Are historical sciences somehow inferior to experimental ones? This attitude, sometimes referred to as "physics envy" (when adopted by non-physicists), is one of the most pernicious myths surrounding science and its practice. Everybody knows that physics is the queen of the sciences and that scientists in other disciplines are simply trying to catch up to the high standards set by Newton and his intellectual descendants. Right?

According to many practicing scientists (myself included), this is just a big load of dingo kidneys, and work done in the philosophy of science by Carol Cleland (e.g., Cleland 2002) has elegantly shown why. Cleland begins by acknowledging that there actually is a continuum between largely ahistorical disciplines--such as fundamental physics--and largely or exclusively historical ones--like paleontology or astronomy. This continuum passes through disciplines such as ecology and evolutionary biology, where one finds a mix of experimental and "detective" work, due to the fact that the objects of study (population of living organisms) do lend themselves to experimental manipulation but the results of such manipulations depend to a large degree on the past (often unknown) history of the objects themselves.

For Cleland, the fundamental distinction here is between sciences that attempt to predict the future and those that focus on "postdiction" of the past (for obviously, one cannot predict the past). A predictive science has to deal with what Cleland calls "the underdetermination of the future by the localized present," while a postdictive science is characterized by "the overdetermination of the past by the localized present." These are crucial concepts, which themselves rely on the asymmetrical arrow of time. Let us consider two illustrative examples from Cleland's paper.

Suppose we predict that a short circuit will cause a fire in a house. This may be a reasonable prediction, but it depends on a variety of other circumstances (co-occurring causes) in order to be true: not only does there have to be a short circuit, but the house has to have nonfunctional sprinklers, for example, or there has to be sufficient flammable material around, and so on. The prediction of the event, in other words, is underdetermined by the individual causes: one cause is not sufficient to guarantee the outcome. If this were a scientific hypothesis, we would have to be very specific about which other conditions ought to hold for the hypothesis to be verified (or rejected); just because the fire didn't occur, we are not authorized to reject the hypothesis that a short circuit causes a fire because the house also happened to have an efficient sprinkler system, without which the predicted outcome would, in fact, have happened. This is why, in physics and other experimental sciences, one has to specify the conditions of an experiment very carefully: without such precaution, too many other things might explain why the prediction was not accurate. The price for our ability to predict the future is that we can do so accurately only under very restrictive conditions; the more we relax such restrictions, the less we can falsify our hypotheses (because factors other than the focal one may have changed) and the less accurate the predictions become.

On the other hand, consider the other example discussed by Cleland: a baseball hitting a window. If you come home and see just a few pieces of glass scattered on the floor or the bill from a company that replaces windows or see a baseball on the floor that couldn't have gotten into the house in any other way because you had locked all entrances when you left, you would be able to infer the past event with a high degree of reliability. This is because the present situation overdetermines the past event; like Sherlock Holmes, you need surprisingly few clues to infer an amazing amount of detail, a situation that allows detectives, archaeologists, and astronomers to make a living.

The asymmetry of the two situations, then, lies in the fact that predictive sciences are attempting to go from causes to effects, while postdictive ones go from (some of) the effects to the likely causes. The irony of Cleland's analysis is that, because of the asymmetry in the determination of causes and effects, postdiction is actually much more powerful than prediction, in some sense turning on its head the classic view of historical sciences as "inferior." This is of course not a call to crown paleontology as the queen of the sciences, but rather a very healthy warning--based on a clever philosophical analysis--about the actual nature of scientific evidence.

Many sciences, as I mentioned at the beginning, in fact feature a complex mix of experimental and historical research, and as such, they may vary greatly in their effectiveness at prediction and postdiction of the phenomena of interest. What is still common to all science as a human activity aimed at understanding the natural world is that all scientific disciplines rely on some type of empirically based hypothesis testing. One may have to specify strict conditions for prediction, or abandon prediction entirely in favor of postdiction, but there have to be ways to test either the predictions or the postdictions by using empirically (i.e., either experimentally or observationally) obtained data. If one cannot do this, then one is not engaging in science--pace the proponents of so-called Intelligent Design theory and other such inanities.