In a work appropriately recognized by many of the contributions to this volume, but still unfortunately neglected in mainstream analytic philosophy, the German philosopher Hans Vaihinger offered the perspective of, in German, als ob, ⁴ which, in its most common English translation, means “as if.” Vaihinger’s philosophy of “as if” was for some years highly influential,⁵ and for our purposes noteworthy for the influence that it had on Hans Kelsen (Bindreiter 2002; Paulson 1992a), and, at more or less the same time, Fuller (1967, pp. 94–137). Vaihinger’s basic idea has both logical and psychological dimensions. With respect to the former, the idea is straightforward: the premise of an argument can typically be presented in hypothetical form, and we are highly familiar with the idea of assuming something for the sake of argument, or with presupposing the truth of one proposition in order that we can make sense of some other proposition. To take the standard example from Bertrand Russell (1905), the statement “The King of France is bald” presupposes that there is a king of France, and we can accordingly make sense of the statement on the assumption that there is a king of France, even though, in fact, there actually is no king of France. Thus, the presupposed factual proposition that there is a king of France is a fiction insofar as it allows the sentence in which it is contained to make sense even if and when the proposition is false. But it is a fiction only contingently, because the logically equivalent statement, “The Queen of England is not bald,” similarly presupposes that there is a Queen of England. But because there actually is a Queen of England, the presupposition in this case, contingently but not necessarily, is not false and consequently not a fiction at all.

Sometimes, however, we assume some fact for the sake of argument, or hypothesize some fact, under circumstances in which it simply does not matter whether the assumed fact is true or false. Yet although the actual fact of the matter is irrelevant, we nevertheless assume the existence of some fact because the assumption makes some other point understandable, or clearer. So consider the following: “Assume that 70 % of some population favors a particular policy. Then, in a democracy, that is the policy that should be adopted.” Here it does not matter whether in this or that instance the 70 % assumption is true or false. Rather, the 70 % majority is an assumption that enables us to understand the idea of democracy, an understanding which is independent of the actual truth or falsity of the factual assumption. In discussing “fictions” as important elements of thought, therefore, Vaihinger was less concerned with the facticity of the “as if” than with the way in which “as if” thinking often facilitated human understanding and comprehension. That Vaihinger was an important Kant scholar and the founder of the journal Kant-Studien underscores the similarities between Vaihinger’s philosophy of “as if” and the Kantian transcendental understandings which are related to but not quite the same as the factual presuppositions in Russell’s example and in the subsequent philosophical literature (Strawson 1952, pp. 174–179).

Vaihinger’s perspective links closely with Kelsen’s Grundnorm.⁶ The Grundnorm under Kelsen’s system is not a factual presupposition as much as it is a Kantian transcendental understanding (Paulson 1992b; Raz 1979), but such a transcendental understanding, or at least the factual components of it, might still turn out to be either factually true or factually false. If factually true—if making sense of a legal system requires that we adopt an understanding that turns out, contingently, to mirror reality, then it would be odd to describe that understanding as a fiction. But if the understanding is not true, or if, more precisely, it simply does not matter whether it is true or not, then we can understand why Kelsen would describe the understanding that is the Grundnorm as a fiction. It is thus the contingent falsity and, even more, the irrelevance of the factual truth or falsity of the Grundnorm that explains why it is commonly thought of as a legal fiction in Kelsenian legal theory.⁷ But although the assumed existence of something that may not exist perhaps justifies (but only barely) the term “fiction,” and although it is the very assumption rather than the factual existence of the truth of the Grundnorm that distinguishes it from Hart’s (derivative) ultimate rule of recognition (Hart 2012),⁸ the use of the term “fiction” in this context seems substantially metaphorical. If we understand a fiction not so much as an assumption (whether a transcendental understanding or not) whose factual truth is not relevant but as a statement that is both literally false and known by the maker of the statement to be literally false—were it otherwise we would call it a “mistake” and not a “fiction”—then the common usage of “legal fiction” to refer to Kelsen’s idea of the Grundnorm is itself at least slightly fictional with reference to the term “fiction” itself.

Both Kelsen and Vaihinger recognized that it is sometimes important to act and speak as if something were true, and that we do so not necessarily only when what we assume to be true is false, but also when we do not want the truth or falsity of the assumption to matter. Vaihinger’s and Kelsen’s ideas were thus not primarily about counterfactual assumptions, but instead about “afactual” assumptions in which the factual truth or falsity of the assumption is largely irrelevant. The connection between Vaihinger and Kelsen thus seems apparent, but what appears as well is the conclusion that to describe the ideas of “as if” or the Grundnorm itself as “fictions” is a usage that, while not plainly wrong, is at least somewhat non-standard. Normally we understand a fiction as being closer to a lie —the intentional statement of a falsity—but lacking the deceptive component of actual lying. Unlike lies, fictions are known by their recipients to be false, and that is why we do not normally think of novelists and screenwriters as liars, even though they are saying things that are literally untrue and known by the novelists and screenwriters to be untrue. Despite the intentional falsity, however, we do not understand playwrights and novelists as liars because the audience is aware of the untruth and thus the element of deception necessary for a proposition to count as a lie is missing. The audience or the readers may be expected to suspend their disbelief in the truth of the fictional narrative while they are reading or listening, and this may be a variant on the kind of transcendental understanding that is the key to the idea of the Grundnorm , but the fictional aspect of the Grundnorm is no more problematic than is fiction in general . The Grundnorm may be a legal fiction, but only in the sense that any assumption or presupposition is potentially fictional, and fictions of this variety are such a large part of our conceptual machinery as scarcely to deserve notice at all, let alone even the mildest of condemnation. More importantly, the use of “fiction” to describe the Grundnorm is sufficiently distant from the traditional way in which “fiction” has been used in “legal fiction” in common law discourse as to suggest that discussions of the fictional aspect of the Grundnorm , while important in themselves, are largely peripheral to the topic of legal fictions itself.

Once we turn from presuppositions to presumptions, the tendency to condemn legal fictions accelerates. Thus it is often said that legal presumptions , or at least many legal presumptions, are legal fictions because such presumptions require judges or juries to accept as true things that are false, and that this is highly undesirable.⁹ Yet although phrasing a presumption as the mandatory acceptance of that which is not true does suggest some degree of falsity or fictionality in the typical presumption, the initial impression—and Bentham’s impression—that a presumption is a fiction, or even that it allows fictions, turns out, as Raymundo Gama explores in greater depth in this volume (Chap. 16), to be more misleading than helpful.

Consider a typical generalization in a non-legal context. “Swiss cheese has holes,” for example. Or “Volvos are reliable.” Such statements are not true of all members of the described class. There is Swiss cheese without holes that is still Swiss cheese, and there are unreliable Volvos. But if that is so, then what is it to make a statement about a class? And what makes a statement about a class true? Or false? Consider, for example, the class of Yugos, a notoriously unreliable automobile manufactured in the former Yugoslavia in the 1970s and 1980s. Assuming (possibly counterfactually) that there were at least two reliable Yugos, it would be true to say that “some Yugos are reliable,” yet the statement “Yugos are reliable” seems false. But if there are some unreliable Volvos and some reliable Yugos, then why is the statement “Volvos are reliable” true and the statement “Yugos are reliable” false? And the answer is that according to the standard pragmatics of standard English, a statement about a class which attributes a property to the class is true if and only if the property is present within the class to a greater extent than the property is present within another class with which the class under discussion is being explicitly or implicitly compared. Note that there is thus no requirement that the property appear within a majority of the members of a class for the statement about that property within the class to be true. It would be accurate to say that “Yugos are unreliable” even if only 45 % of Yugos were unreliable, because 45 % is a rate of unreliability far higher than the rate for cars in general or other models of car. Similarly, it is correct to say that “pit bull dogs are aggressive” even though only a minority of them are, and that is because the rate of aggression for the class of pit bulls is higher than the rate of aggression for most other breeds and higher than the rate for all other dogs. Thus, statements attributing a property to a class—generics (Cheng 2011; Leslie 2008; Schauer 2003) —may be true if the property appears in all members of the class, if the property appears in a majority of members of the class, or if the property appears within the class to a greater degree than the property appears within some implicit comparison class.

The point of the foregoing is to establish that probabilistic statements about a class may indeed be true (or false) without being universally true (or false) for all members of the class. And this analysis can be transferred from descriptive generalizations to the prescriptive generalizations we call rules (Schauer 1991). Prescriptive rules are designed to serve background justifications, as with a “no driving in excess of 100 km/h” rule that is designed to foster the background justification of highway safety. As in this example, rules are also probabilistic generalizations, because rules are generalizations about the forms of conduct that will serve those background justifications. So if we prohibit vehicles in the park because we wish to secure peace and quiet in the park (Hart 1958, 2012, p. 129; ; Schauer 2008) , we have made the assessment that prohibiting vehicles, in general, will advance peace and quiet, again in general and not necessarily in every case. The prohibition is embodied in a rule with the full and necessary acceptance of the fact that the rule can be the right rule even though some vehicles might not be detrimental to peace and quiet (over-inclusion) and even though some non-vehicles might turn out to be detrimental in just the way that vehicles ordinarily are (under-inclusion).

Legal presumptions typically, even if not always or necessarily, operate to precisely the same effect, even if not in exactly the same way. So let us consider the classic historical common law presumption of paternity—the presumption that the husband of the mother is the father of any child born to the mother during the marriage (Glennon 2000). Assuming for the moment—but only for the moment, as we shall see in Section III below—that fatherhood is a physiological state requiring some biological connection between the father and the child and requiring that the father have had sexual relations (or a modern scientific or technological equivalent) with the mother in order to produce the child, then some number of people presumed by the law to be fathers will not in fact be fathers. It is, after all, quite possible to be the biological father of a married woman’s child without being the person to whom the mother was married at the time of conception or the time of birth. But the law, partly because it reflected a traditional social aversion to illegitimacy, and partly to create and enforce obligations of support, presumed that all men married to women at the time the women gave birth were the fathers of the child so born, even though, in fact, only most of them were.¹⁰

The connection between rules and presumptions is now exposed. Most legal rules treat the typical as universal, as when they treat ordinarily peace-and-quiet-hampering vehicles as if they necessarily hampered peace and quiet. And so too with, for example, the ordinary speed limit, which is based on an assessment of the maximum safe speed for average drivers in average vehicles under average road, weather, and traffic conditions, but which is nevertheless the speed limit for all drivers in all vehicles under all conditions. Thus, instrumental legal rules presume that all vehicles will detract from peace and quiet even if only most of them will, and presume that all drivers can drive safely only up to a certain speed, even if in fact some drivers can drive safely at a higher speed. Accordingly, the similarity between the typical rule and the typical presumption is not merely the parallel between the two in treating what is typically or ordinarily the case as if it were always the case. Rather, rules in many cases, and always when rules serve instrumental purposes vis-à-vis their background justifications, just are presumptions, although this feature of rules is sometimes obscured by the way in which presumptions are normally expressed. But the presumption that the husband of the mother of a child born during wedlock could also be expressed in terms of a rule requiring the husband of the mother of a child born during wedlock to support the child, treat the child as legitimate, etc. In that case, the very same consequences flowing from the presumption would be articulated in the form of a rule rather than a presumption, but the legal import of the two would be identical.

We can now see why the common talk of presumptions as fictions is in many important respects the product of confusion. Asking legal decision-makers to assume for the purpose of decision that that which is false is true does indeed often look like a fiction, but no more or less so than asking legal decision-makers to apply rules in the area of their under- and over-inclusiveness. When a police officer stops an experienced and safe driver for exceeding the speed limit under ideally safe traffic and weather conditions, the officer is assuming that the driver and the conditions are average even though they are not (Schauer 2009, Chap. 1) . When the park official keeps out a non-peace-and-quiet-hampering vehicle he is assuming (or effectuating the rule-maker’s assumption) that the vehicle will hamper peace and quiet even though it will not. In these and countless other cases, the application of a rule in the area of its under- and over-inclusiveness can be seen to be a fiction in just the same way that the typical factual presumption is a fiction in the area of its factual falsity. Perhaps rigid rule-following is itself a problem, and this possibility will be the subject of the next section. But if rigid (or even non-conclusive) rule-following itself has its place and has its uses, then there is no reason to believe that there is anything alarming about achieving the same end through the use of a presumption, a presumption that we can see as doing pretty much exactly the same thing in pretty much exactly the same way.

In the previous section we assumed that rigid rule-following might in some contexts be desirable (Schauer 1988, 2009, Chap. 2