Informal Logic X.2, Spring 1988 Upsetters STEPHEN S. CAREY Portland Community College A world of things can happen to upset the inferences we draw. Consider one. Im- agine you and I are standing in line for the next performance at a movie theater. I know the theater holds exactly 100 people; I've been there before and have counted the seats. Counting the people standing in line, I conclude we will be seated for the next performance since there are only 93 peo- ple ahead of us. Is my conclusion justified? The answer here is a resounding, "it depends. " What if several people in line ahead of us are saving places for others? If enough people are doing so, my conclu- sion surely doesn't follow. Thus, if! think it does follow, I presumably believe that no more than five seats are being saved for peo- ple not in line. That more than five seats are being saved I call an upsetter in light of the fact that, if true, it will upset my belief that we will be seated for the next performance. In what follows I want to consider the role upsetters play in the arguments we give. In Section I, I will argue that if upset- ters play any role at all, their denials must be included as premises of the arguments they would, if true, upset. This fact about upsetters is important. If, as I will argue in Section II, we are to accommodate it, we must modify the more or less standard view of what it is for one claim to count as sup- port for another. I No doubt, the temptation is great to think upsetters should not be included as premises of the arguments they purport to upset. On this view, it would be wrong to represent the argument implicit in the ex- ample above as follows: (a) PI: Only 93 people are in line in front of us. P2: The theater seats 100 people. P3: No more than a total of 5 seats are being saved for others by people in line in front of us. C: We will be seated for the next performance. The problem here is with P3. For if we include P3, as a premise of (a), we will have to say that (a) is incomplete, almost hopelessly incomplete. Any of an indefinite- ly large number of things, in addition to that denied in P3, could occur which would upset C: some of the seats might be out of order, people now in the theater could decide to sit through another performance, you might have second thoughts about see- ing this particular film, etc. Despite the fact that some of these things are highly im- probable, it would seem that the denial of each constitutes a premise of (a) if P3 con- stitutes a premise. For if P3 is a premise, it is because C does not follow from PI and P2 in the event that P3 is false. But the same is true for the denials of each of the possibilities mentioned above. So if P3 is a premise of (a), it would seem (a) is com- mitted to an indefinite series of premises corresponding to every possible upsetter of C. If our choices are limited to acknowledg- ing that (a) is in all likelihood of indefinite length or denying that P3 is a premise, our choice is clear: we must deny that P3 is a premise of (a). To do otherwise is to challenge an assumption so deeply imbed- ded in logical theory that it is rarely stated: 98 Stephen S. Carey an argument can be formulated in terms that are complete, i.e., as a series of supporting claims and a conclusion to which no fur- ther claims need be added if the supporting claims are to actually support the conclusion. Nevertheless, there seems to be a perfectly correct sense in which (a) requires the truth ofP3, since ifP3 is false, C's con- clusion does not follow. How, then, are we to accommodate P3 if not as a premise of (a)? Consider some of the moves we might make here. (1) It might be suggested that the argu- ment, (a), presupposes the denied upsetter, P3. One claim presupposes another if the truth of the latter is a necessary condition for either the truth or falsity of the former. Similarly, an argument presupposes a claim if it would make no sense to give the argu- ment in the event that the claim were false. If, for example, I believe there will be no further performances today, there would be no point in my arguing that we will be seated for the next performance. If there are no further performances, it is neither true nor false that we will be seated for the next one. That there will be another peformance is, thus, a presupposition of (a) or, more accurately, something one would presup- pose if one were to argue for C, the con- clusion of (a). P3, however, does not seem to function as a presupposition of (a). Cer- tainly, (a) requires that P3 be true if C is to follow. But the failure of P3 to be true does not render (a) pointless. If P3 is false, C may well be false, but it is not neither true nor false. So, (a) does not presuppose P3 but does seem to depend on the truth of P3. P3 seems to be more like a premise of (a) than a presupposition of (a). (2) It might be suggested that P3, the denied upsetter, is a back-up for a gap-filler of (a). I use Ennis's picturesque terminology to describe this move though other terms could be used. Following Beardsley, for ex- ample, we might say that (a) is a complex argument; PI and P2 along with a further premise support C while P3 in tum supports that further premise. Here is a more detailed account of this move, using Ennis's terms: The argument implicit in my imagin- ed scenario is committed to a gap- filler: if the theater holds 100 people and you and I are 94th and 95th in line, we will be seated for the next performance. (P4 in what follows.) One who believes the upsetter denied in P3 will not occur has a reason to believe P4. So, P3 is a back-up for P4. This is a tempting move in that it not only assigns a plausible role to P3 but also explains why P3 should be important to (a): a crucial premise depends on the truth of P3. However, I am not convinced this move is the right one to make. The effect of adding P4 is to tum (a) into a deductively valid argument, since PI, P2 and P4, taken together, necessitate C. But in the circumstances I have imagined for (a) I see no reason to think someone would give a deductively valid argument. Would some- one who actually argued for C on the basis of PI and P2 believe C couldn't be false given the truth of PI and P2? Realistical- ly, I can imagine someone willing to grant that if nothing goes astray-if none of the things I have called upsetters come to pass-then C will be true if PI and P2 are. But things do go astray and arguers are generally aware of this. A naive arguer might intend, in giving (a), to give a deduc- tive argument, or at least believe the con- clusion of his or her argument to be a sure thing, given the truth of PI and P2. A sav- vy arguer, however, would not. So it is not clear that we need to construe (a) to be deductive. My inclination is to say that most people in a position to give (a) would be shrewd enough to realize that though C is a good bet, it is not a sure thing. Never- theless, I am willing to acknowledge that someone, in the circumstances I have im- agined, might give a deductive argument for C; if such an argument is given there will be no problem about upsetters since there will be no conceivable upsetting conditions or, at any rate, none the arguer is prepared to grant as being at aU likely. But it is surely possible someone would give a non- deductive argument for C and here upset- ters cannot be characterized as back-ups for gap- fillers. A closely related move here would be to suggest that (a) involves a tacit premise to the effect that no upsetting conditions are likely to occur. On this view, specific upset- ters, like that in P3, are not a part of the argument. The problem here is that a tacit premise to this effect adds nothing to the agument. One who believes that nothing can happen to upset the argument simply believes the hypothetical premise, P3, to be true. Thus, though the proposed tacit premise may tell us about the attitude of the argument's author, it adds nothing to the argument's content. Ennis has an interesting reply to my ob- jection to adding validating gap-fillers, like P4: Argument criticism, I believe, can be facilitated by an argument reconstruc- tion process that requires each part of the reconstructed argument to be either a deductively valid step or an explicitly stated proposition (including such propositions as those that claim that the inductive evidence is suffi- cient). This approach to argument criticism does not hold that all arguments are really deductive. Rather, it holds that a useful step in argument criticism is the reconstruc- tion of the argument into a form that has as inference steps only deductively valid ones. (pp. 84-85) This is an intriguing idea but I do not think it provides any solid reason to think P3, the denied upsetter, a back-up for a gap- filler. Maybe it would be useful to think of P3 as though it were a back-up for a gap- filler of the deductively valid argument: "PI, P2 and P4, so C." But even if think- ing this way helps us to evaluate (a), it is not a reason to suppose that (a) is deduc- Upsetters 99 tive and that, as a result, P3 is not a premise. It is one thing to decide how to go about evaluating an argument but quite another to decide what an argument's implicit and explicit parts are and how they relate to one another. (3) It might be suggested that the upset- ter denied in P3 sets forth a condition of rebuttal. This move is made by Toulmin, originally, in The Uses of Argument. On Toulmin's view, corresponding to the arguments we give are warrants- generalized hypothetical statements to the effect that one is entitled to draw the con- clusion from the data contained in the argu- ment's premises. Thus, standing behind the argument, "PI and P2, so C," is the war- rant: any pair of people who are 94th and 95th in line at a theater which seats 1 00 peo- ple will be seated for the next performance. (W in what follows.) A rebutting claim sets forth circumstances under which the general authority of the warrant must be set aside. Thus, on Toulmin's view, C follows from PI and P2 in light of W, unless the upset- ter in P3, or some other upsetter, obtains. Setting aside the question of whether W is to be characterized as a premise of or a war- rant corresponding to (a), there is something odd in the suggestion that the upsetter in P3 constitutes a condition under which W would have to be set aside. To see this, con- sider what happens when circumstances are imagined in which someone might actual- ly give the argument, (a). Imagine, first, that (a) is given by so- meone who believes the upsetter in P3 to be true or, at any rate, given in cir- cumstances where there is good reason to believe the upsetting condition will occur. In these circumstances, W will surely not be given as a premise, for it will be believ- ed to be false. Remember, W is the case unless the upsetter mentioned in P3 occurs and, in these circumstances, there is every reason to believe the upsetter will occur. But to be a rebutting condition, the upset- ter in P3 requires that W be a premise (or warrant). Otherwise, what does it rebut? 100 Stephen S. Carey Now imagine instead that (a) is given by someone who believes P3 to be true, i.e., that the upsetting condition in P3 will not occur, or that it is given in circumstances where there is no reason to think the upset- ting condition will occur. Here, W may well be true. But in such circumstances it is hard to imagine that someone would cite the upsetter in P3 as a rebutting condition or anything else for that matter. After all, the upsetter in P3 is believed to be highly unlikely. What could be made of someone remarking, "W, unless the upsetter in P3 occurs, but there is just no chance that the upsetter will ocur?" Of course, one who argues for C might be unaware of the upset- ter in P3. Here, W might initially be given as a premise of (or warrant corresponding to) the argument. But once the upsetter in P3 is brought to the attention of the arguer-once it is pointed out either that the upsetter is likely or unlikely to occur-the difficulties for W discussed above reappear. In sum, the denials of upsetters are neither presuppositions nor back-ups for gap-fillers of the arguments they purport to upset. Nor do upsetters express conditions of rebuttal. Thus, none of the moves I have considered has shown the denied upsetter, P3, to be something other than a premise of (a). This, coupled with the fact that P3, if false, will undercut the support provid- ed for C by PI and P2, suggests that P3 is a premise after all. If P3 is not a presup- position, backup for a gap-filler, if it does not set forth a condition of rebuttal and if, in addition, its truth is required for the in- ference in (a), what could it be if not a premise? Yet despite the case I have made, I am sure most logicians would want to deny that P3 is a premise of (a) for a powerful reason given earlier. If P3 is understood to be a premise of (a), then (a) is committed to an indefinitely long string of premises, corresponding to all possible upsetters of C. In the next section, I want to consider this objection to the inclusion of P3 as a premise of (a), for I think it can be shown to be without foundation. II Under what conditions does a claim, P, constitute a premise in an argument for another claim, Q? I think the following, by Wesley Salmon, is representative of what most logicians would say about the relevant notions: argument, premise, and conclu- sion. "An argument consists of one state- ment which is the conclusion and one or more statements of supporting evidence. The statements of evidence are called 'premises'." (p. 3) Thus, if P is to be a premise in an argument for Q, P must pro- vide "supporting evidence" for Q. Fine. But under what conditions does P provide such support? Salmon goes on to say the following. "The premises of an argument support the conclusion if the truth of the premises would constitute good reason for asserting that the conclusion is true." (p. 4) So, P supports Q just in case P, if true, would constitute good reason for asserting Q. This, I suspect, is roughly the account most logicians would give of what it is for one claim to support another. But though the account is correct, I think it is also in- complete. More is required of P if it is to support Q. In a context where someone ac- tually advances P as providing support for Q, P must be true or presumed to be true by one who gives or accepts it as support for Q. If a claim is neither true nor believ- ed to be true by the parties to an argument, it makes little sense to speak of the claim as providing support for the argument's conclusion. To point out that such a claim would support the conclusion, were the claim true, is not to show that it actually provides supports for that conclusion. For example, we would have strong sup- port for C, the conclusion of (a) from Sec- tion I, were the theater manager a good friend and were she to have promised to save seats for us. If no such promise were made, the claim that it was cannot be said to support C; certainly, I would not advance it as supporting C nor would anyone who believed it false accept it as supporting C. Thus, the question of whether P supports Q can be answered only by reference to the context in which the argument, "P, so Q, " is advanced, i.e., by reference to the fac- tual background against which the argument is advanced as well as to the beliefs of the parties to the argument about the truth of the claims involved. If P is neither true nor advanced in a context where it is believed true, P cannot be said to support Q, nor even to appear to support Q. Under what conditions does a claim, U, count as an upsetter corresponding to the argument, "P, so Q"? I think the response of most logicans to this question would be analogous to the response above about the conditions under which one claim supports another: U is an upsetter corresponding to "P, so Q" just in case U, if true, would undercut the support otherwise provided for Q by P. In this answer lies the source of the logician's hesitancy to include the denials of upsetters among the premises of an argument. Any argument can be upset by a limitless number of events, given that we construe an upsetter to be anything which, iftrue, would undercut the support provided the argument's conclusion by its premises. And if we add the denial of one upsetter as the premise of an argument, it would seem we must add the denial of every conceivable upsetter since each, if true, would upset the conclusion at issue. I think, however, this is wrong. The thought that every conceivable upsetter must be added if one is, depends on the view I have said most logicians hold of the conditions under which a claim functions as an upsetter. And I think this view is mistaken. The view I have attributed to logicians does not set forther the conditions under which some specific claim, U, actually serves as an upsetter corresponding to "P, so Q." Rather, it sets forth the conditions under which any claim could serve as an upsetter. The distinction here is similar to that drawn earlier with respect to support- ing claims. Much as we can distinguish be- tween conceivable and actual supporting claims, we can similarly distinguish be- Upsetters 101 tween conceivable and actual upsetters. The range of events which conceivably could upset a given argument is no doubt enor- mous and logicians are right in suggesting this. But the number of upsetting events cor- responding to a given argument, in context, is bound to be much smaller. Under what conditions does a specific claim, U, function as an upsetter correspon- ding to "P, so Q"? Here we must consider the context in which "P, so Q" occurs much as we needed to do so to account for the conditions under which P actually sup- ports Q. U functions as an upsetter only if, in the context in which "P, so Q" occurs, U meets two conditions. First, U must be such that, if true, it would undercut the sup- port otherwise provided for Q by P. Second, U must stand a realistic chance of coming to pass or be thought to stand such a chance by the parties to the argument. Imagine, for example, that the theater manager knows you and I tend to behave badly at the movies and so will refuse to seat us if she sees us in line. Were this true, C might not follow even if PI and P2 were true. (PI and P2: Only 93 people are in line in front of us and the theatre seats 100 people.) But in a con- text where this imagined upsetter is false, it cannot function as an upsetter; certainly I would not consider it in giving my argu- ment nor would it be cited by anyone in- tent on criticizing my argument. The restraints imposed by argumentative context on upsetters are not identical to those imposed on supporting claims in general, nor are they as straighforward. P must be true or at least presumed to be true if it is to support Q. By contrast, an upset- ter need only be such that the possibility of its truth is sufficiently likely to pose a real threat to the support provided Q by P. No doubt, there will be borderline cases- possible upsetters whose probability of com- ing to pass is, in a given context, difficult to gauge. But there will also be clear cases-highly improbable and, thus, irrele- vant upsetters as well as plausible upsetters that clearly demand to be taken seriously. 102 Stephen S. Carey In a context where we have every reason to worry about the possibility that people in line in front of us are saving seats, P3, the denied upsetter from Section I, is a clear example, corresponding to the argument, (a), of the former; in a context where we have no such reason, it is a clear example of the latter. The fact that there may well be borderline cases, thus, need not be taken to suggest that the context in which an argu- ment is given does not provide a rationale for rejecting many possible upsetters. Interestingly enough, the problems we are likely to encounter in deciding the relevance of upsetters may also confront us in deciding about other sorts of supporting claims. Suppose someone, in the process of giving an argument, makes a claim, the truth of which is questionable. Suppose also that the claim, if true, would support the argument's conclusion. Do we include the claim in formulating the argument? Certain- ly, we would do so if the claim were clear- ly intended to be a premise by the argu- ment's author. However, in the absence of any contextual clue to this effect, we might well have a difficult time deciding whether to include the claim in formulating the argu- ment. But we nevertheless make such deci- sions, and in making them, the context in which the argument occurs will be our guide. In a context where the claim is like- ly to be true or to be believed true by the parties to the argument, we may decide to include it. In a different context, we may decide otherwise. I think argumentative context can similarly guide us in deciding whether to reject or include possible upsetters. It may seem here that I am ignoring an important point: to decide whether a claim is intended as a premise is not to decide whether the claim actually supports the argument's conclusion. Put another way, it is one thing to reconstruct an argument and quite another to evaluate it; only in evaluating an argument do we need to be concerned with the truth or falsity of a pro- posed supporting claim. However, I think this point is of dubious value, at least when applied to anything other than textbook ex- amples of argument. Can a claim be false yet provide support for another claim? Con- sider the following "argument:" We shouldn't celebrate Lincoln's birthday, since he never really was President of the United States. I should think this is an example not of a suspect argument, but of something that is not an argument at all. Why? Because its purported premise, in being patently false, fails to support the conclusion in defense of which it is advanced. Certainly one who understands that the claim is false would not advance it as a premise nor accept it as such. I see no reason why logicians should not do the same, in reconstructing the arguments people give. If I am right, the hesitancy of most logi- cians to include upsetters as premises stems from a failure to appreciate the extent to which argumentative context constrains us in our choices of what do and do not con- stitute supporting claims. The idea that con- text imposes such restraints would hardly be remarkable-I suspect most logicians, if pressed, would concur in principle-were it not for the fact that logicians tend to ig- nore it in speaking of the very idea of what it is for one claim to support another. Recall the view, attributed earlier to logicians, of what it is for one claim to support another: if P is to support Q, P must be such that it would, if true, constitute good reason for asserting Q. This account makes it look as though the question of whether one claim supports another can be answered in abstraction from any concern with the con- text in which the claims are advanced. To resolve the question of whether one claim supports another, one need only speculate about whether it is possible to imagine cir- cumstances in which the former would con- stitute a reason for asserting the latter. By contrast, I have maintained that the ques- tion of whether one claim actually supports another can be decided only by reference to contextual matters, i.e., by reference to the beliefs and expectations of the people involved and the facts of the case. When upsetters, in particular, are ex- amined in abstraction from any context, one upsetter is bound to seem as likely as another. From such a perspective, it will doubtless seem that if any upsetter is includ- ed in formulating an argument, all must be included. After all, each, if true, would upset the argument's conclusion. But much as the context in which an argument is given determines what is and is not to count as support for its conclusion, so context dic- tates that all possible upsetters are not on an equal footing. The constraints imposed by an argument's context, thus, will often insure that, by including a given upsetter as a premises, we do not automatically com- mit the argument to an endless string of premises. In short, the inclusion of upset- ters need not be understood to challenge what, in Section I, I characterized as an assumption "deeply imbedded in logical theory' ': arguments can be formulated in terms that are complete. Imagine once again that you and I are standing in line at the theater, that I have counted the people in line in front of us and know the theater's capacity. I give the now familiar argument: There are only 93 peo- ple in line in front of us, the theater holds 100, so we will be seated for the next per- formance. But now add a bit more detail. Hearing my argument, you reply, "But what if some of the people in front of us are saving seats?" I know this is not like- ly, for the theater has a policy, plainly stated on the ticket stubs, prohibiting this. And so Upsetters 103 I reassure you. "Don't worry. No more than a few impolite people are going to ig- nore the theater's policy." In this context, given what I believe and given your con- cerns, I think we must include my reassurances as a premise of the argument I have given. To do otherwise is to risk missing the content of, as well as the con- cerns which have prompted me to give, my argument. References Beardsley, Monroe, Thinking Straight, Fourth Edition, Prentice-Hall, Inc., Englewood Cliffs, N.J., 1975. Ennis, Robert H., "Identifying Implicit Assumptions," Synthese, 51: 1, April 1982, pp. 61-86. Salmon, Wesley C., Logic, Second Edition, Prentice-Hall, Inc., Englewood Cliffs, N.J., 1973. Toulmin, Stephen, The Uses of Argument, Cambridge University Press, Cam- bridge, 1958. Toulmin, Stephen, Rieke, Richard, and Janik, Allan, An Introduction to Reason- ing, Macmillan Publishing Co., Inc. New York, 1979. Dr. Stephen S. Carey, Department of Philosophy, Portland Community College, 12000 SW 49th, Portland, Oregon, 97219. D