Stud. Hist. Phil. Mod. Phys., Vol. 31, No. 1, pp.

109 } 116, 2000
( 2000 Published by Elsevier Science Ltd. All rights reserved
Printed in Great Britain
1355-2198/00 $ - see front matter


The Rise of Modern Probability Theory
S. L. Zabell*

Jan von Plato, Creating Modern Probability (Cambridge: Cambridge University
Press, 1994), x#323 pp., bibl., index, ISBN 0-521-44403-9, hardback, $69.95,
0-521-59735-8, paperback, published 1998, $18.95.

Jan von Plato's book Creating Modern Probability is an outstanding contri-
bution to the history of mathematical probability. Part of its importance stems
from the apparent paradox that although the period it covers is central to the
development of the subject, this same period has hitherto been among the most
neglected and least understood parts of it in historical studies. Thus, in order to
appreciate the importance and novel elements of von Plato's book, one must
"rst know something about the curious historiography of mathematical prob-
ability itself.
Points of origin are often arbitrary and capricious; citing a source as being the
"rst to discuss a given topic usually serves only to provoke others to perversely
unearth yet earlier instances. Nevertheless, one must start somewhere, and an
obvious starting point for historical studies of the subject is one by a master of
the subject: Pierre Simon, the Marquis de Laplace's concluding historical
chapter in his Essai philosophique sur les probabiliteH s of 1814 (the Essai in turn
being one of the earliest attempts at popular science). But it is an inauspicious
start: the chapter in the Essai is terse, cursory, and selective; it is certainly
incomplete (Laplace was severely criticised by some of his staunchest propo-
nents, such as De Morgan, for repeated failure to cite the earlier work of

(Received 24 June 1998; revised 4 June 1999)
* Departments of Mathematics and Statistics, Northwestern University, Evanston, IL 60208, U.S.A.

PII: S 1 3 5 5 - 2 1 9 8 ( 9 9 ) 0 0 0 2 9 - 5


so much admired but of whose life so little is known. the Marquis de Condorcet. the history of probability did not initially thrive. 1921. What a cast of characters! And no wonder many later historians never felt the need to go past them. First. perished in the irrational terror of the French Revolution. p. Todhunter was not himself a contributor to the "eld of mathematical probability. depuis ses origines jusqu1à nos jours being merely a turgid exception. and gravitational attraction. landing on his feet to not merely survive but #ourish in the aftermath of the Revolution. . spending his days in Slaughter's co!ee house. But surprisingly. geometry. and the di!erential and integral calculus. all this changed in 1865. 2 The often highly critical John Maynard Keynes had enormous respect for Todhunter. In the English language. and so on. 432). from University College London in 1842. James Bernoulli. Todhunter's remarkable book is distinctive from a number of very di!erent perspectives. however. nor even in any sense a research mathematician. one of England's "nest mathematicians. the ingenious but mysteri- ous Reverend Thomas Bayes. meditating on the consequences of his law of large numbers for twenty years. For from the start the history of the subject has always had a certain romance attached to it: Girolamo Cardano. the calculus of variations.A. he became a don at St John's and spent much of his time turning out popular textbooks on arithmetic. until he. like Lavoisier. algebra. the equally aristocratic but more politically agile Laplace. when Isaac Todhunter published his monumental History of Mathematical Probability from the ¹ime of Pascal to that of ¸aplace. the austere jansenist Blaise Pascal and the respectable jurist Pierre Fermat engaged in the reputable mathematical pursuit of analysing disreputable games of chance. Todhunter's list. going so far as to state in his ¹reatise on Probability that: &Of mathematical works published before the time of Laplace. his famous paper only seeing light of day because it was published posthumously by his friend Richard Price. the Hugenot refugee Ab- raham De Moivre. an enthusiastic advocate of the uses of probability in arriving at a more rational judicial system. beyond criticism' (Keynes. despite this rich load of material.1 Instead Todhunter is remembered today primarily for several technical his- tories: of mathematical probability. until he perished before he published. are complete and exact*a work of true learning. The result was a tremendously useful synopsis of more than a century and a half of mathematical research.110 Studies in History and Philosophy of Modern Physics contemporaries) and it austerely disdains to discuss the personalities that make the history of probability such a fascinating subject. and also his commentary and analysis.2 1 His mathematical career was industrious rather than brilliant: senior wrangler at Cambridge in 1848 after having received an earlier B. The History of Mathematical Probability in particular is an out- standing piece of scholarship: nothing less than a careful and systematic tech- nical review in chronological order of every book and paper on mathematical probability that Todhunter both thought relevant and could lay his hands on. Charles Gouraud's 1848 Histoire du calcul des probabiliteH s. the gambling scholar and mystic. until the Scottish con man Sir Alexander Cuming goaded him into tackling the central limit problem.

Maurice Kendall felt able to claim that Todhunter lacked an imitator or rival (1963. Stigler's ¹he History of Statistics (1986). who has composed the history of his own times without indulging the prejudices and passions which usually a!ect the mind of a contemporary' (Gibbon. 26). Porter's ¹he Rise of Statistical ¹hinking 1800}1900 (1986). and was responsible for stimulating widespread interest in the whole "eld. for the very "rst time. 1981). Ch. Theodore M. 1863. David's amusing 1962 book Games. Andrew Dale's History of Inverse Prob- ability (1991). Nor did the earlier period su!er neglect either: other scholars returned to it to conduct more detailed studies: Lorraine Daston's Classical Probability in the Enlightenment (1988). (For example. p. and for the most part his few successors did not go much beyond this point. then absent in the Western intellectual tradition. including. more specialised monographs have also appeared: for example. who died in 1756. The reason for this may perhaps in large measure be attributed to the appearance of one book. Hacking's beautiful but provocative book advanced the crypto-Foucaultian thesis that mathematical probability emerged precisely when it did because a special concept of evidence. and Anders Hald's A History of Probability and Statistics and ¹heir Applications before 1750 (1990). and Ian Hacking's ¹he ¹aming of Chance (1990). 3. serious attention now being paid to the post-Laplacian nineteenth century: notably Stephen M.3 In addition. the 1982}1983 research group on the probabilistic revolution sponsored by the Zentrum fuK r interdisziplinaK re Forschung in Bielefeld. Ian Hacking's ¹he Emergence of Probability. in 1975. 1979). after that one faced the sober prospect of an unaccompanied trek across the complex landscape of the later nineteenth century. Keith Michael Baker's beautiful and profound study Condorcet: From Natural Philosophy to Social Mathematics (1975). the next "fteen years saw a remarkable #owering of the subject. had just begun to emerge. for example. beginning with Laplace the mathematical prerequisites necessary to understand the mathematical litera- ture substantially increased.N. Although the speci"c thesis that Hacking advanced soon found critics (see. In addition to these important books.) This was in many ways understand- able: up to Laplace. (For example. Karl Pearson's previously unpublished University College lecture notes on the history of statistics (1978). a number of other. the fascinating correspondence between Markov and Chuprov (Ondar.) In any case. Vol. Garber and Zabell. Gods and Gambling ends by discussing Abraham De Moivre. Todhunter existed as an invaluable resource and guide. F. 205).The Rise of Modern Probability Theory 111 Todhunter had limited himself to the period up to Laplace. (Thus. and the unpublished manuscripts of the Marquis de Condorcet 3 One is reminded of Gibbon's famous comment regarding the end of the Roman history of Ammianus Marcellinus &It is not without the most sincere regret that I must now take leave of an accurate and faithful guide. as recently as four decades ago. . his erudite and provocative book placed the discipline of mathematical probability in the mainstream of historical studies. and of particular importance.) But today this picture has dramatically changed.

almost all historical attention has still been largely con"ned to the period prior to 1900. and Herman Weyl's equidistribution theorem and his philo- sophical views on causality. his focus and organisation is primarily on individuals rather than topics. Oscar Sheynin. func- tional analysis. chance. These strengths result both from von Plato's willingness to confront the technical material head-on. Second. natural limitations. for example. uncertainty. W. The book later returns to . and choices must be made. and many other distinguished scholars in this "eld. ED mile Borel and the mathematics of the denumerable (including the strong law of large numbers).) Von Plato attempts the formidable task of organising his material by dividing it into three concurrent streams. Not to speak. Indeterminism has become an integral part of quantum mechanics. Pierre Crepel. randomised double-blind clinical trials of medicine. F. A. his approach is deliberately internalist: &To the circumstances of the individuals and institutions that contributed to the creation of modern probability I have paid little atten- tion. focusing on David Hilbert and his sixth problem (the challenge to "nd an adequate axiomatic foundation for probability and physics). although there is some discussion and citation of the secondary literature. &Pathways to Modern Probability'. The "rst of these. to a lesser extent. in Chapter 2. perhaps for reasons similar to those that earlier caused historians of science to focus on the period up to Laplace. First. but also in von Plato's hands some obvious strengths. discusses the origins of modern measure theoretic math- ematical probability. It is characteristic of von Plato's approach that he focuses on a limited number of key "gures. group theory. is seen primarily through the eyes of Bruno de Finetti. and so forth. But despite this wealth of new material. Monte Carlo methods of applied mathematics. The result is that. Subjective probability. Edwards. Stephen Stigler. of course. and to take as his starting point the original sources. the technical mathematical barriers for this period are even more serious: the history of probability in the twentieth century is largely the history of measure theoretic probability (and. including some of the currently available archival materials. for example. Such an approach has. First. von Plato concentrates on the primary source materials themselves. the chapter on de Finetti is very likely the best survey currently available in English of de Finetti's contribu- tions to mathematical probability. Thus completeness in coverage is impossible. of the many beautiful papers that have come out in the last 25 years by Bernard Bru. Jan von Plato deals with this problem in three ways. Such detailed work has to wait for a general organisation of historical studies on probability' (p. and so on).112 Studies in History and Philosophy of Modern Physics (Bru and CreH pel. an extensive literature on de Finetti in Italian. of course. Ivo Schneider. ergodic theory. Eugene Seneta. stochastic models of theoretical population genetics. (There is. But perhaps even more important is the sheer volume and breadth of the material involved. Concepts of randomness. of course. 4). Andrew Dale. and risk pervade modern science and society so completely that they must be regarded as an essential characteristic of the twentieth century. Third. 1994). and presents a balanced discussion of both technical detail and philosophical perspective.

1894}1912. and the later di!ering probabilistic interpretations of quantum mechanics by SchroK dinger.The Rise of Modern Probability Theory 113 this topic in the chapter on Kolmogorov (who towards the end of his career advanced a theory of complexity for "nite sequences). not inappropriately. 1866. and Einstein. personal. and Heisenberg. and in part to our knowledge'. his concluding reference. an attempt was made to merge the two viewpoints by the use of the method of arbitrary functions.nspeakable in Quantum Mechanics (1987). John Venn. let alone an agreed synthesis. is to Bell's Speakable and . in the work of Poisson and Cournot.) But this limited if clever attempt to resolve the inconsistencies between two such rad- ically di!erent views of the nature of probability (discussed by von Plato in Chapter 5) largely disappeared from view after the 1930s. Gibbs. The "rst of these applications would not have surprised Laplace. referred to it as repre- senting the &Material view of Logic as opposed to the Formal or Conceptualist' (Venn. PoincareH had asked: how can the objective probability of the roulette wheel arise from our purely subjective ignorance of the initial conditions under which the wheel is spun? His answer is ingenious: regardless of the form of this distribution of subjective belief. who wrote at the begin- ning of his Essai: &Probability is relative in part to our ignorance. not decreased. or rational (perhaps representing preferences that are observed. or uniquely . (The argument presupposes that the outcomes are highly sensitive to the initial inputs. the two outcomes red and black are each expected to occur about half the time. The second focus of von Plato's book (the subject of Chapters 3 to 5) is the evolving relationship between probability and physics: the introduction of statistical modes of reasoning into the new discipline of statistical mechanics by Boltzmann. Bohr. whose ¸ogic of Chance (1866) was the "rst book in English to systemati- cally develop a purely frequentist theory of probability. and Heisenberg. To the contrary: the distinctions between di!ering concepts of probability have multi- plied. internally consistent. (To his many other references in this chapter I would add Thomas Kuhn's 1978 Black-Body ¹heory and the Quantum Discontinuity. Maxwell. But this dialectical clash between an idealist and a materialist perspective did not result in the twentieth century in a deH tente. The pragmatic use of the mathematics of chance in statistical mechanics is entirely consistent with such a subjectivist view of the nature of probability. Preface).) The distinction between the subjective and objective concepts of probability "rst arose in the nineteenth century. and it is not inaccurate in the foundations of probability to think of a subjective establishment challenged by an objective alternative. Probabilities today are not just epistemic or physical: epistemic probabilities can be psychological. Beginning with PoincareH . provided only that it is continuous in form. Bohr. but the use of probability in quantum mechanics * as ordinarily interpreted*insists on its purely objective nature (the absence of &hidden variables'). and the introduction of indetermin- ism into quantum mechanics by SchroK dinger. Von Plato's "fth chapter discusses the earlier views of Einstein. In the late nineteenth century the subjective viewpoint remained in the ascendant.

Kolmogorov inequality. One of the strengths of von Plato's book is his chapter on Kolmogorov. and as frequencies can be either in"nite or "nite. As von Plato discusses. The third and "nal part of von Plato's book discusses the establishment of mathematical probability as an independent coherent discipline. and commenting on the interaction between Kolmogorov and contemporaries such as Khinchin.) Kolmogorov's axiomatic mathematical approach. although beginning in the late 1930s he published a number of papers in French (the most important of these being his 1937 &La preH vision: ses lois 4 Shafer (1998) brie#y discusses the French intuitionist background to Kolmogorov. Tornier. contrasting his ideas with those of Hilbert.) Of equal value is von Plato's chapter on Bruno de Finetti. a succession of di!erent approaches (notably those of Copeland. is one of the true masters of the subject: an impressive number of its basic results are due to him. These three individuals nicely represent the diversity of viewpoints that one "nds in the twentieth century. Kolmogorov. but also a recognition of its limitations. was spectacularly successful. and elegantly exploited for maximum e!ect. like Laplace. and almost all measure-theoretic introductions to mathematical probability discuss in detail (among others) the Kolmogorov zero-one law. and Chapman}Kolmogorov equations. has been the least successful of the three. Kolmogorov remains little studied by historians of probability. every argument and technique being deftly. on the other hand. . but reading Kol- mogorov's later proof a!ords somewhat the same pleasure as listening to a symphonic masterpiece. Kol- mogorov three-series and extension theorems. the Russian language and the sheer di$culty of the material must both play a part here. and also either quantitative or qualitative: physical probabilities in contrast represent either propensities or frequencies. Popper. appears to have been entirely independent of this earlier thread.114 Studies in History and Philosophy of Modern Physics determined). Most of de Finetti's earlier papers appeared in Italian (von Plato's bibliography cites fourteen of these alone between the years 1926 and 1934).4 (Their proofs of the law of the iterated logarithm make an interesting contrast: Khinchin's earlier proof has a vigorous. Essential to his concept of a collective displaying a limiting relative frequency is the absence of choice subsequences having a di!ering limiting frequency. cleverly. Ville. Fisher's recourse in the 1950s to an intuitive concept of &recognisable subset'. the Italian probabilist who played a large role in the rise of the modern subjectivist theory of probability. focusing on the three di!ering approaches of Richard von Mises (objective and frequentist). although it represents an informal scienti"c view of probability widely held today. Wald. crude e!ectiveness and originality. (The statistician R. But despite this. and Church) led to a progressive clari"cation of the concept. Bruno de Finetti (subjective and epistemic) and Andrei Kolmogorov (eclectic. A. and if "nite correspond to sequences that are random either because of the process generating them or the patterns displayed. Reichenbach. measure theoretic. in formulating his own ideas regarding the nature of probability. The approach of von Mises. placing the latter's ideas in its proper intuitionistic perspective. and intuitionistic).

Presses Universitaires de France). (1988) Classical Probability in the Enlightenment (Princeton: Princeton Uni- versity Press). K. Savage. (1975) Condorect: From Natural Philosophy to Social Mathematics (Chicago: University of Chicago Press). and it would have been desirable to have had a detailed discussion placing Finetti's subjectivist approach within the broader context of the devel- opment of inductive logic. there still remains even today little discussion in English of de Finetti's earlier papers and initial philosophical views. J. and de Finetti had only a very limited in#uence on philosophers or statisticians until his work was discovered by Savage in the 1950s and given a much wider audience. A. P. In sum. few read the beautiful Italian language. S. No one book could provide a complete panorama of such a landscape. Good. Bell. psychology. Frank Ramsey. The chapter in von Plato's book on de Finetti goes far in providing a picture of the early de Finetti. References Baker. ¹extes Rares ou IneH dits (Paris: Institut Nationale d'ED tudes DeH mographique. J. The chapter in Creating Modern Probability on de Finetti clearly illustrates at once both the strengths and limitations of the approach von Plato takes in his book. Johnson. . conditioning on the past of an exchangeable sequence results in inductively evolving degrees of belief largely independent of the particular prior used to quantify initial belief. Its limitations are that the work of de Finetti did not occur in a vacuum (although of all the modern contributors to subjective probability. and sociology. and their gradual conceptual and technical assimilation into "elds as diverse as physics. and L. the topics discussed in this book represent a crucial but hitherto largely neglected subject in the history of probability: the simultaneous trans- formation of probability and statistics into modern mathematical disciplines. in a manner reminiscent of (and perhaps even in part inspired by) PoincareH 's method of arbitrary functions. but in Creating Modern Probability Jan von Plato provides us with a personal but extremely useful map of the terrain and a guide to many of its points of interest. part of the importance of de Finetti's representation theorem lies in its freeing the probabilistic explication of inductive logic from a need to provide a unique or nearly unique quanti"cation of degree of belief. (1994) Condorcet: ArithmeH tique Politique. Thus. economics. political science. Its obvious strengths have been described. L. M. he is probably the least linked to the others). B. (1987) Speakable and . ses sources subjectives'. given initially as a series of lectures at the Institut Henri PoincareH in Paris). and CreH pel. (1991) A History of Inverse Probability: From ¹homas Bayes to Karl Pearson (New York: Springer-Verlag). For example.nspeakable in Quantum Mechanics (Cambridge: Cam- bridge University Press). Daston. J. Bru. discussing the contributions of such important "gures as W. biology.The Rise of Modern Probability Theory 115 logiques. E. I. Alas. Dale. But despite this wider appreciation of his achievements. I.

Studies in Subjective Probability (New York: John Wiley. Dale. (1866) ¹he ¸ogic of Chance (London: Macmillan) (2nd edn. P. Garber. Porter. (1990) ¹he ¹aming of Chance (Cambridge: Cambridge University Press). (1937) &La PreH vision: Ses Lois Logiques. 516}524. (1863) ¹he History of the Decline and Fall of the Roman Empire (London: J. C. M. I. (1986) A History of Statistics: ¹he Measurement of . M.E. J. Ses Sources Subjectives'. Kyburg and H. M. Gibbon. Kuhn. Pearson. S. (1921) A ¹reatise on Probability (London: Macmillan). Smokler. Ondar. 204}205. L. (1981) ¹he Correspondence Between A. A. de Finetti. Annals of Probability 26. Gouraud. Virtue). Annales de l1Institut Henri PoincareH (Paris) 7. O. A. Reprinted 1987 (Chicago: University of Chicago Press). (1978) ¹he History of Statistics in the 17th and 18th Centuries (New York: Macmillan). T. S. (1848) Histoire du Calcul des ProbabiliteH s.E. B. (1814) Essai Philosophique sur les ProbabiliteH s (Paris: Courcier).ncertainty before 1900 (Cambridge. S. Hald. S. 1964). F. Kendall. Archive for History of Exact Sciences 21. (1990) A History of Probability and Statistics and ¹heir Applications before 1750 (New York: John Wiley). Hacking. (1986) ¹he Rise of Statistical ¹hinking 1800}1900 (Princeton: Princeton University Press). MA: Harvard University Press). Chuprov on the ¹heory of Mathematical Probability and Mathematical Statistics (New York: Springer-Verlag). Stigler. M. Shafer. 33}53. T. 1995). 1888). . (1978) Black-Body ¹heory and the Quantum Discontinuity.116 Studies in History and Philosophy of Modern Physics David. Pierre-Simon ¸aplace: Philosophical Essay on Probabilities (New York: Springer-Verlag. Laplace. (1979) &On the Emergence of Probability'. D. Depuis ses Origines Jusqu1à nos Jours (Paris: Auguste Durand). K. N. (1865) A History of Mathematical Probability from the ¹ime of Pascal to that of ¸aplace (London: Macmillan). Todhunter. 3rd edn. (1998) &Review of Creating Modern Probability by Jan von Plato'. Biometrika 50. 1}68. I. English translation by A. 1876. I. 1965 (New York: Chelsea). J. and Zabell. Hacking. 1894}1912 (Oxford: Oxford University Press). Gods and Gambling (New York: Hafner). (1962) Games. (1963) &Isaac Todhunter's History of the Mathematical Theory of Prob- ability'. A. I. Reprinted 1949. Reprinted in English translation in H. Kh. Venn. Keynes. (1975) ¹he Emergence of Probability (Cambridge: Cambridge University Press). G. Markov and A. E. S.