… we are trying to account for uniqueness of detail that cannot, both by laws of probability and time’s arrow of irreversibility, occur together again.
Continuing through Wonderful Life: The Burgess Shale and the Nature of History by Stephen Jay Gould (1989):
… We talk about the “scientific method,” and instruct schoolchildren in this supposedly monolithic and maximally effective path to natural knowledge, as if a single formula could unlock all the multifarious secrets of empirical reality.
… These procedures are powerful, but they do not encompass all of nature’s variety. How should scientists operate when they must try to explain the results of history, those inordinately complex events that can occur but once in detailed glory? Many large domains of nature — cosmology, geology, and evolution among them — must be studied with the tools of history. The appropriate methods focus on narrative, not experiment as usually conceived.
The stereotype of the “scientific method” has no place for irreducible history. Nature’s laws are defined by their invariance in space and time. The techniques of controlled experiment, and reduction of natural complexity to a minimal set of general causes, presuppose that all times can be treated alike and adequately simulated in a laboratory.
… But the restricted techniques of the “scientific method” cannot get to the heart of this singular event involving creatures long dead on an earth with climates and continental positions markedly different from today’s. The resolution of history must be rooted in the reconstruction of past events themselves — in their own terms — based on narrative evidence of their own unique phenomena.
… Historical explanations are distinct from conventional experimental results in many ways. The issue of verification by repetition does not arise because we are trying to account for uniqueness of detail that cannot, both by laws of probability and time’s arrow of irreversibility, occur together again. We do not attempt to interpret the complex events of narrative by reducing them to simple consequences of natural law; historical events do not, of course, violate any general principles of matter and motion, but their occurrence lies in a realm of contingent detail.
[line break added] (The law of gravity tells us how an apple falls, but not why that apple fell at that moment, and why Newton happened to be sitting there, ripe for inspiration.) And the issue of prediction, a central ingredient in the stereotype, does not enter into a historical narrative. We can explain an event after it occurs, but contingency precludes its repetition, even from an identical starting point.
… But historical science is not worse, more restricted, or less capable of achieving firm conclusions because experiment, prediction, and subsumption under invariant laws of nature do not represent its usual working methods. The sciences of history use a different mode of explanation, rooted in the comparative and observational richness of our data. We cannot see a past event directly, but science is usually based on inference, not unvarnished observation (you don’t see electrons, gravity, or black holes either).
The firm requirement for all science — whether stereotypical or historical — lies in secure testability, not direct observation. We must be able to determine whether our hypotheses are definitely wrong or probably correct (we leave assertions of certainty to preachers and politicians). History’s richness drives us to different methods of testing, but testability is our criterion as well.
[line break added] We work with our strength of rich and diverse data recording the consequences of past events; we do not bewail our inability to see the past directly. We search for repeated pattern, shown by evidence so abundant and so diverse that no other coordinating interpretation could stand, even though any item taken separately, would not provide conclusive proof.