "student's grandmother is far more likely to die suddenly just before the student takes an exam than at any other time of the year." Mike Adams, The Dead Grandmother/Exam Syndrome and the Potential Downfall of American Society, Conn. Rev., Summer 1990, at 70, 70, available at The author facetiously concludes: "Family members literally worry themselves to death over the outcome of their relatives' performance on each exam." Id. at 72.

n42. When there is discussion at a certain level, it may employ items derived from a higher level. A discussion at the decision level will employ rules (that are effectively taken for granted). The discussion at the rules level will employ a theory (that is taken for granted). A discussion at the theory level will employ a research program (again taken for granted).

n43. Presumably the more common research program would include the normative standard of maximizing the joint welfare of students and professors or, less plausibly, the welfare of students alone.

n44. Evidence for this proposition is provided by California's motorcycle helmet law. Starting in 1992, all motorcycle riders have been required to wear helmets in an attempt to reduce the number of injuries and fatalities arising out of motorcycle accidents. One reason the state legislators adopted the law was to save taxpayer money. The taxpayers bore most of the cost of the accidents because fewer than half those hospitalized had private medical insurance. See Wendy Max et al., Putting a Lid on Injury Costs: The Economic Impact of the California Motorcycle Helmet Law, 45 J. Trauma: Injury, Infection & Critical Care 550 (1998) ("During the first 2 years of implementation of California's helmet law, there were reduced costs for injuries and fatalities and large dollar savings to the state and other payers compared with the previous year."). Note also that recent lawsuits against the tobacco and gun industries have relied on the notion that the use of tobacco and guns increases public health costs. See Douglas Glen Whitman, Legal Entrepreneurship and Institutional Change, 12 J. des Economistes et des Etudes Humaines 257 (2002).

n45. Strictly speaking, the premise need not be applied repeatedly. By mathematical induction, the result can be achieved by recognizing that adding or subtracting one grain never matters. See Stephen Read, Thinking About Logic: An Introduction to the Philosophy of Logic 174-75 (1995); see also "Sorites," in The Dictionary of Philosophy 524 (Dagobert Runes ed., 2001) ("In a statement of a sorites all conclusions except the last are suppressed, and in fact the sorites may be thought of as a single valid inference independent of analysis into constituent syllogisms.").

n46. "Vagueness is a widespread feature of our thought. Consider the following list: "child', "book', "toy', "happy', "clever', "few', "cloudy', "pearl', "moustache', "game', "husband', "table.'" R.M. Sainsbury, Paradoxes 26 (2d ed. 1985).

n47. Rescher, supra note 13, at 78-79.

n48. Logicians often call this the "tolerance principle." See, e.g., Mark Sainsbury, Logical Forms: An Introduction to Philosophical Logic 275 (1991).

n49. Philosophers and linguists have attempted to resolve the sorites paradox in a variety of ways. See, e.g., Timothy A.O. Endicott, Vagueness in Law 77-98 (2000); Rescher, supra note 13, at 77-83. We do not attempt to add to this literature. It is worth pointing out, however, that one unsuccessful method of resolving the paradox involves trying to create a third category, the "unsure" or "ambiguous" category. For example, in the case of collections of grains of sand, we might say there are clear heaps, clear nonheaps, and cases that are neither. The problem with this approach, and the reason it does not resolve the paradox, is that the borderline between clear cases and unclear cases is itself vague - and therefore susceptible to sorites reasoning. If you start with a collection of sand that is clearly a heap and remove one grain of sand, you still have something that is clearly a heap. Apply this premise repeatedly, and we eventually conclude that there are no unclear cases. Alternatively, one could insist that there is a definite borderline between the clear cases and the unclear cases - but this is just as problematic as asserting that there is a definite borderline between the cases where the original vague term applies and cases where it does not. Just as there is no specific minimum number of grains of sand that constitutes a heap, there is also no specific minimum number of grains of sand that constitutes a clear heap. The transition from clear to unclear is itself indeterminate. See Walton, supra note 3, at 50-51.

n50. See Walton, supra note 3, at 37-68.

n51. If the cases arise in a mixed way, that is, some near one and others near zero, then the decisive factor may be the location of the burden of proof or standard of persuasion. For more on how burdens of proof or standards of persuasion can impede slippery slopes, see infra Part VII.

n52. For a brief discussion of the nontransitivity of similarity in a related context, compare Sainsbury, supra note 48, at 329.

n53. This problem is not created by a shifting definition of similarity, as in the example of saying a banana is similar to a cigar, because the same similarity relation is used in every step here.

n54. See Rescher, supra note 13, at 15-20.

n55. The resister of a slippery slope argument "should demand that the argument be looked at in a holistic way, and point out that, because of the vagueness of the key term, it is arbitrary to fasten on any particular point in the reapplication sequence." Walton, supra note 3, at 59. There is, however, the important economic question of whether any particular decisionmaker has an incentive to take account of the cumulative effect.

n56. The judgment of "closeness" or "similarity" depends on a theory, including its empirical presuppositions.

n57. For a critical analysis showing that this connection was being made in the late nineteenth century, see Herbert Spencer, The Coming Slavery, in The Man Versus the State 44 (Liberty Classics 1981) (1884).

n58. A.V. Dicey makes a closely related analysis of the growth of government during the latter part of the nineteenth century. See A.V. Dicey, Lectures on the Relation Between Law and Public Opinion in England During the Nineteenth Century 303 (1981). The utilitarian argument for laissez-faire became a utilitarian argument for state intervention because of certain doctrinal and institutional reforms wrought by the Benthamite liberals. They fought to overthrow the doctrine of natural rights, promote the idea of Parliamentary sovereignty, and improve the efficiency of government administration. All of these lowered the cost of state regulation when utility might be (or appear to be) directly enhanced. This, in turn, made the acceptance of such arguments more likely. "The effect actually produced by a system of thought does not depend on the intention of its originators; ideas which have once obtained general acceptance work out their own logical result under the control mainly of events." Id. at 310.

n59. Even if some individuals happen to foresee consequences, they will not have an incentive to act on this foreknowledge if they cannot affect the consequences. But the inability to affect outcomes is a good reason that a rational individual will not even try to anticipate consequences in the first place (unless he is an academician!).

n60. See, e.g., Frederic Bastiat, What Is Seen and What Is Not Seen, in Selected Essays on Political Economy (George B. de Hurszar ed., Seymour Cain trans., 1964).

n61. For a more extensive discussion of "cost-lowering" slippery slope mechanisms, see Volokh, supra note 5, at 1043-47. Note that the change in incentives does not ensure that a particular chain of events will occur, but makes it more likely than otherwise.

n62. In other words, given the initial intervention, the cost of the second is perceived to be lower, perhaps so much lower that it creates a perceived net benefit. Thus, the argument urging acceptance of the second intervention is more likely to be persuasive.

n63. A contemporary illustration of this phenomenon is the vast network of price controls in Zimbabwe under the rule of Robert Mugabe. As each control was imposed it created conditions that the government interpreted as suggesting the desirability of further controls. See Economic Focus: The Zimbabwean Model, Economist, Nov. 30, 2002, at 68.

n64. We are not, of course, claiming that the actual preferences of voters in the United States or any other country are like those described in this example. The example is purely hypothetical.

n65. However, the two slope processes may be combined. For instance, members of Group B, who favor restrictive immigration once generous welfare is in place, may take that position because they believe generous welfare will encourage a larger than normal amount of immigration, leading to an expanded tax burden.

n66. For a discussion of Condorcet's paradox or the "paradox of voting," see, for example, P.A. McNutt, The Economics of Public Choice 61-66 (1996).

n67. In this case, a single-peaked preference means that all three groups are in agreement that a particular alternative is the worst. A multi-peaked preference means that, from the point of view of the individual groups, there are at least two, possibly three, least preferred alternatives. See Michael J.G. Cain, Social Choice Theory, in The Elgar Companion to Public Choice 107 (William F. Shughart II & Laura Razzolini eds., 2001).

n68. Nontransitivity means that if the voting process generates the outcomes "A is preferred to B" and "B is preferred to C," it could, nonetheless, generate the outcome "C is preferred to A," without any change in the utilities of the voters. The latter outcome would be inconsistent with the logical (transitive) implication of the first two because A "should be" preferred to C. For further discussion of transitivity in this context, see Dennis C. Mueller, Public Choice III 586-88 (2003).

n69. Volokh, supra, note 5, at 1048-75.

n70. See Mueller, supra note 68, at 38-49.

n71. For a full discussion of the fetal personhood issue, see Aaron Wagner, Comment, Texas Two-Step: Serving up Fetal Rights by Side-Stepping Roe v. Wade Has Set the Table for Another Showdown on Fetal Personhood in Texas and Beyond, 32 Tex. Tech. L. Rev. 1085 (2001).

n72. 410 U.S. 113 (1973).

n73. Wagner, supra note 71, at 1103.

n74. In Michigan, however, the court of appeals argued that fetal personhood is "not pertinent" to the question of whether a pregnant mother has the right to use deadly force to repel an assault on her "nonviable" (outside the womb) seventeen-week old fetuses. People v. Kurr, 654 N.W.2d 651, 657 (Mich. Ct. App. 2002). It reached this conclusion by interpreting the "defense of others" doctrine in light of Michigan's Fetal Protection Act. Id. at 654 (interpreting Mich. Comp. Laws 750.90a (2003)). The court argued that it is reasonable to extend the meaning of "others" to include a fetus or embryo because the state had enacted protections for the fetus against assault or gross negligence. Id. at 653-54. Thus a mother has the right to use deadly force to protect her fetus, only while still in the womb, whether viable or not, against assault even when her own life is in no danger. Id. at 655. "Any other result would be anomalous given the express policy of this state as declared by the Legislature in the fetal protection act." Id. at 657. The more important anomaly, however, would seem to be with Roe v. Wade. This is because the state is offering a level of protection from assault to unviable fetuses that is, in principle, indistinguishable from that provided to persons under the law. Consider that the pregnant woman is able to use (1) deadly force (2) even when her own life is in no danger to protect her fetus deemed "other" for the purposes of the defense of others doctrine. (One is tempted to ask: What does "other" or "another" mean to the court - another what? Since the defender in this case, the mother, is a person, it seems reasonable to conclude that the "other" is a person as well.) This is in sharp contrast to the settled law that anti-abortionists cannot use even less than deadly force to protect fetuses when they are endangered by their mothers. There may be other, doctrinally more consistent, ways to protect the right of abortion while granting some protection to the fetus against assaults by anyone besides the mother. But the Michigan Court of Appeals chose to provide protection to the fetus to the same degree as a person and chose to consider it within the scope of a doctrine plausibly (and heretofore) applicable only to persons. This belies the court's statement that whether the fetus is a person is "not pertinent" to its decision. It decided the case as if an unviable fetus were a person. Thus, the legal tensions discussed in the text remain.

n75. Systemic consistency is the notion that "rule A, should be adopted in preference to a competing rule, rule B, because neither applicable social propositions [for example, moral norms or policy goals] nor any deep doctrinal distinction would justify adopting rule B while adhering to some other previously announced rule." Eisenberg, supra note 30, at 93 (emphasis added).

n76. Consider a series of cases: A ... m, n ... B where A is innocuous and B is danger. Now suppose:

Different judges ... hold different theories about the correct ground for the distinction [between A and B]. But they have to accept each other's decisions as part of the law. This may be illustrated as follows. Judge X may think that n and B are similar and that the line should be drawn between m and n, while Judge Y thinks that the line should be drawn between n and B. If Judge Y upon this basis has accepted n, then Judge X, respecting the precedent created by Y, will make the further step toward the acceptance of B. Though neither Judge X nor Judge Y would have made the step from A to B directly, their combined activity leads to an acceptance of B.
Wibren van der Burg, The Slippery Slope Argument, 102 Ethics 42, 50 (1991).

n77. On benevolence as a natural virtue, see David Hume, A Treatise on Human Nature 369-70 (David Fate Norton & Mary J. Norton eds., 2000) (1740); on justice as artificial, see id. at 307-11, 319.

n78. For Hume, substantive justice consists, most fundamentally, of the rules that function to preserve existing property rights in a "general, inflexible" manner. See David Hume, An Enquiry Concerning the Principles of Morals 171 (Tom L. Beauchamp ed., Oxford University Press 1998) (1751). But since "possession and property should always be stable, except where the proprietor agrees to bestow them on some other person," rules regarding the transference of property by consent (contract law) are implied. See Hume, supra note 77, at 330.

n79. Hume, supra note 77, at 319.

n80. See id. at 368-71.

n81. See id. at 370-71.

n82. Cf. 1 Herbert Spencer, Principles of Ethics 156 (Liberty Classics 1978) (1897). Spencer argues:

The motive causing a generous act has reference to effects of a more concrete, special, and proximate kind, than has the motive to do justice; which, beyond the proximate effects, usually themselves less concrete than those that generosity contemplates, includes a consciousness of the distant, involved, diffused effects of maintaining equitable relations.

n83. Hume, supra note 78, at 171 (footnotes omitted).

n84. For an analysis of Hume on justice and benevolence, see, for example, James Baillie, Hume on Morality 153-59 (2000). For Hume's claim that neither public nor private benevolence can be the foundation of justice, see Hume, supra note 77, at 309-11.

n85. A number of analysts have modeled judges in this way. See William M. Landes & Richard A. Posner, Legal Precedent: A Theoretical and Empirical Analysis, 19 J.L. & Econ. 249, 273-74 (1976); Thomas J. Miceli & Metin M. Cosgel, Reputation and Judicial Decision-Making, 23 J. Econ. Behav. & Org. 31 (1994); Georg von Wangenheim, The Evolution of Judge-Made Law, 13 Int'l Rev. L. & Econ. 381 (1993); Douglas Glen Whitman, Evolution of the Common Law and the Emergence of Compromise, 29 J. Legal Stud. 753 (2000).

n86. Strictly speaking, the individual is confronted with a situation in which the costs of adhering to an argument such as "lifestyle choice should be unrestricted because it is so important to individual identity" have increased, perhaps greatly so. If he now rejects this argument, his actions will change.

n87. In effect, we are appealing to the internal or psychological costs of accepting an argument. If the human brain has any desire for consistency, it will be psychologically costly for the individual to resist a similar or entailed argument given previous arguments.

n88. Schauer, supra note 7, at 373-76.

n89. This reply is most appropriate, of course, for macroprocesses. But it can also be true of a microprocess involving a sequence of individuals with identical theories.

n90. By "myopic behavior" we mean the phenomenon of excessively discounting future costs relative to the preferences expressed by the agent prior to the decision that constitutes the first step on the slippery slope. For example, an individual may believe and accept an SSA that claims that taking decision A will significantly increase the probability of the danger case D. Further, he may accept the argument that, all things considered, the costs of D will exceed in present value the benefits of A (and any other intermediate steps). Nevertheless, when it comes time to decide A or not-A, the individual is "myopic" and chooses A. For an examination of this apparent "preference reversal," see the literature on "hyperbolic discounting," especially Jon Elster, Ulysses Unbound: Studies in Rationality, Precommitment, and Constraints 29-34 (2000), and David I. Laibson et al., Self-Control and Savings for Retirement, in Brookings Papers on Economic Activity 91, 92-100 (William L. Brainard & George L. Perry eds., 1998). For a compact survey of the empirical evidence, see Shane Frederick et al., Time Discounting and Time Preference: A Critical Review, 40 J. Econ. Lit. 351, 360-63 (2002). Not all economists, however, believe that the assumption of inconsistent intertemporal preferences is useful or warranted by the more basic assumption of rationality. See, e.g., Gary S. Becker & Casey B. Mulligan, The Endogenous Determination of Time Preferences, 112 Q.J. Econ. 729, 736-37 (1997).

n91. In other words, simply because there is an undesirable change in context does not imply that there has been a change in values.

n92. There is an alternative way to model our future selves that makes no reference to meta-values. A single individual decisionmaker can be modeled to have a change in values. In this analysis, rationality obliges him to be unbiased between present and future values (or present and future selves). He thus simply discounts future utility by a rate reflecting its uncertainty. Therefore, the decisionmaker's actions will strongly favor the present only when future utility has a large uncertainty discount. See Richard A. Posner, Rational Choice, Behavioral Economics and the Law, 50 Stan. L. Rev. 1551, 1568 (1998) ("What is true is that any personal discount rate higher than necessary to adjust for the risk of death is suspect from the narrowest rational-choice standpoint, as it implies an arbitrary preference for present over future consumption.") (emphasis added). But see Rubenfeld, supra note 11, at 118-19.

n93. See generally Jon Elster, Ulysses and the Sirens: Studies in Rationality and Irrationality (1979).

n94. But cf. Eisenberg, supra note 30, at 76 ("The legal standing of every rule announced in a binding precedent depends not simply on the fact that it was announced, but on whether the rule is congruent with [current] applicable social propositions, considered either explicitly or tacitly."). Applicable social propositions include current widely shared moral norms. To the extent that Eisenberg is correct in his characterization of the common law process, present courts will be less able to bind future courts. Thus initial decisions perceived as likely to lead to undesirable results later may not be taken in the first place.

n95. James M. Buchanan, Cost and Choice: An Inquiry in Economic Theory 42-44 (1969).

n96. In addition to finding cases similar according to inconsistent theories, courts may also distinguish cases (that is, make exceptions) according to inconsistent theories.

n97. Joseph Raz recognizes the possibility of conflicting analogies in the law. His emphasis is on a single court choosing sides, as it were, in a conflict over policy goals, for example. He does not seem to appreciate that the existence of incompatible analogies can produce a chain of similarities (or differences) across different courts or in the same court at different times. See Joseph Raz, Law and Value in Adjudication, in The Authority of Law: Essays on Law and Morality 180, 205 (1979).

n98. Some legal scholars think that this is always the case. "According to the moderate thesis [of American realism] ... in reality, judges at every level are able to select or disregard precedent to suit the conclusion already arrived at." Mark Tebbit, Philosophy of Law: An Introduction 31 (2001). More radically, the Critical Legal Studies movement believes "All rules will contain within them, deeply embedded, structural premises that clearly enable decision makers to resolve particular controversies in opposite ways." Mark Kelman, A Guide to Critical Legal Studies 258 (1987).

n99. Bryner provides an example:

Some laws provide competing objectives that give administrators broad latitude. Under the Emergency Petroleum Allocation Act [15 U.S.C. Sec. 753(b)(1)], for example, regulations were to be issued for the allocation of petroleum products that "protected the public health, maintained public services and agricultural operations, preserved a sound and competitive petroleum industry, allocated crude oil to refiners to permit them to operate at full capacity, resulted in an equitable distribution of supplies to all parts of the country, promoted economic efficiency, and minimized economic distortion."
Gary C. Bryner, Bureaucratic Discretion: Law and Policy in Federal Regulatory Agencies 7 (1987).

n100. However, there are, within the common law, resources to resist the proliferation of theories. For example, Raz believes:

[A] modified rule can usually be justified only by reasoning very similar to that justifying the original rule. Not only will its justification show the reason for applying the ruling to a subclass of the cases to which it was originally applicable, it will also show the relevance of all the operative conditions set out by the original rule.
Raz, supra note 97, at 187-88. This point is by no means uncontroversial. See, e.g., Eisenberg, supra note 30, at 52. To the extent that Raz is incorrect, multiple theories will be more widespread and the "problem" noted in the text more severe.

n101. One of the vaguest (and hence one of the most elastic) terms in ethical and legal discussions is "addiction." This is not simply a consequence of the way the word is used but also of the lack of clarity in the underlying theory of "autonomous behavior" that addiction is supposed to overwhelm:

The mere fact that a person is physiologically dependent and uses a drug to relieve withdrawal symptoms does not entail that his use of that drug is nonautonomous. In addition, the pain of withdrawal must be sufficiently severe so that it is unreasonable to expect him to endure it. Only then might it be said that an addict is powerless to quit ...

Social conventions create a vague and imprecise threshold of pain or discomfort that adults should be able to withstand for the sake of avoiding an evil.

Douglas N. Husak, Drugs and Rights 108-09 (1992). Even more troubling is the inconsistency in the application of the threshold. On the one hand, heroin use is considered sufficiently painful to stop that it is labeled "addictive." On the other hand, the level of pain experienced by heroin users who do not get their fix is not sufficient to excuse them from even minor crimes committed to sustain their addiction. Id. at 113. To see the possibilities for a large expansion in the use of the term "addictive," consider the following news report:
Caroline Goddard, director of the Obesity Treatment Center Medical Group in Sacramento, is among those who contend there is an addictive quality to foods high in fat and sugar because of their ability to stimulate pleasure centers in the brain ... .

Among her patients is Virginia Lee, who has struggled for years to stay under 300 pounds. Lee says she feels incapable of ordering a cheeseburger without the fries, and that when she gives in and goes to a McDonald's, she finds herself compelled to go back the next day. And the next.

Will Evans, Vets of Tobacco Wars Take Aim at Fast Food: Lawsuits Blaming Restaurants for Obesity Appear to Represent a New Trend, Sacramento Bee, Feb. 24, 2003, at A1, available at

n102. It is relatively well defined, but not perfectly. The Scitovsky objection is a well-known source of indeterminacy in the Kaldor-Hicks (wealth maximization) approach. See Mark Blaug, Economic Theory in Retrospect 589-90 (4th ed. 1985). In the context of law and economics, see, for example, Mario J. Rizzo, The Mirage of Efficiency, 8 Hofstra L. Rev. 641, 649-51 (1980).

n103. We abstract from problems relating to the activity levels of plaintiffs and defendants. See Steven Shavell, Economic Analysis of Accident Law 21-32, 41-46 (1987).

n104. See generally Gillian K. Hadfield, Bias in the Evolution of Legal Rules, 80 Geo. L.J. 583 (1992).

n105. The clash of research programs or paradigms characteristic of great transitional phases in the law accentuates the proliferation of incompatible theories. In the "Progressive" and post-World War I period the first wave of Legal Realists sought to balance and partially integrate the competing research programs of formalist liberalism and pragmatic welfarism. No coherent synthesis was achieved. See David Ingram, The Sirens of Pragmatism Versus the Priests of Proceduralism: Habermas and American Legal Realism, in Habermas and Pragmatism 83-98 (Mitchell Aboulafia et al. eds., 2002). The Realists reconstructed "judicial reasoning as an impartial process of reconciling or balancing different perspectives, values and interests through open and public "conversations' with scientific experts, affected parties and the broader community." Id. at 98. In this context tentative theories abounded and the myth of deducing the one appropriate rule for a situation took its final blow.

n106. For example, the decisionmaker could accept that allowing physician-assisted suicide (PAS) for terminal illnesses would likely lead to PAS in severe but nonterminal cases like "Lou Gehrig's disease." But the latter, while undesirable, might not be so bad as to make the disadvantages of prohibiting PAS in terminal cases worth bearing.

n107. It is quite interesting to note that the U.S. Supreme Court in Atkins v. Virginia, 536 U.S. 304 (2002), did not take the route of instructing states to follow a clear IQ rule. In its holding that the execution of "retarded" persons violates the Eighth Amendment to the U.S. Constitution, the Court did not specify a single (or even multiple) sharp criterion to distinguish the "retarded" from the "normal." The Court seems to quote with approval various psychiatric standards. Nevertheless, these are quite vague, both theoretically and empirically. See id. at 309 n.3. From a theoretical perspective: "Clinical definitions of mental retardation require not only subaverage intellectual functioning, but also significant limitations in adaptive skills such as communication, self-care, and self-direction that became manifest before age 18." Id. at 318. From an empirical perspective:

To the extent there is serious disagreement about the execution of mentally retarded offenders, it is in determining which offenders are in fact retarded... . Not all people who claim to be mentally retarded will be so impaired as to fall within the range of mentally retarded offenders about which there is a national consensus.
Id. at 317-18. Furthermore, the underlying theory about why mental retardation is relevant in a criminal context is also vague.
Mentally retarded persons frequently know the difference between right and wrong and are competent to stand trial. Because of their impairments, however, by definition they have diminished capacities to understand and process information, to communicate, to abstract from mistakes and learn from experience, to engage in logical reasoning, to control impulses, and to understand the reactions of others.
Id. at 318. As a result of this three-fold vagueness, any court attempting to apply legislation that embodies these criteria or standards will face slippery slope problems emanating from analogies with previously decided cases. The class of retarded may narrow or widen depending on the predilections of judges or other decisionmakers (for example, on their attitudes toward capital punishment in general). Furthermore, the capital punishment limitations for the retarded may extend to limitations on other forms of punishment as well if the rationale of the decision is generalized: Retarded persons' "deficiencies do not warrant an exemption from criminal sanctions, but they do diminish their personal culpability." Id. at 318. Limitation of personal responsibility, by the logic of this theory, cannot be limited to acts of murder.

n108. See, for example, Endicott's argument:

Higher-order vagueness is a threat because the [underlying] theory needs a notion of "admissible" sharpenings [bright line distinctions or rules]. The meaning of "tall" does not allow you to sharpen it so that no one less than nine feet tall is tall. So clearly tall people must be those who are tall on all admissible sharpenings. But "admissible" seems to be vague, just as "clearly tall" is vague ... . We could [then] formulate a new form of the sorites paradox ... .
Endicott, supra note 49, at 80.

n109. H.L.A. Hart, Positivism and the Separation of Law and Morals, 71 Harv. L. Rev. 593, 606-07 (1958).

n110. 413 U.S. 15, 24 (1973).

n111. We leave it to the reader to decide whether that has indeed happened, and if so, whether the slippage was desirable or not.

n112. This seems to be the implication of Schauer's characterization of Harry Kalven's First Amendment views: "He applauds over-protection of speech as the only alternative to under-protection." Frederick Schauer, Harry Kalven and the Perils of Particularism, 56 U. Chi. L. Rev. 397, 407 (1989) (reviewing Harry Kalven, A Worthy Tradition: Freedom of Speech in America (1988)).

n113. Id. at 403-04.

n114. See, e.g., John Stuart Mill, On Liberty 65 (Stefan Collini ed., Cambridge Univ. Press 1989) (1859). Mill stresses the difficulty the general public will have in estimating the social value of original ideas. "Originality is the one thing which unoriginal minds cannot feel the use of. They cannot see what it is to do for them: how should they? If they could see what it would do for them, it would not be originality." Id.; see also id. at 67 ("In other times there was no advantage in [exceptional individuals acting differently from the masses], unless they acted not only differently but better. In this age, the mere example of nonconformity, the mere refusal to bend the knee to custom, is itself a service."). Karl Popper observes that even wrong ideas will have social value. See, e.g., Karl R. Popper, On the Theory of the Objective Mind, in Objective Knowledge: An Evolutionary Approach 153 (rev. ed. 1981). "Avoiding error is a poor ideal: if we do not dare to tackle problems which are so difficult that error is almost unavoidable, there will be no growth of knowledge. In fact, it is from our boldest theories, including those then there which are erroneous, that we learn most." Id. at 186.

n115. Bernard Williams distinguishes between "reasonable" and "effective" stopping points. See Bernard Williams, Which Slopes Are Slippery?, in Making Sense of Humanity and Other Philosophical Papers 214-15 (1995). Not everything reasonable is effective and not everything effective is reasonable. For an example of a reasonable but possibly ineffective stopping rule, suppose the issue of euthanasia is tentatively resolved by a distinction between acts of the patient and those of the doctor. For many purposes this distinction will be clear enough to prevent a slide from suicide to murder. But this reasonable stopping point might become ineffective when it is understood that some patients lack the physical ability to effect their own decision. Is physician assistance, under these circumstances, simply an aid to suicide or is it murder? For an example of a rule that may be effective but unreasonable, the law might impose a fourteen-day cut off point for fetal experimentation. "Fourteen days" is a quite clear but largely arbitrary stopping point. Is there much difference between a fourteen-and fifteen-day fetus in terms of morally relevant human characteristics? See also John D. Arras, Slippery Slope Arguments, in Encyclopedia of Ethics 1594, 1594-95 (Lawrence C. Becker & Charlotte B. Becker eds., 2d ed. 2001). In terms of our framework, however, the problem is not a reasonable stopping point versus an effective one. The real difficulty is that a distinction at one level of analysis may be sharp but at a more theoretical level, it becomes vague. In the first case, the standard that people ought to be able to effect their desires regarding their own life and death is only implicit in the simple cases. Reference to it in more complex cases clouds a previously sharp distinction based on the external observation of acts. In the second case, the standard that fetuses that are also "persons" should not be experimented upon is only implicit as long as we do not question the fourteen-day rule. A sharp stopping point becomes vague by reference to the underlying standard.

n116. The concept of duress has become increasingly vague over the past three centuries:

In Blackstone's time relief from an agreement on grounds of duress was a possibility only if it was coerced by actual (not threatened) imprisonment or fear of loss of life or limb... . Today the general rule is that any wrongful act or threat which overcomes the free will of a party constitutes duress. This simple statement of the law conceals a number of questions, particularly as to the meaning of "free will" and "wrongful."
John D. Calamari & Joseph M. Perillo, The Law of Contracts 261-62 (2d ed. 1977) (citation omitted).

n117. One could include under "physical intimidation" only violence, threats of violence, imprisonment, and threats of imprisonment. This, absent the threat of imprisonment, is more or less the rule Blackstone believed was in effect in the seventeenth century. See id. at 337.

n118. Of course, there is some residual vagueness even in this rule, since the meaning of "physical intimidation" has fuzzy boundaries. Even when a relatively restricted Blackstonian conception is involved, terms such as "violence" and "threat of violence" are vague at the boundaries.

n119. Daniel A. Dombrowski & Robert Deltete, A Brief, Liberal, Catholic Defense of Abortion 11-12 (2000).

n120. In the courts of chancery prior to the seventeenth century equity decisions were made in personam. See F.W. Maitland, Equity: A Course of Lectures 8 (1949). In these circumstances, an exception would not establish a precedent for further exceptions.

n121. Compare, for example, Hayek:

In certain conditions, namely when some basic principles of the law have been accepted for some time, they will indeed govern the whole system of law, its general spirit as well as every single rule and application within it. At such times it will possess great inherent stability. Every lawyer will, when he has to interpret or apply a rule which is not in accord with the rest of the system, endeavour so to bend it as to make it conform with the others ... .

The situation is entirely different, however, when a general philosophy of the law which is not in accord with the greater part of the existing law has recently gained ascendancy. The same lawyers will, through the same habits and techniques, and generally unwittingly, become a revolutionary force, as effective in transforming the law down to every detail as they were before in preserving it.

F.A. Hayek, Law, Legislation and Liberty, Rules and Order 66 (1973) (emphasis added).

n122. We do not wish to imply that this is the only function of precedent.

n123. The distinction between a rule and its exceptions is troublesome. The difficulty of maintaining the distinction gives exception-making its precedential value. Consider the argument made by Paul Ramsey:

The effort to locate a justifiable exception can only have the effect of utterly destroying its exceptional character. The deed is found to be morally doable, it is repeatable, it is one of a kind. How rare or frequent is of no consequence to the moral verdicts we render. The same justifying features, the same verdict, the same general judgment falls upon the alleged exception, if it is justified; and so that act falls within our deepened or broadened moral principles.
Paul Ramsey, The Case of the Curious Exception, in Norm and Content in Christian Ethics 67, 78 (Gene K. Outka & Paul Ramsey eds., 1968).

n124. Schauer, supra note 112, at 403-04.

n125. "Nazis became political speakers, a suburban community populated by Holocaust survivors became a public forum, and popularly inspired restrictions became government censorship." Id. at 408.

n126. Note we say "inhibits," not "completely prohibits."

n127. "A principle of free speech, according to which the mode of analysis shifts when an occurrence can be categorized as "speech,' is incompatible with a principle of maximally contextual evaluation of all aspects of situations in which speech is present." Schauer, supra note 112, at 397-98.

n128. Obviously this is a simplification. Courts would mostly likely decline to use only an IQ score to make such a determination. See elaborations on the concept of mental retardation cited by the Supreme Court in Atkins v. Virginia, 536 U.S. 304, 309, at nn.3, 5 (2002).

n129. Christopher B. Mueller & Laird C. Kirkpatrick, Evidence 3.8, at 134-35 (2d ed. 1999).

n130. See, e.g., Cal. Fam. Code 7120 (West 1994). The conditions for emancipation in California include being at least fourteen years of age, living separately from a parent or guardian, managing one's own financial affairs, and not having an illegal source of income. Id.

n131. See, e.g., Turner Broad. Sys., Inc. v. FCC, 512 U.S. 622, 642 (1994).

n132. James Buchanan has emphasized choice among rules, as opposed to choice within rules, as the essence of constitutional-level thinking. See, e.g., James M. Buchanan, The Limits of Liberty: Between Anarchy and Leviathan (1975); James M. Buchanan, The Constitution of Economic Policy, 77 Am. Econ. Rev. 243-50 (1987).

n133. This is a position most often associated with Justice Brennan and the Warren Court. See Bruce Ackerman, A Generation of Betrayal? 65 Fordham L. Rev. 1519 (1997); Michael Les Benedict, Constitutional History and Constitutional Theory: Reflections on Ackerman, Reconstruction, and the Transformation of the American Constitution, 108 Yale L.J. 2011 (1999); Jeffrey Goldsworthy, Dworkin as an Originalist, 17 Const. Comment. 49 (2000). Our intention here is not to enter the debate over constitutional interpretation on the side of originalism, but rather to observe how the inevitable vagueness of constitutional terms creates a fuzzy boundary between the choice of rules and the application of rules.