Charles Darwin as a statistical thinker.

Recently historians and philosophers of science have been interested in the role of statistics and probability in investigating population variation. The focus is typically on investigators applying statistics and probability to explain large scale phenomenon that arise out of the collective behavior of numerous and varied individuals. The case studies that inform this work come mostly from molecular physics and 20th century genetical versions of evolutionary theory. Charles Darwin's work on evolution is rarely mentioned in this context except to point out his shortcomings-he made evolutionary theory "ripe" for statistical investigations, but he was not a statistical thinker. But this is a mistake, Darwin was a statistical thinker. In this essay I describe two instances where Darwin utilized statistical methods to investigate evolution. In the light of these cases, we ought to revise our views about Darwin's scientific methodology, in particular, how he came to develop his ideas about evolution and about the nature of his "population thinking". Furthermore, Darwin's cases provide us with an expanded view about what constitutes "statistical thinking" in the biological sciences. In the examples we will find Darwin using statistical measures of type frequencies to detect large scale ensemble effects, confirm hypotheses by comparing between expected and observed averages, and applying the astronomer's law of error to explain evolutionary trends.

Natural Selection beyond Life? A Workshop Report.

Natural selection is commonly seen not just as an explanation for adaptive evolution, but as the inevitable consequence of "heritable variation in fitness among individuals". Although it remains embedded in biological concepts, such a formalisation makes it tempting to explore whether this precondition may be met not only in life as we know it, but also in other physical systems. This would imply that these systems are subject to natural selection and may perhaps be investigated in a biological framework, where properties are typically examined in light of their putative functions. Here we relate the major questions that were debated during a three-day workshop devoted to discussing whether natural selection may take place in non-living physical systems. We start this report with a brief overview of research fields dealing with "life-like" or "proto-biotic" systems, where mimicking evolution by natural selection in test tubes stands as a major objective. We contend the challenge may be as much conceptual as technical. Taking the problem from a physical angle, we then discuss the framework of dissipative structures. Although life is viewed in this context as a particular case within a larger ensemble of physical phenomena, this approach does not provide general principles from which natural selection can be derived. Turning back to evolutionary biology, we ask to what extent the most general formulations of the necessary conditions or signatures of natural selection may be applicable beyond biology. In our view, such a cross-disciplinary jump is impeded by reliance on individuality as a central yet implicit and loosely defined concept. Overall, these discussions thus lead us to conjecture that understanding, in physico-chemical terms, how individuality emerges and how it can be recognised, will be essential in the search for instances of evolution by natural selection outside of living systems.

Galton, reversion and the quincunx: The rise of statistical explanation.

Over the last six decades there has been a consistent trend in the philosophy literature to emphasize the role of causes in scientific explanation. The emphasis on causes even pervades discussions of non-causal explanations. For example, the concern of a recent paper by Marc Lange (2013b) is whether purported cases of statistical explanation are "really statistical" or really causal. Likewise, Michael Strevens (2011) argues that the main task of statistical idealizations is to distinguish between the causal factors that make a difference to the phenomenon to be explained and those that do not. But, the philosophy literature poorly reflects the history of the development of statistical explanation in the sciences. Francis Galton's (19th century) explanation for the laws of heredity is our case. Galton's statistical explanation was both innovative for his time and influential to our contemporary sciences. The key points to understanding Galton's statistical explanation for reversion is that it is autonomous from the real-world biological properties that make up an instance of reversion while still being approximately true of many real-world biological phenomena. Ours is an expanded discussion of ideas originated in Hacking (1990) and Sober (1980). We will articulate these features and compare our account with that of Lange and Strevens.

Under the influence of Malthus's law of population growth: Darwin eschews the statistical techniques of Aldolphe Quetelet.

Charles Darwin, James Clerk Maxwell, and Francis Galton were all aware, by various means, of Aldolphe Quetelet's pioneering work in statistics. Darwin, Maxwell, and Galton all had reason to be interested in Quetelet's work: they were all working on some instance of how large-scale regularities emerge from individual events that vary from one another; all were rejecting the divine interventionistic theories of their contemporaries; and Quetelet's techniques provided them with a way forward. Maxwell and Galton all explicitly endorse Quetelet's techniques in their work; Darwin does not incorporate any of the statistical ideas of Quetelet, although natural selection post-twentieth century synthesis has. Why not Darwin? My answer is that by the time Darwin encountered Malthus's law of excess reproduction he had all he needed to answer about large scale regularities in extinctions, speciation, and adaptation. He didn't need Quetelet.