There Are Causes, and There Are Causes
Let's us say we are shooting cannon balls into the East River from Brooklyn. Suddenly, we see one of them soar off and land on the FDR Drive on the Manhattan side of the water. What happened?
"What caused that one ball to cross the East River?" you ask.
There are different possible answers, but they generally should take the form of offering a difference in the mechanical forces acting on that ball.
"Well," I might say, "I doubled the amount of powder I used for that shot," or perhaps "That ball only weighed half of what the others did," or something of the sort.
But let's say you see me driving across the Brooklyn Bridge. You turn to our mutual acquaitance, Sanford, and ask him, "What caused Gene to cross the East River today?"
Sanford answers, "Well, Gene wants to go to Pennsylvania today, and to do so, he must get off of Long Island somehow. He prefers the Brooklyn Bridge as it is toll free and relatively easy to access."
What Sanford offers is a reason, and not a set of mechanical forces. If he had answered, "Oh, you see, he is located inside a vehicle being impelled westward due to the numerous explosions occurring in its engine," you would deem him a wise ass.
The difference is crucial in terms of how one can go about testing one's answer. In the first case, let us call it "cause1," it makes perfect sense to demand the explanation be made more rigorous by repeatedly varying the conditions and seeing what happens. We try shots with lighter cannon balls and heavier, with more charge and less, and so on, and record what happens. If 99% of lighter balls make it across, but only 1% of the heavier balls, then we will endorse the answer, "The shot made it across because I used a lighter ball." (But what if the percentages are 51% and 49%?)
In the second case, for "cause2," such procedures are nonsensical. My decision to take a particular route may never be repeated under any conditions but only occur under precisely those of that moment, and yet the explanation may be as sound, indeed, more sound, than a cause1 explanation. Nor does this sort of explanation involve anything like naive belief in the agent's own statements. If Sanford offers the above explanation, but, as you watch me through binoculars, you see me get off the bridge onto local roads, park, meet a woman, kiss her, and walk arm-in-arm with her into an apartment building, and, after a few hours, come out and drive back to Brooklyn, you will logically conclude that my purported reason was just a cover story, and the real reason is that I am having an affair with someone in Manhattan. You might dig deeper to confirm your suspicion, but note what you don't need: You don't need to do a statistical study of all the times I cross the East River to see what "correlates" with such crossings. Perhaps the woman and I usually meet in Brooklyn, and that was the sole exception. That would not count against this explanation in the least.
The upshot of all this? History is not a statistical science, and would not be improved, but instead changed into a different thing, by enforcing statistical methods upon historians. Whatever worth statistical social science has, it is not a replacement for history, and its existence does not condemn history as "pre-scientific."
"What caused that one ball to cross the East River?" you ask.
There are different possible answers, but they generally should take the form of offering a difference in the mechanical forces acting on that ball.
"Well," I might say, "I doubled the amount of powder I used for that shot," or perhaps "That ball only weighed half of what the others did," or something of the sort.
But let's say you see me driving across the Brooklyn Bridge. You turn to our mutual acquaitance, Sanford, and ask him, "What caused Gene to cross the East River today?"
Sanford answers, "Well, Gene wants to go to Pennsylvania today, and to do so, he must get off of Long Island somehow. He prefers the Brooklyn Bridge as it is toll free and relatively easy to access."
What Sanford offers is a reason, and not a set of mechanical forces. If he had answered, "Oh, you see, he is located inside a vehicle being impelled westward due to the numerous explosions occurring in its engine," you would deem him a wise ass.
The difference is crucial in terms of how one can go about testing one's answer. In the first case, let us call it "cause1," it makes perfect sense to demand the explanation be made more rigorous by repeatedly varying the conditions and seeing what happens. We try shots with lighter cannon balls and heavier, with more charge and less, and so on, and record what happens. If 99% of lighter balls make it across, but only 1% of the heavier balls, then we will endorse the answer, "The shot made it across because I used a lighter ball." (But what if the percentages are 51% and 49%?)
In the second case, for "cause2," such procedures are nonsensical. My decision to take a particular route may never be repeated under any conditions but only occur under precisely those of that moment, and yet the explanation may be as sound, indeed, more sound, than a cause1 explanation. Nor does this sort of explanation involve anything like naive belief in the agent's own statements. If Sanford offers the above explanation, but, as you watch me through binoculars, you see me get off the bridge onto local roads, park, meet a woman, kiss her, and walk arm-in-arm with her into an apartment building, and, after a few hours, come out and drive back to Brooklyn, you will logically conclude that my purported reason was just a cover story, and the real reason is that I am having an affair with someone in Manhattan. You might dig deeper to confirm your suspicion, but note what you don't need: You don't need to do a statistical study of all the times I cross the East River to see what "correlates" with such crossings. Perhaps the woman and I usually meet in Brooklyn, and that was the sole exception. That would not count against this explanation in the least.
The upshot of all this? History is not a statistical science, and would not be improved, but instead changed into a different thing, by enforcing statistical methods upon historians. Whatever worth statistical social science has, it is not a replacement for history, and its existence does not condemn history as "pre-scientific."
I'm still not sure how distinct social science and history are/ought to be. Certainly there are some historical questions amenable to statistical evaluation, even if this one isn't. It seems to me the relation between history and social science is roughly comparable to that between natural history and natural science. Different to be sure, but not in quite as much conflict as these seem to imply. The boundaries strike me as a little blurrier.
ReplyDeleteIt shouldn't surprise us if the relation between history and social science is similar to that between natural history and natural science. After all - we are just smart apes. The whole natural/social distinction is the truly artificial distinction.
But - I'll admit - probably usefully artificial.
Daniel, I never said they were in conflict!
ReplyDeleteWith the "smart apes" remark, you are just trying to get my goat. The issue has nothing to do with any transcendental nature of humans, but with the distinction of explanation as a reason and explanation as pointing to a mechanical force.
As someone said at the Oakeshott conference, "If I push you hard enough, you fall over whether or not you understand what I am up to. But if I ask you to go to the store, you can't respond as I wish unless you understand me."
*That* distinction is fundamental, and has nothing to do with apeness or souls or anything of the sort. Apes, to, can only respond to our requests when they understand them.
Oh, and I also never claimed that statistics are never appropriate in history. No, the claim is that the explanation of why I crossed the river would not become more rigorous by somehow incorporating a regression analysis.
ReplyDeleteWhen is statistics appropriate in history?
ReplyDeleteIf...you see me...meet a woman, kiss her, and walk arm-in-arm with her into an apartment building, and, after a few hours, come out...you will logically conclude that my purported reason was just a cover story, and the real reason is that I am having an affair with someone in Manhattan.
ReplyDeleteActually Gene, if I saw you come out of the apartment after a few *minutes*, I might conclude you were having an affair. In the above scenario, I would conclude you were taking cooking lessons from your aunt.
I think you meant to say "a few seconds," Bob. Touche!
ReplyDeleteWell, Avram, let's say the question at hand is "To what extent did the Celtic residents of Britain adopt Roman architecture?" Then we might survey the known sites, determine what percentage of the buildings in those sites were built in Roman style, and establish some confidence interval as to what the overall percentage was across the island.
ReplyDeletere: " No, the claim is that the explanation of why I crossed the river would not become more rigorous by somehow incorporating a regression analysis."
ReplyDeleteRight - I'm thinking we're in more agreement than I originally thought based on comparisons you had been making with social science.