Toulmin’s Problematic Notion of Warrant

Abstract

We argue that it is wrong to try to identify a statement in an argument text as expressing a warrant. Indeed, it is wrong to apply the data/warrant/backing distinction to argument texts. Although Toulmin might arguably agree about some warrants, in presenting a “layout of arguments” he has been taken by some as presenting an account of analyzing argument texts. His indication that propositions expressing warrants are conditionals (or generalized conditionals) reinforces this reading. Furthermore, in applying his layout to quasi-syllogisms (syllogisms with a general conditional premise and a particular premise), Toulmin allows that the general premise may serve as a warrant. We detail many problems Toulmin encounters in taking quasi-syllogisms as a paradigm for arguments. He allows that the general premise may in this context also present a statistical report or both a warrant and a statistical report. Statistical reports summarize observations of particulars, and serve as backing for a corresponding warrant of the form “An A may be taken to be a B.” The general premise, then, does not express a universal proposition about all A’s, if the class of A’s is open ended or unlimited. Toulmin understands conditionals in general this way. But this is problematic. How can Toulmin maintain that there are no universal descriptive generalizations going beyond observed data, extrapolating from observed data, and that when open-ended generalizations appear in arguments, they always function as warrants? Toulmin does not offer a satisfactory argument for this view.

Toulmin was influenced by Ryle’s view of conditionals as inference licences, but Ryle’s argument is so fraught with difficulties that we conclude it does not make its point. Mill has presented an account of the quasi syllogism which remarkably anticipates Toulmin’s. A general categorical proposition may be understood first as asserting that whatever has a given attribute has some further attribute. But it also may be taken as an inference licence and backed by observation of particulars. As an inference it is not part of our knowledge but a rule enabling us to pass from something known or learned to some further proposition. We compress all our observations of particulars and possible inferences into one memorandum or register licencing inferences from particulars to particulars. In syllogistic reasoning, we are inferring the conclusion from the backing observations by analogy. In reasoning just from data and inference licence to a claim, we are deciphering what is registered in that memorandum. But although a proponent make take a universal proposition as an inference rule in his reasoning, when he presents that statement in a syllogism to a challenger,in effect in an argument text, it is more plausible to interpret it as a general statement according to Mill’s first interpretation, given the proponent’s intent to convince the challenger. The challenger is being invited to reason from two statements to the conclusion. So neither in Toulmin, Ryle, or Mill have we found a successful case for identifying universal conditional and related statements as warrants in argument texts.

In addition, there are further disadvantages to this view. Although in quasi-syllogisms, distinguishing warrant from data may seem straightforward, in many other arguments making such a distinction may not even seem possible, witness instances of the rule of conjunction, disjunctive syllogism, pure hypothetical syllogism, syllogism in Barbara, forms of immediate inference, arguments involving suppositions, inductive generalizations. Toulmin’s apparent insistence that warrants be included in argument diagrams encounters even further problems. Since deductive arguments do not state the inference rules they instance, we should in many cases have to supply those rules as warrants in diagramming these arguments. But why should we have to supply the warrant just to satisfy the requirements of our diagramming system? Furthermore, if the warrant corresponding to a quasi-syllogism is its universal premise while a warrant corresponding to a syllogism in Barbara is a formal inference rule, then there will be a distinct disparity in the warrants of these two syllogisms, even though the argument forms seem parallel. Such considerations lead us to conclude that warrants are not parts of arguments and should not be represented as such.