On Sun, May 25, 2008 at 1:39 AM, <wokshevs@xxxxxx> wrote: > Didn't Aristotle maintain there are (at least) two different categories of > inquiry: one which permitted of truth of the unchanging via demonstration > of > necessity, the other permitting judgement only in the qualitative and > scalar > terms of "for the most part" and "usually, in such circumstances" ? The > former > category he named "epistemce," the latter "deliberation." Is it that the > intelligence required to categorize inquiry in such terms "thinks in > degrees > instead of categories"? > Over on anthro-L I have been engaged in a discussion with someone who might (only slightly tongue in cheek) described as a cultured despiser of statistics. The following exchange includes some pertinent material. ---- My debating partner writes, > > I know that puts me at odds with all of you who had to take statistics in > college, and for that > I'm sorry. But I find it really hard to take any of that stuff seriously > after I've added the mandatory "except when they're not" to every sentence > about how people in general or members of a population "are". I reply, To me the sort of problems you are encountering with "are....except when they are not" occur because, while data are often statistical, most of us, most of the time, try to interpret them in terms of classic scholastic (originally Aristotlean) logic. The scholastic assumes a world in which things fall into natural categories. The members of a category must, by definition, share a set of uniform properties. These constitute the category's essence. Any other properties the category's members possess are, again by definition, accidents. They have no bearing on inferences deduced from category membership, which depend entirely on the uniform properties that define the essence of the category. Start from this way of looking at the world and scholarship becomes a search for definitions. "Are...except when they are not" is frustrating evidence that you haven't found the essence yet. If you plug along and keep running into one "Are...except when they are not" over and over again, you may well decide, after a while, that this business of chasing essences is fruitless. It may occur to you that, "Whoa! There are nothing but accidents here!" So you change your metaphysics and become a nominalist. You go about telling people that categories are just made-up ways to summarize accidents. Very convenient, of course, since without them we'd have trouble agreeing on anything. At the end of the day, however, they are just "social constructions." And, since, in your view of knowledge, there is nothing but categories to know, there you are stuck. But let's go back to the scientist. From her perspective, the critical components of knowledge aren't scholastic categories, with their essences composed of uniform properties. Instead they are property spaces in which cases with multiple properties can be distributed in all sorts of interesting ways. Instead of Aristotle, her hero is Descartes, not for his "Cogito ergo sum," but his main contribution to science—analytic geometry. The simplest kind of property space has only two dimensions. It is represented by graphs like those in which, for example, stock prices or population size are tracked over time. In its raw form, the data on which the graphs are based are just points in the space, defined by the intersection of two measurements, one on the x axis, the other on the y axis. They form what statisticians call a scatter plot. Already, however, we can see a whole new class of questions arising. It could be that all the points fall neatly into two clusters and, further, that all members of cluster A overlap at exactly the same point {x(A), y(A)}, while all members of cluster B overlap at another point {x(B), y(B)}, which would imply that A and B are perfectly formed scholastic categories. In real science, however, this result is incredibly rare. A scatterplot gets its name from the fact that the points in the graph are scattered about (like, in a moment of whimsy, I imagine, the freckles on a pretty girl's face, not two perfectly positioned beauty marks). So what does our scientist do? Does she throw up her hands and conclude that generalization is impossible. No. Instead of concentrating on points and trying to divide them into categories, she steps back and looks at their distribution: Are they tightly or loosely clustered? Do they fall within a specific range? Are there outliers that fall outside the zone in which most of the points are concentrated? Is it reasonable to fit a curve to the data, so that most of the points fall close to the positions predicted if the formula defining the curve were correct? What I want to emphasize here, however, is the way the questions have changed. Instead of a fruitless search for essences, continually frustrated by "are....except when they're not," we have a search for interesting patterns in distributions and how to explain them. Finally, stop and imagine what happens when someone who habitually thinks of the world in terms of scholastic categories reads a scientist's report. The scientist reports the mean of a distribution. The scholastic assumes that the mean is a property claimed to be the essence of a category; but, then, whoops an "are....except when they're not" pops up. The problem isn't in the variation. The problem is in the assumption that the mean is a property that defines a category. The disappointed scholastic turned nominalist says, "You see, it's all arbitrary; it's all social construction. The scientist replies, "No, it's all distributions. "Are....except when they're not" is predictable. Here is a possible explanation. Would that things were so simple. But as Frank Luntz says, "It isn't what you say, it's what they hear." And as long as people hear what the scientists have to say about distributions as claims about the essences of natural categories, the kind of debate illustrated in this thread will go on in pretty much the same fruitless way. -- John McCreery The Word Works, Ltd., Yokohama, JAPAN Tel. +81-45-314-9324 http://www.wordworks.jp/