Steve Fuller, Philosophy, Rhetoric and the End of Knowledge (Wisconsin, 1993), pp. 367-376. (New edition to be published by Praeger in 2002.)

Consequential Theory:

Presumptions are normative instruments; for injecting some make-believe into an all too real world, with the long-term hope that reality may become more like make-believe. A scientific community may never know whether a given theory is really true, but by granting the theory paradigmatic status, members of the community are forced to act as if it were so, which causes them to frame their positions in terms of that presumption. Since a presumption is established on explicitly normative grounds, it is often difficult to claim that, just because a community presumes certain things, it follows that most of the community's members actually believe those things to be true. In fact, if the presumption is doing any real normative work, and hence correcting people's prior beliefs, then individual members of a community should suspect the truth or appropriateness of the presumption in particular cases, but nevertheless believe that the presumption should be upheld so as to force the relevant countervailing arguments to be mustered. The presumption of innocence in Anglo-Saxon law seems to work this way, but it also captures the attitude that disciplinary practitioners have toward a "widely held" theory in their field.

One way to understand this attitude is in the Durkheimian sense of reinforcing a collective identity, for the presumptively true theory in a field exemplifies what it is to excel at the methodological standards that confer epistemic legitimacy on the field as a whole. This point stands, even if the theory is eventually superseded and currently suspected by a large portion of field. Moreover, it may be argued (pro Popper and contra Polanyi) that what distinguishes science from a community of faith is precisely that leading theories are presumed rather than believed. As a result, scientists are professionally mandated to treat presumptions not as positive accomplishments in their own right but as way stations to be superseded on the road to inquiry.

A critic typically takes aim at a presumption, whether it be of a defendant's innocence (in the case of a trial prosecutor) or a belief's orthodoxy (in the case of an innovative scientist). The critic functions as an agent of rationality insofar as she clearly distinguishes the presumption's ability to be defended (its real probative strength) from its ability to prevent attack (its mere conventionality). The critic succeeds in overturning the presumption (and thus in fully "bearing" the burden of proof) if she can show that the presumption's unassailability has served only to mask its indefensibility. A situation of this sort arises if the presumption can be defended on no grounds other than the fact that it has traditionally been presumed. Among the more obvious candidates for overturned presumptions would, therefore, be various "folk" beliefs that may have been warranted when first introduced, but now have only habit on their side against our current background knowledge.

A presumption overturned in one case is not overturned once and for all. In other words, the effects of criticism appear to be purely local, confined to the single challenged case. (This is why one false prediction does not refute a theory–a point that Lakatos realized, but Popper did not.) For example, if a particular defendant is proven guilty, the general presumption of innocence remains unaffected. Indeed, the presumption would not be overturned even if defendants were always proven guilty, since the reason why innocence is presumed in trials is conceptually unrelated to the police's success rate at apprehending guilty parties. Likewise, if we extend this legalistic model to epistemic matters, to prove in one case that folk psychology does not offer the best explanation for someone's behavior does not diminish the presumption in favor of folk psychological explanations. However, it might seem that in epistemic matters the frequency with which the presumption is overturned should play a role in determining the presumption's fate–that the law should resemble more closely how science is intuitively thought to operate. Thus, Nicholas Rescher (1977) has developed an account of presumption closely tied to a claim's probability, with each defeat of the claim increasing its burden of proof.

I, too, believe that the difference between legal and epistemic presumptions has been exaggerated. In my view, both are normative correctives to widespread beliefs, which may, in the long term, cause those beliefs to change, but which do not depend on that prospect for their validity (cf. Fuller 1988a: chap. 4). Not surprisingly, then, the critic has her work cut out for her! Let us consider how this applies in both the legal and the scientific cases (cf. Fuller 1985).

Richard Whately (1963: pt I, chap. 3, sec. 2), the nineteenth-century Anglican bishop and rhetorician, clearly modeled the modern theory of presumption on what he took to be the conservative grounds for presuming innocence in Anglo-Saxon legal procedure. It is common for rhetoricians nowadays to interpret this conservatism as a strategy for risk-averse institutional action (cf. Goodnight 1980: 304-32). Thus, the judge would presume the defendant innocent in order to minimize the worst possible trial outcome–a safeguard against a hasty judicial decision that would needlessly ruin the defendant's life, were it subsequently shown to be based on a faulty understanding of the facts. But even given this interpretation of presumption's "conservative" function, it does not follow that presuming innocence is the most conservative course of action, especially if we are judging that action according to its consequences rather than its intentions, and moreover, if we expand the scope of the parties potentially under risk to include not only the individual brought to trial but the society at large. Both sorts of judgment are empirical in character, not merely the products of conceptual analysis performed upon "innocence" and "conservatism." It may be, for example, that the presumption of innocence has the effect of encouraging individuals to be more reckless in their actions–to do the sorts of things that superficially resemble crimes–because they know that even if they are brought to trial the burden will be with the plaintiff to show that the superficial resemblance is anything more than that. Even more likely is the increase in risk that the rest of the society will have to absorb as a result of the fact that the presumption of innocence will permit unconvicted felons to roam free. Thus, the social function served by a presumption of innocence probably has little to do with whatever risk-averse impulses may have prompted the good bishop (cf. Goodnight 1980: 322).

However, the presumption of innocence may discourage illegal activities in a somewhat different way. For example, in American civil procedure, there are two senses in which the burden of proof must be borne in a case. Whereas the plaintiff must always bear the burden of persuasion in demonstrating the defendant guilty of the alleged wrongdoing, the defendant may have the burden of producing evidence that shows that the plaintiff has not interpreted the defendant's actions in the most natural manner. The idea behind the defendant's bearing the burden of production is that if the defendant is indeed innocent, then he or she will likely have access to some fact that recontextualizes the case sufficiently to defeat the plaintiff's charge (Conrad et al. 1980: 840-43).

In fact, the problem so far with our analysis of the presumption of innocence is that the considerations that we just raised to show its possible risk-enhancing consequences are the very ones invoked by French jurisprudents in justifying the presumption of guilt as the appropriate stance for the judge to take toward the defendant (cf. Abraham 1968: 98-103, for a comparison of the role that presumption plays in Anglo-American accusatorial and European inquisitorial legal systems). In other words, both presumptions–of innocence and of guilt–have been legitimated on the same conservative basis, even though there are probably no empirical grounds for believing that either presumption especially contributes to a well-ordered society. In that case, why should the legal system presume anything at all about the defendant?

As it turns out, considerations of risk do not play nearly as big a role in Whately's thinking about presumption as the need to ground the persistence of a form of social life in a principle of sufficient reason: that is, there is a prima facie reason for believing that anything that has been the case ought to continue being the case. It might not be clear exactly what social good is served by the persistent social form, but simply the fact that the form has persisted is evidence for its serving some such good. Thus, the defendant is presumed innocent of this wrongdoing because he was innocent of other wrongdoings prior to his appearance in court. Consequently, given the good inductive evidence against the defendant's being guilty, the Anglo-Saxon judge is instructed to proceed cautiously in his inquiries so as to ensure against an unnatural understanding of what took place. The work of Alfred Sidgwick (1884: brother of the utilitarian Henry Sidgwick) marks the transition from Whately's "sufficient reason" analysis of presumption to the more modern "risk" analysis. Sidgwick argues that an inductive regularity probably points to some well-founded phenomenon that it would be risky to act against. This line of argument is generally based on John Stuart Mill's work on the relation between induction and utility. Still, the principle of sufficient reason can cut either for or against the defendant, depending on the form of social life the persistence of which the presumption is supposed to underwrite. It need not be the image of a particular defendant as a law-abiding citizen; instead, it could be, as in French juristic reasoning, the image of law enforcement agencies as consistently doing a job that needs to be done, which would then justify a presumption of guilt.

Whately's intentions notwithstanding, we should hesitate before embracing the sufficient reason interpretation of presumption. After all, the traditional appeal of sufficient reason approaches has been their psychologically compelling character. "Things just don't happen for no good reason," we are prone to say–but about what exactly, in the average legal proceeding? Are we not more likely to think that the defendant, and not the law enforcement agencies, has done something socially deviant (criminal or otherwise) to bring about the need for a trial in the first place? Taking psychology as our guide, then, sufficient reason would seem to weigh on the side of a presumption of guilt rather than a presumption of innocence. That is, if we assume that presumption must operate to conserve our intuitions about our fellow persons rather than to correct them. If we go the latter route, as I now suggest, then we must turn our gaze from the short-term role that the presumption of innocence plays in impeding the judge's actions against particular defendants to its long-term role in revising the judge's (and society's) attitudes toward defendants in general.

If nothing else, the presumption of innocence implies that the law enforcement agencies which bring an individual to trial are likely to be in error, and that the defendant, unless proven otherwise, has acted within the confines of the law. The level of fallibility attributed to the legal system on this view not only runs against our ordinary intuitions but is also quite foreign to the considerations which lead, once again, the French to presume guilt of the defendant. Yet it would be equally misleading to say that the French presumption merely reinforces the intuitions that the Anglo-Saxon presumption seeks to correct. For in the French system, the presumption of guilt licenses the judge to suppose that, regardless of whether the defendant is indeed in the wrong, something strange has been afoot worthy of further examination. What follows, then, is an exhaustive inquiry into the facts of the case, which continues until the judge feels that he has achieved an accurate understanding of what took place and can therefore subsume the case under the appropriate law. Indeed, the investigative powers of the judge are so extensive that he may freely suspend the rights of citizens (e.g., by wiretapping or opening their mail) in pursuit of crucial bits of evidence. By so conferring a greater value on the thoroughness than on the swiftness of the legal proceedings, French law supports an attitude of objectivity, impartiality, and ultimately the sort of certainty classically associated with an unhampered search for the truth. In the long term, then, the presumption of guilt manages to tinge the workings of legal conventions with the aura of scientificity, which has the effect of commanding greater respect of the citizenry than such conventions might otherwise do.

As I have earlier suggested, the presumption of innocence in Anglo-Saxon law serves a strikingly different purpose, namely, as a constant reminder of the mere conventionality and hence likely fallibility of law enforcement agencies. This acts against the natural psychological tendency, on the part of both the judge and the onlooking citizenry, to make too easy an identification between being a defendant and being a wrongdoer, which has the long-term effect of increasing the judge's capacity for fairness. The capacity for fairness, however, is not tied in any empirically clear way to increased social stability or even more correct verdicts. In that case, the presumption of innocence is perhaps best interpreted as simply a mental discipline undertaken by the judge and onlookers to correct what is taken to be an inherently bad psychological tendency that can, in various indirect ways, inhibit the administration of justice and, more generally, the wholesomeness of the defendant's subsequent interaction with his or her fellows.

The upshot of this part of the discussion, then, is that instead of seeing presumption as a conservative force in legal reasoning and hence an object of criticism, our reinterpretation of the Anglo-Saxon presumption of innocence illustrates a sense in which presumption may function as a tool for criticizing and revising beliefs that are widespread among judges and other legal functionaries, and yet not conducive to promoting the goals of the legal system. At this point, I would like to extend this new view of presumption to epistemic matters, where it helps to spell out much of what is involved in radical conceptual change in science.

In keeping with the new view of presumption, radical conceptual change directly brings about a change in the orthodox beliefs of the scientific community, and only indirectly a change in the actual beliefs of individual members of that community. The difference suggested here between orthodox beliefs and individual beliefs is reflected in the different answers that would be given to the following two questions:

Indeed, I maintain that radical conceptual change is possible because the answer to question (2) places no necessary constraints on the answer to question (1), while the answer to question (1) can be deployed as part of a strategy for altering the answer to question (2).

Question (1) considers how members of the community think that the burden of proof should be distributed among their beliefs. For even if the vast majority of members of the community happen to hold a certain belief, that is no indication of whether they would allow it to pass in open forum without strenuous argument. As in the case of legal functionaries wanting to inhibit their natural tendency toward believing that all defendants are probably lawbreakers, there are methodological and ideological reasons why the members of a scientific community might have an interest in keeping certain of their widely held beliefs from achieving the status of an orthodoxy by holding those beliefs accountable to standards of proof that they clearly cannot meet (cf. Harman 1986: 50-52, for the only recent attempt in analytic philosophy to deal with the difference between beliefs "held for oneself" and those "held for others").

Among the methodological reasons may be that the belief, though widely held, is held only on the basis of indirect evidence or on the purely pragmatic grounds that such a belief is implicit in the assumption that the standing beliefs of the community form a maximally coherent set. Elevating a belief of this sort to the status of an orthodoxy, tempting as it may be, would inhibit any further testing and thereby prevent the scientific community from discovering whatever element of falsehood it may contain. Indeed, Popperians would be especially suspicious of such a belief, since its official acceptance promises to close critical inquiry on the set of beliefs it renders maximally coherent. Another methodological reason for restricting the set of orthodox beliefs is to keep domains of inquiry separate. For example, it would not be surprising to learn that most natural scientists believe not only in the existence of God but also in the occurrence of supernatural causation at some point in the history (most likely at the origin) of the universe. Yet the extent of the burden of proof that such a belief must bear (especially as measured by the number of alternative orthodox explanations that must be first ruled out) ensures that it will never again become part of the scientific orthodoxy. Indeed, Jeffrey Stout (1984) has epitomized the Scientific Revolution in the seventeenth century as marking, not the beginning of a decline in the belief in God, but a decline in the social recognition of such a belief as rational, thus shifting the burden of proof from nonbelievers to believers. Less exotic examples of the same phenomenon include the presumption against folk psychological explanations on the part of social scientists as a means of keeping their disciplines separate from common sense, and the presumption against explanatory appeals to the unconscious and class interests on the part of classical humanists as a means of keeping their disciplines separate from the social sciences.

As for the ideological reasons, a commonly held belief may remain unorthodox because it instills a "bad" attitude toward scientific inquiry. For example, few scientists would probably deny that sociologists have accurately captured the extent to which scientific research agendas are opportunistic, indeed to the point that scientists openly admit to manipulating ex post facto the significance of research findings to fit currently popular theoretical debates. Yet, these scientists would also probably resist the suggestion that they henceforth justify knowledge claims in terms of their opportunistic agendas, even if the sociologists turn out to succeed in showing that "the logic of opportunism" better explains (predicts) their behavior than appeals to the allegedly univocal relation in which evidence stands to theory (cf. Knorr-Cetina 1981). Indeed, more than just setting high probative standards, the scientific community has erected many purely conceptual barriers that serve to make the sociologist's stance difficult even to articulate. Many of these barriers appear as distinctions that have been canonized by positivist philosophy of science, especially reasons versus causes and theoretical versus practical. While these distinctions are drawn precisely enough so that each pair of terms is jointly exhaustive, their range of applicability is sufficiently malleable so that anything "inherently" a feature of scientific reasoning can always be made to appear on the "reasons" or "theoretical" sides of the distinction. In this way, scientists are able to project to themselves and to the nonscientific community an image, and ultimately an attitude, of detachment from thoughts of career advancement and other forms of self-interest.

So far we have looked at the use of presumption (and, correlatively, burden of proof) in scientific reasoning as a means of arresting change that the canon of orthodox beliefs would naturally undergo if all commonly held beliefs were granted orthodoxy. However, perhaps the more interesting use of presumption is in facilitating a change in the canon, granting orthodoxy to beliefs that have yet to be widely held by members of the scientific community. For if there is an analogue to the presumption of innocence in Anglo-Saxon law, it is in the attempt to facilitate such a change in the scientific community. However, my model for the structure of this change will be drawn from a determinedly nonscientific source: Pascal's Wager on the existence of God. In a crucial respect, Pascal's problem is similar to that faced by a scientific revolutionary such as Galileo: both want to believe something that, for the moment at any rate, is unwarranted. It would seem, therefore, that they must scotch either their criteria of rationality (which says to have only warranted beliefs) or their desired belief (as rationality would demand). But there is a third way out: they can cause themselves to arrive at a situation in which their desired belief is warranted.

Now, as Bernard Williams (1973: chap. 9) has noted, this third situation can arise either because the belief is indeed true (the change in situation would thus have resulted from some improvement in our cognitive powers) or because appearances have been manipulated so as to make the belief seem true (the change in situation would thus have resulted from some form of deception). Clearly, the former is preferable to the latter, but if Feyerabend's (1975) account of Galileo is to be believed, the latter will occasionally do as well. Williams goes on to make the interesting point that a simple sentence-uttering mechanism may have knowledge without having beliefs, in that whereas what one knows can be read off what one says, what one believes cannot be similarly read off, since it is up to the individual to decide what he will say given what he believes. This allows a possibility for intentional falsehood that is lacking in the machine. If we regard the change in presumption that occurs during a scientific revolution as a decision to say what one does not necessarily believe, at least in the short term, then Williams provides us with a way of identifying a "collective will" of the scientific community which adopts a presumption in favor of certain knowledge claims that its members do not yet individually believe.

In any case, Pascal or Galileo will have to reorganize their environments in such a way that the sort of reasons that would be needed to warrant his desired belief could become available. Again, this may simply involve fabricating some evidence that is tailor-made to the belief, or it may require an extended crucial experiment that is especially designed so that if the evidence warranting the belief does not arise under those circumstances, then the belief is probably false. Whereas the first situation is set up to be foolproof, the second clearly is not, since it is dependent on the state of the world regardless of one's own beliefs. And whatever Galileo may have had in mind, Pascal thought of his wager, with its attendant requirement that he conduct a thoroughgoing Christian life, as a crucial experiment in the above sense. Thus, Pascal feels the risk of the wager, as the strength of God's signs to him vary on a day-to-day basis. Interestingly, this second strategy of presumption formation also approximates Charles Sanders Peirce's use of the term "presumption" (otherwise called "abduction" or "hypothesis"), which, as Ann Holmquest (1986) has observed, operates for Peirce as the motor of scientific progress.

Of course, the distinction between merely fabricating evidence and positioning oneself to acquire genuine evidence can be easily erased from the prospective believer's mind, if she is also able to cause herself to forget her interest in wanting to hold the belief, as Jon Elster (1979: chap. 2) suggests in his account of presumption as "precommitment." Annette Baier (1985: chap. 4) offers an interesting slant on this topic, in that she accepts that changing one's mind consists of rethinking the evidential relations of what one already knows, and is thus not tied to a specific piece of evidence or argument, as ordinary belief revision is. However, Baier argues that this rethinking occurs by remembering what was earlier forgotten–à la Plato's Meno–instead of forgetting what was recalled. Needless to say, either proposal–but especially Elster's–is a tall order, since forgetting is something done not deliberately but only as a by-product of some other activity. A good candidate for an activity of this sort is the restructuring of discourse that is necessary for articulating any presumptively new relation between language and the world: the introduction of new terms, new meanings for old terms, and new inferential moves within the language in general. As one plays this new language game, it becomes so natural that the player becomes convinced that she has been implicitly playing all her life. This describes not only Pascal's adopted Christian lifestyle (in, e.g., Pensées 252), but also the Whiggish characterization in terms of which revolutionary scientists tend to see their predecessors once they have presumed a new paradigm.

Moreover, there is good social psychological evidence that the mere articulation of the new language game, on a regular and elaborate enough basis by enough people, will have the long-term effect of changing the beliefs of individual scientists. The evidence, drawn admittedly from studies of how political factions tend to gain dominance, suggests that at first any splinter group (say, an inchoate paradigm) is presumed by the public (say, the scientific community at large) to be a minority voice that would not have needed to speak up had its views been adequately represented by the dominant party. As Whately himself put it, under those circumstances, only he who asserts must defend. However, as time goes on, the presumption starts to shift away from the dominant party if it refuses to answer the claims made by the splinter group. For in that case, the public tends to interpret the silence as tacit acceptance of the claims and hence ideological capitulation to the splinter party, which, in turn, often ends up leading the majority of voters to move to its side as well, creating a bandwagon effect." The spiral of silence" (Noelle-Neumann 1982) is the expression that public opinion researchers use to characterize this frequent phenomenon, but readers of Kuhn may recognize it as the "Planck Principle" which baldly claims that a new paradigm triumphs once voices of opposition from the old are silenced. If this link is apt, then we see the beginnings of a theory of long-term rational conceptual revision that incorporates much of what is distinctive in Kuhn's work into a general account of presumption. Fuller (1985) made some opening moves in that direction.