121 How Many Examples is an Induction? Dale Hample, University of Maryland Examples have a peculiar status in argumentation theory. A single example offered in support of a claim is susceptible to the charge of fallacy, often that of overgeneralization. In fact, one instance is not always regarded as proof at all: Perelman and Olbrechts-TLJteĐa ;ϭϵϲϵͿ distiŶguish ďetǁeeŶ edžaŵple, ǁhiĐh is pƌoďatiǀe, aŶd illustƌatioŶ, ǁhiĐh is Ŷot. OŶ the otheƌ haŶd, Aƌistotle considered example and enthymeme to be the two main structures of rhetorical proof, indicating that he thought that a single example has some persuasive value. More pointedly for this paper is the consideration that if one has enough examples they are somehow collectively transformed into a sample. Then instead of proof by example one has proof by induction. And induction, of course, has a clearly approved place in argumentation theory. So how many examples constitute an induction? Probably the best way to approach this question theoretically is to study sampling theory. Given assumptions about the distribution of quantitative examples, we can calculate measures of central tendency and their variability. One sample size might give us a confidence interval of ± 2% and another might give ± 5%, and we can choose the number of instances (i.e., data points, examples) on the basis of how much variability or uncertainty we are comfortable with. Even for non-quantitative examples, sampling theory might offer ballpark estimates of when we have enough examples to constitute an induction. Normative approaches, either rooted in sampling theory or some other source, would of course be valuable. Here, however, I take a different course. I want to know when ordinary arguers begin to treat examples as though they were inductions. To understand why people should make this transition, it is important to observe that a single instance is not completely vacuous. Even if one example were clearly insufficient to sustain a general claim, the example is still a step toward such a proof. This contrasts with fallacies that are so awful that they have no probative value at all. An undistributed middle term might be an illustration of a flawed argument scheme that could multiply indefinitely without ever cumulating into anything approaching good support for a conclusion. Examples, then, should be understood as convergent arguments (Henkemans, 2000). A convergent argument is one in which a series of independent lines of proof each converge directly on the same conclusion. Hample (2011) has studied converging but flawed causal arguments, and reported that as the flawed individual proofs cumulate in a conversation, ordinary arguers are less critical of them.