We all know that much in our thought and language, as well as much in the law, is vague. We are also reasonably good at recognizing cases of vagueness, even though most of us would be hard pressed to say exactly what vagueness is. In recent decades, there has been a flowering of work in the philosophy of logic and language attempting to do just that. Much of this work focuses on what it is for a word or phrase to be vague. The aim of this effort is to clarify what it is for a claim, question, command or promise expressed using such a term to be vague, as well as what it is to reason with such terms. Different logico-linguistic theories have different conceptions of the scope of putative laws of classical logic, including bivalence (which states that every declarative sentence or proposition is either true or false) and excluded middle (which asserts all instances of “A or ~A”). In addition to this work in philosophical logic, recent decades have seen a growing interest in vagueness among legal scholars and philosophers of law. Here the focus is not so much on what legal vagueness is, which is generally assumed to be readily recognizable. Rather, it is on the extent and sources of vagueness in the law, the implications of vagueness for interpretation and adjudication, the systemic effects of vagueness and the function—i.e., important positive value—of vagueness in certain areas of the law, as opposed to its disutility in others (Endicott 2000, 2005; Soames 2011).
To date, these two investigations of vagueness—in philosophical logic and the philosophy of law—have been largely independent of one another. This independence gives rise to a natural line of questioning. Can work in one domain contribute to work in the other? Does a commitment to one philosophical theory of what vagueness is carry with it lessons for vagueness in the law? If so, might the need to make good sense of legal vagueness play a role in deciding which philosophical theory of vagueness is correct? Conversely, might one be misled about the pros and cons of vagueness in the law by a faulty conception of what vagueness is? These are the questions to be investigated here. This will be done by comparing two leading philosophical accounts of vagueness and exploring their implications for understanding the value of vagueness in the law and the issues at stake in interpreting vague legal texts.
Vagueness and Borderline Cases
In ordinary life, a remark is often considered vague if the information it provides is insufficiently specific to advance the accepted conversational purpose (especially when the speaker is expected to possess that information). Philosophical logicians have focused on one particularly interesting sub-case—involving the notion of a borderline case—of this more general phenomenon. Vague predicates—like “old,” “bald,” “rich” and “red”—are those for which a range of borderline cases separate things to which the predicate clearly applies from those to which it clearly does not apply. These are cases in which there is no clear answer to the question of whether a predicate is, or is not, true of an object. In such cases, we are pulled in both directions—being inclined to resist definitive verdicts in favor of equivocal remarks like “It sort of is and sort of isn’t,” “It’s not clearly one or the other” or “Call it what you like, but neither choice is definitely correct.” In some contexts it may be acceptable to treat the predicate as applying, while in others it may be fine to treat it as not applying. But no investigation into the facts in virtue of which the predicate means what it does could ever identify one of these uses as definitely correct and the other as incorrect. In situations that call for a verdict, this means that a decision is required that is not dictated by knowledge of the nonlinguistic facts under discussion plus the linguistic rules governing vague terms. This is the sense in which “philosophical vagueness”—susceptibility to borderline cases—is an instance of ordinary vagueness: insufficient informativeness (Soames 2011). If a theater director’s assistant has been told to cast a bald character and the chief candidate for the role is a borderline case of baldness, the director’s request plus the assistant’s knowledge of the candidate are insufficiently informative to determine whether or not to offer the role. Some further basis for decision is needed.
Two Philosophical Theories of Vagueness
According to one theory, vague predicates are both partially defined and context sensitive. To say that P is partially defined is to say that it is governed by linguistic rules that provide sufficient conditions for P to apply to an object and sufficient conditions for P not to apply, but no conditions that are both individually sufficient and disjunctively necessary for P to apply or not to apply. Because the conditions are mutually exclusive, but not exhaustive, there are objects not covered by the rules for which P is undefined. In the case of vagueness, this, in turn, gives rise to context sensitivity. Since the rules of the common language, plus all relevant nonlinguistic facts, don’t determine P-verdicts for every object, speakers using P in certain contexts have the discretion of extending its range to include some initially undefined cases, depending on their conversational purposes. Often they do so by predicating P of an object o, or denying such a predication. When they do, and other conversational participants accommodate their conversational move, the class of things to which P does, or doesn’t, apply is contextually adjusted to include o, plus objects similar to o (in certain respects). In such cases, P is (partly) “precisified” by narrowing the range of items for which P is undefined (Tappendon 1993; Soames 1999, ch. 7; Endicott 2000; Shapiro 2006).
Since what counts as a rule of the language (governing the use of a particular predicate) is also vague, higher-order vagueness arises when one considers the predicate ┌is determinately P┐, where for o to be determinately so-and-so is for the claim that o is so-and-so to be a necessary consequence of the rules of the language governing “so-and-so” plus the (relevant) nonlinguistic facts about o. Because of this, the range of application for an ordinary vague predicate P can be divided into five regions as follows:
Let “red” be P. Items in R1PDP are determinately red, items in R3PDP–R5PDP are not determinately red and it is unsettled whether items in RPDP2 are determinately red or undefined for “red.” Similar characterizations hold for “not red,” working from R5PDP and moving left. Iterating “determinately” doesn’t change things (Soames 2003).
Next consider the proposition p expressed by “It’s red” relative to an assignment of o as referent of “it” and a context C including a set of standards governing “red.” We are not here considering the proposition asserted by an agent who utters “It’s red” in C, referring to o. The issue is semantic (the proposition semantically expressed relative to a context and an assignment), not pragmatic (the proposition asserted by an utterance). If, given C’s standards for “red,” o is in R1PDP (R5PDP), then p is true (not true) in C; if o is in R3PDP, p is undefined for truth in C. (“False” and “not true” are interchangeable when applied to propositions.) If o is in R2PDP, it is unsettled whether p is true or undefined in C; if o is in R4PDP, it is unsettled whether p is not true or undefined. When a proposition p is not true, it is a mistake to assert p, but it may be correct to deny p—i.e., to assert its negation. However, when p is undefined for truth, it is a mistake to either assert or deny p because neither p nor its negation can be known to be true (Soames 2010). When it is unsettled whether p is true or undefined it is unsettled whether one who accepts it has made a mistake.
Now consider the related case in which an agent A says “It’s red” of o in a context in which the standards governing “red” prior to A’s utterance place o in regions 2 or 3, but the audience accommodates A by adjusting the contextual standards to render A’s remark true. In such a case the proposition q that A uses “It’s red” to assert is different from the proposition p that the sentence semantically expresses, relative to the context and prior to accommodation (plus an assignment of o as referent of “it”). After accommodation, the partially defined property contributed by “red” to the asserted proposition has o in its region 1. If o was in region 3 originally, A’s remark will be true by stipulation, in the sense that it is only because A’s sentence has been taken to assert q, rather than p, that A’s remark counts as true. By contrast, if o had been in region 2 by previous standards, A’s remark will again be judged true, but this time it will be unsettled whether it is true by stipulation, because it will be unsettled whether the proposition p that A’s utterance would have asserted without accommodation is itself true. These instances of smooth accommodation contrast with an attempt to extend the extension of “red” to an item x in region 4 prior to A’s remark. In such a case, A’s remark will be problematic and may not be accommodated, since prior to A’s utterance it was unsettled whether o was undefined for “red” (and so open for inclusion under the predicate) or definitely not red (and so outside the range of legitimate speaker discretion).
That, in a nutshell, is one philosophical theory of vagueness. Another important theory is the epistemic theory, according to which vague predicates are always totally defined, with sharp boundaries separating items to which they apply from those to which they don’t—e.g., a single second separating moments when one is young from those when one is not, and a single penny separating one who is rich from one who is not. Borderline cases are those of which we can never know the vague predicate P to be true, or to be untrue of a given item. So, whereas the previous theory takes borderline cases to be those for which P is undefined, the epistemic theory takes them to be cases for which one can never know how, in fact, P is defined (Williamson 1994). Here I will be concerned with the standard version of epistemicism, which does not take vague terms to be context sensitive, as opposed to the version in Fara (2000), which does.
According to this theory, bivalence and the law of the excluded middle hold without exception, even for sentences containing vague language. Sorites paradoxes are blocked by denying the major premise of paradoxical arguments like the following:
Minor: A newborn baby is young at the moment of birth.
Major: For every number n, if one who is precisely n seconds old is young, then one who is n + 1 seconds old is also young.
Conclusion: Everyone is young.
Whereas the previous theory of vagueness rejects the major premise while also rejecting its negation (since both are undefined), epistemicism claims the major premise to be false and its negation to be true, which it asserts:
~ Major: There is a number of seconds n such that anyone who is precisely n seconds old is young, but anyone who is n + 1 seconds old is not young.
What epistemicism doesn’t do is identify any number n as the number in question. Unlike still another theory—supervaluationism about vagueness (Fine 1975)—which also preserves the law of the excluded middle, and asserts (~ Major), epistemicism acknowledges every instance of the quantified major premise to be either true or false, despite the fact that some of the truths are unknowable.
Higher-order vagueness arises for the epistemicist when one considers the predicates ┌is an object that can be known to be P┐ and ┌is an object that can be known not to be P┐. When P is vague in the epistemicist’s sense, these predicates are also vague. This means that although both predicates are totally defined, and although there are sharp lines separating things to which they apply from things to which they do not, the precise location of these lines is unknowable. Thus, the range of application of P can be divided into four regions as follows:
|P & so knowable||P but unknowable||Not P but not so knowable||Not P & so knowable|
Let “red” be P. Items in R1E are red and can be known to be so; those in R2E are also red, but cannot be known to be red. Similarly, items in R4E are not red, and can be known not to be, while those in R3E are not red but cannot be so known. Since the norm of assertion is knowledge (Williamson 1996), this means that to assert of an item x in R2E–R4E that “It is red,” as well as to assert of an item y in RE1–RE3 that “It is not red” is to violate the norms governing our linguistic practices, and so to make a kind of mistake. Of course, some of these mistakes are worse than others since when x is in RE3–RE4, and y is in RE1–RE2, what one asserts is also false (in addition to being unknowable). However, all are violations.
This creates a prima facie difficulty. Together, epistemicism plus the view that knowledge is the norm of assertion direct us not to assertively predicate either a vague predicate P, or its negation, of any item in its unknowable range R2E–R3E. In many conversational settings this is unproblematic, since there is often no need to provide definite P-verdicts for particular borderline cases. In some settings, no judgment whatsoever is required, while in others a hedged judgment—e.g., ┌That may be P┐, ┌That is probably P┐, ┌That is unlikely to be P┐—will do. However, if there are situations that do require definite P-verdicts, such hedges will not serve. In these cases, the demand for an unequivocal verdict conflicts with the epistemic theory of vagueness plus the conception of knowledge as the norm of assertion. Since there appear to be legal contexts of this sort, they may provide good test cases for evaluating the dispute between the epistemicism and other theories of vagueness.
Vagueness in the Law
Since vagueness in the law comes in different forms with different consequences, some preliminary distinctions are needed to narrow our focus. Three domains of legal vagueness are particularly important: vagueness in the content of the law, vagueness in the allowable evidence and prescribed procedures used in reaching a legal verdict, and vagueness in the enforcement or effect of the laws. A good example of the latter is the enforcement of the 65 miles per hour speed limit on freeways in southern California. Though the content of the law is precise, the practice of enforcing it includes a range of speeds of roughly 66–70 miles per hour at which whether or not one is stopped is (under normal conditions) a matter of substantial discretion on the part of the highway patrol. The effect is to create a range of borderline cases in which it is vague whether, and to what extent, drivers are in legal jeopardy. This sort of vagueness—which has no effect on the content of the law—is valuable and necessary both to allow law-abiding citizens a reasonable margin for error in their attempts to comply with the law, and to allocate the resources of law enforcement and the judiciary reasonably and efficiently.