Adrian Currie, Rock, Bone, and Ruin: An Optimist’s Guide to the Historical Sciences (Cambridge MA: MIT Press, 2018), 376 pp., ISBN: 9780262037266, $35.00, £27.00.

Some of the best scientific theories that had enjoyed universal acceptance, such as Newtonian classical mechanics, were falsified. The “pessimistic induction” from the history of science as Larry Laudan called it, may be that contemporary science will also suffer such fate.¹ Yet, some sciences have had a history that supports an “optimistic induction” because they had only one, first and last, founding scientific revolution, and have since progressed, expanded, and refined, without a second “Einstein-like” scientific revolution overthrowing the first. Scientists still believe in Darwinian evolution, random mutations and natural selection, though they know much more about the history of life than Darwin, and learned to use genetic evidence. Early 19th century historical philology that inferred language families and descent from information preserving correlations between languages is still valid. Lyell’s scientific revolution in geology was complemented, not replaced, by continental drift theory. Ranke’s methods for inferences from archival documents and his conclusions have been retained, even as historians expanded their scope of inferences from other primary sources such as material remains, the shapes of fields, econometric evidence, artistic representations, and so on. Interestingly, all these “optimistic sciences” are historical, they infer probable knowledge of the past.

Another optimistic aspect of the historical sciences is their progressiveness. Though the past cannot change, the historical sciences have been successful in inferring progressively more knowledge of it. In less than a century we have learned that the universe began in the Big Bang, the continents have been drifting, some dinosaurs had feathers and all became extinct following a meteor impact—one of five such episodes of biological mass extinction, the Neanderthals did not disappear but interbred with homo sapiens into extinction, Biblical era Israelites believed their Yahweh God had a female consort, and so on and on. Adrian Currie’s new book presents reasons for this progressive induction.

Physics assumes that information is never lost from the universe. An all-knowing Laplacian demon should be able to infer from a sufficiently detailed description of the contemporary universe and the laws of nature the lost books of the Pre-Socratic philosophers, the spiritual beliefs of Cro-Magnon cave dwellers, and what our seventh century ancestors had each day for breakfast, and when they went hungry. However, by the time information signals from the past reach the present they are corrupted, decayed and mixed with noise. Lesser epistemic agents than Laplacian demons cannot infer such interesting details. Arguably, historians cannot generate new evidence, but depend on the cruel censorship of time. This leads to pessimistic deduction about the progressive prospects of the historical sciences.

Currie set to undermine such pessimism from the bottom up, by analyzing fascinating inferences in paleontology, geology, and archaeology, from the inference of extinct giant platypus from a single fossilized tooth to the inference of frozen “snowball” earth 500,000,000 years ago from rock morphology. This meticulous analysis introduces to philosophy fascinating new case studies, though none from cosmology, human history, and historical linguistics.

Present traces infer past events. Traces may preserve information from the past, or are the premises in a conditional that infers the past on the basis of known probabilistic dependencies or causal regularities. Causal, conditional, and information preserving relations between a present and its past are distinct but not mutually exclusive. The present is not always useful for inferring history. For example, it is impossible to infer the perpetrators of a perfect crime.

Currie shows that evidence in the historical sciences is not confined to traces, but includes analogies, simulations, experiments, and speculative theoretical “scaffolds” that can direct fruitful research initially and be discarded later. Inferences from traces in the historical sciences have been discussed by philosophers like Sober, Kosso, Cleland, Turner, Jeffares, and this reviewer, among others.² Currie adds an interesting range of examples, but not much theory. He endorses a conditional dependence view of the relation between traces and history, but in the course of case analysis Currie is forced to resort to causal and informational analysis. He characterizes the theories that mediate between traces and history as “midrange,” though the theories cosmology uses to infer the Big Bang are grand, while the generalizations historians use are often local and narrow, for example the bureaucratic norms that governed forms of recording in archived documents. Currie does not attempt to fit his philosophy to these historical sciences, Cosmology and Human Historiography, but there is no philosophic reason for this limitation of the analysis.

Currie rejects models of probabilistic inference of common cause events from independent, otherwise unlikely correlated, and information preserving traces, such as biological homologies, grammatical mistakes in copied manuscripts, and historical testimonies, as characteristic of the historical sciences. But he does not engage much with the literature on the subject and mischaracterizes (138–148) the inference of common cause as a “unifying” explanation of patterns, whereas in fact the Bayesian inference is of hypotheses of low prior probability, high informational value, from the unlikelihood that independent units of evidence would agree on them by coincidence or as a result of separate causes. The independence of the evidential sources allows to infer the vanishing likelihood of their coherence given separate types of sources. For example, if two independent witnesses report that the same number won the lottery, the number had a low prior probability and their testimonies are information rich. The independent testimonies do not confirm the hypotheses nor allow it to unify them, rather they reduce to vanishing probability the hypothesis that the witnesses, by coincidence, stumbles on the same number because they wished to deceive or were confused. The independent testimonies may still be misleading because their common origin may not be the lottery draw, but some other common source, for example, a phone number of a common acquaintance.³ Detailed multiple independent testimonies to the same events operate like this in historiography and for this reason historians search for them assiduously.

Currie’s true main argument is that known traces are not the exclusive evidential foundation of the historical sciences. Coherence with preexisting probable knowledge affects the prior probabilities of historical hypotheses. Historical scientists discover new traces, and extract more information from known traces by developing or applying new theories. For example, correlations between teeth, jaws, cranium, and body size infer vast information from the discovery or analysis of a single fossilized tooth. The book ingeniously demonstrates this model by examining diverse case studies in natural historiography, in “deep history.”

Scientific experiments serve as “smoking guns” that choose between competing hypotheses. Currie shows that simulations and analogies can fulfil a similar function in the historical sciences. Archaeologists experiment with replicas of ancient tools and simulate construction methods. Computerized modeling of complex processes can generate surprising results, for example about how dinosaurs could and could not have moved. Analogies can serve as theoretical “scaffolding” that infer the history of a trace from the history of an analogous but different type of trace. For example, Martian historical geology and natural history may be inferred from analogies with terrestrial geology; the traits of extinct animals may be inferred from those of analogous similar living animals. Analogies are probable. It is likely that early Mars with atmosphere and flowing water resembled early earth and that dinosaurs resembled their avian descendants, at least in certain respects.

Philosophers, at least since the Neo-Kantians, have attempted to distinguish historical from other sciences. Currie demonstrates that use of controlled experiments cannot be the sought-after shibboleth. Perhaps the distinction is between theoretical sciences of types and historical sciences of tokens: The social sciences are interested in revolutions and economic recessions, while historiography is interested in the French Revolution and the 1929 Great Recession.⁴ Generative linguistics is interested in language; historical linguistics is interested in Sanskrit and Lithuanian. Epistemically, the historical sciences infer from traces that preserve reliably information signals from past token events. Currie demonstrates that the evidential basis of the historical sciences is not limited to traces. Yet, they are indispensable: No traces, no history.

This is an impressive debut monograph for Currie, who wrote a dissertation with Kim Sterelny at the Australian National University. I look forward to read more of his philosophy in the years to come. This is an optimistic induction from the quality of his current writings, but the frequencies of good quality arguments in his work permit it to be probable.