You are on page 1of 173

( ( (

_____
THE RISE OF ,

STATISTICAL THINKING

Theodore M. Porter

PRINCETON UNIVERSITY PRESS


-,.
( ( (

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

Porter, Theodore M., 1953-


The rise of statistical thinking, 1820-1900.

Includes index.
1. Mathematical statistics-History. I. Title.
QA276. 15.P67 1986 519. s'o9 85-43306
ISBN o-691-08416-5 (alk. paper)
Copyright© 1986 by Princeton University Press
Published by Princeton University Press,
41 William Street, Princeton, New Jersey o8540 This book is dedicated to my parents
In the United Kingdom:
Princeton University Press, Guildford, Surrey

ALL RIGHTS RESERVED

Library of Congress Cataloging in Publication Data will be


found on the last printed page of this book

This book has been composed in Electra

Clothbound editions of Princeton University Press books


are printed on acid-free paper, and binding materials
are chosen for strength and durability

Printed in the United States of America by


Princeton University Press
Princeton, New Jersey

(''

' (
CONTENTS

IX
List of Abbreviations
XI
Preface
Introduction 3

PART ONE
THE SociAL CALCULUS

Chapter 1. Statistics as Social Science


The Politics of Political Arithmetic
The Numbers of a Dynamic Society
Chapter 2. The Laws That Govern Chaos
Quetelet and the Numerical Regularities of Society
Liberal Politics and Statistical Laws
Chapter 3· From Nature's Urn to the Insurance Office

PART TWO
THE SUPREME LAW OF UNREASON

Chapter 4· The Errors of Art and Nature 93


100
Quetelet: Error and Variation
110
Chapter 5· Social Law and Natural Science
1 11
Molecules and Social Physics
Galton and the Reality of Variation 128

PART THREE
THE SCIENCE OF UNCERTAINTY 149

Chapter 6. Statistical Law and Human Freedom 151


The Opponents of Statistics 152
Statistics and Free Will 162
The Science of Diversity 171
Statistik: Between Nature and History 177
Chapter 7· Time's Arrow and Statistical Uncertainty in Physics
and Philosophy 193

VII
( (
---·· Contents.-.
Buckle's Laws and Maxwell's Demon 194
Boltzmann, Statistics, and Irreversibility 208 ABBREVIATIONS
Peirce's Rejection of Necessity 219

PART FOUR
POLYMATHY AND DISCIPLINE
231 AOB Annuaire de l'observatoire de Bruxelles.
AQP Adolphe Quetelet Papers, Bibliotheque nationale de Bel-
Chapter 8. The Mathematics of Statistics 233 gique, Brussels, Belgium.
Lexis's Index of Dispersion 240 ATBM Wilhelm Lexis, Abhandlungen zur Theorie der Bevolke-
Edgeworth: Mathematics and Economics 255 rungs- und Moralstatistik (Jena, 1903).
Chapter 9· The Roots of Biometrical Statistics Archive Archive for History of Exact Sciences.
Galton's Biometrical Analogies BAAS Reports of the Meetings of the British Association for the
Regression and Correlation Advancement of Science.
Pearson and Mathematical Biometry BCCS Bulletin de Ia commission centrale de statistique (of Bel-
Conclusion gium).
BJHS British Journal for the History of Science.
Index 321 Cournot A. A. Cournot, Exposition de la theorie des chances et des
probabilites (Paris, 184 3).
DSB Charles C. Gillispie, ed., Dictionary of Scientific Biog-
raphy (16 vols., New York, 1970-1980).
ESP Karl Pearson, Early Statistical Papers (Cambridge,
1948).
FGP Francis Galton Papers, the Library, University College,
London.
Galton Karl Pearson, The Life, Letters, and Labours of Francis
Galton (3 vols. in 4, Cambridge, 1914-1930).
HSPS Historical Studies in the Physical Sciences.
JAI Journal of the Anthropological Institute of Great Britain
and Ireland.
Jbb Jahrbiicher fiir Nationalokonomie und Statistik.
JCMP James Clerk Maxwell Papers, Cambridge University Li-
brary, Cambridge, England.
JHB Journal of the History of Biology.
JRSS Journal of the Royal Statistical Society (since 1887).
Journal of the Statistical Society (1873-1886).
Journal of the Statistical Society of London (1838-1872).
MacKenzie Donald MacKenzie, Statistics in Britain, 1865-1930:
The Social Construction of Scientific Knowledge (Edin-
burgh, 1981).
viii ix
( ( (
Abbreviations
Maxwell Lewis Campbell and William Garnett, TheLife oflarnes
Clerk Maxwell (London, 1882). PREFACE
NMAB Nouveaux memoires de /'academie royale des sciences et
belles-lettres de Bruxelles (after 1840, de Belgique).
Oeuvres Pierre Simon de Laplace, Oeuvres completes ( 14 vols., This study is a history of statistical thinking as it developed among social
Paris, 1878-1912). scientists, biologists, and physicists. Their aim was to extend the reach
Phil Mag London, Edinburgh and Dublin Philosophical Magazine of exact science into the social and biological domains by analyzing
and Journal of Science. large-scale phenomena quantitatively in terms of the collective behavior
Phil Trans Philosophical Transactions of the Royal Society of Lon- of numerous individuals. Those individuals were essentially unknow-
don. able, either because they were too small to be seen or because they were
PPE Francis Ysidro Edgeworth, Papers Relating to Political exceedingly numerous and diverse. Statistics hence served not just as an
Economy (3 vols., London, 192 5). adjunct to observation, but as the basis of theory, and the growth of sta-
PRJ Proceedings of the Meetings of the Members of the Royal tistical thinking accompanied the rise. of several new areas of science
Institution of Great Britain. during the nineteenth century. Most notable among these were demog-
Prob Rev Lorraine Daston, Michael Heidelberger, Lorenz Kri.iger, raphy and social statistics, statistical mechanics, and population ge-
eds., The Probabilistic Revolution, vol. 1, Ideas in His- netics. The study of variation within these fields was the direct source of
tory (Cambridge, Mass., 1986). modern mathematical statistics, which in fact grew out of the biological
PRSL Proceedings of the Royal Society of London. study of heredity around 1900.
PS Ludwig Boltzmann, Populiire Schriften (Leipzig1 1905). This year should mark a real advance in the appreciation and under-
SHSPI Egon S. Pearson and Maurice Kendall, eds., Studies in standing of statistical thinking and its place in the development of mod-
the History of Statistics and Probability, vol. 1 (Lon- ern thought, because five works on the history of probability and statis-
don, 1970). tics are currently in press or have recently appeared in print. The others
SHSP2 Maurice Kendall and R. L. Plackett, eds., Studies in the include books by Lorraine Daston, Ian Hacking, and Stephen Stigler,
History of Statistics and Probability, vol. 2 (New York, and a two-volume collection from a 1982-83 research project directed
1977). by Lorenz Kri.iger and Michael Heidelberger at the Zentrum fi.ir inter-
Stigler Stephen M. Stigler, The History of Statistics: The Meas- disziplinare Forschung of the University of Bielefeld, in which Profes-
urement of Uncertainty before 1900 (Cambridge, sors Daston, Hacking, and Stigler and I participated. Had I known how
Mass., 1986). much work was going on concerning the history of statistics and prob-
SP W. D. Niven, ed., Scientific Papers oflames Clerk Max- ability when I began my dissertation in 1979, I would have become
well (2 vols., Cambridge, Eng., 1890). timid and chosen another topic. As it happened, there is no problem.
TPSC Transactions of the Philosophical Society of Cambridge. These works are all quite different, reflecting the rich diversity of sources
WA Ludwig Boltzmann, Wissenschaftliche Abhandlungen, and effects of statistical thinking. I have nevertheless learned a great deal
Fritz Hasenohrl, ed. (3 vols., Leipzig, 1909). from my fellow probabilists, often in a form too subtle to be properly
ZGSW Zeitschrift fur die gesammte Staatswissenschaft. credited in footnotes. I must be content to offer my deep appreciation to
the ZiF, and to my colleagues there during that splendid year.
A number of people have read all or part of this manuscript in one
form or another and offered helpful suggestions or criticisms. These in-
clude Diane Campbell, I. B. Cohen, Lorraine Daston, Gerald Geison,
Charles Gillispie, Ian Hacking, Bruce Hitchner, Robert Horvath, Mark
X XI
( ( (
Pr-eface
Kac, Daniel Kevles, Thomas Kuhn, Mary Morgan, JoAnn Morse,
James Secord, John Servos, Stephen Stigler, Geoffrey Sutton, and Nor-
ton Wise. To them I am most grateful. I wish also to thank my teachers,
especially Harold Bacon, Robert Fox, Gerald Geison, Thomas Kuhn,
Paul Robinson, Carl Schorske, and John Servos. For professional guid-
ance, and also, as it happens, for much intellectual stimulation, I thank
also Daniel J. Kevles (who has recently published a book on a closely
related subject), Norton Wise, and Glenn and Rita Ricardo Campbell.
THE RISE OF STATISTICAL THINKING
By much my greatest scholarly debt is to Charles Gillispie, my graduate 1820-1900 '
advisor, who was extraordinarily generous with his time, and who never
relaxed his standards. This book has benefited mainly from his criticism,
but his humor and frequent encouragement were also greatly appreci-
ated.
Parts of this book have appeared in different form in Historical Studies
in the Physical Sciences, British Journal for the His tory of Science, and
in Lorraine Daston et a!., eds., The Probabilistic Revolution (Bradford
Books/MIT Press). For assistance in putting the manuscript into its pres-
ent form I thank Gail Peterson, who typed it, Robin Good, who checked
references for me, and M. Roy Harris, for computer translation. I thank
also the secretarial staff of the University of Virginia History Depart-
ment, especially Lottie McCauley and Ella Wood. This work was sup-
ported by a Princeton University fellowship and an Andrew Mellon
postdoctoral instructorship at the California Institute of Technology.
Acknowledgment is due to the Academic royale de Belgique, Cam-
bridge University Library, and the Library, University College London,
for permission to quote from the Quetelet, Maxwell, and Galton papers,
respectively.
Geoffrey Sutton and JoAnn Morse offered valuable friendship while
we were graduate students and since. Many of my ideas about the history
of science in general, and much else, are the result of informal discus-
sions with them. It is not easy to know how to thank my parents. Ded-
icating this book to them is no substitute, but will perhaps be accepted
as a token of much more. Diane Campbell has been one of my most
helpful critics, and also an ideal companion in the challenging task of
making two academic careers within one home.

THEODORE M. PORTER
19 September 1985

xii
(
(

INTRODUCTION

Statistics has become known in the twentieth century as the mathemat-


ical tool for analyzing experimental and observational data. Enshrined
by public policy as the only reliable basis for judgments as to the efficacy
of medical procedures or the safety of chemicals, and adopted by busi-
ness for such uses as industrial quality control, it is evidently among the
products of science whose influence on public and private life has been
most pervasive. Statistical analysis has also come to be seen in many sci-
entific disciplines as indispensable for drawing reliable conclusions from
empirical results. For some modern fields, such as quantitative genetics,
statistical mechanics, and the psychological field of intelligence testing,
statistical mathematics is inseparable from actual theory. Not since the
invention of the calculus, if ever, has a new field of mathematics found
so extensive a domain of application.
The statistical tools used in the modern sciences have virtually all
been worked out during the last century. The foundations of mathe-
matical statistics were laid between 1890 and 1930, and the principal
families of techniques for analyzing numerical data were established
during the same period. The achievement of the founders of mathe-
matical statistics-Pearson, Spearman, Yule, Cosset, Fisher, and
others-was formidable, and is justly venerated by their scientific heirs.
There is, however, another story to be told, of the background that made
this burst of statistical innovation possible. The development of statistics
was necessarily preceded by its invention. This was the contribution of
the nineteenth century, the culmination of a tradition of statistical
thinking that embraced writers from a variety of backgrounds working in
areas which were otherwise unconnected.
The invention of statistics was the recognition of a distinct and widely
applicable set of procedures based on mathematical probability for
studying mass phenomena. Statistics was and continues to be seen as es-
pecially valuable for uncovering causal relationships where the individ-
ual events are either concealed from view or are highly variable and sub-
ject to a host of influences. The identification of statistics as a category
of knowledge was first of all a scientific accomplishment, and not a
purely mathematical one. To be sure, the central role of probability the-
ory in the history as well as the logic of statistics is plain. But most of the

3
(
(
In traduction Introduction
probability mathematics needed by the founders of statistics had been includes a chapter on "the statistical view of nature. ')• Merz's phrase re-
available for almost a century, since the time of Laplace and Gauss. In- ferred not only to the mathematical techniques for analyzing data de-
deed, practical techniques for the use of such mathematics in the anal- rived from probability theory, but also, and principally, to the strategy
ysis of numerical data were worked out with great sophistication during of investigation derived form the numerical science of society and of
the early decades of the nineteenth century in the form of error theory, states. It was this science for which the term "statistics" had been
which was used widely in geodesy and observational astronomy. In ret- adapted early in the nineteenth century. Its practitioners were initially
rospect, the history of error theory seems to abound in precursors to the called "statists," and only late in the nineteenth century did they assume
principal accomplishments of mathematical statistics. If statistics were the title of statisticians.
just mathematics, the "anticipations" of the error theorists would leave \It is no accident that a quantitative, erJ1pirical social science of the
little basis for the claim that Quetelet, Lexis, and Calton were original nineteenth century should have given its name to a branch of applied
thinkers in this field. mathematics, for the contribution of "statists" to the style of reasoning
The identification of precursors, however, is almost always mislead- associated with the new mathematics was a fundamental one.\ To begin,
ing, and it is no less so here. Identical mathematical formulations must statists familiarized the scientific world and the educated public with the
yet be viewed as different if they are interpreted differently, especially use of aggregate numbers and mean values for studying an inherently
when their purpose is scientific and not purely mathematical. As Ste- variable object. Statistical writers persuaded their contemporaries that
phen Stigler shows in his new book, the effective use of probabilistic systems consisting of numerous autonomous individuals can be studied
techniques for estimating uncertainty in astronomy and geodesy was by at a higher level than that of the diverse atomic constituents. They
no means sufficient to enable social scientists to apply similar analysis to taught them that such systems could be presumed to generate large-scale
the problems of their disciplines. 1 Decades after a sophisticated theory order and regularity which would be virtually unaffected by the caprice
of errors had been developed, there r~mained the great problem of find- that seemed to prevail in the actions of individuals. Since significant
ing points of contact between the mathematical formulas of error theory changes in the state of the system would appear only as a consequence
and the scientific objects of the social and biological sciences, where of proportionately large causes, a science could be formulated using rel-
variation was genuine and important. Once Galton and Pearson had in ative frequencies as its elemental data.
some measure solved that problem, the formulations of the error theo- \Practicing statists, of course, did not identify their enterprise with any
rists could readily be seen to be applicable. In this new context, how- doctrine so abstract as this. Theirs was a world of progress and discon-
ever, the analysis of error had become something quite different, a tent, of surveys and census figures, into which advanced mathematics
method for studying the causes of variation, and not just for measuring and abstruse philosophy rarely intruded. They were, for the most part,
it. Only through their successful application to the refractory but rich reformers and bureaucrats. As nineteenth-century liberals, they were
problems of the social and biological sciences did the probabilistic tech- impressed by the power and dynamism of that complex entity, society,
niques of error analysis grow into the powerful and flexible method of and were pleased to find evidence that it exhibited a stability which
analysis that we know as mathematical statistics. seemed not to be dependent on the intermittent wisdom of governing
The study of variable mass phenomena had different origins from authorities. Hence they were delighted by the uniformity from year to
mathematical probability and error theory. It began instead with the de- year which was found to characterize not only natural events like births
velopment of numerical social science and the formation of what was and deaths, but also voluntary acts such as marriages and even seem-
ingly senseless and irrational phenomena like crime and suicide. From
regarded during the late nineteenth century as the characteristically sta-
this was born Adolphe Quetelet's doctrine of "statistical law," which
tistical viewpoint. John Theodore Merz recognized this viewpoint in his
held that these regularities would continue into the future because they
1904 History of European Thought in the Nineteenth Century,
'John Theodore Merz, A History of European Thought in the Nineteenth Century (1904-
' See Stigler. 1912;4vols., New York, 1965), vol. 2, pp. 548-626.

4 5
··---- lntroduetien Introduction
arose necessarily from an underlying stability of the "state of society." sition at a given time, the existence of variation among observations was
Using statistics, it seemed to be possible to uncover general truths about quite naturally interpreted as the product of error, and the error curve,
mass phenomena even though the causes of each individual action were accordingly, was conceived as descriptive of the imperfections of instru-
unknown and might be wholly inaccessible. ments and of the senses. The practical task of fitting curves of various
The doctrine that order is to be found in large numbers is the leitmotif sorts to astronomical observations, as well as the search for rigorous
of nineteenth-century statistical thinking. The regularity of crime, sui- foundations for the method of least squares, led to much sophisticated
cide, and marriage when considered in the mass was invoked repeatedly mathematical work; but since the object of the exercise was always to
to justify the application of statistical methods to problems in biology, manage or estimate error, there was little incentive to study variation for
physics, and economics by writers like. Francis Galton, James Clerk its own sake.
Maxwell, Ludwig Boltzmann, Wilhelm Lexis, and F. Y. Edgeworth. A major transition in thinking about the error law was initiated by the
Indeed, the use of probability relationships to model real variation in Belgi-an Adolphe Quetelet. Quetelet journeyed to Paris in 1823 to learn
natural phenomena was initially made possible by the recognition of observational astronomy, but while there he was introduced to mathe-
analogies between the objects of these sciences and those of social sta- matical probability and was infected by the belief in its universal appli-
tistics. We find them not only in veiled allusions; they are openly and cability, as championed by Laplace, Poisson, and Joseph Fourier. The
explicitly developed in both popular and technical writings. new social science of statistics became for him a branch of the "social
Thus this book is, in one.sense, a history of the influence of ideas de- physics," patterned closely on celestial physics, for which he wished to
veloped within social statistics. It is a study of the mathematical expres- lay the foundation. Every possible concept from physics was given a so-
sion of what Ernst Mayr calls population thinking, 3 a phrase which cial analogue in this gradualist metaphysic of society, and the error law
points no less clearly than statistics to sources in the human sciences. As finally found its place in 1844 as the formula governing deviations from
in all noteworthy cases of intellectual influence, however, the benefi- an idealized "average man." Quetelet interpreted the applicability of
ciaries here were no mere passive recipients of social-scientific dogma. this law as confirmation that human variability was fundamentally er-
The leading characters in this story were "moral statisticians," econo- ror, but the effect of his discovery was to begin the process by which the
mists, kinetic gas theorists, and biometricians. The objects of their work error law became a distribution formula, governing variation which was
required them to find some way of studying mass phenomena profitably itself seen to have far greater interest than any mere mean value.
without first having to attain detailed knowledge of the constituent in- The further development of statistical mathematics was, until the end
dividuals. They were successful precisely because they were able to of the nineteenth century, largely the result of work in other natural and
adapt existing methods and concepts to new objects. In doing so, they social scientific disciplines. Quetelet's belief in the widespread applica-
contributed as much to the statistical method as to their particular fields. bility of the error law to variation of all sorts, although not his interpre-
Just as statistical reasoning was closely associated with the idea of tation of it, won acceptance by the ablest workers on statistical mathe-
large-scale regularity during the nineteenth century, the history of sta- matics of the late nineteenth century. James Clerk Maxwell, who
tistical mathematics before 1890 is, for the most part, a history of the learned of Quetelet's use of the error law from an essay review of one of
normal or Gaussian distribution. This is the familiar bell-shaped curve, Quetelet's books by John Herschel, proposed that the same formula gov-
known to nineteenth-century writers as the astronomical error law. Al- erned the distribution of molecular velocities in a gas. He, along with
though it had earlier been used in connection with the classical "doc- Ludwig Boltzmann, made of statistical gas theory one of the great
trine of chances," it became closely associated with astronomy as a con- achievements of late nineteenth-century physics and formed an impor-
sequence of its incorporation into the method of least squares for tant part of the background to the new quantum theory as well as to Wil-
reducing astronomical observations. Since stellar objects have a real po- lard Gibbs's statistical mechanics. In social science, Wilhelm Lexis used
3 Ernst Mayr, The Growth of Biological Thought: Diversity, Evolution, and Inheritance similar formalism to provide a measure of the stability of statistical se-
(Cambridge, Mass., 1982), pp. 45-47. ries, and Francis Edgeworth showed how error analysis and related tech-
6 7
(
(
---~ Introciucflon
Introduction
niques might be applied fruitfully to problems in economics such as in- oneers of statistical thinking were widely read generalists, interested in
dex numbers. Francis Galton, who was introduced to Quetelet's ideas historical, philosophical, or soCial issues as well as in their research
by the geographer William Spottiswoode, employed the error curve in areas. As Karl Pearson himself pointed out in some lectures he gave on
his study of heredity and, in the end, found that his index of hereditary the history of statistics:
regression was in fact applicable as a general tool of statistical mathe-
matics to the study of variation in data of all sorts. I ought, perhaps to apologise for carrying you so far afield in these
At the same time as the mathematics of statistics was being advanced lectures. But it is impossible to understand a man's work unless you
through its application to new objects, the statistical approach began to understand something of his environment. And his environment
be seen as distinctive and even as challenging for conventional views of means the state of affairs social and political of his own age. You
science and of natural law. Quetelet's contemporaries held no uniform might think it possible to write a history of science in the 19th cen-
view as to the nature of statistical science, but they were all in accord that tury and not touch theology or politics. I gravely doubt whether you
their discipline consisted of the application of the tried and true method could come down to its actual foundations, without thinking of
of natural science to a social object. The assertion of what may be called Clifford and Du Bois Reymond and Huxley from the standpoint of
statistical determinism in relation to man and society, however, pro- theology and politics. What more removed from those fields than
voked a backlash of opposition to statistics and a critical analysis of the the subject of differential equations? What more removed from
nature of statistical reasoning. On one level, this reassessment led to morality than the theory of Singular Solutions? Yet you would not
wide adoption of the view that statistics could provide little of scientific grasp the work of De Saint-Venant or Boussinesq unless you real-
worth so long as attention was focused on mean values rather than on ised that they viewed Singular Solutions as the great solution of the
variation. More abstractly, critics of the idea of statistical law put forward problem of Freewill, and I hold a letter of Clerk-Maxwell in which
by Quetelet and the historian Henry Thomas Buckle began to argue that he states that their work on Singular Solutions is epoch making on
the statistical method was inherently an imperfect one, applicable pre- this very account!4
cisely because the remoteness or intrinsic variability of the constituent
Many contributed to a great range of scientific specialties, and all were
objects rendered exact deterministic knowledge inaccessible. The elab-
alert to developments outside their own fields. The general development
oration of this view, especially by kinetic theorist and social thinkers, in-
of a statistical method required effective communication between di-
augurated what Ian Hacking calls the "erosion of determinism," which
verse studies, and this investigation reveals much about the links that
has profoundly influenced the scientific world view of the twentieth cen-
have given science a certain measure of unity, making it more than a
tury. collection of isolated disciplines.
The modern field of mathematical statistics arose from the diverse ap-
Statistics has been prominent not only on the margins between dis-
plications to which statistical ideas and methods were put during the
ciplines, but also in the nebulous and shifting border region that sepa-
nineteenth century. It became, under Pearson, Fisher, and others, a
rates science from nonscience. Statistics has contributed essentially to a
mathematical resource for a variety of disciplines for which numerical
considerable expansion of the scientific domain, but it has been for two
data could be obtained through experiment or observation. Before 1890,
centuries or more a singularly problematical method of science. Prob-
however, and indeed for some decades thereafter, statistical methods
ability was a suspect area of mathematics almost from the beginning,
and concepts were developed not by mathematicians but by astrono-
though from a purely technical standpoint its accomplishments were al-
mers, social scientists, biologists, and physicists. The development of
ready impressive in the time of Montmort and De Moivre. Influential
statistical thinking was a truly interdisciplinary phenomenon for which
writers on error theory, such as Augustin Cauchy and James Ivory, ac-
mathematics had no priority of position; new ideas and approaches arose
4 Karl Pearson, The History of Statistics in the 17th and 18th Centuries Against the Chang-
as a result of the application of techniques borrowed from one or more
ing Background oflntellectual, Scientific, and Religious Thought, E. S. Pearson, ed. (London,
disciplines to the very different subject matter of another. The great pi- 1978), p. 360.

8 9
( ( ~ .

---·-Introduction Introduction
cepted the method of least squares but refused io rest content with the been applied and the context of ideologies and philosophical attitudes
probabilistic assumptions in its underpinnings. s Laplace himself was in- within which statistics had been pursued. Mathematics and physics
spired by his ambitions for probability to redefine scientific certainty- may, as the Comtean hierarchy of the sciences suggests, have logical
to limit perfect knowledge to an omniscient but imaginary demon, and primacy over biological and social science, but historically the situation
to insist that while events in the world are completely determined by is much more complex and interesting.
preexisting causes, our knowledge of their outcome is necessarily subject There are, to be sure, special reasons why statistical theory should be
to a certain domain of error. The statistical approach also presented se- more powerfully affected by considerations external to its specific sub-
rious problems for the kinetic gas theory, which reduced the determin- ject matter than other areas of mathematics or science. Its main task,
istic laws of thermodynamics to mere regularities. While Maxwell and after all, has been to provide analytical methods by which practitioners
Boltzmann labored to incorporate the necessary refinements into the ki- of other disciplines can analyze their numerical data, and the statisti-
netic theory, others, such as Planck's student Zermelo, argued that this cian's quest is as much for useful techniques as for timeless truths.
connection with statistics invalidated atomism altogether. 6 Nevertheless, the history of statistics should not be seen in this respect
That probability seemed to imply uncertainty clearly discouraged its as utterly unique, but as an ideal type for one aspect of the historical
use by physicists. From the standpoint of social science, on the other process through which modern science has evolved.
hand, statistical method was synonymous with quantification, and
while some were skeptical of the appropriateness of mathematics as a
tool of sociology, many more viewed it as the key to exactitude and sci- Some changes of terminology almost always accompany the emergence
entific certainty. Most statistical enthusiasts simply ignored the depend- of new areas of science, especially when, as in the present case, a sig-
ence of statistical reasoning on probability, and those who acknowl- nificantly new style of thought is involved. A preliminary discussion of
some key terms ought therefore to be helpful.
edged it generally stressed the ties between probability and that most
ancient and dignified among the exact sciences, astronomy. The social \"Statistics" as a plural means to us simply numbers, or more partic-
science of statistics, in the hands of Quetelet and his admirers, consti- ularly, numbers of things, and there is no acceptable synonym. That
tuted a self-conscious attempt to imitate the successful strategy of natural usage became standard during the 183os and 184os. It seems almost im-
possible now to talk about such numbers and numerical tables published
science. Statistical quantification in social science was more commonly
before that time without using this anachronistic term. That all gener-
seen as exemplary than as problematical, and the aspirations of statistics
ations previous to the 182os managed to get by without it reveals dimly
reveal much about what were taken to be the essential features of science
how different was the world they lived in-a world without suicide rates
during the nineteenth century.
unemployment figures, and intelligence quotients. To be sure, thi~
Finally, the history of statistics sheds light on the relations between
prenumerate age was not entirely deprived of statistical tables, but the
abstract science and what are often seen as its applications. In truth,
great explosion of numbers that made the term statistics indispensable
practice was decidedly ahead of theory during the early history of statis-
occurred during the 182os and 183os. The demands it placed on people
tics, and "pure" or abstract statistics was the offspring, not the parent, of
to classify things so that they could be counted and placed in an appro-
its applications. The statistical techniques and approaches that were in-
priate box on some official table, and more generally its impact on the
vented by Lexis, Edgeworth, Galton, Pearson, and their successors re·
character of the information people need to possess before they feel they
fleeted at once the particular problems to which statistical methods had
understand something, are of the greatest interest and importance.
s James Ivory, "On the Method of Least Squares," Phil Mag, 65 (1825), 1-10, 81-88, 16!- In the nineteenth century, statistics designated an empirical, usually
168. On Cauchy, see lvo Schneider, "Laplace and the Consequences: The Status of Proba- quantitative, social science. Before that, it was an ill-defined science of
bility Calculus in the 19th Century," in Prob Rev.
6 Ernst Zermelo, "Uber mechanische Erklarungen irreversibler Vorgange," Annalen der states and conditions. The term only came to be applied commonly to
Physik, 59 (1896), 793-801. a field of applied mathematics in the twentieth century. "Statistics"
10 11
(
--Introduction· Introduction
gained a wide meaning, by which it could be used to refer to mass phe~ larity of statistical aggregates. More recently, Poisson's phrase has been
nomena of any sort, mainly by analogy with its social object. After the used to denote the rule that errors of mean values conform to the normal
mid-nineteenth century, it became common to investigate collective distribution. I know of nobody who used the phrase in this way before
phenomena using what came to be called the statistical method, the Emile Dormoy in 1874, whose work is mentioned in chapter 8.
method of reasoning about events in large numbers without being trou- There are, incidentally, a number of terms for the normal distribu-
bled by the intractability of individuals. tion, but, fortunately, there is little opportunity for confusion. The
The forms of "probability" are less troublesome for the nineteenth standard nineteenth-century phrases were "error curve" and "error law";
century, though probability did not come into its own as a branch of the eponymous "Gaussian" became common in the late nineteenth
pure mathematics until quite recently. Laplace's substitution of "cal- century, and "normal law" was coined by Pearson in 1894, as was
culus of probabilities" for the traditional "doctrine of chances" at the "standard deviation." The nineteenth-century measure of the width of
end of the eighteenth century was intended to make clear that rational a distribution was the "probable error," the magnitude of error which
belief or expectation rather than the outcome of games of chance was its precisely one-half of the measurements would, over the long run, ex-
proper object. Indeed, this view lingered on, but the mathematics of ceed. These few terms, fortunately, nearly exhaust the technical vocab-
chance was almost unfailingly referred to as probability or calculus of ulary of nineteenth-ceniury statistical thinking, apart from error theory
probability after Laplace, and I see no reason to depart from the modern and insurance mathematics. The technical and mathematical content
usage. in the following chapters is minimal, and, except for a few pages here
"Determinism" was until the mid-nineteenth century a theory of the and there, should be readily comprehensible even to readers with no
will-a denial of human freedom-and some dictionaries still give this mathematical training.
as its first meaning. Partly as a result of statistical discussion, it assumed
during the 185os and 186os its modern and more general meaning, by
which the future of the world is held to be wholly determined by its pres-
ent configuration. It differs from fatalism in that it rests on natural laws
of cause and effect rather than on some transcendant force. Its opposite,
we may note, underwent a similar change. "Indeterminism" now refers
to the view that some events in the world are not wholly determined by
natural causes, but are, at least to some extent, irreducibly random or
stochastic. Indeterminism may be contrasted with probabilism, which
implies simply that our knowledge does not permit perfect prediction,
though there may be no exceptions to complete causality in the world.
The phrase "law of large numbers" was coined by Poisson in 1835.
To him, it referred to the proposition that the frequencies of events
-
must, over the long run, conform to the mean of their probabilities
when those probabilities fluctuate randomly around some fixed, under-
lying value. Virtually everyone else who recited the phrase in the nine-
teenth century made no distinction between Poisson's theorem and the
one in Jakob Bernoulli's 1709 Ars Conjectandi, according to which the
frequency of events must conform over the long run to the fixed prob-
ability governing each trial. It is most convenient here to adopt this un-
discriminating usage, which really expresses simply the observed regu-
12
( (

PART ONE
( ( (

THE SOCIAL CALCULUS

The modern periodic census was introduced in the most advanced states
of Europe and America around the beginning of the nineteenth cen-
tury, and spread over much of the world in subsequent years. Records
of population, health, and related matters, however, had been collected
intermittently in a variety of territories since ancient times. Most often,
the chief purpose of this statistical activity has been the promotion ofbu-
· th' but reaucratic efficiency. Without detailed records, centralized administra-
"In this life we want nothing but Facts, m, no mg .
Thomas Gradgnnd tion is almost inconceivable, and numerical tabulation has long been
Facts."
recognized as an especially convenient form for certain kinds of infor-
"So many people are employed in situations of trust; so mation. Until about 18oo, the growing movement to investigate these
of so many will be dishonest. I have
many peop1e, Out ' . . 1 H numbers in the spirit of the new natural philosophy was likewise justified
heard you talk, a hundred times, of Its bemg ~ aw. .ow. as a strategy for consolidating and rationalizing state power.
, Tom Gradgnnd (Jumor)
can 1 h e1p laws.7 The importance of bureaucratic efficiency was by no means forgotten
-CHARLES DICKENS, Hard Times (1854) when political arithmetic gave way to the new social science of statistics
at the beginning of the nineteenth century. Increasingly, however, sta-
Humanity is regarded as a sort of volume of Genna~ tistical writers became persuaded that society was far more than a passive
. of the kind described by Carlyle, and aU It recipient of legislative initiatives. Always dynamic, often recalcitrant,
memoirS, . . d X -ANONYMOUS (1866) society evidently possessed considerable autonomy, and had to be
wants ... IS an m e ·
understood before the aims of the state could be put into effect. Statistics
I can scarcely conceive any more wholesome study. f~r a was for the most part a liberal enterprise, pursued by business and profes-
. T ry than the examination of statistics, sional men who favored a narrower definition of the function of the state
stern unb en d mg o . bl 1
for he cannot fail to recognize the grand mecusa e aw, even while working to enlist it in some particular reform. The most fer-
as true in politics as in everything else, that ~ovement vent advocates of systematic bureaucratic expansion still conceded that
must always be progressive and never retrogressiVe. 8 ) the state could act successfully only within the constraints defined by the
-L. L. PRICE (ca. 18 3 nature of society.
Hence, statistical authors of the scientific persuasion set themselves
to uncover the principles that governed society, both in its present con-
dition and, especially, as a historical object. The concept of "statistical
law" was first presented to the world around 1830 as an early result of
this search. As a social truth it was propagated widely and refined or dis-
puted by decades of writers. Shortly afterwards, statistical regularity
came to be seen as the basis for a new understanding of probability, the
frequency interpretation, which facilitated its application to real events
in nature as well as society. The idea of statistical regularity was thus of
signal importance for the mathematical development of statistics.
17
'JffK.•
iiP 77 :. . (
( /
\

Statistics as Social Science


ray, head of the "Knights and Burgesses" that "sit in the Parliament of
---------Chapter O n e - - - - - - - - Nature,"> as well as to the high official, Lord John Roberts, who might
be moved to execute some of its recommendations.
William Petty, who invented the phrase "political arithmetic" and is
STATISTICS AS SOCIAL SCIENCE thought by many to have had a hand in the composition of Graunt's
work, was in full accord with his friend as to the purpose of these studies.
Political arithmetic was, in his view, the application of Baconian prin-
ciples to the art of government. Bacon, he wrote, had drawn "a judi-
THE POLITICS OF POLITICAL ARITHMETIC cious Parallel ...·between the Body Natural and Body Politick," and it
was evident "that to practice upon the Politick, without knowing the
The systematic study of social numbers in the spirit of natural philoso- Symmetry, Fabrick, and Proportion of it, is as casual as the practice of
. d duri'ng the 166os and was known for about a century
ph y was p10neere ' d h Old-women and Empyricks."4 Petty sought always to bring "puzling
and a half as political arithmetic. Its purpose, when not c~nfine to t de and perplext Matters ... to Terms of Number, Weight and Measure,"s
calculation of insurance or annuity rates, was the p:om?tion of sou~ , so that official policy might be grounded in an understanding of the land
ll-informed state policy. John Graunt observed m his pathbreakmg and its inhabitants.
~;bservation upon the Bills of Mortality" of 1662: Implicit in the use by political arithmeticians of social numbers was
That whereas the Art of Governing, and the tru~ Politicks, is how the belief that the wealth and strength of the state depended strongly on
to preserve the Subject in Peace and Plenty; that men study only the number and character of its subjects. Accordingly, the sovereign was
that part of it which teacheth how to s~pplant and ~ve:-reach one frequently enjoined to take measures to protect the lives or health of his
another, and how, not by fair out-runmng but by tnppmg up each subjects. For example, Petty pointed out that considerable expenditures
other's heels, to win the Prize. . were required to raise a man or woman to maturity, and he advised the
Now, the Foundation or Elements of this honest harml.ess Pol- king that money spent to combat the plague could bring higher rewards
icy is to understand the Land, and the Hands of the. Terntor~, to than the most lucrative of investments, since it would preserve part of
be governed according to all their intrinsick and accidental differ- the vastly greater sum wrapped up in the lives of those who might other-
ences. • wise perish. Still, it was the standpoint of the sovereign that was almost
always assumed, and the members of society typically appeared as ob-
Graunt's scholarly claims were modest. "I hope," he ~egan, that readers jects that could and should be manipulated at will. Petty's philosophy
"will not expect from me, not professing Letters, thmgs d~mo~strate? appeared sometimes as authoritarian as that of his old master, Thomas
with the same certainty' wherewith Learned men determme. m then Hobbes. In his Treatise o{Ireland, Petty proposed that all Irishmen, save
Schools· but will take it well, that I should offer at a new thmg, and
a few cowherds, should be forceably transported to England, for since
could f;rbear presuming to meddle where any of the Learned Pens have
the value of an English life far surpassed that of an Irish one, the wealth
ever touched before, and that I have taken the pains, and been at the
of the kingdom would thereby be greatly augmented. 6
charge of setting out those Tables, whereby all men may ~oth correct my
Petty envisioned political arithmetic as embracing a variety of
Positions, and raise others of their own.,, Still, ~e desned to conduct
schemes through which number and calculation could be applied to
his inquiry philosophically, and his work was dedicated to Robert Me-
Observations upon the Bills of Mortality (1676; 1St ed., 1662), repr. in C. ' Ibid., p. 325· Moray was president of the newly created Royal Society of London.

. 396. On po11bca1 ant me 1c, see


~~r :;:u;~~ Economic.V~n'tin~shof~~r Wil!i~:!~~cr ~~~~~~:~~~~~~~~~·P~l~~~~i
William Petty, "The Political Anatomy of Ireland" (1691), in Economic Writings (n. 1),
4
H.'
vol 2 pp •
vol. 1, p. 129.
395 8 s Petty, "A Treatise oflreland" (1687), in ibid., vol. 2, p. 554·
Arifu~eti~: Civil Strife and Vital Statistics," Isis, 68 (1977), 67- 4· 6 Ibid., pp. 555-561. Seealsovol. 1, pp. 108-109.
> Graunt, Observations (n. 1), P· 334·

I9
/
( ( \

Statistics as Social Science


Stallstfcs as Socia!Science
Similarly, presumed influences on population were routinely in-
subjects. The Irish resettlement project, while exceptionall~ high-
voked to attack customs, creeds, and circumstances of every description.
handed was not uncharacteristic methodologically. Another Illustra-
Cities almost always showed a surplus of deaths over births, obviously
tion of Petty's computational rationalism is provided by his calculation
attributable to the idleness, luxury, and corruption that they nourished,
of the number of clergy requisite for the English people, based solely on
and were the bane of political arithmeticians. Thomas Short called
the land area. Petty's successors made political arithmetic a ~ore sober
them ''Golgothas, or Places of the Waste and Destruction of Mankind
discipline, closely tied to the empirical collection of population records
but seldom of their Increase, and often least Prolific ... ,, British and Ger~
and especially to the preparation of accurate life tables for the purpose
man authors rarely failed to castigate the Catholic Church, whose ad-
of calculating insurance and annuity rates.
Even so, political arithmetic continued to inspire vigorous co~tro­ herence to priestly celibacy was blamed for the alleged depopulation of
~sy, for eighteenth-century mercantilists generally_ re~arded either popish lands. "This >uperstitious and dangerous tenet," wrote Robert
population size or its rate of growth as the supreme_cntenon of a pros- Wallace, "most justly deserves to be esteemed a doctrine of those devils
perous and well-governed country. Thus Montesq_meu employed a du- who are the seducers and destroyers of mankind, and is very suitable t~
bious comparison of ancient and modem populatwns to affirm the su- the views and designs of a church, which has discovered such an enor-
periority of ancient morals, 7 while David Hume observed that, after mous ambition, and made such havock of the human race, in order to
taking account of climate and geography, "it seems natural t~ exp~ct t_hat raise, establish and preserve an usurped and tyrannical power. " 12 Al-
wherever there are most happiness and virtue, and the Wisest mshtu- cohol, gaming, promiscuity, and bad air were likewise condemned by
tions there will also be most people. " 8 Jean-Jacques Rousseau exalted the political arithmetician as moralist.
the ;numerator as the preeminent judge of policy: "Th~ gov_er~me~t The most ambitious eighteenth-century work on population was Jo-
under which ... the citizens do most increase and multiply, IS mfalh- hann Peter Siissmilch's treatise on the "divine order" in these demo-
bly the best. Similarly, the government under which ~people di_mi~­ graphic affairs, which went through four editions between 1740 and
ishcs in number and wastes away, is the worst. Experts m calculation. I 1798 and grew to three thick volumes by the final printing. God's first
leave it to you to count, to measure, to compare.
"9
. commandment was to "be fruitful and multiply" (Genesis 1), Siissmilch
Obligingly, the experts in calculation strugg.led .t~ find a reha?le wrote, although he conceded that the human institution most condu-
method for measuring population size and deCJdmg If 1t was exp~ndmg cive to population increase, marriage, was, "not so much a fruit of re-
or contracting. Rousseau's infallible guide, however, proved evas1~e, for ligion and of Christianity, as a necessary consequence of nature and
census data were nonexistent, birth registers were imperfect, and It was sound reason. But is it not good, if Christianity confirms inculcates
difficult to decide upon a community for which the ratio of total pop- and blesses the command of nature?"'' Since this "first fundamental
ulation to annual births would be representative of an entire state. It was law of earth" was grounded in human welfare and th~ prosperity of
more practical to turn the equation around and cite evidenc~ of changes states, every institution, creed, habit, and law could be judged against
in prosperity or general happiness as proof tha.t the popul.a~wn must b_e one universal standard-did it promote or inhibit the growth of popu-
increasing or declining. Accordingly, the social and political determi- lation? Siissmilch undertook to discover the conditions which most en-
nants of population growth were widely discussed and debated long be- " Thomas Short, A Complete History of the Increase and Decrease of Mankind (London
fore Mal thus composed his famous essay. ' 0 '767), p. i. '
" Robert Wallace, A Dissertation on the Numbers of Mankind in Antient and Modern
-See Montesquieu, Persian Letters (1721; New York: Penguin, 1973), PP· 202 ff, and also T imes in Which the Superior Populousness of Antiquity is Maintained (Edinburgh, 1753), p.
8
tht· first chapter in Book 2 of De I'esprit des lois (Pans, 1746)., . . . d 7
> David Hume, "Of the Populousness of Ancient Nahons, m Essays: Moral, Po/Jtrcal an ' 3 Johann Peter Slissmilch, Die giittliche Ordnung in den Veriinderungen des menschlichen
Liraary (1741-42; Oxford, 1964), P· 385. . f d 6 ) Geschlechts aus der Geburt, dem Tode und der Fortpflanzung desselben erwiesen (2 vols., Ber-
., Rousseau, "The Social Contract," in Ernest Barker, ed., Socwl Contract (Ox or , 19 o , lm, 3rd ed., 1765), vol. 1, p. 446. On Slissmilch see Jacqueline Hecht, "Johann Peter Si.iss-
ltl-·307, p. 28o. D 'd V Gl N b ·ng the mllch: Point alpha ou omega de Ia science demographique naive," Annales de demographie
,,, On the population debate in England, for example, see av1 · ass, um en hrstorrque, 1979, 101-134.
1\"ple: The Great Demography Controversy (London, 1978).
21
20
(
/
\
Statistics as Social Science
Statistics as· Social Science
couraged population increase by comparing the available information
. marriage and to discourage emigration Pola d d . .
from a variety of countries, both past and present. He made extensive
population decline because they ignored th n ~n Sbm had suffered
use of parish records and maintained an extensive correspondence. He holding them "could becom h . ese va ues; ermany, by up-
also relied heavily on the publications of such writers as Antoine De- land. "•4 e t e happiest, most powerful, and richest
parcieux of France; Pehr Wilhelm Wargentin of Sweden; Nicolas
Struyck, Willem Kersseboom, and Bernard Nieuwentyt of the Nether- th Si.issmilch, like most political arithmeticians, advocated expansion f
lands; and British authors Graunt, Petty, Gregory King, and John Ar- . e government apparatus for collecting population numbers and o
Important for acting 0 th C . , more
buthnot.
offspring ~ot
only of th: sci:~~ifi~o:~~t~!~ ~~!~~c~n~~ithhtmetic
wasban
Si.issmilch's theology led to a system of ethical and political maxims a1so of enhght d d · 1g enment, ut
founded on the same desideratum of maximizing population. A Prot- was to be put a~~~e s:~i~~s;k~;dae:~~inl~reat Britain this knowledge
of exte~siv~ statistical informatio~ was t~rf:::e~:~~~l;z:~?cate tdhebuse
estant pastor, he naturally decried alcohol, gambling, prostitution, ur-
ban life, priestly celibacy, and similar ills. He disapproved equally of eaucrahzation and h F IOn an ur-
war, which, he explained, was based simply on princely misperception. Condorcet mo~t earne:~~eh:sed r~n~h progressives s~ch as Turg?t and
istic interests like church anS n~b~i;~~s~~onser:vahve and pa:h~ular­
It was perfectly plain to this apostle of procreation that the inclination of
princes to covet one another's territory atose from the universal desire to this rationalization ought to i I d th. e ulti~ate benefic1anes of
gain new subjects for their states. Hence all the unpleasant marching immediate effect was to be th~~~n:oli'de t~ass off pnvate citizens, but its
and shooting of war, of which Si.issmilch had direct experience, was a IOn o state power.
unnecessary, for the desired object could be attained directly by remov-
ing the obstacles to natural population increase. The principle of geo-
metric groWth, through which a single couple had produced hundreds THE NUMBERS OF A DYNAMIC SOCIETY
of millions of descendants in a few thousand years, insured that the
number ofloyal subjects would multiply at a rate far surpassing the most Political arithmetic was s 1 td b
Britain around . .upp an e Y statistics in France and Great
optimistic expectations for victory in combat. . l the begmnmg of the nineteenth century The shift . t
Princely success, then, was to be measured by achievement in the do- :~~oa~~n;a~f~~ceompan!ed by a su~tle mutation of con.cepts that ~~n ~~
mestic sphere, and Si.issmilch was ready with voluminous advice for There was little in mthoestbiamckportantdm fthhe history of statistical thinking.
achieving it. He proposed certain measures, such as state support of groun o t e na "t f . "
such developments but t . me s a Ishcs to presage
medical care and systematic inoculation for smallpox, to reduce the , a cer am measure of 'ty .
death rate, but his campaign was aimed principally at the supply side. chang~s ~hen they are considered in their conte~;' appears m these
Taxes, he remarked, must be kept low, for a heavy burden of taxation Statistics derives from a Germa t S . ..
stantive by the G ··tr fi n erm, tatzstzk, first used as a sub-
robs men of the subsistence necessary to enter married life, and thereby o mgen pro essor Gottfried Achen 11 ·
leads to the diminution rather than increase of the true wealth of the perhaps an unfortunate choice for its etymol wa ~~ I 749· It was
definition remained a matter ~f d b t fi ogy was am Iguous and its
state. Distribution of land was crucial, for the unavailability of agricul- its pro er sub ·ec . e a e or more than a century. Even
tural territory obliged many couples to delay marriage for years. The Ro-
man empire had been built by thrifty, industrious farmers inhabiting
ni_nete~nth c~nt~~:~~:::t::~:~~~:; ~~~·ou~h 1~ost w.riters of the early
With states, or at least with th IhnsiCa y a science concerned
plots ofland just large enough to support their families. By imitating the ose matters t at ought to be known to the
ancient Romans, and eliminating obstacles to the subdivision of tracts "Ibid., p. 557 .
of land such as primogeniture, modern states could likewise achieve ''See Keith Baker, Condorcet: From Natural PhJ h .
1975), chaps. 4 and 5· In England pol't' I 'th l osop y to Socral Mathematics (Chicago
greatness. Above all, it was essential to defend "Liberty and Property"- by Richard Price, sought to ally their fie~~a~~h metbl~ans of the late eighteenth century, led
Si.issmilch used the English words-for these served both to promote traits~; see Peter Buck, "People Who Counted· ~plu . r~am~m rather. than bureaucratic cen-
tury, ISIS, 73 (1982), 28-4 5. · 0 thea Anthmetic m the Eighteenth Cen-
22
( ( (

Statistic.~ as Sodal Science __ _ Statistics as Sociql SciellCe


1~1eri~al definition. As late as 1830, statistical articles were scarcely dis-
"statist." Statistics initially had_ no more to do with the collection or
tmgmshable from geographical ones-and were classed among the sci-
analysis of numbers than did geography or history. Its task was to de-
ences geographiques-in the Parisian Bulletin univer.~el des sciences et de
scribe, and numerical tables were involved only to the extent the author
l'industrie. The durability of the old German definition of statistics
however, is probably less important than the circumstance that by th~
found them available and appropriate to the particular subject matter.
The anglicized form of this German term was introduced by John
late 182os a tradition of numerical social studies had become established
Sinclair, hub of a network of Presbyterian pastors whose collective labor
made possible his 21-volume compilation, the Statistical Account ~f
in France and Belgium and had laid claim to the title "statistics."
The quantitative science of statistics inherited from its German
Scotland. Sinclair, who claimed that he had deliberately adopted thts
n~mesake an exceptionally wide scope, extending from geography and
foreign expression in order to draw attention to his project, sought to dis-
climate to government, economics, agriculture, trade, population, and
tinguish his enterprise from the German one by noting th~t ":h~reas_ t~1e
culture. In its extended sense it included also facts from medicine as well
latter dealt with "political strength" and "matters of state, hts mqumes
as the natural history of man. Statists conducted surveys of institutions
were designed to ascertain the "quantum of happiness" enjoyed by the
of all sorts and maintained records of trade, industrial progress, labor,
inhabitants of a country, "and the means of its future improvement."' 6
poverty, education, sanitation, and crime. Although their science came
\It is difficult to establish just when people came to ~ee s_tatistics ~sa sci-
to be identified most closely with the censuses that Great Britain
ence that was specifically defined in terms of numencalmformatiOn_, for
the tradition of statistical surveys and statistical accounts permttted
~ranee, Prussia, the United States, and other countries began conduct~
mg around the beginning of the nineteenth century, its scope was far
numbers to assume ever greater prominence without forcing any dis-
br~a?er than had been envisioned for political arithmetic, except by
continuity. The transition seems to have taken place in Great Britain al-
Wilham Petty himself.
most subconsciously, and with a minimum of debate. Bisset Hawkins
Political arithmetic, as I have suggested, was associated with central-
was probably not too far ahead of his countrymen in 1829 when he de-
izing bureaucracy. The information that numbers could provide was vi-
fined medical statistics as "the application of numbers to illustrate the
tal for controlling the population, and especially for augmenting tax rev-
natural history of man in health and disease," and pointed to an anal-
enue. More fundamentally, however, the ideal of enumeration was one
ogous statistical field in political economy. 17 v . w~1ich few other than agents of the crown would seriously have enter-
Statistique evolved in the same direction, albeit along a less contm-
tamed, at least on the Continent, under the Old Regime-and mon-
uous path. Statistics was identified in France with numerical informa-
archs typically regarded demographic figures as a state secret too sen-
tion about society as early as 1820, when Charles Dupin defended the
siti~e to publish. Implicitly, at least, statistics tended to' equalize
placement of statistics among the sciences mathematiques rather than
~ub1ects. It makes no sense to count people if their common personhood
the sciences morales et politiques in the Parisian journal Revue ency-
IS not seen as somehow more significant than their differences. The Old
clopedique. Similarly, the great official compilation of population a.nd
Regime saw not autonomous persons, but members of estates. They
mortality figures for Paris and the Seine was published unde~ t~e :ttle
possessed not individual rights, but a maze of privileges, given by his-
Recherches statistiques beginning in 1821. The Genevan Bzblwteque tory, identified with nature, and inherited through birth. The social
universelle, however, published purely descriptive contributions under
world was too intricately differentiated for a mere census to tell much
this title until 1828, when it switched abruptly to an exclusively nu-
about what really mattered
,6 Sir John Sinclair, ed., The Statistical Account of Sc?tland ~-21 vols., Edinburgh .. 1791.-
Such Old Regime hierarchies had already broken down to a consid-
1799), vol 20 , Appendix, p. xiii, to which August Ludwtg Schlozer responded that Smclat~ erable extent in eighteenth-century England, and the French Revolu- '
must never have seen a handbook of statistics tf he thmks tt concerns only pohttcal power,
tio~ was instrumental in destroying them on the Continent, although
Theorie der Statistik nebst Ideen tiber das Studium der Politik uberhaupt (Gottmgen, :So4), P·
17 . The Oxford English Dictionary, my source for the Sinclair quote, shows an earher use of
s~ctal. heterogeneity remained a major concern of statisticians, espe-
the term in an English translation of a German work. ctally m Germany, throughout the nineteenth century. Needless to say,
'7 Bisset Hawkins, Elements of Medical Statistics (London, 1829), p. 2.

25
24
(
Statistics as Social Science
Statistics as Socia!Science
royal absolutism was not always the beneficiary of these changes; per-
haps bureaucratic centralism was, but its victory was an ambiguous one. into the internal structure of hunian society"•9 which could be derived
from statistical investigations.
If statistics provided bureaucracies with some of the knowledge that is
indispensable to power, they also suggested certain limitations to this Malthus's views are characteristic of a set of attitudes that emerged in
power. The limitations in question are not constitutional ones, but con- the wake of the French Revolution and that underlay the statistical
. straints that now seemed to exist independently of any particular formal mov~n:ents of the second quarter of the nineteenth century. As Stefan
Colltm has argued, both the historicism and the new interest in "social
arrangements of government. For the expansion in the scope of nu-
science" of the early nineteenth century were manifestations of a new
merical investigations was accompanied by an important change in the
sense among European thinkers and writers that "society" was in reality
conception of their object.
a mor~fund~~ental dimension of_existence than either state or govern-
That shift can be characterized through a comparison of the most em-
ment. Stahshcs, one of the earliest of the would-be social sciences
inent students of population before and after the French Revolution,
Siissmilch and Thomas Robert Mal thus. Siissmilch, always presuppos-
can:~ to be conceived in France and England as the empirical arm of
political economy. Although the laws of economic behavior could be
ing that population increase was the highest aim of every leader, devoted
discovered deductively, their very universality rendered them unsuitable
for c_apt~ring the particular circumstances of a country in any given state
his work to showing what the state could do to promote the growth of
population. Malthus, who held that a high population density is the
~f ~Istoncal development. One of the principal tasks of numerical sta-
principal cause of misery and poor health in a country, stressed instead
hshcs from the beginning was to fill this gap, to chart the course of eco-
the constraints imposed by laws of population on the possible forms of nomic and social evolution.,,
governmental and social organization. Population was no longer some-
The "era of enthusiasm" in statistics was thus inspired by a new sense
thing pliable, to be manipulated by enlightened leaders, but the product of the power and dynamism of society. Society was regarded both as a
of recalcitrant customs and natural laws which stood outside the domain source of progress, revealed by the beginnings of industrialization and
of mere politics. Government could not dominate society, for it was it- a_s a .cause of i~stability, typified by the French Revolution and by,con-
self constrained by society. hnumg unrest m Great Britain as well as France. Statistical investigation
According to Malthus, the principle of population contradicted the was not the product of sociological fatalism, however, but of cautious
utopian schemes of Condorcet and Godwin by showing "that though hopefulness for improvement. Statistics reflected a liberal temperament
human institutions appear to be, and indeed often are, the obvious and and a ~earch for reform that flourished not during the years of repression
obtrusive causes of much mischief to society, they are, in reality, light followmg the Congress ofVienna, but the late 182os and especially the
and superficial, in comparison with those deeper-seated causes of evil, 1.83os. The statists sought to bring a measure of expertise to social ques-
which result from the laws of nature and the passions of mankind. "' 8 t~ons, to replac~ the contradictory preconceptions of the interested par-
Mal thus held society to be a dynamic and potentially unstable force, an ties by the certamty of careful empirical observation. They believed that
incipient source of turmoil that threatened to confront the lover of Eng- the confu_si~n o_f politics could be replaced by an orderly reign of facts. 22
lish freedoms with a choice between revolution and repression. To avert The ongms, 1f not the zenith, of the great statistical enthusiasm of the
this, Mal thus upheld the need for public education to acquaint the peo-
'0 Ibid., vol. 1, p. 20.
ple with the true causes of their misery. Political leadership was not un- 0
'.Stefan Collini, "Political Theory and the Science of Society in Victorian Britain " His-
important, for a wise government could perhaps chart a safe course tonca/ Journal, 23 (tg8o), 203-231, pp. 203-204. · '
through these troubled seas. For this, however, it required a familiarity " ,~ee VIctor L. Hilts, "Ali is Exterendum, or, the Origins of the Statistical Society of Lon-
don, Is1s, 69 (1978), z 1-43 .
with the principles of political economy, and also that "cleared insight " See [William Cooke Tavlorj, "Objects and Advantages of Statistical Science " Foreign
Quarterly Revzew, 16 (183 5), 205-229. (For identification of anonymous authors in 'British re-
,s Thomas Robert Malthus, An Essay on the Principle of Population (2 vols., London, 6th v~cws I have used Walter Hough:?n, ed., Wellesley Index to Victorian Periodicals [4 vols.,
ed., 1826), vol. 2, p. 20. 1 oronto, ~966 ct seq.].) The label era of enthusiasm·· was introduced by Harald Westergaard,
Contnbutwns to the Hzstory of Statistics (London, 1932).
26
(
(
Statistics as Social Science
Statistics as Social Science
early nineteenth century are to be found in France. The passion for in-
me~~ir on the abolition of the d~ath penalty by the Societe fran~aise de
formation reared its head already during the Consulate and Empire, al-
stabsbque universelle: "By attaining statistical certainty that fewer
though the ambitions of the early Bureau de statistique so far exceeded
crimes are committed where the penalty of death has been abolished
its resources that the extraordinarily detailed census it undertook to carry
t~e influence that a gentler and more genuinely philosophicallegisla~
out through the prefects yielded virtually no usable data. 23 In the end,
bon exerts on the criminality of human actions can be better appreci-
an impatient Napoleon suppressed the bureau in 1811, as he had abol- ated. " 2 ;
ished the second class of the Institute, moral and political sciences, in
Alth~u~h the occasional appearance of numerical results directly
1803. The restored monarchy had little desire to resuscitate it, although contradicting preconceptions of this sort obliged statistical authors to
the diverse records that were'"collected and published under Louis XVIII maintain. a ce:t~in degree of flexibility in interpreting their figures, the
and Charles X assisted the private statistical writers who began to flourish gene.ral d1spos1b~n of these writers was to present their findings as direct
around 1820. Among these records were the accounts of the health of and .~~controvertible proof of the propositions they seemed to support.
army recruits, which first appeared in 1819; the comprehensive French Stahshcs, wrote Alphonse De Candolle, "has become an inexhaustible
judicial statistics, beginning in 1827; and the pathbreaking Recherches ars~nal of do~ble-edged weapons or arguments applicable to everything,
statistiques sur la ville de Paris et le departement de la Seine, published :Vh1ch have sd:nced many by their numerical form. "z6 Less cynically,
in 1821, 1823, 1826, and 1829 under Chabrol and the physicist Joseph ~~~or~ mys.tenously: Moreau.de Jonnes claimed that the figures of sta-
Fourier. Finally, the census was reconstituted under the July monarchy b~bcs are hke the h1eroglyph1cs of ancient Egypt, where the lessons of
and placed under Alexandre Moreau de Jonnes .. history, the precepts of wisdom, and the secrets of the future were con-
The statistical initiative in France was taken chiefly by advocates of cealed ~n mysterious characters. They reveal the increase in the power
public health, particularly by army surgeons released from service at the of empues, the progress of arts and of civilization, and the ascent or re-
conclusion of the Napoleonic wars. Initially, these writers focused their trogression of the European societies. "z7 Statistics "does not have the
attention on the salubrity of orphanages, prisons, and poorhouses, usu-
ally with the specific aim of instigating reform of the institution in ques-
tion. Other public-spirited mutations of the old descriptive statistics fol- I power to ac.t,': he wrote, "but it has the power to reveal, and happily, in
our day, th1s IS practically the same thing. "zB v
'Although an attempt was made in 1830 to form a French statistical
lowed shortly thereafter. Among the most noteworthy and influential
was the use of statistics in the campaign for public education. A magical
significance was attached to the proposition that education leads to a re-
duction of crime, which was conclusively demonstrated again and again
through the shameless manipulation or misinterpretation of numerical
II society, the result was unsatisfactory, and consequently the leading writ-
ers ~o~s~ssed no institution or journal dedicated specifically to statistics.
~tahsbc1ans were prominent in the classe des sciences morales et poli-
bques of the Academy when it was revived under the July monarchy
but the~r leading organ was probably the Annales d'hygiene publique:
records. A. Taillandier was able to establish, without any information a~d the1r thought retained a medical cast even in reference to purely so-
on literacy rates in the population at large, that "the definitive result of Cial problems. The editors of that journal announced in their 1829 pro-
these researches on the instruction of prisoners reveals that 67 out of 100 s~ectus t~at ~ublic hygiene "can, by its association with philosophy and
are able neither to read nor write. What stronger proof could there be Wl.t~ legislatwn, exert a great influence on the march of the h,Uman
that ignorance, like idleness, is the mother of all vices?" 24 In the same spmt. It should enlighten the moralist and cooperate in the nobl~ task
vein, the patient forebearance of statisticians from admitting precon-
'' "Introduction," Bulletin de Ia societe franqaise de statistique universelle vol. 1 ( 18 30.
ceived opinions was revealed in the announcement of a prize for the best 1831), p. 33- '
6
" Alphonse De Candolle, "Considerations sur la statistique des delits," Bibliotheque uni-
,, See Marie-Noelle Bourguet, "Decrire, compter, calculer: The debate over Statistics dur-
ing the Napoleonic Period," in Prob Rev.
ver:;lle des sczences, be/les-lettres et arts, 104 ( 1830), Litterature, 15 9 - 186, p. 1 60 .
•• A. Taillandier, Review of"Compte general de !'administration de Ia justice criminelle en Alexandre Moreau de Jonnes, "Tableau statistique du commerce de Ia France "Revue
encyclopedique, 31 (1826), 27-46, p. 27. '
France," Revue encyclopedique, 40 (1828), 6oo-612, p. 612. 8
' Jonnes, Elements de statistique (Paris, zd ed., 1856), p. 5·

28
___ ---
........_

(
( ,I
\
Statistics as Social Science Statistics as Social Science - ·
of diminishing the number of social infirmities. Scarcities and crimes
Tory named John Rickmann, refused to collect occupational records of
are maladies of society that it should work to diminish. ">9 Statistics the sort desired by political economists. The great burst in official, as in
could provide an understanding not only of the prevailing causes of private, statistical activity occurred during the 183os, when a statistical
death and disease, but also of crime and revolution, respectively the
office was set up under G. R. Porter at the Board of Trade (1832), and
chronic and epidemic disorders of the human spirit.
the General Register Office was created (1837) to collect vital statistics
Michelle Perrot characterized the French moral statisticians as bour-
and to supervise a greatly expanded census beginning in 1841. This was
geois reformers, seeking to control deviant behavior of all sorts _with
the period of the "condition of England question," of heightened fears
numbers. Similarly, William Coleman has argued thatthese quantifiers
about social dislocation and the prospect of revolution, but it was also a
sought "to replace the long reign of opinion, party interest, and poli_tical
time when the strategy of conservative repression was largely rejected.
confusion with a secure core of well-established social facts and ngor-
The study of social numbers embodied hope as well as fear, and it is not
ously deduced truths" based on social science. Although there was som~
by chance that statistics became "the favourite study of the present
connection between public hygiene and the St. Simonian movement,
age" 3 ' in Great Britain precisely during those years of upheaval that
most of these writers were reluctant to endorse deep and systematic state
brought forth the Reform Act of 1832, the Factory Act of 1833, and the
intervention in private concerns. Perhaps their subservience to classical new Poor Law of 1834.
political economy was, as Coleman intimates, a failure of the diagnos-
The characteristic institution of statistical enthusiasm in Great Britain
tician to countenance the cure-whatever that might have been. More
was the private statistical society. The first important association of Brit-
significant for our purposes, however, is that these soci~medical inv~s­
ish "statists" was the statistical section of the British Association for the
tigations of 181 5-1848 represented "the earliest systematic ~~ort to ~e1ze
the complex interrelations of social structure and change, mvolvmg a Advancement of SJ:jence, formed in 1833, just two years after the
search for the regularities that underlay the evident disorder of French founding of the parent body. Statistics represented the British Associa-
social development during the early nineteenth century. 30 tion's foray into social science, a move not universally applauded. The
The desire to understand contemporary social transformations, and to creation of the new section was inspired by the presence at the Cam-
establish a scientific basis for social policy, was also at the heart of the bridge meeting ofMalthus and, especially, Quetelet, who discussed his
statistical movement in early Victorian Britain. An official census had own work on the statistics of crime and suicide, As Lawrence Goldmann
been set up in 1801 in response to manpower needs during ~he Napo- has recently shown, the inspiration for its founding came primarily from
leonic wars but the first four decennial censuses were considered un- Richard Jones, who opposed the abstract, deductive style of Ricardian
satisfactory ~ven at the time, since information was collected primarily economics and sought to replace it with a historical and empirical ap-
through the networks of the established church and because the head, a proach. Charles Babbage subverted the rules of the B. A.A. S. in order to
gain approval for the new section.
>9 "Prospectus," Annales d'hygiene publique et de medecine legale, I (182~). v~vii. See ~!so
Bernard-Pierre Lecuyer, "Medecins et observateurs sociaux: Les Annales d hyg.zene.~ublzque The founding members, especially Malthus, Babbage, Richard
et de medecine legale, 182o-185o," in Pour une histoire de Ia statzstzque (/ourn~es, d et~de sur Jones, W. H. Sykes, and J. E. Drinkwater, formed the original core of
/'histoire de Ia statistique), vol. 1, Contributions (Paris, ~977), P~· 445-476; -~ hyglen~ en
France avant Pasteur," unpublished paper; "Demogr~ph1e, statlshque, et hyg1ene pubhqu: the Statistical Society of London-progenitor of the present-day Royal
sous Ia monarchic censitaire," Annales de demographze hzstonque, 1977, P~·. 215-z45; Rene Statistical Society-which was organized in March 1834. During the
Le Mee, "La statistique demographique officielle de 181? a 1870 en France, rbzd., 1~79, PP;
251 . 279. On the attempted formation of a French stahsllcal soc1ety, see l~:t~rs from\ 1llerme interim, a different group of reforming physicians and industrialists had
to Quetelet, 1 Jan. 1830 and 2.1 Jan. 1831 in cahier 2500 AQP. The Soc1~te fran~a1se de sta- formed a statistical association in Manchester, which was the most ac-
tistique universelle had little or no support from the leadmg French stallshc1ans. .
> William Coleman, Death is a Social Disease: Public Health and P~lttzcal _Economy m
0
tive and successful of these societies during the first decade of their ex-
Early Industrial France (Madison, 1982), p. 275; also M1chelle Perrot, Prem.~eres mesures ,, [Herman Merivale], "Moral and Intellectual Statistics of France," Edinburgh Review, 69
des faits sociaux: les debuts de Ia statistique criminelle en France (1790-183o), m Po~~ une
(1839), 49-74, p. 49· On the response of political economists to this political instability, see
histoire (n. 29), pp. 125-137; Ian Hacking, "Biopower and the Avalanche of Numbers, Hu-
Maxme Berg, The Machinery Question and the Making of Political Economy (Cambridge,
manities in Society, 5 (1982.), 279-295. Eng., 1980).

31
( ( (
Statistics as Social Science
Statistics as Social Science
few.w~re willing to undertake the arduous labor necessary to prepare real
istence. A host of similar organizations-perhaps twenty-were formed stahshcal re.p~rts. Moreover, the society was a bit nonplussed by the
or publicly contemplated in various provincial cities during the ensuing va~ue defimtwn and uncertain frontiers of the field it had so enthusi-
two decades, but all soon collapsed. J> Hence British statistics was prin- astically entered. The leading nat.ural scientist among its founding
cipally identified during the Victorian period with the statistical societies members, Charles Babbage, quickly lost interest. William Whewell
of London and Manchester, Section F of the British Association, the who was a close friend a~~ correspondent of Jones, and who had invited
General Register Office and Board of Trade, and with the masses of data 9uetelet to the 1833 Bnhsh Association meetings, was one of the ear-
compiled into Blue Books under various parliamentary directives. liest members of the council of the Statistical Society of London b t b
The active members of the early statistical societies were far less likely ~ugust 1834 his portrait of the society's activities was hardly en~o~ra ~
to be seriously interested in natural science or mathematics than to be mg: He wrote to Quetelet: "You will find that the Statistical Secti:n
politically engaged. Most of the leading members of the Manchester or- ':h~ch spru_ng ~p under your auspices at Cambridge is grown into a Sta-
ganization, such as James Phillips Kay, William Langton, Benjamin hsh~al SoCiety l.n London, with many of our noblemen and members of
Heywood, and Samuel and W. R. Greg, were industrialists, or at least Parliament for Its members. Our Committee has had several meetings
had close family ties to industry. More particularly, they were, as Mi- ~ut we are still somewhat embarrassed by the extent of our subject i
chael Cullen remarks, "improving" employers, seeking to ameliorate ave no doubt, however, that we shall do something." Six months lat~r
sharp class divisions and to prevent social upheaval through kind treat- Whewell begged Quetelet to suggest a project for the society he h d
ment of workers, the inculcation of morality, and the provision of ed- helped to ~ound.'and indicated that the society was attempting to dr:w
ucation. Although they were philosophically opposed to government ~p a questwnnalfe for distribution. He was himself, he confessed 1 .
mterest. 34 , osmg
meddling, especially in the domain of commerce, they agitated for an
active state role in sanitation and, like their French contemporaries, Later in 1~35 a questionnaire was at last completed and sent out un-
tended to view public education as a panacea for crime and social un- der the a.usplces of the society by George Richardson Porter, one of its
most achve me,~bers. ~nd also head of the statistical office at the Board
rest.
The Statistical Society of London perhaps fits even better than does ~f Trade. The Inqumes on Education in Belgium," sent to Quetelet
Manchester Michael Cullen's characterization of statistical work as "a :c!uded under the heading "Re~u~ts ~f Education" a set of question~
movement of the reforming establishment, Whig to Liberal in poli- h~lch su~est the degree of sophistication in social science attained by
t IS fledglmg organization:
tics."n Active researchers, however, were distinctly more radical than
the great body of political worthies who offered the new organization lit-
:~;, Wh~t has been the effect of the extension of education on the
tle more than the prestige of their names and two guineas. The mem-
ha.blts of the People? Have they become more orderly, abste-
bership skyrocketed, reaching 318 within a year of its founding, but very
mious, contented, or the reverse?
'' See Michael Cullen, The Statistical Movement in Early Victorian Britain: The Foun- 2. W~at is the proportion of crime under the two heads of offenses
dations of Empirical Social Research (Hassocks, 1975); Hilts, "Aiiis exterendum" (n. 21); Victor agai~st prop~rty and against the person, to population and prop-
L. Hilts, Statist and Statistician: Three Studies in the History of 19th-Century English Sta-
tistical Thought (New York, 1981); T. S. Ashton, Economic and Social Investigations in Man- erty m the different Provinces?
chester: A Centenary History of the Manchester Statistical Society (London, 1934); David 3· What is the proportion of crimes to education? Are the educated
Elesh, "The Manchester Statistical Society: A Case Study of Discontinuity in the History of
Empirical Social Research," Journal of the History of the Behavioral Sciences, 8 (1972.), 280- found to be.m~re exempt than the uneducated, or the reverse?
301, 407-417; Philip Abrams, The Origins of British Sociology 1834-1914 (Chicago, 1968), 4· ~hat de~cnphon. of crime prevails most in the educated prov-
chap. 3; Jack Morrell and Arnold Thackray, Gentlemen of Science: Early Years of the British
Association for the Advancement of Science (Oxford, 1981), pp. 2.91-296; Berg, Machinery mces, cnmes agamst property or against the person?
Question (n. 31), pp. 291-314; Lawrence Goldmann, "The Origins of British 'Social Science':

Lo~don
Political Economy, Natural Science and Statistics, 1830"1835·" Historical Journal, 26 (1983), ,. See l~tters of 4 Aug. 1834 and 3 Feb. 1835. Whewell to Quetelet cahier 6
2 44
The letter IS repnnted m Isaac Todhunter, William Whewell (2 vols. ' ' 1 876 )'p.
' AQ
1 8P.

587-616.
n Cullen, Statistical Movement (n. 32), p. 82.
33
( ( (

Statisrtcs-as SociaFScience Statistics as Social Science


nishes the legislator with materials on which to found remedial
5. What proportion of criminals, especially in the grosser class of
crimes could read and write, in the returns of 18~3. or ~834? measures for social derangement, and plans for increasing the mass
6. What is the political effect of the diffusion ofhel ducat;on, tas of social happiness; and though its conclusions often consist in a
evinced by the increase of newspapers, pamp ets, e c. e c. bare numerical statement of aggregate results, yet they come home
with all the authority of stubborn facts, and often tell more than the
etc.? 1 d most elaborate moral appeal. From information thus furnished, it
7. What is the number of Books published during the ast year an
cannot be questioned that the public attention has been fastened,
how classified? 35
with an intensity never before given to the subject, upon the phys-
This was hardly the type of questionnaire that could be answered.w~th­ ical and moral degradation of the poorer classes in the metropolis
out bias, and Porter divulged a few of his O\Vn in a subseq~ent missive. and manyQfour large towns. The appeal thus made has been nobly
The level of crime, he wrote, "is disgraceful to us as a nation and pro- responded to; the dry facts have been interpreted; and means have
ceeds from the deplorable ignorance in which the great b~lk of the peo- been adopted for carrying the blessings of education, order, and vir-
ple are kept." The cause, he explained, was the estabhshe.d church, tue into those dark recesses where ignorance, vice, and misrule ap-
whose power and wealth would be reduced if the peo~le we~e mstructed. peared to have fortified themselves in impenetrable obscurity. 38
Its leaders accordingly had sought to keep education m thelf own hands
and to render the knowledg~ impa~ed "as little ~urtful. to ~hen;,s,~lves as \\Notwithstanding their obsession with filth and criminality, and their.
possible by adulterating it w1th theu own sectanan preJudices. . . deaication to public health and education, British statists were not con-
Porter's tirades against Anglicanism are by no means .de~mng attn- tent to be simply activists and reformers. Theirs was to be the empirical
butes of the statistical movement as a whole, but they do mdicate some- science of society, yielding the certainty of science and deserving there-
thing of the extent to which statistical facts were sought to bolste: ~ar­ spect appertaining thereto. The philosophical justification they gave for
ticular social and political programs. Cullen ar.gues that stabsb~al the scientific character of their activity was problematical, however, and
research was aimed primarily at vindicating industnal progress by laymg the London society received much notoriety for it The statistical section
the blame for social ills on other causes, including alcohol, moral de- of the British Association had been suspect from the beginning because
generacy, and the growth of cities.37 More charitab~y, one can at least it was feared that investigation into the affairs of man would engender
see that the statistical enthusiasts held strong commitments .. Ignorance controversy and might even lead to the breakup of the larger organiza-
and filth were seen as responsible for the prevalence of .disease, the tion. 39 Partly as a defensive move, and partly to reassure interested po-
rampant growth of crime, and the.thr~at o~ domestic turmml a.mong th~ litical leaders that their support of statistics would not embarrass them,
working classes. Statistical investigatiOn, 1t was pr~sumed, would pro the statists adopted the position that they were concerned exclusively
vide empirical support for the reforms that were obvwusly I~eeded. Thus with facts. The council of the London Statistical Society announced the
the Royal Cornwall Polytechnic Society issued the followmg statement ,following general principle:
in 183T
, The Science of Statistics differs from Political Economy, because,
If"the proper study of mankind be man," the value of statistical in- although it has the same end in view, it does not discuss causes nor
formation can no longer be doubted. It stimulates the ben~volence
,s "Extract from the Annual Report of the Royal Cornwall Polytechnic Society for the Year
and gives aim and effect to the energies, of the philanthropist; 1t fur- !837," JRSS, I (!838), 190.
' 9 See, for, example, the report of Adam Sedgwick's address of 1834 in "Proceedings of the
" Porter to Quetelet, 28 May 1835, cahier 2041, AQP. British Association," The Edinburgh New Philosophical Journal, 17 (1834), 369-374, p. 372,
• 6 Ibid., 4 June 1838. d sim Bernard-Pierre Lecuver indi- which was recognized as a criticism of statistics, and perhaps also phrenology. See also "Phre-
n Cullen, Statistical Mf ovemen~ (nF 32), p. :e4~~~ !:tadie~ profcssionnellcs dan~ les An- nology and the British Association," The Phrenologicalfournal and Miscellany, 9 (1834-1836),
cates that the same was o ten true 111 ranee, s ·. h d l'usure au tra- 120-126; cf. Roger Cooter, The Cultural Meaning of Popular Science: Phrenology and the Or-
nales d'hygiime publique et de medecine legale, ou une prem1ere approc e e ganization of Consent in Nineteenth-Century Britain (Cambridge, Eng., 1984), esp. chap. 3·
vail," Le Mouvement Social, no. 124 (1983), 45-69.
35
34
-- - - - - - - - - ~

(
(
(
Sfatistics as Social Science--
-Statistics as Social Science
reason upon probable effects; it seeks only to collect, arrange,. and
compare, that class of facts which alone can for~ .the baSIS of . truth of facts, not to subserve some immediate purpose of administration
correct conclusions with respect to social and pohhcal govern- or legislation." At the same time, he held statistics indispensable for the
ment. ... practical statesman: "It is by comparing the numbers of the subject un-
Like other sciences, that of Statistics seeks to deduce from well- der consideration ... that his practical judgement ought to be formed."
Without numbers, legislation is ill-informed or haphazard. 44
established facts certain general principles which interest and affect
Although an American census was mandated by the Constitution and
mankind; it uses the same instruments of comparison, calculation,
carried out decenially beginning in 1790, a decade before any other
and deduction: but its peculiarity is that it proceeds wholly by the
country except Sweden had instituted such a periodical tabulation,
accumulation and comparison of facts, and does not admit of any
American statistical work seems to have had little influence on Euro-
kind of speculation. 40
pean. The tone of early American statistical writing is scarcely distin-
The motto of the new society was Ali is exterendum-"to be threshed out guishable from that of Britain. Already in 1811 we find James Mease
by others"-and the council declared t~~t "the first and ~o.st ess.~ ~tia]
rule" of the conduct of the society was to exclude all opmions.
4 As proclaiming the virtues of pure numerical facts, and others suggesting
that when these facts were fully known, differences of opinion must dis-
late as 1861, William Farr wrote to Florence Nightingale: "We do .not appear. In the United States, as in Britain, it became popular during the
want impressions. We want facts . . . . Again I must repeat. ~Y. obJec- 183os and 184os to produce great compilations showing the impressive
tions to intermingling Causation with Statistics .... The stahshc1an has progress of the nation, and implying that lands less fully blessed with
nothing to do with causation; he is almost certain in the present state of freedom, including the slave South as well as most of the Old World,
knowledge to err. . . . You complain that your report would be dry. The were destined to fall hopelessly behind if they did not reform. Battles be-
dryer the better. Statistics should be the dryest of all rea?ing. "4 •. · tween North and South were fought with numbers long before they were
The "most essential rule" of statistics was, of course, Ignored m prac- fought with soldiers, most instructively, perhaps, in the form of a debate
tice. Statistical investigation-the study of school attendance records inspired by some errors in the 1840 census over the question of whether
and of death rates in relation to sanitary conditions, or the conduct of insanity was in fact ten times more prevalent among free northern blacks
surveys of factories and neighborhoods to learn about diet-required than among southern slaves. Finally, crime, intemperance, and the
other familiar objects of moral statistics were as much at issue among
more time and energy than busy people were willing to invest u?,le~s
American statists as in Britain and France.4;
they had a real interest in the outcome. Doubtless the search for a sci-
Nowhere else was statistics pursued with quite the level of enthusiasm
ence of government, the principles of which would be discovere~ out-
as in Britain, and numerical statistics developed more slowly outside the
side the realm of partisan dissention and arise from the accumulation of
politically and economically most advanced states ofwestern Europe.
simple, irrefutable social facts," as John Eyler characteriz.es it, 43 was ~u­
To be sure, disciplines of that name flourished in Italy and, especially,
tile from the beginning. Nevertheless, the facts helped to mform official
Germany, during the early decades of the nineteenth century, but this
policy and the disavowal of speculation pe~ha~s a~ded weight to the con-
was the old university statistics. While it can hardly be faulted for failing
clusions of statists in the view of the public. It 1s of the essence of sta-
to anticipate what statistics would become in other countries some years
tistics," wrote the political philosopher George Cornewall Lewis,4'lthat
later, this variety of statistics lacked a clearly defined scope, purpose, or
its object is scientific, not practical; that it is intended to represent the
method. German statisticians debated these matters endlessly and failed
•a "Introduction," JRSS, 1 (1838), pp. 1, 3· ,
44
., Cullen Statistical Movement (n. p), p. 85. See also Hilts, "Aliis Exte:endumS (n.:n). George Cornewall Lewis, A Treatise on the Methods of Observation and Reasoning in
•' See M;rion Diamond and Mervyn Stone, "Nightingale on Quetelet, ' JRS , A, 144 Politics (2 vols., London, 1852), vol. 1, pp. 133-134.
(1981), 66-79, 176-213, 332-351, P· 7°· od fW"/1" F (B 1 •
45
Se~ Patricia Cline Cohen, A Calculating People: The Spread ofNumeracy in Early Amer-
., John M. Eyler, Victorian Social Medicine. The Ideas and Meth so 1 zam arr a- zca (Chzca~o, !982), chaps. 5 and 6. On statistics in America, see also James Cassedy, De-
timore, 1979), p. 16. mography m Early Ame_rz~a: Beginnings of the Statistical Mind (Cambridge, Mass., 1969);
Cassedy, Amencan Medzcme and Statistical Thinking, 18oo-186o (Cambridge, Mass., 1984).

37
(
(
(
Statistics as Social Science
Statistics as SocialScience
to reach agreement on anything. Statistics was variously seen as the de-
scriptive science of states, or the study of the state or condition (Zustand) society of public-spirited persons from all of G
Freih~rrn von Reden, a Hanoverian bureaucrat ;~~any was set u~ by
oF~:;:asnoml
of any object at a given time-both interpretations being consistent with
the Latin root status, though only the former could be read into the Ger- beedn mvolved in seeking solutions to the problems eb hme
man Staat.
an craftsmen H. v ·
r· d
· IS erem ur eutsche Statistik49 did n t
a orers
· h
Government statistical agencies had been set up in four German-lan-
guage states prior to 1848--Prussia, Austria, Bavaria, and Wiirttem- E~;~;; ~~; ~~:::,;n~~~ ·~:~;~~~~;~:~:, ~~~~~.~;;:~:.:~~:~
berg. The great surge in official statistics, however, came in the after- ~the ~umencal method of statistics early in 1848 he d'd
math of the revolution, during the early years of German ~7:t basisdof It~ all~ged superiority for inyestigating the c;nditioin ~~ s~~
industrialization. Statistical bureaus were already operating in virtually y, a~ mamtamed that the descriptive method of staf f
appropnate for studying the state. ;o Is Ics was more
every German state a decade before the unification brought a greater
measure of standardization to these things. At the same time, numerical The nume~ical social science of statistics never became as popula
statistics became increasingly influential among social and political sci- a movement m Germany as it had in Britain and France erha b r
entists at the universities, and it appears that here, as in France and Brit- cause t~e fi~ld remained under the control of governmen~ ~urea~~ra~~
ain, the sudden growth of interest in statistics can be attributed to an in- ~~~ s~~::~~;t professors. It did, however, effectively supplant descrip-
crease of anxiety about social change and instability.
Numerical statistics entered the German universities as a conse- ~ne;a',wad~:::d~h:~~~o~~~~;~;~~;.;~;~:~~;~:;~;:l :~~;:::;
quence of the successful effort tq redefine the discipline in terms com- SCI~n IS s an government administrators was produced durin the
patible with the work of census bureaus and with current usage in mai~der o.f the.nin~teenth century. Statistics was pursued in cl~se c:~~
nectwn With histoncal economics especially by the " d . . I
French and English. The use of numbers had occasionally been cham- I · t " f th v · fl" . ' aca ernie socia -
~~ ~n~ers;an~rei~ ui So~alpol.itik, and was a principal tool in the effort
I
pioned by university statisticians such as Schlozer, who thought them
especially suitable for rendering descriptions concise and systematic, . j· . an so vet e social problems associated with German in
d us t na Izahon. -
and, during the 184os, by Christoph Bernoulli, descended from the
great Swiss scientific family, who advocated a new science of Popula- 49 See F. W. Reden ed Zeitschrift de V . [i· d
also]. ~- W., "Die Errlcht~;ng statistische; e.rems ur eutsche Statistik (z vols., 1848-1849);
tionistik. 4 6 Generally, however, numbers were regarded during the Vz~;telJahrsschri{t, 9 (1846), no. 3, 95-128Bureaus und stahshscher Pnvatvereine," Deutsche
early nineteenth century as secondary and even superficial. Such was C. G. A. Kmes, Die Statistik als selbstiind· w· h
Eberhard A. Janak, Theorie der Statistik in G zged '_zssensVc aft (Kassel, 185o), p. 2 3· See also
the opinion of the Tiibingen professor Johannes Fallati, leading aca- "A Ph ons
. t.zsc h e Betrachtungen tiber St t" t"k .h run Bzugen
h ( tenna ' 18s6)
. ; Leopo ld N eumann,
demic statistician in Germany during the 184os. 47 F allati, however, be- /s
Oesterreichische Vierteliahresschrift (I"' a~ 1 h.t 1 re andlung und Jhre neueren Leistungen,"
ur ec s-un taatswzssenscha{t, 3 (18 59 ), 87_114 _
lieved numerical statistics to be essential for practical administration and
reform, and he championed the development of statistical organizations
on the model of the societies that had sprung up in London, Man-
chester, Bristol, and Ulster to collect, organize, and disseminate this
useful if shallow variety of statistical information. 48 In 1847 a statistical
• 6 Schlozer, Theorie (n. 16), p. 20; Christoph Bernoulli, Handbuch der Populationistik,
oder der Volker- und Menschenkunde nach statistischer Erhebnissen (Ulm, 1841), and "Vor-
wort," Schweizerisches Archiv fiir Statistik und Nationalokonomie, 1 (1827).
4 7 Johannes Fallati, Einleitung in die Wissenschaft der Statistik (Ti.ibingen, 1843), pp. 104-

106.
48 Fallati, Die statistischen Vereine der Englaender(Ttibingen, 1840); "Gedanken tiber ?\lit-
tel und Wege zu Hebung der praktischen Statistik mit besonderer Ri.icksicht auf Deutschland,"
ZGSW, 3 (1847), 496-557.

39
!

The Laws That Govern Chaos


That the vaunted science of statistics was merely a method and the
statistical society composed of "hewers and drawers to those engaged on
- - - - - - - - - Chapter Two
any edifice of physical science" was wholly unacceptable to the Victo-
rian statists. "Statistics, by their very name, are defined to be the obser-
vation necessary to the social or moral sciences, to the sciences of the
THE LAWS THAT GOVERN CHAOS
statist, to whom the statesman and the legislator must resort for t[Ie prin-
ciples on which to legislate and govern.''; Statists were by no means in
agreement, however, that the exclusion of opinions was forever to be the
The council of the Statistical Society of London was ~e~hap.s a ~it ex- defining attribute of their science. More common was the view of J. E.
.t d that brute facts were at once the dtstmctJve feature Portlock, organizer of the short-lived statistical society of Ulster, who be-
treme wl1en 1 argue . . . . , · t.
. . d the foundation of the soCial dtscip1me stabs tcs, lieved that statistics represented the empirical stage of a social science
0 f rna d ern sCience an · . . .
but the standard nineteenth-century apology .for statistical science :V~s that had previously been based on mere conjecture and would soon yield
based on the utter reliability of its results. Nmeteenth-century statist!- reliable laws of the sort attained through the same procedure by "astron-
. 11 d 1·1.1 the existence of government records such as census omy, zoology, botany, chemistry, and geology." 6 Portlock, however,
cwns reve e . d · ,
data, which liberated them from the need for su~m~se an con~ecture:- never characterized in any detail the nature of this universal scientific
both standard equipment for political arithmetic m the previOUS cen- procedure, and in practice the task of the numerical science of statistics
tury. The special merit of statistics was its insistence on acc~rat~ and ex- remained the collection and presentation of tabular information on sub-
haustive enumeration, its exclusion of guesses and approxnnahm~. b jects of political or social interest.
This science of pure facts was of doubtful standmg even from t e e-
. . h R John Robertson reviewing the first volume of the
gmnmg, owever. . - .' . · \V
Journal of the Statistical Society of London for the Benth~mite est- QUETELET AND THE NUMERICAL REGULARITIES OF SOCIETY

minster Review, mocked the pretense of the council that opmiOil was to
be excluded from this science. Opinion, he observed, IS JUS.t a L~tm Adolphe Quetelet was among the few nineteenth-century statisticians
word for thought: "It is not needful in the present day to discourage who pursued a numerical social science of laws, not just of facts. Like
thinkers they are not too numerous."' Robertson argued that facts can Comte and St. Simon, he believed in a dynamic social entity endowed
. ' d l . th r ht of theories or indeed that theones are with properties and tendencies that would not be significantly altered or
be mterprete on v m e Ig ' h ·
;eciselv "facts as ~iewed by the most powerful minds~"z and t at statists impeded by the whimsical acts of political leaders. Quetelet even pirated
had wa~ted much labor trying to prove silly or even m~oherent clam:s Comte's phrase, physique sociale, as the title of his new science. His pro-
because they had failed to analyze concepts and establish clear defim- posed method, however, departed radically from the positivist way, for
f . "No mere record and arrangement of facts can constitute a scl- whereas Comte insisted on the distinctiveness of each level on the hi-
:~~1~. "3. he wrote and numerical statistics ought to be understood as a erarchy of the sciences and eschewed the use of mathematics or even of
1 ' ' d · f t h. i belong to vanous
method--" a mode of arranging an statmg ac s w IC 1 numbers in the physiological and social disciplines, 7 Quetelet main-
sciences." 4
' "Sixth Annual Report of the Council of the Statistical Society of London," IRSS, 3 ( 1840),
' [R. John Robertson], "Transactions of the Statistical Society of London, vol. 1, part 1,"
1-13, pp. 1-2.
Westminster Review, 29 (1838), 45-72, P· 49· 6 j. E. Portlock, "An Address Explanatory of the Objects and Advantages of Statistical En-

' Ibid., p. 48 quiries," JRSS, 1 (1838), 316-317, p. 317.


' Ibid., pp. 68-69. . . l llv well received· see also Thomas 7 Sec Auguste Comte, "Plan des travaux scicntifiques m'cessaires pour rcorganiser la so-

. 4 Ibid.' p. 70. The exclusion of opmiOns ~sr~~ng~r~~:~ation Die kirchliche Armenpflege, ciete" (1822), Opuscules de philosophie sociale, 1819-1828 (Paris, 1883), pp. 159-163, 172;
Chalmers' essav, wh•ch I have seen only 1111 • "A Grumble/ ,Anon], "Statistics," Fraser's Cours de philosophie positive (6 vols., Paris, 183o-1842), vol. 2 (1835), p. 371; val. 4 (1839),
Otto von Gerlach, trans. (Berlm, 1847), c 1ap. I2, · pp. 7, 511-5!6 .
.\1agazinc, 52 (18551, 91- 95·
41
(
I
'
The Laws That Govern Chaos
The Laws That Govern Chaos
tained that a single method was appropriate for every science and that
erlands. So:m afterwards he wrote a doctoral dissertation under the
the social physicist could do no better than to imitate the celestial one.
mathematiCian and_ old revolutionary J. G. Garnier, who regaled his
Quetelet came to statistics from astronomy, and his commitment to pupil With tales of lite among the great savants of Paris. A few years later
the use of mathematics in the social sciences distinguished his approach the ambitious Q~Ietelet gained tentative approval from the government
from that of his reform-oriented statistical contemporaries. At the same for the constructiOn of an observatory in Brussels, where he had taken a
time, his active interest in social policy and even in certain concrete re- new position and where he was already scheming to build a scientific
forms as well as the wide scope of his ambitions for statistics separated
empu~. As a result~ h.e was authorized to make a pilgrimage to the great
his work from that of the various astronomers and mathematicians of his SCientific metropolis 111 order to learn enough astronomy to equip and
dav who wrote demographic models and computed life tables. Quetelet run the proposed observatory. ln Paris he was taken in by Bouvard at the
w~s almost unique in the early nineteenth century in combining the Royal Observatory and given the necessary instruction in the instru-
characteristic concerns of the statistical movement w·ith the technical
ments and methods of scientific observation. Among these was the
tools of astronomers and probabilists. His contribution to statistical
method ?f least squares, by then used routinely to reduce astronomical
thought and to the mathematics of statistical analysis was a characteristic
o~servabons, which in the Laplacian tradition was closely associated
if not unsurprising product of his syncretic approach. Social physics was
With the general field of mathematical probability and even with popu-
an elaborate metaphor that integrated Quetelet's genuine concern for
latwn and mortality studies. By the time he returned to Brussels, Que-
the advancement of scientific knowledge with his desire to turn science
telet had become mfected with an enthusiasm for statistics, the social
to the promotion of sound government and social improvement. I\ !ath- science of the observatory.
ematics would bring order to the apparent social chaos.
Q~etelet stressed the similitude of statistics with astronomy through-
out h1~,career, and he always attributed his acquisition of a "taste for sta-
In many ways, Quetelet's work in statistics reflected the experiences
of a whole career. Born in 1796, soon after the conquest and annexation
t~st~cs, specifically to this venture to Paris of 182 3· He identified his sta-
of Austria's Belgian provinces by France, Quetelet was educated in a hsbcal teachers as "the great French school of mathematics and social
French lyc;ee and retained close cultural as well as scientific ties to science," noting especially Fourier, Laplace, Lacroix and Poisson and
France throughout his life. Although he was more devoted to art and
~e claimed to have "had the good fortune of enjoying the lessons" ~f the
literature than to mathematics during his youth, he did not fail to notice
two g;eatmastcrs" Laplace and Fourier. 9 Although he did correspond
the exalted role assigned to science under the Napoleonic empire. With Founer, who once sent him a letter containing a quotable line
It was known that the sciences were honored, that their power had a~out the need for statisticians to be well-versed in mathematics jf their
never been greater: it was known that the man whose military glory SCier:ce was to progress, it seems unlikely that he actually received for-
resounded then throughout Europe took to heart the goal of reflect- malmstructwn from these men. Nevertheless, it is clear that he studied
ing onto them a part of the prestige that surrounded him, and that :heir written work carefully, that he admired them greatly, and that he
he had raised the most illustrious savants to the dignity of princes Identified his own goals with their ideas and with the remarkable
and of the highest functionaries of the empire. Such munificence, achiev~ments of contemporary French science generally.
which honored him who exercised it as much as the savants \vho Havmg familiarized himself with the relations of statistics to the most
\Vere its object, maintained that source of illustrious men that had advanced mathematics ofprobability theory under the guidance of these
arisen in the midst of the revolutionary exaltation. b august personages, Quetelet was little impressed bv the ordinarv statis-
tical enthusiasts of his generation. He consistent!; denied that .his sci-
After the defeat of France, Quetelet became a teacher of mathematics
ence was the province of physicians and amateur reformers, insisting in-
in Ghent at a college set up by the government of the newly united Neth-
Adolphe Quetelet, His loire des sciences mathematiques chez les Belges !Brussels, 1864). p. H 9 See ~r~nk Ha~kins, ~dolphe Quetelet as Statisticzan, iu Columbia University Studies m
Istory, E,conmmcs.and fublic Law, ~1, no. 4 \New York,
8
19 - 20 ; also Quetclct, "'Des
3 J 2. lms concernant lc developpemcnt del homme." AOB, 38 ), 1 os-zo6.

43
···The Laws.That Govern Chaos
The Laws That Govern Chaos
stead that the true foundation for statistics had been established by
be based on exhaustive measurement, the quantitative form uf natural
mathematicians and astronomers. "It is not to doctors that we owe the
history that Susan Faye Cannon has calied Humboldtian science. •3 "In-
first tables of mortality, they were calculated by the celebrated astrono-
stead of words facts are wanted," Quetelet wrote, "and sage observations
mer Halley," he wrote. Indeed the astronomer's love of natural order mstead of vague hypotheses and unfounded systems .... This manner
provided the foundation for statistical science: "The laws that concern of ~nly proceeding scientifically characterizes the nineteenth century,
man, and those that govern social development, have always had a spe- which lS destmed to occupy one of the highest places in the annals of the
cial attraction for the philosopher, and perhaps most especially for those human spirit. " 14 The diversity and complexity of these phenomena-
who have directed their attention to the system of the universe. Accus- the volume of facts required-was not a disadvantage but an asset, for
tomed to considering the laws of the material world, and struck with the Quetelet_ was an energetic organizer and indefatigable correspondent.
admirable harmony that reigns there, they can not be persuaded that Already m the early 182os he had decided that the newly reconstituted
similar laws do not exist in the animate world. "!0 Statistics was a science Brussels Academy could labor most effectively as a group, working pre-
like other sciences, to be presented in scientific periodicals such as his Cisely on the class of objects forming the heart of Quetelet's res~arch
own Correspondance mathematique et physique. ' 1 proJect, whose variety was so great that it would never yield to the in-
The focus of Quetelet's earliest work in statistics was the study of pe- dividual scientist. 15 Quetelet was also a leader of the international net-
riodic phenomena. This was already evident in his first statistical mem- work of quantitative natural historians that rose to prominence during
oir, presented to the Brussels Academy in 1825, although his an- the 183os.
nounced purpose there was the improvement of Belgian insurance The young Quetelet was highly impressed by the rationalistic French
tables. Arguing that analogy with the regularities of plant and animal life pr~gra1~ for applying mathematical probability to such problems as the
"entitles us to believe that the influences of these laws extend even to the rehabdtty of testimonies, the correctness of judicial decisions and the
human species," Quetelet proposed to examine the relationship of validity of electoral results. He proclaimed the universality of the rule of
births and deaths to time of year in order to determine "whether it is pos- numbers-Mundum regunt numeri--and argued that "all that can be
sible to ascertain in this regard some law of nature. "• 2 As it turned out, numerically expressed" falls under the jurisdiction of proba.Oilitv the-
the natural laws of mortality and natality proved to be statistical gener- ory. ' 6 Placing statistics under the domain of mathematical prob;bility
alizations of the form y = a + b sin x, where y was the number of births was his highest aim. Quetelet was disdainful of the mass of statistical re-
or deaths, x the time of year, suitably normalized, and a and b arbitrary formers who flourished during the early nineteenth century. He in-
constants, to be determined empirically. Such formulas, which Que- veighed against the "strange abuses" of statistics perpetrated by dilet-
telet called statistical laws, were typical results of his early investigations. tantes and scientific illiterates who wished only to gain support for
They were based partly on curve-fitting, partly on analogy, and were not preconceived ideas, and argued that the contemporary infatuation with
interpreted concretely or grounded in physical models. statistics had "rather retarded than accelerated its advance. "17
The motif of periodic phenomena supplied also the connection be- Perhaps these complaints were justified, but his confidence in there-
tween social and natural science in Quetelet's early work. Quetelet demptive power of probability theory was at best prophetic. The avail-
maintained that tides, weather, flowering of plants, terrestrial magnet-
. " Susan F~ye Cannon. "'Hurnboldtian Science,"' in Science in Culture: The Early Victo-
ism, and events in the life of man constituted a unified set of phenom- nan Penod (New York, 1978J. chap. ;.
ena, which could be studied by a single method. These studies were to ' 4 Quetelet. "Recherches statistique> sur le Royaume des Pays-Bas,"' J\'MAB,,,5 ( 1 8 29 ) sep-
arate pagmahon. p. H. •

w See Quetclet"s '"Notice scientifique" for his book Sur 1"/wmme, AOB, 7 (1840), 230. •, ' 5 Qudelet, Sciences mathematiques et physiques che.z les Belges au commencement du XIXe
" See Quetelet, "Avertissement et observations sur les recherches statistiques inserces dans
stec!e (Brussels, 1866), p. 9; Histoire (n. ~). pp. 375-376.
. •o Quetelct, Popular Instructions on the Calculation of Probabilities ( 1828), Richard Beam-
ce recueil." Correspondance mathematique et physique, 4 (1829), 77-82. Ish, trans. !London, 1839), p. 108.
" Quetclet, "Memoire sur les lois des naissances ct Ja Ia mortalitc 3 Bruxellcs," NMAB, 2
(1826), 493-512, p. 496. . "Quetelet, "Sur !,'ap~rec_iation des documents statistiques, et en particulicr sur l'appreci-
atJOn des moycnncs, BCCS, 2 ( 1844), reprint, p. 1.

44
45
The Laws That Govern Chaos The Laws That Govern Chaos

able techniques of enor analysis were of little use in analyzing official tific plan.s. Some ofhis proteges left their posts to join the military, and
aggregate tables based on complete enumeration, and this was the stand- m.any scientific positions at universities, colleges, and museums were
ard material used by nineteenth-century statisticians If probability the- ehmmated. ' 9 His cherished observatory, by then nearly complete after
ory was to contribute to the statistical work of Quetelet and his contem- years of hard work and planning, was occupied by the radical Liege vol-
unteers to asSJSt 111 the defense of Brussels; "shots were fired wildly
poraries, it was the genius to create new techniques, not the skill to
through the wmdows, blood spilled in many places," and the structure
employ existing ones, that was required. In fact, Quetclet almost never
was then "converted into a fortress" and "surrounded with ditches and
used mathematics in his empirical work on statistics, and he developed
ramparts. "zo In the end, the observatory and Quetelet's position as royal
no new evaluative tools that might have been put to use by others. Al-
astronomer were salvaged--after serious discussion of using the observ-
though he incessantly exhorted his colleagues to furnish error estimates
atory as a magazine-but the "intellectual movement" which he felt to
along with mean or composite numbers, he never did so in his own writ-
have been taking hold, mostly as the result of his own efforts, was re-
ings. The most sophisticated test Quetelet ever applied in published
duced to nought. 21 If the eloges Quetelet wrote as permanent secretary
work to assess the reliability of his numbers was to divide the data at ran-
of the Brussels Academy are to be believed, virtually no Belgian savants
dom into two or three groups and then compare their respective means./
passed through the revolution without serious damage to their careers. 22
Lacking, as he did, the genius to formulate a usable mathematical
He l~ter pronounced it a universal truth that the abrupt changes char-
procedure for analyzing statistical information, Quetelet had recourse to
actensbc of political revolutions invariably bring about a loss of the del-
a strategy especially suitable for deeply ambitious but fundamentally
ICate "living force" of science,,
sensible persons of moderate ability. In his practical work, he compiled
Mecarzique sociale, the direct ancestor of physique sociale and social
and arranged statistic2l data as best he could in order to learn something
correlate of r:zecanique celeste, was originally announced in a paper read
of the composition and perhaps the causes of aggregate phenomena like by Quetelet m ~ar~h 1831. This, Quetelet's earliest paper after the Sep-
natality, mortality, marriage, crime, and suicide. He gave tables of these tember revo~ubon, mtroduced the first of what soon became a battery of
events according to age, sex, profession, and place of residence, and dis- metaphors lmkmg the social order to planetary astronomy. Prominent
cussed the implications of his numbers in a coherent and temperate among these was the distinction between the constant forces of nature
manner. At the same time, virtually independent of this practical work, and perturbational forces, generated by the conscious decisions of man.
he developed an extravagant system of metaphors and similes linking the Although Quetelet maintained that in social mechanics, as in astron-
social domain to the theories and even the mathematics of physics and omy: the effect of t~e perturbational force must at first be disregarded,
astronomy. This was the much-vaunted science of social physics. It em- he d1d not delay posmg what he evidently viewed as the critical question
bodied Quetelet's bid to become the Newton of statistics, and not regarding perturbations of the social system. Laplace had claimed tore-
merely-as his friend Villerme kept offering-its nineteenth-century fute the ol.d v~ew that planetary movements \Vcre unstable, and required
Siissmilch. ,s ProVIdential mtercession to restore their natural orbits. "Can the forces
If social physics is to be conceived in part as a testament to the con-
'9 _Quetelet was evidently upset by his own loss of income as a result of the termination of
fidence and ambition of the astronomer, it must also be recognized as a ~~~c lectures at the Brussels Museum; see Villerrne to Quetelet, 7 Dec. 1 834, cahier 25 60 ,

paean to social order in the spirit of gradualist liberalism. The promi-


. w Quc;elet, "Lcttrd M. k Bourgmestre, 15 dec. 1831," AOB. 1 ( 1834), 285; letter to Bou-
nence of antirevolutionary metaphor in its basic structure can perhaps vard, 5 N~v. _1830~ quoted lfl },"seph Lattin .. Quete/et: Statisticien et Sociologue (Louvain,
be partly explained in terms ofQuetelet's experience as a scientific em- 1912), p. ,?2; also Ld. M<nllv, Essal sur la v1e et les omTages de Lambert··Adolphe-)acques
Quetdet~ Annua1re de l'Academie royale des sciences, des /ettres, et des beaux arts de Belgique
pire builder in unstable timeo. The September 1830 Belgian revolution, 41 (1875;, 109-279, pp. !86-18;. ,
which secured the independence of the southern provinces from the " ~ce Qudclet, "Apcr~.u de !'Eta! actud des Sciences Mathematiqucs chez les Belges,"
BAAS ( 18391, p. 58; H1slmre (n. 8:, p. 370 .
united Netherlands, seemed also to wreak havoc with Quetelet's scien- . " Sec Quctelct, Sciences mathematiques (n. 15), esp. chapter on Quetelet's friend Dande-
lm.
,s See letters of 31 Aug. 1835 and 30 April 1837 in cahier 2560, AQP.

47

The lAwsThat GovemChaos


The Laws That Govern Chaos
of man," Quetelet asked, "compromise the stability of the social sys-
After having viewed the course that has been followed bv the sci-
ten1. "zJ
ences in regard to worlds, can we not attempt to follow tl~em with
Notvvithstanding the evident implications of this metaphor, pertur- regard to men; would it not be absurd to believe that among all the
bational forces were no mere secondary effects on the course of the so- things that occur in accordance with such admirable laws, the hu-
cial system. These were the moral forces of man, the living forces of man species alone rests mindlessly abandoned to itself, and that it
progress, providing the impetus to overcome the dead hand of nature possesses no principle of conservation? We do not fear to say that
which threatened to render society "stationary and incapable of amel- such a supposition would be more injurious to divinity than there-
ioration. ">4 Quetelet invested in these forces all the ambivalence evoked search itself that we propose to carry out. '7
by the idea of rapid social change in a cautious, nineteenth-century lib-
eral. He wrote: Quetclet's fascination with the possibility of subjecting ostensibly un-
controlled social phenomena to scientific order was at the heart of his
These forces that characterize man are the living forces of his na- dedication to the concept of statistical laws. Characteristically, he gave
ture, but . . . does there exist something analogous to the principle this idea its fullest development in reference to such events as crime-and
of conservation of living force in nature? What, moreover, is their suicide, the immoral materials for that "moral statistics" which was cen-
destination? Can they influence the march of a system or imperil tral to the early statistical movements in Britain and France as well as
its existence? Or, indeed, are they, like the internal forces of a sys- the Low Countries. In 1827 the French government began publishing
tem, entirely unable to modify its trajectory or the conditions of its records of criminal activity in its Compte general de !'administration de
stability? Analogy would lead one to believe that in the social state la justice criminelle. They revealed, to the amazement of numerous
one may anticipate finding in general all the principles of conser- readers, that criminal activity varied little from year to year. Quetelet re-
vation that one observes in natural phenomena."' ,../ ported in 1829 his shock at the "frightening regularity with which the
same crimes are reproduced. " 28 Andre-Michel Guerry, the Parisian
Although social physics was intended to promote social improvement lawyer and astute analyst of social numbers generally placed alongside
by serving as a guide to the legislator, it served most prominently as an Quetelet as a founder of the science of "moral statistics, " 2 9 was also as-
allegory of stability and lawlike certainty. That Quetelet's new science tonished that crime could display such regularity. "If we consider now
was inspired by his experience during an age of revolution is suggested the infinite number of circumstances that can cause the commission of
most concretely by the preface to a particularly extravagant essay on the crime, ... we will not know how to conceive that in the end result
"Analogies between physical laws and moral laws" which he published their conjunction leads to such constant effects. "3o '
in a book of 1848. He wrote: "In a moment when passions were acutely That a writer like Quetelet, well informed about mathematical prob-
excited by political events, I sought, in order to distract myself, to estab- ability and conversant with the leading works of political arithmetic,
lish analogies between the principles of modern mechanics and what could have been startled by the regularity of these nmnbers seems itself
was taking place in front of me. " 26 More generally, he exalted lawfulness to be a bit surprising. There was, after all, a rich tradition of admiration
in every domain, holding it quite literally as tantamount to godliness. for the constancy and stability of population statistics dating back to John
This appears clearly in his standard defense against the charge of mate- Graunt's amazing discovery that the ratio of male to female births in
rialism and fatalism:
Quetelet, "Recherches sur le penchant au crime aux differens ages." Ntv1AR, 7 (1832),
separate pagmatton, p. 4.
'' Quctelct, "Recherches sur Ia loi de la croissance de !'hom me," NMAB, 7 ( 18-p), separate 8
' Quetelet, "Recherches statistiques" (n. 14), p. z8.
pagination, p. 7·
' 9 In fact, this phrase, like the concerns that underlay it, had already been current in France
'• Quetelet, "Recherches sur lc poids de l'homme aux differcns ages," N;'vfAB, 7 (1832),
for at least sever~! years. ~ce, for example, Charles Dupin. ''Effet.s de l"emcignement popu!aire
separate pagination, p. 11.
sur lcs, prospcntes de Ia hance," reviewed by Aubert de Vi try in Bulletin universe/ des ~-ciences
'5 Quetelet, '"Loi de croissance·· (n. 2 3), p. 2.
et de I rndustne, 6th s_ec .•. Bulletin des sciences geographiques, 8 (1826), 329-336.
z6 Quetclet, Du systeme social et des lois qui /e regissent {Paris, 184Rl, p. w4.
A. M. Guerry, Essa: sur Ia statistique morale de Ia France (Paris, 18 33), p. 1 1.

49
- - ~- - --- - - - - ~~ ~ - ----- -- ~

(
The Laws That Govern Chaos
-The Laws That GovemChaos
which are human actions, like every other natural event are deter-
London exceeded unity by a small proportion every year, without ex-
mined .by universal laws. However obscure their causes, history,
ception. Jakob Bernoulli's demonstration that p:obabilistic events of any
which IS concerned with narrating these appearances, permits us to
sort can be expected to give rise to stable ratios If repeated often enough
hope that if we attend to the play of freedom of the human will in
was intended as an explanation for regularities such as this, and it was
repeated in most subsequent general works on mathem~tical probabil-
the large, we may be able to discern a regular movement in it, and
that what seems complex and chaotic in the single individual may
ity, including that of Laplace. Quetelet' s reaction makes 1t clear that the
be seen from the standpoint of the human race as a whole to be a
extension of this principle to events so irrational, disorderly, and anti-
steady and progressive though slow evolution of its original endow-
social as crime had not been anticipated. ~lent. Si~ce the free will of man has obvious influence upon mar-
In fact, demographic regularities were generally viewed as exemplars
na~es, bnths, and deaths, they seem to be subject to no rule by
of natural theology, or at least of the harmonies of nature, even by wnt- which the number of them could be reckoned in advance. Yet the
ers who accepted and understood Bernoulli's argument. The mathe- annual. tables of them in the major countries prove that they occur
matician De Moivrc, like John Arbuthnot and William Derham, held accordmg to laws as stable as [those o~ the unstable weather which
up the regularity of births, marriages, and even of deaths by age as c~m­ we !ike:vise cannot determine in advance, but which, in th~ large,
pelling evidence of the wisdom and benevolence of the Creator. Suss- mamtam the growth of plants, the flow of rivers, and other natural
milch pointed to the birth differential and to the mortality rates of boys events in an unbroken, uniform course. Individuals and even
and girls to establish that the totality of these processes led to a perfect w~1ol~ peoples think little on this. Each, according to his own in-
balance of the sexes by the time the age for marriage was reached, thus clma~wn. f?llows his own purpose, often in opposition to others; yet
facilitating the great goal of all human activity: maximal population in- each md!Vldual and people, as if following some guiding thread, go
crease. He rejoiced also in the stability of death tables, noting that tt toward a natural but to each of them unknown goal; all work toward
yielded "extraordinarily great advantages for the state and other arrange- ~urthering it, even if they would set little store by it if they did know
ments in the life of man, for it contains the basis and points to the rules, 1t. 33
according to which annuities and tontines may be arranged and deter-
mined. "v It also guarantees that young men and women wtll ex 1st m Until the nineteenth century, then, statistical regularity was generally
equal numbers at the age of ~arriage. N~ blind accident, he r:~~arked, sc.cn as pertammg to the natural history of man, and as indicating divine
"could place goal and means m so beautiful a umon together. w1sdom and planning. The first well-publicized instance of statistical or-
Si.issmilch's table attracted the interest of Immanuel Kant, who was der which could not be plausibly interpreted in this way"was Laplace's
deeply impressed by these instances of the emergence ~f larg~-scale reg- announcement in the Philosophical Essay on Probabilities that the
number of dead letters in the Paris postal system was constant from year
ularity from local chaos. Kant used this truth of pohbcal anthmetJc to
illustrate how, despite the undirected behavior of every md1vl~u~l,_ the
to year. The uniformity of murder, theft, and suicide was even more dif-
philosopher might hope to "discover a natural p~rpose in this Jdwttc
ficult to explain in natural-theological terms. Quetelct was, in some
~en~e, able to do so, but only by embracing a cosmology at once phys-
course of things human . . . . a history of a defimte plar.1 for creah:res
Icalist and theological that made mass regularity the expected outcome
who have no plan of their own." Teleology could preva1l111 h1story, JUSt
of natural processes in every domain.
as order emerges in political arithmetic: Quetelet interpreted the regularity of crime as proof that statistical
\Vhatever concept one may hold, from a metaphysical point of laws .may be true when applied to groups even though they are false in
view, concerning the freedom of the will, certainly its appearances, relation to any particular individual. Beyond that, he implied that the
1• j. P. Si.issmilch, Die giittliche Ordmmg in den Verdnderungen des menschlichen Ce-

schlechts (2 vols., Berlin, 3d ed., 1765), vol. 2, pp. 366-367. 12 ~' Immanuel Kant, On His tory ( 1 784), Lewis White Beck, ed. (Indianapolis, 196 3), pp, 11 _

l2 Ibid., p. z89.
The Laws That Govern Chaos
Thetiiws That Govern Chaos
ulties. "3 6 Quetelet allowed his average man a temporal dimension as
obliteration of the particular by the general was responsib~e for the very well as physical and moral ones, thus giving him a mean rate of growth
preservation of society. The prevalence of m.ass regulan.ty, ,he \~rote, and moral development over his average life.
"teaches that the action of man is restramed Ill such a cnclc, that the The calculation of l'homme moyen physique was no problem for Que-
great laws of nature are forever exer~pt~d from his i~fluence; 1 ~ ~~~~ telet, since it merely involved collecting measurements of height,
shows that laws of conservation can ex1st m the moral \\orld, JUSt as tne. weight, and the dimensions of the various limbs and organs. L'homme
are found in the physical world. ">4 As a consequence. of the fundamen- moyen moral, on the other hand, offered certain obstacles, for human
tal truths revealed by the uniformity of statistics, soCie~y could be seen individuals do not present themselves to the scientist endowed with
to be an entity in its own right, independent of the wlmns and Jdwsyn- measurable quantities of courage, criminality, or affection. In this re-
cracies of its constituent individuals. . spect, the average man was a far more tractable problem than the con-
Indeed statistical regularity seemed to Quctelet to provtde the key to crete individual. In principle, wrote Quetelet, the courage or criminal-
· 1 ·'
SOCia SCience.
I-I e cast the "law of large numbers," .Poisson's
. form of ity of a real person could be established if that person were placed in a
Bernoulli's theorem, as the fundamental axiom of soCial physiCS. It con- great number of experimental situations, and a record kept of the num-
firms, after all, that general effects in society are always produc~d by ber of courageous or criminal acts elicited. This would be interesting,
general causes, since chance, or accidental causes, can have. no mflu- but it was wholly unnecessary for social physics. Instead, the physicist
ence on events when they are considered collectively. In p.artlcular, so- need only arrange that courageous and criminal acts be recorded
cial facts can never be generated by the arbitrary, unmotivated, sp~n­ throughout society, as the latter already were, and then the average man
taneous, or otherwise inexplicable acts of the hum.an free will. could be assigned a "penchant for crime" equal to the number of crim-
However, "capricious" and "disordered" may _be the a~t10n of the ~u­ inal acts committed divided by the population. In this way, a set of dis-
man will, "whatever concerns the human species, constdered en ma .. se, crete acts by distinct individuals was transformed into a continuous mag-
belongs to the domain of physical facts; the greater the number of t.n- nitude, the penchant, which was an attribute of the average man. >7
dividuals, the more the individual will is submerged benea_th the se~Ies It seems wrong to argue, as Paul Lazarsfeld did, that Quetclct's
of general facts which depend on the general causes accordmg to whtch method for assigning numbers to social attributes vvas independent of his
society exists and is conserved. " 35 idea of the average man. 38 His "penchants" were, after all, traits of
Quetelet' s most celebrated construct, l' hom me moyen,. or the a\:er~ge l'homme moyen, and it is far from clear that the assignment of numerical
man was similarly dependent on his belief in the regulanty o~ statist! cal values from aggregate statistics to this abstract being involves a genuine
even~s. Quetelet maintained that this abstract being, defined m terms of "contribution to quantification in sociology" except insofar as the con-
the average of all human attributes in a giv~n country'. could be ~rea ted cept of the average man proved to be a useful tool. Quetelet thought this
as the "type" of the nation, the representative of a soClety m soc~al SCI- a valuable exercise largely because l'homme moyen could be used as an
ence comparable to the center of gravity in physics. After all, dev:at10ns analogue to certain concepts in celestial physics. In addition, he found
from the average must necessarily cancel themselves out whenev~: a it appealing to distribute responsibility for crime over the whole com-
great number of instances is considered. Hence for the aver.age man all munity. To speak of the tendency to crime within a given age class was
things will occur in conformity with the mean resu~ts obtam~d for~ so-
ciety. If one seeks to establish, in some way, the basiS of a soctal ~h}SJCS,
' 6 Quctclet. Sur l'homme et le developpement de ses facultes, ou essai de physique socia/e
(1835; 2 vols., Brussels, 1836), vol. 1, pp. zr-zz.
it is he whom one should consider, without disturbing oneself w_tth p~r­ r Quetelet. "Penchant au crime" (n. 27) and "Sur la statistiquc morale et les principes qui
doivent en former la base," N,lv1AB. 21 ( 1848). Quctelet was not the first to seek statistical in-
ticular cases or anomalies, and without studying wheth~r some g!V~n m- dices of moral attributes. The prominent Italian statistician Melchiorre Gioja, Filosofia della
dividual can undergo a greater or lesser development m one of hts statistica (1826; Mendrisio. 1839). p. 570, wrote: "The most certain measure of immorality
is, as everyone recognizes, the ratio of the number of crimes to the number of inhabitants."
,. Quetelet, "De !'influence de libre arbitre de J'hommc sur les faits sociaux," BCCS. 3 ' 8 Paul Lazarsfeld, "Notes on the HIStory of Quantification in Sociology-Trends, Sources,
and Problems," Isis, 52 (1961), 164-181.
(184;), 135-155. p. 136. . "
"Quetclet, "Penchant au cnmc (n. 27). p. So.
53
,
( (
Tltc.Laws ThaLGovern CbqQs_ _ The LawsTI;_at Govern Chaos
more compatible with Quetelet's reformist impulse and his sociological realization of widespread hopes for an exact social science throughout
aims than to discuss crime in terms of the wicked acts of a specifiable his life, not a single trajectory calculation is to be found in all of Que-
group of malignant persons. telet's works. Quetelet the practicing statistician was always more sen-
Crime, then, was a social phenomenon, whose frequency, like that sible than Quetelet the social physicist.
of marriage, should be attributed "not to the will of individuals, but to Quetelet's lasting contribution to science, though related to his social
the customs of that concrete being that we call the people, and that we ideas, was more abstract. He succeeded in persuading some illustrious
regard as endowed with its own will and customs, from which it is dif- successors of the advantage that could be gained in certain cases by turn-
ficult to make it depart. "39 The cause of crime, he explained, may be ing attention away from the concrete causes of individual phenomena
identified precisely as the "state of society." Since Quetelet's often prolix and concentrating instead on the statistical information presented by the
discussions of crime contain only the vaguest hint of what aspects of the larger whole. Quetelet implied that the viability of this approach was not
social condition were responsible for it (he was better on the causes of uniquely tied to society, but was an immediate and general corollary of
fluctuations in the marriage rate), his analysis hardly seems a landmark the universal truth that constant causes must yield constant effects. His
in the history of criminology. His object, however, was not so much to metaphorical science of social physics became an important source of
propose an immediate cure for crime as to establish the need for statis- analogies for scientists in other fields, for it seemed to reveal plainly that
tical analysis as part of a campaign to remove its causes gradually. 40 statistical laws can prevail for a mass even when the constituent individ-
Crime, he wrote, is like a "budget that is paid with frightening regular- uals are too numerous or too inscrutable for their actions to be under-
ity," or "a tribute that man acquits with greater regularity than that stood in any detail.
which he owes to nature or the state treasury.". Indeed, "every social
condition presupposes a certain number and a certain arrangement of
LIBERAL POLITICS AND STATISTICAL LAWS
offenses as a necessary condition of its organization. This observation,
which may appear discouraging at first, becomes, on the contrary, con-
soling, when it is examined closely, because it shows the possibility of Notwithstanding Friedrich Engels' disdainful remarks on "utopian so-
improving men, by modifying their institutions, their qJstorrls, the state cialism," few if any of the pioneer social scientists of the early nine-
of their enlightenment." Really, he concluded, "it only presents us an teenth century believed that the great reorganization of society fore-
extension of a law already well known to all philosophers who have oc- shadowed by their theories was to be a pure triumph of their own
cupied themselves with society in relation to physics: that so long as the inventive genius. Comte and the St. Simonians, for instance, argued
same causes subsist, one may expect the continuation of the same ef- explicitly that social and scientific developments had both made possible
their ideas and prepared the way for their actualization. They conceived
fects. " 4 '
In retrospect, Quetelet's importance as a social scientist seems mod- themselves as discovering, not inventing, the social condition of the fu-
est, for his social physics was and remained completely unworkable. His ture.
great goal was to measure the changes experienced by l'homme moyen Quetelet, unlike the St. Simonians, was no dialectician; there were
over time in order to ascertain the general law of social development. no "critical" epochs required in his version of the historical process.
That is, by plotting the trajectory of the average man through history, he Progress, he believed, was smooth and continuous, and he saw no ad-
hoped to discover the forces acting on the "social body" and ~ence to vantage in a radical change of social organization. The course of the so-
predict its future course. Although he continued to speak of this as the cial body through history was along a path that was inherently progres-
sive, towards the eminently desirable goal of effacing l'homme physique
39 Quetelet, "Libre arbitre" (n. 34), p. 14~. . • , , . . by l'homme intelectuelle. ·P To get there it is most efficient and least trou-
"" See Quetelet, "Sur le poids de l'homme aux d1fferens ages, Annales d hygzene pubilque
et de medecine legale, 10 (1833). s-:n.
•• Quetelet, Sur /'homme (n. 36), vol. 1, pp. 8-ll. " Ibid., vol. z, p. 285.

54 55
The Law.>That Govern Chaos TheLaws That Govern Chaos
blesome to follow a straight line; Quetelet was a "rigorous partisan of the autonomous domain of society. Statistics supported a liberal, mildly bu-
principle of least action that the Supreme Being follows in all things. " 4 ' reaucratic politics, even as it exalted the practical hmction of the social
The cause of this secular progress, according to Quetelct, was the liv- scientist and assured him that society was on course.
ing "moral force" of science and learning, which tended alw~ys to in- The notion of statistical law achieved its fullest expression in Great
crease. Since this force was conserved whenever it acted contwuously, Britain during the 185os, the decade when laissez-faire liberalism
social change could not be averted by even the most skillful reactionary reached its intellectual apogee. Liberals and even progressive conser-
government. Any attempt to block the course of progress could only re- vatives became increasingly sympathetic to the idea that the role of gov-
sult in an accumulation of grievances, a deep well of repressed forces ernment could and would be sharply reduced, while radicals propagated
that must sooner or later escape in a violent explosion. This would lead the claim that the historic role of government had been to protect re-
to a wastage of some of the moral force, accompanied by much misery. gressive interests from the forces of change, to impede the natural course
Even revolutions, fortunately, could do no more than inflict a tempo- of progress, and to obstruct the general tendency towards prosperity and
rary setback. When Quetelet wrote to Prince Albert of Britain in June freedom. Social science began at this time to flourish in the rationalistic
form of conjectural history," 6 which, in its British guise, regarded his-
1 8 4 8 lamenting the "veritable moral cholera, which spreads its ravages
over the whole of Europe," he was able to find solace in the laws of so- torical development as intrinsic to society and largely beyond the influ-
cial physics: "Scourges strike the moral as well as the physical structure ence of kings and legislators. The recent successful campaign to repeal
of humanity and however destructive their effects may be, it is at least the Corn Laws suggested that the enlightenment of society would at last
consoling to think that they cannot in any way alter the external laws compel government to retreat into its proper sphere.
that guide us. Their action is transitory and time will soon have healed Expressions of amazement and admiration at the impressive regular-
the wounds of the social body; but it is not the same with individuals. " 44 ity uncovered through statistical research can be found in a varietv of
'- Given that society was governed naturally by statistical laws, its polit- mid-century writings. Lord Stanley, a leader of the Conservative p;rty,
claimed as an "axiom" of social science the maxim that "the moral and
ical government was constrained to an ancillary role. The wise legislator
physical condition of the human race" was governed by constant statis-
would not try to impose his will on the social system, but \vould seek
tical laws. 4 7 The economist Nassau Senior noted the remarkable result
first to determine the direction and magnitude of secular social evolu-
that "the human will obeys laws nearly as certain as those which regulate
tion-that is, of the average man (constancy was a characteristic of sta-
matter. " 48 Robert Chambers observed in the Vestiges of Creation: "Man
tisticallaws only in the short term). Once the contribution of the con-
is seen to be an enigma only as an individual, in mass, he is a mathe-
stant causes was known, that of the accidental causes could be found
matical problem. " 49 Henry Holland proposed that the "law of averages"
through a process of vector subtraction. It would then be possible to re-
had "acquired of late a wonderful extension and generality of use; at-
store a just state of equilibrium, or to minimize the perturbations, by
taining results, from the progressive multiplication of facts, which are
opposing this resultant with an equal and oppositely directed force.
ever more nearly approaching to the fixedness and certain tv of mathe-
matical formulas. ";o Even Charles Dickens was sufficient!~ impressed
"The whole art of governing," he proclaimed, "resides in estimating the
nature and direction of this resultant. It is necessary to know perfectly
by a recent report of the Registrar General on marriage to print the fol-
the forces and tendencies of the factions that ordinarily divide a state, in
40 Sec J. W. Burrow, Evolution and Society: A Study in Victorian Social Theory (London,
order to judge the most appropriate means for combating and paralyzing
1970!, and J.D.Y. Peel, Herbert Spencer: The Evolution of a Sociologist (New York, 19"'1).
them."45 Certainly Quetelet's social physics did not eliminate the need 47 Lord Stanley, "Opening Address ... as President of Section F ... , "TRSS, 19 ( 1,856).

for an active state, but neither did it give the state much control over the 305-310, p. 305.
4 ' Nassau W. Senior, "Opening Address ... ," JRSS, 23 (t86o), 357-362, p. 359·

., Quetelet, Du sysieme social (n. 26), p. 110. . . 4" Robert Chambers, Vestiges of Creation !New York. 1846), pp. 333-334, quoted in C. C.
44 Quoted in Harriet H. Shoen, "Prince Albert and the Application of StallstlCS to Problems Gillispie, Genesis and Geology (New York, 1959), p. 1 57·
50 [Henry Holland], "Human Longevity," Edinburgh Review, 105 (1857), 46-77, pp. 54-
of Government," Osiris, 5 ( 1938), z86-z87.
45 Quetekt, Du systeme social (n. z6), p. 289. 55·

57
The Laws"l'hat CovernChaos The Laws That Govern Chaos

lowing reflections by Frederick Knight Hunt in Household Words: ''Not and open thought, where every man, we firmly believe, obtains the
content with making lightning run messages, chemistry polish boots, reward which his merit deserves, [the society should not seek gov-
and steam deliver parcels and passengers, the savants are superseding the ernment support]. . . [W]e certainly act upon this maxim, that if
astrologers of old days, and the gipsies and wise women of modern ones, our society cannot support itself by its own intrinsic merits, and by
by finding out and revealing the hitherto hidden laws which rule that its own usefulness, we arc quite ready to close it tomorrow."
charming mystery of mysteries-that lode star of young maidens and gay
Ncwrnarch interpreted the laws of statistics in the same spirit. Writing
bachelors-matrimony.">! In the same vein, William Fan, effective
in the Economist newspaper, he explained how governments had begun
head of the General Register Office, modestly remarked:
to learn that legislation could only accomplish its aims if it was formu-
It would formerly have been considered a rash prediction in a mat- lated in accordance with the natural principles of these things. Official
ter so uncertain as human life to pretend to assert that 9000 of the support for statistics, he proposed, embodied a desperate attempt to un-
children born in 1841 would be alive in 1921; such an announce- derstand the condition of society, deriving from
ment would have been received with as much incredulity as Hal-
ley's prediction of the return of a comet, after the lapse-of 77 years. nothing less than the necessity under which all Governments are
What knew Halley of the vast realms of aether in which that cornet rapidly finding themselves placed, of understanding as clearly and
disappeared? Upon what grounds did he dare to expect its re-ap- fully as possible the composition of the social forces which, so far,
pearance from the distant regions of the heavens? Halley believed Governments have been assumed to control, but which now, most
in the constancy of the laws of nature; hence he ventured from an men agree, really control Governments. The world has got rid of a
observation of parts of the comet's course to calculate the time in good many intermediate agencies, all of them supposed originally
which the whole would be described; and it will shortly be proved to be masters, where in truth, they were even less than servants.
that the experience of a century has verified quite as remarkable The rain and the sun have long passed from under the administra-
predictions of the duration of human generations.,, tion of magicians and fortune·· tellers; religion has mostlv reduced
its pontiffs and priests into simple ministers with ver~ circum-
The political viewpoint that underlay much of this talk is more readily
scribed functions; commerce has cast aside legislative protection as
apparent in some remarks made by William Newmarch. When the In-
ternational Statistical Congress met in London in r86o, Newmarch a reed of the rotten est fibre; and now, men are gradually finding out
greeted the assembled native and foreign dignitaries on behalf of the Sta- that all attempts at making or administering laws which do not rest
tistical Society of London with the following observations, which, it is upon an accurate view of the social circumstances of the case, are
recorded, were received with great applause. neither more nor less than imposture in one of its most gigantic and
perilous forms ....
On the part of the Statistical Society, which is an entirely voluntary Crime is no longer to be repressed by mere severity; Education
association of individuals, not connected with the State, and l is no longer within the control of the maxims which preceded
think I may say, not in the smallest degree desiring to be connected printing,-Law is found to be a science perhaps the most difficult
with the State, we pride ourselves upon our entire independence, of any-Justice means more than tricks and plausibilities of pro-
we feel that if we are to maintain our ground in this country of free cedures;-Taxation, Commerce, Trade, \Vages, Prices, Police,
s• [F. K. Hunt]. "A Few Facts about Matrimony."' in Charles Dickens, ed., Household Competition, Possession of land.-every topic from the greatest to
Words, 1 (1850), p. 374 (for authorship, see Anne Lohrli, comp., House/wid Words :Toronto, the least which the old legislators dealt with according to ... ca-
1973]. p. 319).
''William Farr Iwritten under the name of the titulary head of the C.R.O .. George Gra-
" William Newmarch, Opening Address. in of the Proceedings of the Fourth Session
ham]. "Report,'" in.Fi{th Annual Report of the Hegistrar General of Births. Deaths. and Mar-
of the 1ntemational Statistical Congress, 186o 186r i, p. ll6.
riages(1843). p. 21.
59
The Laws That Gvwrn Chaos
The La'lvs That Govern Chaos
price . . have all been found to have laws of their own, complete
from the true way because its people were excessively dependent on the
and irrefragable. 54
state, while Scotland and Spain revealed in their histories the distortions
The most determined and influential effort to harness the laws of his- caused by the excessive power of religious institutions. The perturba-
tory in opposition to government meddling was Henry Thomas Buckle's tions of Germany and the United States, involving respectively an excess
ambitious but uncompleted project, the History of Civilization in Eng .. In the.concentration and in the diffusion of knowledge, were to make up

land. Buckle was the child of a prosperous Cambridge-educated mer- the thnd volume of the introductory material, but Buckle died before he
chant father and a devout Yorkshire Calvinist mother, and if his repu- could write it. Liberal England, that model of healthy development and
tation had remained what it was during the nineteenth century, the the ostensible object of the whole exercise, was discussed only inciden-
childhood of this lifelong bachelor would doubtless have been memo- tally in the volumes that were actually completed.
rialized by a psychobiographer. As it is, we know that he was a spoiled Buckle's avowed goal was to elevate historv to the rank of the other
child, raised with little direction or discipline, and that he became rad- sciences. Although history "deserves the high~st faculties of science," he
icalized in politics as in religion around the time of his grand tour of the w.rote: the pu~;ic had become accustomed to such mediocrity among
Continent, in 1841. The influence of the maternal religion, however, h1stonans that any author who from indolence of thought, or from nat-
can perhaps still be seen in the deterministic cast of mind revealed by ural incapacity, is unfit to deal with the highest branches ofknowledge,
his early enthusiasm for phrenology and by the explicit arguments in his has only to pass some years in reading a certain number of books, and
History. The introductory section of his work-the second volume of th~n he is ~ualified to be an historian. "56 Buckle denounced the presen-
which, incidentally, was dedicated to the memory of his mother-in- tation of history as a chronicle of kings and battles and insisted on the
terpreted belief in strict scientific causality as derived from religious fa- need to look beyond the surface confusion of particular events in order
talism. 55 t~ establish the simple, general, underlying principles. Such investiga-
Buckle was an archetypical autodidact, conversant with if not master tion wo.u.ld demonst~ate, he proposed, that the substance of history was
of a variety of fields and authors. His study originated as a complete his- not polttJcs, but society-not court intrigues, declarations of war, and
tory of the world, constructed on entirely new principles, but he soon issuance of church decrees, but the slow, continuous advancement and
found that a radical new interpretation required new research and that diffusion of knowledge. Government activity might seem whimsical and
one lifetime would not suffice to finish such a project. He decided, re- unpredictable, but the regular course of social development confirmed
luctantly, to confine his ambitions and to write instead a history of Eng- that "in the moral world, as in the physical world, nothing is anomalous;
land. The wisdom of this choice he justified in terms ofQuetelet's met- nothing is unnatural; nothing is strange. All is order, symmetry, and
aphor of social physics. England, he explained, had been the least law."s7
affected of all countries by destructive perturbations, and its progress re- Buckle's fondness for an intellectual and social approach to history re-
flected most clearly the natural evolution of society. All other societies flec.ted his political viewpoint. His book carried on a sustained polemic
had somehow been displaced from the preordained path of progressive agamst corporate entities such as governments and religious institutions.
liberalization. This appeared clearly in his general introduction and in Church and court were mere superstructures, lacking any autonomous
the diverse studies of social pathology which were to fill the first three principles of development, whose only contribution to history had been
volumes of the general history, thereby permitting him to salvage there- to hold society to the institutional forms of previous ages. Buckle was
sults of his first ten years of research. France, it seemed, had departed consistently cynical of the motives that determined the activities of
~ublic institutions. He wrote of Scottish preachers, for example, that
s• [William Newmarch], "Some Observations on the Present Position of Statistical Inquiry
hke every corporation, which has ever existed, whether spiritual or
with Suggestions for Improving the Organization and Efficiency of the International Statistical
Congress," jRSS, 23 (186o), 362-369, pp. 362-363.
" There is no good modern work on Buckle's career. On his life, sec A. H. Huth. The Life , 56 Henry Thomas Buckle, History of Civilization in England (2 vols., 1 8 57_1 86 1 ; New
and Writings of Henry Thomas Buckle (New York, 188o). )ork, 1913), vol. 1, p. 3;secalsop. 671.
57 Ibid., vol. 2, p. 25.

6o
~The LawsThat Govern Chaos The Laws That Govern Chaos
temporal, their supreme and paramount principle was to maintain their introspection the rules that operate in individual minds, they prove
own power. ">8 But it mattered little whether rulers were generous or themselves wholly accessible to the historian who has learned the essen-
self-interested. Charles III, he observed, had the best interest of his peo- tial lesson of statistics. Buckle cited the proven regularities of statistical
ple in mind when he sought to free them from Spain's superstitious and social science in order to persuade his readers that mankind was no ex-
bigoted clergy. In the absence of enlightenment, however, such efforts ception to the universal reign of natural order. The instances he seized
were "worse than futile" for they served only to strengthen the people's upon were the very ones that had most impressed Quetelet, and were
sympathy for their clerical oppressors. 59 taken directly from the work of the Belgian statistician-murder, sui-
On occasion, Buckle thought, unwise policy had led to lamentable cide, the misaddressing of letters. As Quetelet himself had marvelled,
consequences. By blocking the development of Protestantism, the lead- not only do murder and suicide occur at a constant rate in any given so-
ers of France, Spain, and Italy had locked their nations into a system of ciety, but so do the categories of those committed with guns, knives, or
beliefs that could only confuse and retard their development, producing poison, and by asphyxiation, hanging, or drowning. However diverse
disequilibrium and stimulating social turmoil. In the long run, how- and irrational the behavior of individuals may seem, the collective reg-
ever, institutions were powerless, and Buckle scoffed at the "folly of ularity of human behavior proves that each act is a necessary conse-
lawgivers" to think that their enactments could affect any phenome- quence of unvarying social laws.
non-such as suicide-that was governed by general laws. 60 Legisla- Indeed, in his insistence on the universality oflaw, Buckle even went
tion, always derivative in its development, was at best a measure, and beyond Quetelet. The latter, ever faithful to the principle of modera-
never a cause, of progress. "No great political improvement, no great tion, aimed to promote the influence of science on legislation, and he
reform, either legislative or executive, has ever been originated in any had no desire to entangle statistics in sectarian controversy. To this end,
country by its rulers," 6 ' Buckle wrote. Moreover, "every great reform he worked out a compromise position, a reconciliation of statistical law
which has been effected has consisted, not in doing something new, but with free will that would exonerate statistics from the stigma of materi-
in undoing something old. " 62 "Lawgivers are nearly always the obstruc- alism and fatalism without giving up the possibility of a rigorous social
tors of society, instead of its helpers. " 6 > . science patterned after celestial mechanics. The heart of his argument
Statistics was more critical to the justification of Buckle's project than was a distinction between the laws of society and those ofindividuals.
to its execution, and social numbers were not cited in the substance of Society, he held, is the product of a social contract, into which individ-
his historical discussion. Buckle invoked the results of statistics as com- uals have invested a portion of their freedom. Even the laws of society
pelling evidence for the existence of social laws, rather than as instances are not so rigid that they cannot be altered through the activity of the
of the general principles governing history; indeed, his rhetoric of per- legislator-who seems somehow to stand outside the universe oflaw. In-
fect statistical regularity was not easy to square with an approach that dividuals participate in the laws of society in the sense that they are all
presupposed the inherent dynamism of society. Nevertheless, the sue~ accidental variants of the average man, and hence possess, as a first ap-
cess of statistical science provided, in Buckle's view, a crucial lesson to proximation, identical penchants for crime, suicide, marriage, and so
the historian. Isolated events always seem unpredictable and confusing, on. Still, the "moral causes that leave their mark on social phenomena
and the historian must not lose himself in their intricacies. Instead, he are ... inherent in the nation, not in individuals, " 64 and the individ-
must take a broad view and seek general principles. Although social phe- uals have been accorded a limited domain for the exercise of their free
nomena have proven intractable to metaphysicians, who seek through wilL The combined effect of a multitude of free decisions is always to
neutralize them. Accordingly, the will, though free, can have only a tri-
<8 Ibid., p. 193. fling effect on society as a whole. Quetelet relegated it to the domain of
;o Ibid., p. 91.
6o Ibid., vol. 1, p. JC)ll. accidental causes, and insisted that the constant causes which were
5, Ibid., p. 198.
6' Ibid , pp. 199-zoo.
64 Quetelet, "Statistique morale" (n. 37), p. 6. Sec also "De l'inHncncc" (n. 34), and Du
"' /bid., vol. z, p. 2.;4. systiime social (n. z6), pp. 71·73.
( (
The Laws]'hat GovernChaos
The Laws That Govern C71cios
bee~ ~artially arrested. He expressed admiration and astonishment that
alone of interest to social physics were quite uninfluenced by human de~1abons from the "moral laws" revealed by statistics were so small, de-
freedom. spite the complicated condition of the social world. "From the circum-
Buckle accepted this distinction between levels of human existence, stance that. the discrepancies are so trifling," he remarked, "we may
but denied that freedom pertained to individuals any more than it did to form some 1dea of the prodigious energy of those vast social laws which
societies. The autonomy of the will is a mere fiction, appropriate for cul- tho.ugh constantly interrupted, seem to triumph over every obst;cle, and
tures dependent for their maintenance on hunting and gathering or sub-
which, when examined by the aid of large numbers scarcely undergo
ject to numerous tornadoes and earthquakes, but not for a prosperous any sensible perturbation. "67 '
and educated modern society that was increasingly taking charge of its
. Buckle was possibly the most enthusiastic and beyond doubt the most
own fate. In Buckle's view, the possibility of scientific history would be
I~fl~ential popularizer of Quetelet's ideas on statistical regularity, but
contradicted if some "mysterious and providential" force, such as divine
similar remarks may be found in publications ranging from the Journal
authority or unfettered human will, could act on society. The funda-
of.the. Institute of Actuaries to the Revue des deux mondes. 68 Quetelet;s
mental prerequisite of scientific history was the absolute universality of
pnncipal books were translated into English and German, while Buck-
law, the denial of"chance or supernatural influence. " 6 s The regularities
l~'s two completed volumes went through edition after edition in Eng-
of statistics, he thought, were compelling not just as evidence of the law-
lish, French, German, Russian, and various other languages. Quetelet
fulness of social evolution, but also of individual actions. If human be-
havior were subject to free will, we could expect to see no relation be- and ~uckle we~e read, moreover, by a generation of Europeans who
tween the quantity of vicious behavior and the state of society. Indeed, w.ere I.mbued .with a sense of society as a fundamental and preeminently
nothing would seem more disorderly or incomprehensible than h1stoncal entity that was capable of having its own laws. Statistics ac-
transgressions against the moral order. Yet crimes occur in the same quainted these readers with a new form in which natural laws could re-
numbers yea~ after year, obeying ostensible social laws that left no place veal themselves, one whose distinctive features were noted by writer
for divine guidance of history, metaphysical free will, or individual after writer. ~he individuals are so numerous, and subject to so complex
moral responsibility. Suicide, for example, an. ~rray o~ cucumstances, that it is impossible to foresee with any reli-
ability then future behavior. Yet whenever a large number of individ-
is merely the product of the general condition of society .... The uals is considered at once, "the influence of contingencies seems to dis-
individual felon only carries into effect what is a necessary conse- appear before that of general laws. "69
quence of preceding circumstances. In a given state of society, a The principle of statistical regularity, or of the stability of the mean
certain number of persons must put an end to their own life. This ~as invo~ed incidentally in a variety of works on social and political sub~
is the general law; and the special question as to who shall commit Jects dunn~ the nineteenth century. Charles Morgan, following Que-
the crime depends of course upon special laws; which, however, in telet, explamed how legislation need not take account of numerous in-
their total action, must obey the large social law to which they are dividual natures, but could instead be founded upon a knowledge of the
all subordinate .... [T)he offences of men are the result not so
much of the vices of the individual offender as of the state of society 67 Ibid., p. Z3.
68 Samuel Brown, :·on \?e Uniform Action of the Human Will, as exhibited by its Mean
into which that individual is thrown. 66 Results m Social Stat~shcs? Assurance Magazine and Journal of the Institute of Actuaries z
(185z), 341-351; Louis Etienne, "Le positivisme dans l'histoire," Revue des deux mondes '74
The existence of statistical laws also demonstrated the impotency of (1868), 375-408. '
government to sustain the malicious influence that it had always worked 6o [Herman Merivale], "Moral and Intellectual Statistics of France "Edinburgh Review 6 9
numb~rs "was also invoked by
(1839), 49-74, P· 51. Statistical regularity, or "the law oflarge
on society. In the long run, Buckle maintained, the progressive laws of the la1ssez-faue economist Frederic Bastiat, who argued that the spont~ncous order of a free
social development must prevail even in nations where their action has econm_ny, reveal~d m the existence of stable wage and interest rates, is wholly analogous to the
operation of an msurance company. See Bastiat, Economic Harmonies (1851), W. Hayden
6s Buckle, History (n. s6), vol. 1, p. 6. Bayers, trans. (Pnnceton, 1964), chap. 14, pp. 361-406.
66 Ibid.' pp. Z0-22.
The Laws That Govern Chaos The Laws That Govern Cha()s
"abstract being 'man' " which can be attained through statistics.~' by citing the regularity of dead letters. 7 ; M. L. Wolowski, who held that
George Cornewall Lewis proposed that the law of averages could be em- "the law of large numbers governs the moral world, as it governs the
ployed to determine the "prevailing character" of a government, the physical," argued that statistical laws provided essential empirical con-
"medium state between opposite extremes." He wrote: "\Ve may be un- firmation for economic deductions. 76 Mark Pattison, in his review of
able to predicate any invariable and universal tendency of a form of gov- Buckle, asserted that "social history can only be composed upon statis-
ernment, just as we may be unable to say that all men live a certain tical data," for in "dealing with the individual human being, everything
number of years. But as we can say of men, that the average duration of is uncertainty; it is only of man in the aggregate that results can be cal-
their life is a certain number of years-so we may say of a form of gov- culated with accuracy. "77 Perhaps the most enthusiastic statistician of all
ernment, that it has a certain prevailing average character. " 7 ' John was Florence Nightingale, who was enthralled by the laws Quetelet dis-
Stuart Mill mentioned statistics as characteristic of his "Inverse Deduc- covered and who sought to integrate statistics into administration not
:ive Method," and began in 1862 to cite Buckle in his Logic for the prin- only by collecting and popularizing pertinent numbers, but also by en-
ciple that the "very events which in their own nature appear most ca- dowing statistical research or setting up a tripos at Oxford based on Que-
pricious and uncertain, and which in any individual case no attamable telet' s Physique sociale. 78
degree of knowledge would enable us to foresee, occur, when consJ_d- The leaders of the new social and behavioral sciences of the late nine-
erable numbers are taken into the account, with a degree of regulanty teenth century were generally more attracted to biological, historical,
approaching to mathematical. "7 1 He speculated on the number of years and anthropological ideas than statistical ones, but statistics was not ab-
it would take to average out the effect on history of a Napoleon or Alex- sent from their programs. The young Wilhelm Wundt dismissed Buckle
ander, and argued that statesmen may rely on probable statements re- with a long, critical footnote in his programmatic book of 1862, but re-
garding multitudes, for "what is true approximately of all individuals is peated many of Buckle's pronouncements and expressed hope that use
true absolutely of all masses. "73 of numbers might enable him to penetrate to the unconscious. Accord-
Similarly Karl Marx explained how Quetelet' s doctrine of the average ingly, he called for a natural history of human society based on the laws
man could be used to define a uniform standard of labor, and hence to of statistics, arguing: "It can be stated without exaggeration that more
furnish an exact and defensible interpretation of the labor theory of psychology can be learned from statistical averages than from all philos-
value.74 William Whewell illustrated the "method of means" in science ophies, except Aristotle. ''79 Wundt later became a critic of statistical
70 Charles Morgan, Review of Quetelet's On Man in Athenaeum (1835), 593-595. 61!- law, as he shifted to a less naturalistic, more holistic position. Although
613, 658-661, p. 593· . . .. the mature Wunclt thought the existence of historical laws inconsistent
'' G. C. Lewis, A Treatise on the Methods of Observation and Reasonmg m Polzllcs (2 vols.,
London, 1852), vol. 2, pp. 84-85. . . .. .
with the primacy of psychology among the mental sciences, he at least
7' John Stuart Mill, A System of Logic: Ratiocinative and lnduct1ve, vanorurn editiOn m
promoted error analysis in experimental psychology.
Collected Works, vols. ;-8, ). M. Robson, cd. (Toronto, 1973), p. 932. On the other hand,
Mill asserted in his Principles of Political Economy with Some of thezr Apphcatwns ~o Socwl
75 William Whewell, The Philosophy of the Inductive Sciences (z vols .. London, 2d ed.,
Philosophy ( 184 8), W. J. Ashley, cd. (London, 1923), p. 704, on the authonty of Thomas
Tooke, "that even so long a period as half a century may mclude a much greater proportion of 1847), vol. 2, p. 405.
76 M. L. Wolowski, Etudes d'economie politique et de statistique (Paris, 1848), [>p. 397,
abundant and a smaller of deficient seasons than is properly due to it. A mere average, there-
fore, might lead to conclusions only the more misleading, for their deceptive semblance o~ 401.
77 [Mark Pattison], "History of Civilization in England," Westminster Review, 12 !1857),
accuracy. There would be less danger of error in taking the average of only a small number ot
years, and correcting it by a conjectural allowance for t~; character of the seasons, than ill 375-399. p. 392
,s Marion Diamond and Mervvn Stone, "0/ightingale on Quctelct," JRSS, A, 144 (19lll I,
trusting to a longer average w1thout any such correctwns.
66-79, 176-213, 332-351; see also I. Bernard Cohen, "florence Nightingale," Scientific
n Ibid., p. 6o3
74 Karl Marx. Capital (3 vols., New York, 196;), vol. 1, p. 323; see also his remarks on
American, 250 (1984), /\·!arch, 128-137.
79 Wilhelm Wundt, Beitriige zur Theorie der Sinneswahrnehmung (Leipzig, 1862), p. XX\.
average prices in vol. 2 , pp. 86o-861. Marx also cited Edmund Burke, about whom he other-
wise had little favorable to say, on the point of average labor. Robert Owen, who was not Cited See Carl F. Craumann, "Experiment, Statistics, History: \\'undt's first Program of Psychol-
here but whose work Marx knew and praised (with qualification), made preosdy the same ogy," in W. C. Bringman, R. D. Tweney, eds., Wundt Studies, A Centennial Collection (To-
point, also in defense of a strict labor theory of value, in Report to the County of Lanark (Glas- ronto, 1980), pp. 33-41, or, better, Solomon Diamond, "Buckle, \:Vundt. and Psychology's
Use of History," lsi.>, 75 (1984!, 143-i52.
gow, 1821), p. 7·

66 6;
The Laws That Govern Chaos
The Laws That Govern Chaos
Herbert Spencer's sociology was first of all evolutionary, and was ex--
other,?and, was more sophisticated than ~irtually any of his prcdeces-
principally in the form of biological analogies. As part of his
s~rs. I he relevant point here, however, is that even at the end of the
campaign to discourage government meddling, Spencer expressed dis-
mnctcenth century the regularities of statistics bore impressive tcsti-
dain for those parts of statistics that seemed to yield easy answers, such
mony_to the power and reality of the social domain. In the Rules ofSo-
as sanitary statistics, and he was particularly unenthusiastic about the ev-
~zologzcal Method (1895), Durkheim identified the "average type" with
ident tendency of the medical profession to organize itself "after the
r~: normal, ~s o~posedto the pathological, and with the "group mind
fashion of the clergy"So and take over the domain of public health. Ed-
(lame collectzve), argumg that statistics furnished the means to isolate
ucational statistics, the results of which remained monumentally incon-
those_ "currents of opinion" that "impel certain groups either to more
clusive, suited his temperament better. Spencer, however, was by no
marnages, or to more suicides, or to a higher or lower birthrate etc. "84
means an obscurantist, and though he thought the main contribution of
In Suicide (1897) he used statistics to illustrate how "the indi:idual is
social science would be to educate people on the matter of unintended
domina~ed by a m~ral reality greater than himself," one so pressing that
,-,equences, he was firmly persuaded that its claim to science was just
Durkhe1m felt obliged to reconcile "liberty with the determinism re-
" o this end he cited the results of statistics, even though he made no
vealed by the statistical data. " 8 5 For this, tle argued in the manner of
effort to found theory on them. He argued for the beneficial effects of
Quetelct that the social force "does not determine one individual rather
natural selection with an invocation of Quetelet, contending that the
than another," but only "exacts a definite number of certain kinds of ac-
death of many is redeemed by the survival "in the average of cases" of
~ions."86 F?llowing his German predecessors, Durkheim preferred to
one "perfect specimen. " 8 ' Though he claimed to be little impressed by
mterpret hts figures as expressing a "collective impulse" that could not
Buckle, he defended the possibility of social science against Buckle's
be ~educed to ~he mean of individual dispositions. 81 The power of these
critics Charles Kingsley and J. A. Froude with statistical arguments.
soc1~l tendencies, however, was nowhere more compellingly presented
Granting that "the results of individual will are incalculable," he in-
to v1ew than m the collected acts of numerous individuals driven to su-
sisted "Mr. Froude himself so far believes in the doctrine of averages as
ici~e or marriage not by the particular circumstances that they cite to
to hold that legislative interdicts ... will restrain the great majority of
rationalize their deeds, but by deep reasons "unknown to conscious-
men in ways which can be predicted. ''82 Social science may not be as
ness" arising in the collective social being. SR
exact as astronomy, but "if there is some precision, there is some sci-
The evident success of statistics as an approach to social science was
ence. " 8 3 not interpreted by contemporaries as vindication of a metaphysic which
Emile Durkheim holds the rare distinction among early sociologists
regarded the laws governing certain domains as only probable. On the
of actually having written a statistical work, his study of suicide. This
contrary, statistical laws were deliberately formulated to extend the cer-
was, of course, the most statistical of topics, having engaged the serious
tainty of sciences like astronomy and mechanics to knowledge of phe-
attention of many major European statisticians and legions of minor
n~mena which hitherto had resisted exact scientific investigation. As
ones since the time of Quetelet. Durkheim in fact cited a fair number
will appear in part 3, statistics carried sufficient prestige at the time of
of them-Quetelet, Adolph Wagner, M. W. Drobisch, Georg Mayr,
Buckle that his arguments were able to set off a considerable debate over
and Alexander von Oettingen. He also attempted a critique of statistical
knowledge, though the result was among the most confused of all such 84 Emile Durkhcim, The Rules of Sociological iVIethod ( 1895). George E. G. Catlin, ed., S.
efforts. His use of numbers to draw sociological conclusions, on the A}ol,ov~v and). H. Mueller, trans. (Glencoe, N. Y , 1964), pp. 8, s6 .
.. ' bmle Durkh,eim. Suicide: A Study in Sociology (1897), J /\. Spaulding and George
so Herbert Spencer, Social Statics, Abridged and Revised (1st ed., 18)0, 1892; Osnabruck, Simpson. trans. (New York, 1951), pp. 38-39. -
% lbid., p. 325.
1966), p. 199· 8 7 Ibid., p. 30 6.
s, Ibid., p. 233; alsop. 203.
, 8~ Ibid.,, P-~ 297 On these matters, see also. Stephen Turner, "The Search for a Methodology
So Spencer, The Study of Sociology (9th ed., 188o; Osnabrock, 1966), p. 46. On Buckle, see
m Social Scknce: r;_urkheml, Weber, and the Nmeteenth-Centurv Problem of Cause, Prob-
Spencer, An Autobiogwphy (1904; 2 vols., Osnabri.ick, 1966), voL 2, p. 4·
ability. and /\chon, Ill Boston Studies in the Philosophy of Science, voL 92. 1985. This work
il1 Spencer, Study of Sociology (n. 82), p. 39-
was not yet available to me.
68
(

-- The Laws That GoveriiChaos


the possible incompatibility of statistical laws of m?ral beha:ior_with tr~­
ditional notions of free will. To be sure, the stahst recogmze<1 that hts - - - - - - - Chapter Three - - - - - - - - - -
adoption of a numerical approach constituted abandonment of the
search for laws of individual behavior, but this was not taken to ;I.nply
that chance had any objective role there. At worst, the action ot mdl- FROM NATURE'S URN TO THE
viduals might be like a coin toss-too complex and ~~stable. to foretell
reliably, but nonetheless wholly determined by cond1h~nmg mAuences INSURANCE OFFICE
and natural laws. Indeed, it was not universally recogmzed that the .e~­
act number of crimes or suicides in a society was dependent even tm!l-
allv on the whims of individual decisions. Buckle seemed to believe, and
wa-s certainly read by some to imply, that the root cause of crime.was a During the eighteenth century, as Lorraine Daston has shown, proba-
fixed sum of immorality arising from the ignorance and degr~dati~n m bility was customarily interpreted as the calculus of reasonableness for a
world of imperfect knowledge. Enlightenment thinkers applied the
SOC!·ety , which must be exhausted every year. If a particular· h mchvtdual
did not yield to temptation, others would be impelled w1t yet greater mathematics of chance to an implausibly rich variety of issues. They
force un.til the annual quota of crime had been reached. . used it to demonstrate the rationality of smallpox inoculation, to show
During the second half of the nineteenth century, confidence m the how degrees of belief should be apportioned among testimonies of var-
value and reliability of statistical laws reached the pomt that the soCial ious sorts, and even to establish or preclude the wisdom of belief in bib-
science of statistics could become a model for certain areas of the phys- lical miracles. Condorcet, followed by Laplace and Poisson, furnished
ical and biological sciences. Analogies and similes of social science were calculations of the probability that juries of a given form and size would
used repeatedly to justify the application of what came to be known as
statistical reasoning or the statistical method to problems of thermody-
I
i
reach a just verdict in any given case. These, however, were not the only
calculations presented to the world as evidence of the power of what De
namics, heredity, and price fluctuations. In the end, this form ofknowl- I Moivre called the "doctrine of chances." Probabilists also stressed the
f
edge would indeed become associated with uncer~ainty and. even rnde- !
I
applicability of their subject to actuarial and demographic matters.
terminism, but the initial use of statistics in relatiOn to s~b)ects wh1ch Probability calculations based on mortality records had been used in-
nobody had any reason to think behaved whimsically or me.gularly re- creasi~&IY to set rates for life insurance and annuity purchases-though,
quired first of all that statistics be deemed capable ~f producmg knowl-
edge no less certain or lawlike than the rul.es by wh1ch nature Itself was
I! intcrestlhgly, not for maritime and fire insurance-since Edmond Hal-
ley published the first life table in 1693. IVIathematicians all over Eu-
governed. Statistical principles bore suffiCien~ pr~shge to be applied ,to rope, bu;t.especially in the great comrncrcial states, the Netherlands and
economics, to biology, and even to the most d1gmfied of sc1en~es, ph) s-
.cs where they were used to work out the implications of the kmebc gas
~h~ory. Maxwell, Boltzmann, Galton, and Edgeworth interpreted Buck-
I Great Britain, applied their skill to political arithmetic during the eight-
eenth century. Some of Laplace's most important contributions arose
from his work on population estimates and other demographic prob-
lems.'/
le's statistical observations as justification for the presumption that any
phenomenon composed of numerous independent events could be ex- \Vritings by eighteenth-century probabilists on insurance were plen-
pected to exhibit impressive regularity in the mass. ~he greatest confu-
sion at one level is not only consistent with but 1mp!tes remarkable sta-
bility at another, which manifests itself in the form of statistical laws.
I tiful, but they evidently had little influence on the interpretations given
to probability. The mathematical structure of probability theory was de-
' Sec Lorraine Daston. "The Reasonable Calculus: Classical Probability Theory, 1650·
184o"" !Ph.D. Dissertation, Harvard University, 1979); also Charles C. Gillispie. "Laplace,"
in DSB, vol. 15; Gillispie, "Probability and Politics: Laplace, Condorcet, and Turgot," Pro-
ceedings of the American Philosophical Society, 116 (1972). 1-20: Gillispie, Science and Polity
in France at the End of the Old Regime (Princeton, 198o), chap. 1.

71
( (

FromNature's Um·trrlnsurance--- From Nature' sUm to Insurance

rived fr~m its earliest object, games of chance. Probabilities of single numbers of eventsr introduced the term possibilite for the actual ratios
events were supposed to be known a priori, and calculation was used to of events presented over the long run by nature. He was convinced that
reach formulas for compound or multiple events of various sorts. Gam- these ratios were generally stable, and he developed analytical tools for
ing seemed too vile and contemptible, however, to serve as the real determining the degree of convergence of subjective probabilites to ob-
interpretation for so profound a study, and the eighteenth-century frame jective possibilites. 4 Classical probability writers saw no contradiction
of mind did not permit a domain of mathematics to exist apart from its between the view that probability statements represent the uncertainty
interpretation.• Not chance events, but degrees of belief, were taken to of the rational mind about a particular event and the claim that proba-
be the fundamental object of probability theory. The inescapable games bility estimates should reflect long-run distributions of events in nature
of chance, the casting of dice and drawing from urns, were explained or society. Most subscribed, after all, to a relatively passive form of as-
away as heuristic tools, analogies to the apportionment of belief. "It is sociationist psychology, which made belief almost a direct reflection of
remarkable," wrote Laplace, "that a science, which commenced with experience. Probability assessments measured the grooves etched by
the consideration ofgames of chance, should be elevated to the rank of impressions over time in the tabula rasa of the mind. s
the most important subjects of human knowledge. " 3 Nevertheless, there was much in the classical treatment of probability
Beyond doubt the most famous of Laplace's ideas, in his own time as that could not be justified in terms of experience or even of expectation
in ours, was the imaginary being whose perfect vision and unlimited of stable frequencies. Laplace, following a tradition that originated with
powers of calculation would enable it to know both past and future with Jakob Bernoulli, assumed that the mind naturally distributes belief
complete certainty. Laplace was not seeking to explain the ways of God. equally among complementary alternatives about which it knows noth-
He had need of this hypothesis, the "infinite intelligence," to make clear ing. This "principle of indifference," as it came to be called, was in-
his conception of probability and to explain why he assigned it so prom- voked as required to justify an assumption of equipossibility in relation
inent a place in human thought. Chance meant for Laplace not the ir- to such chance events as a single toss of an untested coin. It also figured
reducibly random, but the fortuitous production of patterns through the prominently in various calculations of a more daring and ambitious
interaction of a multitude of independent causes. If all the events of na- character, such as the method of inference developed by Laplace and
ture could be perceived simultaneously, and if our techniques of cal- termed the "probability of causes." One of the classic instances of this
culation were sufficiently advanced, we would have no more need of genre was his determination of the probability that a fixed cause, rather
probability than the infinite intelligence. Finite humanity, however, than chance, was responsible for the revolution of all the planets in the
had no prospect of attaining omniscience. Probability was thus essential same direction, and roughly the same plane, around the sun. The prob-
for estimating, and above all for reducing, the level of human error. Its ability that the planets would so align themselves if distributed at ran-
application to nature was predicated on a certain level of knowledge, but dom had been given in 1734 by Daniel Bernoulli, and could be com-
its proper object was human ignorance. puted straightforwardly, but the inverse probability of the existence of a
Even as philosophy, the classical theory of probability did not stand fixed cause depended on certain assumptions. One of these was an a
wholly apart from the objective domain of frequencies. When pressed priori probability for the existence of a fixed cause. Laplace, applying the
to give an ultimate justification for basing belief on probability calcu- principle of indifference, chose one half.
lations, Condorcet had recourse to Jakob Bernoulli's theorem about the All told, however, Laplace and Condorcet were far more interested in
conformity of an indefinitely long sequence of similar events to the un- the applications of probability than in its epistemological foundations.
derlying probabilities. Laplace, whose reputation as a probabilist rested
largely on the techniques he developed for extending the theory to great 4 See Condorcet, Essai sur /'application de /'analyse a Ia probabilite des decisions (Paris,
1785), in Keith M. Baker, ed., Condorcet: Selected Writings (Indianapolis, 1976), p. 39; La-
place, "Memoire sur les Approximations qui sont fonctions de Ires grands nombres" (1785),
'Daston, "Reasonable Calculus" (n. 1), p. 78. Oeuvres, vo_l. 10, p. 310.
'Pierre Simon de Laplace, Philosophical Essay on Probabilities, F. W. Truscott and F. L.
'Daston, "Reasonable Calculus" (n. 1), pp. 174-218.
Emory, trans. (New York, zd ed., 1917), p. 195.

72 73
· - - - - -- ----

( (

-From Nature'-$-Vmto Insunmce From Nature's Urn to Insurance


One of the first thinkers to examine atlength the pertinent philosoph.i~al served seven ships, all bearing flags, should properly conclude with a
issues was Augustus De Morgan, whose work on the logic of probability probability of 8/9 that the next vessel to appear would also bear a flag
helped to establish it as a worthy philosophical problem. De Morgan :vas became notorious. 8 De Morgan also upheld subjective probability as a
the most influential of a group of members of the Royal Astrono_m.Ical logic of tolerance which, if understood, would serve as an antidote to
Society who sought to do for probability what John Hersch~!, Wilh~m religious bigotry. Although he was not wealthy, he resigned from Uni-
Whewell, Charles Babbage, George Peacock, and George AHy, all ong- versity College, London, in protest against religious interference with
inally at Cambridge, had begun to do for the algebraic method~ of La- faculty appointments, and he insisted that all knowledge except pure
place's Mecanique celeste and Fourier's essay on heat. 6 lt was evident t? mathematics and logic was probabilistic-that is, uncertain. He wrote
De Morgan, Herschel, Thomas Galloway, John Lubbock, ~~d Francis in his book of 1838:
Baily that England had fallen far behind France in probability theory,
Probability is the feeling of the mind, not the inherent property of
and that it would be a great achievement simply to introduce t~etr c~u~­
a set of circumstances .... [R]eal probabilities may be different to
trymen to the profound mathematical reasoning of Laplace s Theorze
different persons. The abomination called intolerance, in most
analytique des probabilites. Unlike the algebraists, they were less ~?n­
cases in which it is accompanied by sincerity, arises from inability
cerned to establish an autonomous mathematical research tradition
to see this distinction. A believes one opinion, B another, C has no
than to learn those parts of probability that were needed by scienc~ ~~d
commerce. In particular, they were interested in the probab1hstic opinion at all. One of them, say A, proceeds either to burn B or C,
method ofleast squares-which, by 1830, was indispensable for obser- or to hang them, or imprison them ... ; and the pretext is, that B
and C are morally inexcusable for not believing what is true. Now
vational astronomers-and also in the mathematics of insurance. lnte~­
substituting for what is true that which A believes to be true, he
estingly, the principal founder of the Astronomical S.ociety, Francis
either cannot or will not see that it depends upon the constitution
Baily, had originally made his fortune in the insurance mdustry, and at
of the minds of B and C what shall be the result of discussion upon
least a few of these men derived income from positions on the boa~~s of
them.9
insurance institutions at different times. 7 As mathematical probabilists,
they made no claim to important advances over Laplac~ and. ~oisson. For this and other reasons, De Morgan maintained that no measure
They sought instead to make Laplace accessible, and theH wntm~s ca~ of probability can ever be objective. "I throw away objective probability
be found in the Encyclopaedia Metropolitana, the Encyclopaedza Brz- altogether," he wrote, "and consider the word as meaning the state of
tannica, Lardner's Cabinet Cyclopaedia, the Penny Cyclopaedia, and the mind with respect to an assertion, a coming event, or any other mat-
the Library of Useful Knowledge. ter on which absolute knowledge does not exist." Probability, he main-
De Morgan was no frequentist. On the contrary, he adopte~ ~n un- tained, was indistinguishable from psychology, and a perfect measure of
compromisingly subjectivist framework for the logic of probab1lity. He probability would depend on weighing the force of sundry impressions
was no less enthusiastic than Laplace and Condorcet about the proba- on the human mind. ' This, however, was not to endorse pure irration-
0

bilistic study of testimonies and judicial decisions, and his ?wn dem- alism, or to give up all hope of a mathematical theory, for the judging~,.,,.,,,,.,,,""'"~·
onstration that a mathematical Robinson Crusoe who had hitherto ob- ·~-;g;,, •. mind was not impervious to influence from the real world. De Morgan
6 See Susan Faye Cannon, "The Cambridge Network," in Science and Cul~ure: The ~arly refuted d'Alembert's famous objection to the assumption of equiprob- '"'"""""""""""
Victorian Period (New York, 1 97 8), chap. 2, ~nd Maurice C~~sland and Crosbte Smtth, The ability-which for De Morgan was absolutely central-by noting: "If
Transmission of Physics from France to Bntam: 180~-1840, HSPS, 9 (1978), ~-61., ._
1 See Francis Baily The Doctrine of Life-Annuztzes and Assurances Analytzcally Investz 8 See john V. Strong, "The Infinite J;lallot Box of Nature: De Morgan, Boole and jevons on
gated and Practically Explained (2 vols., London, 1813); John Herschel, "Memoir of Francts Probability and the Logic of Induction," in PSA: Proceedings of the Philosophy of Science As--
Baily " in Journal of a Tour in Unsettled Parts of North America in 1796 and 179~, Augustus De sociation (1976), vol. 1, 197-211, pp. 200-201.
Mor~an, ed. (London, 1 856). The election of De Morgan and Gall~way to posthons wtth m- o Augustus De Morgan, An Essay on Probabilities and on Their Application to Life Contin-
surance offices is mentioned in Sophia Elizabeth De Morgan, Memozrof Augustus De Morgan gencies and Insurance Offices (London, 1838), pp. 7-8.
(London, 1882), pp. 6o-61, 110, 181, 279-280, 363. w De Morgan, Formal Logic (1847; London, 1926), p. 199.

74 75
From Nature's Um to Insurance
FmniNature's Utnlo Insurance-
renee of events of the same kind, enables us to apply the calculns of
any individual really feel himself certain, in spite of authority and prin- probabilities to many of the most interesting questions connected with
ciple as here laid down, that the preceding case~ [in :osses of a coin- our social and political institutions; and to determine the average result
namely H, TH, TT] are equally probable, he is fully JUstified~~ ~dopt­
of a series of coming events with as much precision as if their chances
ing zh instead of 3/4 [for the probability of 'heads at least. once m two were determinate, and known a priori_ "•6
tosses]." However, let him spend an evening betting agamst an oppo-
In France, as in Britain, continued adherence to a fundamentally
nent on that supposition, and he would rapidly come around to a more
subjectivist understanding of probability was accompanied during th~
orthodox view." early nineteenth century by an attenuation of interest in this aspect of it.
- The circumstance that experience was permitted to correct deviant
S. F_ Lacroix, though enthusiastic about Condorcet's work on testi-
notions about probability suggests that even De Morgan allowed some-
monies and judicial probabilities, denied his claim that probability was
thing in probability which was not entirely subjective. Indeed, ~e Mor-
gan was an insurance enthusiast, exalting it as "the most enhght~n~d
exclusively an expression of imperfect knowledge. Whether a judgment
is announced before or after the drawing from an urn, he explained, the
and benevolent form which the projects of self-interest ever took. tt IS,
long-run tendency will be for the results to conform to the ratios of balls
in fact, in a limited sense and a practicable method, the agreement of a
within the urn, just as the most complex social phenomena manifest a
community to consider the goods of its individual members as com-
mon. "12 He was especially attached to friendly societies, whose benefits remarkable order when considered in large numbers. He wrote: "H is
could "raise working men to an unknown degree of independence as the truly upon this connection that all legitimate applications of the calcu-
poor Jaws are removed. "'3 Insurance, the focus of De Morgan's only full lus of probability are based. "'7 Simeon-Denis Poisson upheld the entire
book of probability, required attention to its objective aspect ..He ex- spectrum of applications of probability set forth by his master, Laplace,
plained: "The word probability has two different senses, the colhswn of against challenges by Poinsot and others, and he cast his main work on
which is a grand source of confusion: it is used to refer both to the state the subject in the form of an essay on the probabilities of judicial deci-
of the mind and to the external dispositions which are to regulate the sions. Much of his work, however, was devoted to an extension of Ber-
long run of ~vents: to our strength of prediction, and also to the capacity noulli's law of large numbers to cases where no single, fixed, underlying
of circumstances to fulfil our prediction." 14 De Morgan thought that the probability exists, and the part on judicial decisions involved a statistical
word probability should be reserved for degrees of belief: and presup- analysis of actual court records. The law oflarge numbers, he proposed,
posed no capacity to make reliable predictions. !he obJ~divc ratws ~~n­ is "the foundation of all applications of the calculus of probabilities,"
erated by nature and society he preferred to des1gnate wtth the term fa- applicable equally to the physical and the moral sciences. ,s
cility." He made frequent use of this statistical idea, and h~ stress~d :hat The main break with the classical interpretation of probability oc-
probability calculations could take on a measure of o?JectJve rehabthty curred during the early 184os. In 1842 and 1843, four writers from three
when applied to long series of events. "Uniformity IS the law of l~rg.e countries independently proposed interpretations of probability that
masses compared with each other," he wrote, citing Quetelet. 15 Sum- were fundamentally frequentist in characteL So perfect a temporal co-
larlv his friend Thomas Calloway marvelled that the "constant approx- incidence can only be regarded as the outcome of chance, but the gen-
im~t,ion to fixed ratios, which is proved by all experience, in the recur- eral timing of this change is not entirely surprising. Applications like the

" De l\1organ, "Theory of Probabilities," Encyclopaedia ~etropolitan_~: voL z (Lond~n, ' 6 Thomas Galloway, A Treatise on Probability (Edinburgh, 1839; reprinted from Encyclo-

393 . 490 . P- 401 ; on cquiprobability, see De Morgan, Probabthty, m The Penny Cy-
1 8 45 ), paedia Britannica, 4th ed_ ), p. 4-
clopaedia of the Society for the Diffusion of Useful Knowledge, val. 19 (London. 1841), 24-3°. '7 ~- f'_ Lacroix, Traite elementaire dt1 calclll des probabilites (Paris, 3d cd., r833), P- 69-
'8 s_ !)_ POisson, Recherches Sllr Ia probabilite des jugements (Paris, 183?), p. 12_ See also
p. 27-
" De Morgan, Essay (n. 9), P- xv. his "Recherches sur Ia probabilite des jugemcnts, principalement en ;.atiere criminal,"
c;?omptes rendus hebdomadaires des seances de l'Acadernie des Sciences, 1 (r835), 473-494;
"'4 Ibid_, p. 295 , - " J>SC' (186 4,) 407
. - ~
De Morgan, "On the 1heory of Errors of Observation, T , IO 427, IJ- No~e :~rIa !01 des grands nombres, ibid_, 2 ( 1836), 377-382; "Note sur le calcul des proba-
q

brhtcs, 1brd., PP- 395·400, wtth cntJque bv Poinsot.


407-
'' De Morgan, Essay (n_ 9), p. 119.
77
( (
'
From Nature's Urn to Insurance
From~Nflture's Umt{) Insurance
nebulous, and wondered if the passage of a skiff would suffice. This am-
probability of causes had come increasingly under attack during the
biguity,_l~e maintained, also vitiated Laplace\ argument about the plan-
early nineteenth century, and were looked upon with a level skepti-
etary orbtts; there was no valid a priori reason for excluding from his cal-
cism bordering on outright hostility by all four authors. At the same
culation the comets, which revolve in a of planes and in both
time, the growing prominence of the social science of statistics offered a
directions around the :;un.
new justification for the mathematics of gaming that would eventually
A f~w years later, a controversy dealing with the same general issue
liberate probabilists from the need to invoke, in defense of their craft,
arose m the context of another astronomical problem. In his Outlines
applications like the probability of causes or judicial decisions that had
of Astronomy, John Herschel presented a calculation deriving from a
come to seem implausible and even dangerous. Richard Leslie Ellis,
1767 paper by Michell in which he used the inverse method to deter-
John Stuart Mill, A. A. Cournot, and Jakob Friedrick Fries offered rea-
mine t~e probability that the large number of optically adjacent stars ob-
sons for a frequentist interpretation of probability that were as diverse as
served m the sky implied that some at least were true doubles. The Edin-
their national origins. Yet they all chose to explain probability in terms
burgh professor of natural philosophy, J. D. Forbes, challenged this
of the regularities produced by chance phenomena, exemplified by the
argument i1; 1850. While conceding the qualitative plausibility of the
results of statistics. ~vidence, Forbes maintained that no numerical probability could be
One line of thought leading to a frequentist view of probability was
JUStified, for any calculation of probabilities regarding a configuration
based on distaste for the ostensible arbitrariness of most a posteriori, or
that had already been observed involves a hypothesis especiallv framed
inverse, probabilities. An inverse probability is one involving the for-
to accord with the data, and hence to yield a high probability ~fa fixed
mula of Thomas Bayes, rediscovered independently by Laplace and ap-
cause. Probabilities, he wrote, "have no absolute signification with ref-
plied by him in a variety of ways, among them his determination of the
erence to an event which has occurred, such as the distribution of stars
probability that a fixed cause had produced the observed alignment of
on the ce!estial sphere . . . . They represent only the state of expectation
the solar system. In the more mundane terms of games of chance, a di-
of the mmd of a person before the event has occurred, or having oc-
rect probability might be exemplified by the chance that a fair coin
curred before he is informed of the result. "zo In addition, Forbes argued
tossed ten times would yield eight heads and two tails; the corresponding
that the alternative hypothesis-distribution of stars over the celestial
inverse problem would give the probability that the coin is unfair (to any
globe at random--was inadequately defined, and he was uncertain that
specified extent), supposing that eight heads and two tails appeared in
it could even be given a precise definition.
the first ten tosses. All the prominent frequentists of the nineteenth cen-
These papers by Ellis and Forbes were answered bv Herschel and W
tury distrusted a posteriori calculations, especially when they purported
F. Donkin, 21 and soon afterwards George Book ente;ed the fray. Boole:
to yield truths about the real world. The most extensive arguments were
who. had. already published an outline of his ideas on symbolic logic,
made, however, by a group of British writers.
readily discerned the applicability of his methods to this controversy over
This critique was initiated in 1842 by R. L. Ellis, who argued bluntly
the foundations of probability. His proposed resolution of the conflict
that the "estimates furnished by what is called the theory posteriori of a fi.rst pub~ished in a series of papers for the Philosophical Magazine be~
the force of inductive results are illusory. "'<J Ellis pointed to inconsis-
gmnmg m 1851, 22 was generalized in his important work of 1854, The
tencies in the examples of this genre given by previous authors. De Mor-
gan's lost seaman might have seen seven ships sail by with red flags, in .'"J.D. Forbes, "On the Alleged Evidence for a Physic" I Connexion between Stars Forming
Bmary ~r Multtple Groups, Deduced fran:': the Doctnne of Chances,·· Phill'v1ag [3], 37 ( 1 8so),
which case he would have been obliged to assign the same probability 401-42;, p. 406. See also Barry Gower, Astronomy and Probabilitv: Forbes versus Nlichell
to the proposition that the next ship would display a red Hag as that it on the DrstnbutJOn of the Stars." i\nnals of Science. 39 (19821. 14 5-160.
' ' Jlohn !lers;;hd], "Quetelet on Probabilities," Edinburgh Review. 92 ( 185ol. 1-57, p. 37'
would fly any Hag at alL By this reasoning, green Hags would seem to be 1
v\ F. Donkm, _On Certain Questions Relating to the Theory of Probabilities," Phil tv!ag i4l.
excluded entirely. Ellis also argued that the concept of "next event" was 1 (1851J, 353-308,458-466:2 (1851), ss-6o · ·
"These papers are reprinted in George Boolc, Studies in Logic and Probabilit}' I London
'" R. L. Ellis, "On the Foundations of the Theory of Probabilities:· TPSC, 8 (1849!. 1-6. 1952). . •
p. 6.
79
;8
( (
From Nature's Urn to Insurance From Nature's Urn to Insurance
Laws ofThought. Boole explained that the conversion of a direct into an
Boole arrived also at a frequentist interpretation, but from idealism
inverse probability always involved two undetermined parameters. In
rather than Ellis's "sensational philosophy." He wrote, in the tradition
the case of the "probability of causes," it was always necessary to assign
of Scottish Common Sense, that logic and probability "set before us
some value to the a priori probability that a fixed cause exists, and to the
what, in the two domains of demonstrative and of probable knowledge
probability that this unspecified cause would suffice to produce the ob- are the essential standards of truth and correctness,-standards not de~
served effect. Such assignment must always be arbitrary, and therefore
rived from without, but deeply founded in the constitution of the hu-
the eventual solution necessarily involves some questionable assump- man faculties. " 2 5 Yet he denied probability could be defined in terms of
tions. By tacitly assigning values to these parameters-one half to the anything so capricious as "the strength of that expectation, viewed as an
first and one to the second-Laplace and De Morgan had given their emotion of the mind." Regularity of masses is a property of nature, he
analyses of causation a misleading appearance of precision. To assume observed:
these values on the basis "of the equal distribution of our knowledge, or
rather of our ignorance," was, in Boole's view, quite arbitrary, and The rules which we employ in life-assurance, and in the other sta-
amounted to nothing more than "the assigning to different states of tistical applications of the theory of probabilities, are altogether in-
things of which we know nothing, and upon the very ground that we dependent of the mental phenomena of expectation. They are
know nothing, equal degrees of probability." 2 3 founded upon the assumption that the future will bear a resem-
R. L. Ellis inferred from his objections to the ideas of Laplace and De blance to the past; that under the same circumstances the same
Morgan that something was fundamentally wrong with the prevailing event will tend to recur with a definite numerical frequency; not
interpretation of probability. Positing "the inconsistency of the theory of upon any attempt to submit to calculation the strength of human
probabilities with any other than a sensational philosophy," he proposed hopes and fears.
for the first time that probability statements were properly applicable to Now experience actually testifies that events of a given species
series of events rather than to mental uncertainty about individual trials. do, under given circumstances, tend to recur with definite fre-
Hence he held that the convergence oflong-run frequencies to their re- quency, whether their true causes be known to us or unknown. Of
spective probabilities must be regarded as tautologous: "I have been un- course this tendency is, in general, only manifested when the area
able to sever the judgment that one event is more likely to happen than of observations is sufficiently large. 2 6
another, or that it is to be expected in preference to it, from the belief
Thus, although Boole was a "conceptualist" rather than a "materi-
that on the long run it will occur more frequently." Objecting, not quite
ali~t:' in Jo?n Venn's terminology, 2 7 he based his understanding of prob-
accurately, that the Napoleonic functionary Laplace had wished to
make even the most elementary truths appear in abstruse scientific garb lI ability on msurance and social statistics rather than on a model of un-
certain knowledge derived from the classical urn. He conceded that in
as the result of calculation, Ellis proposed that the idea of probability is
certain games of chance the probabilities could be determined directly
spontaneously produced by the continuous endeavor of the active mind
from knowledge of the composition and balance of coins and dice but
"to introduce order and regularity among the objects of its perceptions."
I even for these he insisted that a measure of probability "derived fro~ the

l-·:
Our deep-seated trust in the constancy of nature assures us that stable
observation of repeated instances of the success or failure of events" must
averages will gradually be generated as "the action of fortuitous causes
disappears." Probability must be associated with observation, not igno-
... in J. C. Shairp, P. G. Tail, and A. Adams-Reilly, Life and Letters of fames David Forbes
F.R.S. (London, 1873), p. 480: "Ave~ de.s chiffres on peut tout demontrer, ought to beth~
rance, of phenomena and allied with notions of order and statistical reg- motto of most of the philosophical apphcahons of the theory of probabilities-which in its own
,~ ularity, not chance. >4 ~ature and accordmg to the plai.n ~iew of it, i~ only a development of the theory of combina-
bons. To attempt to constitute 1t mto the philosophy of science, is, in effect, to destroy the
" ph1los~ of science altogether."
1'George Boole, An Investigation of the Laws ofThoughton Which are Founded the Math- 1 ' Boole, Laws of Thought (n. 23), p. 2.
ematical Theories of Logic and Probabilities (1854; New York reprint, 1958), p. 370. 16 Ibid., pp. 244-245.
1• Ellis, "Foundations" (n. 19), pp. 1-3. See also Ellis's letter to J.D. Forbes, 3 Sept. 185o, 17 John Venn, The Logic of Chance (3d ed., 1888; New York reprint, 1962), pp. ix-x.

8o 81
From Nature'sHm to Insuranee From Nature's Urn to Insurance

be regarded as no less fundamental. To other chance phenomena, in- ous error by clothing it in the garb of knowledge."<> Knowledge,
cluding all those of legitimate practical interest, the gambler's a priori whether probable or certain, \\·as yet knowledge, and as such could onlv
reasoning was wholly inapplicable. Boolc maintained that those rare be acquired in one way. "Conclusions respecting the probability of a fa~t
and useless instances for which results may be obtained by the direct op- rest not upon a different. but upon the very same basis, as conclusions
eration of the binomial distribution on primitive probabilities had dom- respecting its certainty; namely, not our ignorance, but our knowledge:
. inated the attention of mathematicians for too long. The theory of prob- knowledge obtained by experience of the proportion between the cases
ability was properly conceived as "the mode in which the expected in which the fact occurs and those in which it does not occur. "B
frequency of occurrence of a particular event is dependent upon the Mill's objections to the Laplacian view of probability were moderated
known frequency of occurrence of any other events. " 28 for the second edition of 1846. During the interim, John Herschel had
Logic, long central to the Scottish university curriculum and increas- written him a long letter arguing that subjective probability provided··~.,.
ingly influential in England after Richard Whately restored it at Oxford satisfactory measure of the expectation we arc justified in forming with
around 1820, "9 provided also the framework for John Stuart Mill's cri- respect to an event, and that empirical information was required only for
calculating probable ratios of multiple events in the real world. 34 Mill,
tique of classical probability. Mill, seeking a via media between his fa··
whose footing was never too sure in actual questions of science and
ther's rationalism and Macaulay's conservative Baconianism, denied
mathematics, acceded to this argument without really accepting it. Al-
that even mathematics was possible independent of experience. He
though he took out his denunciation of the indifference principle, he let
could not countenance the separation of knowledge from the means of
stand the claim that improved information was more useful than the
attaining it, and he argued that all reasoning proceeds from particulars
most sophisticated mathematical analysis. Every edition contained a re-
to particulars. Intermediate generalizations were useful, but they had no
mark that the probabilities arising from the toss of a die are better known
privileged status above the concrete observations upon which they were
from experience than from mere conjecture based on indifference, al-
based. Laplace's probability of causes, with its argument from igno-
though Herschel's letter would seem to dictate that the two types of prob-
rance, looked to Mill like the kind of a priori reasoning which sup-
ability are not fully commensurable."
ported, on the one hand, Bentham's inflexible deductions and, on the
Indeed, Mill remained suspicious of any application of probability to
other, the intuitionist's exaltation of unreasoning emotion and tradition.
a single case, and preferred always to rely on the regularity of long-terrn
The first edition of Mill's Logic, published in 1843, contained one of
frequencies. He argued that insurance tables, however useful to the
the harshest denuniciations of classical probability written in the nine-
companies that compile them, have scarcely any value for individuals,
teenth century. He argued, with specific reference to Laplace, that those since each will certainly be either above or below average. He joined the
who preferred to make elaborate calculations rather than to seek im- chorus of contemporary statists who made the point that science and
provement of the data had fostered "misapplications of the calculus of certainty emerge on a large scale whenever numerous unsystematic
probabilities which have made it the real opprobrium of mathemat- causes act at the local level. Beginning with the edition of 1862, he cited
ics. "3o Mill dismissed the assignment of probabilities based on indiffer- Buckle for the principle that "ft]he very events which in their own nature
ence as "entirely wrong," and deemed it a "strange pretension ... that appear most capricious and uncertain, and which in anv individual case
by a system of operations upon numbers, our ignorance can be coined no attainable degree of knowledge would enable us t~ foresee occur
into science. "3' Without an adequate base of empirical information, "to when considerable numbers are taken into the account, with a degree of
attempt to calculate chances is to convert mere ignorance into danger- regularity approaching to mathematicaL "3 6

'8 Boolc, Laws o{Thought (n. 23), p. 13. ''Ibid., p. 545 (1843 and 1846 editions).
' 0 See Whately's textbook, Elements of Logic (London, 2d ed., 1827). "Ibid., p. 1142 ( 1843 edition onh ).
14 See John V. Strong, "john Stua;t :Vlill. John Herschel, and the 'Probabilitv of Causes,' "
'"John Stuart Mill, A System of Logic: Ratiocinative and Inductive, variorum edition in).
:'\1. Robson, cd., Collected Works, vols. 7-8 (Toronto, 1973), p. s38. This quote appc3rcd in in PSA: Proceedings of the Philosophy of Science Association (1978), vol. 1, pp .. 31-41.
all editions. " Mill, Logic (n. 30), p. 540
'' Ibid., pp. 1 t4o. 1142 ( 1843 edition only). ,c, Ibid., p. 932. See also pp. 518, 597.847.

8:.?.
From Nature's Um to insurance
FmmNature's-Um to Insurance
hardly surprising that he emphasized so strongly the aspect of probability
A. A. Cournot, now known primarily for his \Vork in economics, was
connected with stable frequencies. "The calculus of probabilities," he
a bit more circumspect in criticizing the classical interpretation of prob-
wrote, "has real importance only inasmuch as it is applied to numbers
ability, but the tendency of his analysis was largely the same as the Brit-
suf~'ciently large that one must have recourse to approximation formu-
ish frequentists. Applications of probability to testimonies, judicial de-
las. 4o
cisions, and the causes of natural phenomena had been denounced in
In Germany, the frequentist viewpoint was introduced in 1842 by a
France by writers of a positivist bent like Destutt de Tracy, Poinsot, and
promment Kantian philosopher, Jakob Friedrich Fries. Like the British
Auguste Comte. Cournot handled these subjects more delicately, allow-
Fries objected strongly to the probability of testimonies and causes, and
ing that calculation of subjective probabilities was not wholly invalid but
indeed to "a large part of the doctrine of probability a posteriori " for
denying that it had any use. The ambiguities of a posteriori probability,
"that which cannot be calculated should not be subjected to pseud~-cal­
he proposed, would "be rectified once the distinction is brought out be-
culations."41 The object of his book was to secure the foundations of
tween probabilities that have an objective existence, and give the meas-
probability and to define its relation to political arithmetic, insurance,
ure of the actual possibility of things, and subjective probabilities, which
and observational astronomy, and his interpretation was based primarily
relate partly to our knowledge and partly to our ignorance. "37 Neither
on confidence in stable statistical frequencies. His argument was not
sort of probability was of much interest unless based on wide experience:
grounded in an exaltation of experience, however, but in a defense of
The mathematical theory of chance would offer no more than the the higher human faculties and an antipathy to the sensualism which he
charm of speculation if it could only teach us the number of imputed to Enlightenment philosophy and to the French approach to
chances for and against an isolated event which will never recur, mathematics.
or which recurs only under very unusual circumstances; as the fol- Fries, though competent if not original in mathematics, wanted to
lowing will show, however, it assumes a very great importance even dispense with most of the combinatorial formalism of the pure calculus
for practice when tests of the same risk can be repeated a great num- of probability. Since the notion of equipossibilitv was merelv a mathe-
ber of times under similar conditions. 38 matical abstraction, with no bearing on the inve;tigation of ;1ature, the
standard deductive approach was applicable to little more than games of
Cournot set himself two objectives in his book on probability: to pro- chance. Moreover, the French mathematicians, and especially the rev-
vide a clear exposition of the mathematics of chance, and to clear up the olutionary Condorcet, had ignored the rational faculties that are nec-
philosophical errors that seemed to prevail in so many works. He did not
css_ary fo~ all judgments; they had confused proper assessments of prob-
present probability as chiefly of mathematical interest, but as an indis-
ability with the mere repetition of similar cases, or the blind influence
pensable tool of the observational sciences, statistics, and practical eco- of habit. The true object of mathematical probability, he explained, is
nomic studies. Cournot regretted the customary association of this field the discernment of general laws whose action is obscured by special
with games of chance, though he also thought that the events of social laws. Hence the arbitrary assumptions and sensualistic procedures of a
statistics were too complicated for probability to be applied fruitfully un- posteriori calculation could best be discarded, and the rational investi-
til after its methods had been better developed. Astronomy seemed to gation of the causes revealed by statistical regularities put in their place.
him a more useful study for perfecting the techniques of numerical
Fries, then, looked to political arithmetic and insuranc~"lls the model
analysis by probabilities. 39 Cournot's general goal was to promote the for scientific studies involvmg probability. Like Ellis, he objected to the
emergence of statistics as a method, and it was this that he saw as the claim that statistical regularity could be explained in terms of Bernoulli's
fruition of mathematical probability. The measurement of belief based principle or Poisson's Ia\\ of large numbers. This was to put the cart be-
on the principle of indifference seemed much less promising, and it is
Ibid., p. ii.
,- Comnot, p. 155. Sec alsop. iv. Jakob Friedrich Fries, Versuch einer Kritik der Princi(Jien drr Wahrscheinlichkeitsrech-
4'
,s Ibid., p. 46 mmg (Brunswick, 1842), in Siimt/iche Schriften, vol. 14 (Aalen. 1974), pp. v-vi.
14 ibid., pp. i, 261-262.
(
From Nature's Urn to Insuram:e From Nature's Urn to Insurance
fore the horse. The uniformity of statistics was to be explained in terrns In Great Britain, the situation \Vas more complicated. De Morgan
of the fundamental truth that constant causes always yield proportionate and Herschel continued to defend Laplace's vievv. The Oxford mathe-
effects. Statistical regularity does not emerge from the chaos of chance; matician W. F. Donkin depicted "a perfect mathematical idenhtv be-
order exists prior to events in the form of natural laws. Hence the im- tween the fundamental laws of the equdibrium of belief as to the .posi-
mediate object of probability as applied to nature was not the individual tion of a point in space and those of the equilibrium of a perfect elastic
event, but the mean value, which alone could be used to uncover the fluid ... [O]ur whole belief expands itself uniformlv over infinite
laws of nature. Mathematical probability, wrote Fries, "has absolutely space, just as a finite mass of perfectly elastic Auid, if
perfectly free,
no significance for the predetermination of individual events. Such sig- would expand itself throughout all space and have a uniform densitv. "45
nificance can only emerge if we make an estimate for a sufficiently ex- He even derived the astronomical error function from a model base.d on
tensive series of events of a certain sort" 42 the analogy between the force of testimony and physical attraction.
It should not be supposed that the frequency interpretation immedi- Stanley Jevons's Principles of Science was evidently uninfluenced by the
ately----or ever-eclipsed completely the subjective interpretation of argument of the frequentists, and even Francis Edgeworth adhered to a
probability. Something like it did, however, become the common-sense subjectivist interpretation, albeit a more nuanced one.
view during the late nineteenth century. The most prominent French The most influential nineteenth-century work on the philosophy of
and German writers on the subject were not won over to the view that probability, however, was resolutely frequentist, and was in fact the
probability was intrinsically objective in character, but neither were they book that gave the frequency interpretation its first full exposition. John
much interested in the allocation of uncertainty among possible out- I Venn argued that quantitative belief cannot be justified with respect to
comes of a single trial, and none seem to have upheld the calculation of
probabilities for causes and testimonies. The German Johannes von
Kries published a subjective interpretation ofprobability in 1886, but he
argued that the rational estimation of real probabilities almost invariably
I individual nonrepeatable events, and insisted that the ideas of probabil-
ity depend on a postulate of ultimate statistical regularity. To determine
the probability of an event, he argued, requires that the event be placed
in a series. The probability value then applies to the series, and not to
required an extensive statistical experience. He dismissed equipossibility the individual occurrence, for since a single event can be placed in dif-
as a mathematical artifice, which generally was not completely valid ferent series, it might otherwise be assigned several contradictory prob-
even for games of chance, and he thought mathematical probability abilities. Venn offered the example of the consumptive Englishman in
should b~ stripped of its pretension of being a general logic of uncer- Madeira, a species too rare to have its own life tables. An insurance
tainty. 43 In France, Joseph Bertrand argued that the principle of indif- company, which is concerned only with averages, can classify this man
ference was not fully coherent, and that "at random" could sometimes either among consumptive Englishmen or among English residents of
be interpreted in different ways to yield utterly disparate results. He fur- Madeira, but no unique probability of death can be assigned to the in-
ther maintained that probability could only be useful in relation to large dividual.
numbers of similar events, and that its application to man and nature Venn maintained that the ultimate basis for probability calculations
was justified exclusively by statistical regularity. "Le hasard corrige le is experience, and its justification the "notorious fact" that a "wide and
hasard " he wrote-"Chance corrects itself. General experience reveals somewhat vague kind of regularity that we have called Uniformity"
'
the justice of this maxim even to those who fail to recognize its rigor. characterizes a vast range of phenomena. 4 6 He deemed it unfortunate
The word rigor is no exaggeration. The results of the free action of that the convenience for mathematical purposes of a priori assumptions,
chance can be predicted with certainty, without impeding in the least its and the association of probability with gambling problems, had "pro·
caprices. "44 duced the impression in many quarters that they are the proper typical
•' Ibid., p. 23. Sec also pp. 135-136, 144. . . 45 W. F. Don kin, "On an Analogv relating to the Theon· of Probabilities and on the Prin-
"Johannes von Krics, Die Principien der Wahrscheinlichkeitsrechnung: cme log1sche Un-
ciple of the Method of Least Squares," Quarterly Journal of Pure and Applied :'vlathematics,
tersuchung (Freiburg, 1886). '(18s;), 152-16z, p. '53·
.,_,Joseph Bertrand. Calcul cles probabi/ites (Paris, 1888), pp. 68-69; 3lso pp. 4-5 ·• 6 Venn, Logic o{Chance ln. 27), p. 240 .

86
(
-Prom Nat11ris Urn tolnwrance
instances to illustrate the theory of chance. Whereas, had the science
been concerned with those kinds of events only which in practice are
commonly made subjects of insurance, probably no other view would
ever have been taken than that it was based upon direct appeal to ex-
perience. "47
•7 Ibid., p. 76.

PART TWO

88
THE SUPREME LAW OF UNREASON

Although the most sophisticated statistical mathematics of the early


nineteenth century grew out of the use of error theory in astronomy and
related fields, the most fruitful applications of statistical reasoning
proved to be elsewhere. So long as probability functions were held to
represent the imperfections of measurement and of observation, there
was little reason to do more with the variation that emerged than to find
While revolutions are taking place with a frightening
ways to estimate it and to eliminate as much of its effect as possible. The
speed, empires are collapsing, and passions ravaging the
source of modern statistics is to be found less in error analysis than in the
world there are savants and philosophers who are follow-
use of probability as a modeling tool to capture and analyze real varia-
ing at~entively the progress of events, analyzing them and
tion in nature and society.
striving to subject them to general laws :eryrnearly as con-
Significantly, the usc of probability to analyze genuine variation orig-
stant as those of astronomy and physiCS. [ 0 prove that
inated in the social science of statistics. In the mid-nineteenth century,
man is placed under the empire of fixed laws which direct
statistics seemed the most promising field of application for the calculus
his will without obstructing his free agency, such IS the
of chance. Charles Couraud evinced especially high expectations for
goal of these ... works. -Due BE CARAMAN (1849)
the statistical work of Quetelet in his 1848 history of probability.' This
shift of interest within probability from error to variation is already to be
The masses seem to me worthy of notice in only three re-
found, in a general way, in the frequency interpretation. Much more
spects: first as blurred copies of great men, produced on
central to the development of a mathematical statistics, however, was
bad paper with worn plates, further as resistance to the
Quetelet's insight concerning the distribution of traits in human popu-
great, and finally as the tools of the great; beyond that,
lations. His idea presents a specific connection of the greatest impor-
mav the devil and statistics take them. tance between social and mathematical statistics. It both illustrates and
. FRIEDRICH NIETZSCHE (1874)
confirms the more general argument of this book concerning the role of
social science in the development of statistical thinking.
There are three kinds of lies: lies, damned lies, and statis-
The object of Quetelet's attention was the probabilistic error func-
tics. MARK TwAIN (attributed to Disraeli)
tion, what Calton called the "supreme law of Unreason." The history
of this curve, now known as the Gaussian or normal distribution, is
practically coextensive with the history of statistical mathematics during
the nineteenth century, and its reinterpretation as a law of genuine
variation, rather than of mere error, was the central achievement of
nineteenth-century statistical thought. That reinterpretation took place
only gradually, through what in retrospect looks like a process of creative
misunderstanding. Once again the lead was taken by social thought, and
' Charles Gouraud. Histoire du calcul des probabilites depuis ses origines jusqu'a nos jours
(Paris, 1848), pp. 12, 139.
" ---~---------·--- ·------ -·'WI':

( ( (

--The Supreme-Law of Unreason


in this matter its influence on the natural sciences is demonstrable. The
main line of development proceeded from Laplace to Quetelet to Max- ---------Chapter F o u r - - - - - - - - -
well and Galton, and from the error of mean values in demography as
well as astronomy to the deviations from an idealized average man to the
distribution of molecular velocities in a gas and the inheritance of bio- , THE ERRORS OF ART AND NATURE
logical variation in a family. Ultimately, even the analysis of error was
transformed by this line of statistical thinking.
1
The expohential function ~ e-''1' , later to become known variously
as the ast~onomer's error law, the normal distribution, the Gaussian
density function, or simply the bell-shaped curve, was introduced to
probability theory by Abraham De Moivre. Like most early probability
mathematics, it first arose in the context of games of chance; it appeared
as the limit of the binomial distribution. Because of its usefulness in
combination and permutation problems, the binomial had become the
heart of the doctrine of chances, but while it could be used to solve, in
principle, a great range of problems, a superhuman computational ef-
fort would have been required to find, for example, the probability that
1,ooo tosses of a fair coin would yield 480 heads and 520 tails. In 1730,
James Stirling, with help from De Moivre, derived an exponential ap-
proximation for factorials. De Moivre then showed in a paper of 173 3,
reprinted in 1738 in the second edition of his Doctrine of Chances, that
the exponential error function gave a very good approximation to the
distribution of possible outcomes for problems like the result of 1,ooo
coin tosses. Now, for the first time, it was practicable to apply probability
theory to indefinitely large numbers of independent events. De Moivre
was highly enthusiastic about his new technique, but at the same time
thought it represented the hardest problem in the whole Doctrine of
Chances.'
De Moivre's discovery received little attention until Laplace began
writi~g on probability during the 1770s. Laplace saw this method of ap-
proximation as invaluable, for he hoped to use it in conjunction with
the new technique of a posteriori probability-which he developed in-
dependently of Bayes around 1774'-to predict distributions of events in
the future, or to infer the existence of real causes, from a record of past
' On De Moivre, see lvo Schneider, "Der Mathematiker Abraham De Moivre," Archive, 5
(1968-69). 177-317. "
• See Laplace's classic memoir, "Memoire sur Ia probabilitc des causes par les evenements"
(1774), in Oeuvres, vol. 8, pp. 27-65.

93
( (

Tfie Errors o{ltn and Nature- The Errors of Art and Nature

occurrences. In a paper published in 1781, Laplace putthis exponential doing so. Quetelet, among others, found a particularly valuable exem-
approximation to work to show that the observed surplus of male over plar in Laplace's calculation of the probability that atmospheric pressure
female births in Paris every year without exception required no provi- was subject to a regular cause of diurnal variation. Since the level of the
dential intercession, since, in view of the average male to female ratio barometer was also subject to a host of irregular causes of variation, im-
and the total number of births, the probability that female births would plying that the "errors" exceeded in magnitude the effect itself, there was
exceed male in any given year was only 1l259. Using the same methods no alternative but to compile a large number of observations and apply
he confirmed with probability 410457l410458 that the male birth ratio the techniques of probability. "As events multiply themselves, their re-
in London was genuinely greater than that in Paris, and calculated the spective probabilities become more and more developed," wrote La-
number of citizens who must be counted to assure a certain standard of place, and indeed the mean value for morning and afternoon barome-
reliability in a population figure derived from a complete register of tric readings revealed a small difference. Only after an infinite number
births and an estimated ratio of annual births to population. 3 The ap- of observations would the discrepancy point with certainty to a fixed
plicability of the mathematics of chance to demographic problems re- cause, so it was necessary to use error analysis to calculate the probability
affirmed Laplace's view that probability theory was generally usable as a with which the effect was indicated. The probability that chance alone
means for narrowing the range of uncertainty, and he began speaking of would produce a mean effect as large as that observed could be expressed
De Moivre's exponential function as the law of facility of errors. He in terms of an integral of the error function. That probability proved to
hailed the exponential integral as a tool of the probability of causes, and be minuscule, and accordingly, Laplace inferred that the effect indi-
proposed that investigators should use error analysis to demonstrate that cated the existence of a constant cause. 6
an obse~t-re~yexists before concocting arguments to explain The incorporation of the same function into observational astronomy
it. 4 at the beginning of the nineteenth century guaranteed that it would be-
Laplace was able to generalize De Moivre's mathematics, making it come known to virtually all physical scientists and many others as well.
unnecessary to assume that individual events distribute themselves sym- Astronomers had been debating for several decades the best way to re-
metrically about the underlying mean like fair coin tosses. He also de- duce great numbers of observations to a single value or curve, and toes-
veloped the general method of error analysis, and applied it to several timate the accuracy of this final result based on some hypothesis as to
problems in a wide variety of fields. Perhaps he was too easily convinced the occurrence of single errors. Thomas Simpson, in 1798, and then
of the quality and independence of the numbers with which he worked, Lagrange furnished expressions for the probability of a given magnitude
as when he calculated odds of one million to one that the mass of Jupiter of error in a mean value, and Laplace also wrote on this subject during
fell within bounds subsequently proved to be mistaken by George Airy. 5 the eighteenth century.7 Then in 1807 Legendre announced a general
Still he exhibited more skill and resourcefulness in the use of error anal- method for reducing multiple observations of an object, such as a star or
ysis ~han would be seen again for more than half a century after his planet. It involved choosing the curve of some specified type that min-
death. imized the sums of the squares of deviations of the individual measure-
Laplace defined for subsequent generations both the class of problems ments from it-a procedure which, if the object were stationary, was no
to which error analysis could be fruitfully applied and the method for different from taking the mean. This, the so-called "method of least
1 See Laplace, "Memoire sur les probabilites" (1781), in Oeuvres, val. 9, 383-485, PP· 437,
squares," was presented by Legendre without probabilistic justification.
466; "Surles naissances, les mariages, et les morts a Paris depuis 1771 jusqu'en 1784, et dans Almost immediately, Carl Friedrich Gauss announced that he had
toute l'etendue de Ia France pendant les annees 1781 et 1782" (1786), in Oeuvres, voL 11, 35-
46. Daniel Bernoulli had applied De Moivre's result to the proble~ of ~trth rahos m 1no- been using this method since 1798. At the same time, he presented a
1771, in a work that Laplace doubtless knew. See 0. B. Sheymn, Dame! Bcrnoulh on the 6 Laplace, Theorie analytique des probabilites (Paris, 2d ed., 182o), in Oeuvres, vol. 7, pp.
Normal Law," in SHSPz, pp. 199-202.
• Laplace, "Memoire sur les approximations qui sont fonctions de tres grands nombres, 280, 3 55-3 58.
7 See Stigler; also Isaac Todhunter, A History of the Mathematical Theory of Probability
Suite" ( 1786), in Oeuvres, voL 10, 295-338, p. 319· (New York, 1949), chap. 15.
s See Cournot, pp. 242-243.

94 95
The Errors of Art and Nature The Errorso{Art and Nature
forma! derivation of it--or rather, worked out the conditions under One more extension of the growing domain of the error law was made
which it was valid. He concluded that this method yielded the most by Joseph Fourier. Fourier became involved in social statistics earlv in
probable value for the actual position or path of the object in question lm c.areer as the result of his administrative as well as scientific activities.
if it was supposed that errors from the true value were distributed ac- Havmg serv~d as permanent secretary of Napoleon's Institute of Egypt,
cording to the already-familiar error curve of De Moivre and Laplace. he wa.s appomted prefect of Iscre upon return to France-in vvhich
Gauss's derivation, like those that followed it, remained controversial, capaCity he supervised for his departement the great compilation
for Gauss offered no justification for his premise, the applicability of the planned by the Bureau de statistique. Although his participation in this
error law, beyond its necessity if the conclusion were to be upheld. His ~nsuccessful project was evidently halfhearted, it was a statistical posi-
argument was also found exceptionally obscure and difficult, but never- tion that h.e secured to support himself under the restored monarchy. He
theless it was enormously influential. In 1810 Laplace provided an al- became dtrector of the Bureau de statistique of the Departcment de !a
ternative derivation of least squares, giving at the same time his familiar S.,eme under his friend and former pupil at the Ecole polytechnique, the
argument on the limit of the binomial to establish that errors of mean Count ~e Chabrol, prefect of the Seine. In that capacity, his involve-
results in astronomy, like those in population studies, ought to be dis- ment Wtth the collection and presentation of statistical tables was inti-
tributed according to the error law. 8 mate. The publications of his office were exemplary, and were indis-
Nineteenth-century astronomers and mathematicians produced an pensable for the first generation of statistical enthusiasts in France 10
enormous volume of literature on the method of least squares, some Fo.urier published five papers on statistics and demography aft~r his
merely to instruct readers in the techniques for using it, but much that appom:ment to the statistical bureau in Paris, including four weighty in-
questioned the foundations of the method or sought to clarify its relation troductions to the offi~ial statistical compilations. The first of these pa-
to the error law. By the late 183os a consensus seemed to be established pers, a lengthy mem.mr on what he called the analytical theory of in-
that the error curve applied not only to the errors of calculated mean val- surance, used Dame! Bernoulli's logarithmic function for moral
ues, but also to the distribution of individual errors. In 1837 G. H. L. expectation to show how both parties to an insurance contract can derive
Hagen argued that every particular measurement was a composite phe- bene~t from it. Tl:e next two dealt with population, one a pioneering
nomenon, and that measurements were subject to the error law for the work m mathematical demography, showing how life expectancies and
same reason that the binomial distribution converged to it. The next other q~antities could be calculated from age and mortality tables for a
year, his teacher, F. W. Bessel, specified eleven different classes of ran- populat~on not in equilibrium, the second an empirical study of the
dom errors that occurred in every telescopic observation, including ex- population ofParis since the end of the seventeenth centurv. These were
pansion and contraction of the telescope, errors of the observer, irreg- f~llowed by two memoirs on the value or precision of me-an results de-
ularities in the atmosphere, and so on. This plurality of sources of nved from numerous observations.
deviation, he explained, accounted for the dominion of the error curve ~ou~ier was highly impressed by the regularity of statistics, and he
over their results.9 m~t.ntat2ed as "t~e foundation of most statistical researches" the prop-
os:twn that the mdefimte repetition of events that appear to us as for-
8 See Stigler; also articles on Gauss (by Churchill Eisenhart) and Laplace (bv Stephen Stig-
ler) in William H. Kruskal and judith M. Tanur, eds., International Encyclopedia of Statistics tUitous must make everything irregular in them disappear; in a series of
(2 vols., New York, 1978), vol. 1, pp. 378-386, 493-499; R. L. Placket!, "The Discovery of an tm~ense number of facts, only the constant and necessary relations,
the Method of Least Squares," in SHSPz, pp. 230-251.
0 See G.H.L. Hagen, Grundziige der Wahrscheinlichkeits-Rechnung (Berlin, 1837); F. W.
determmed by the nature of things, endure. " 1 ' At the same time, he em-
Bessel, "Untersuchungen uber die Wahrscheinlichkeit dcr Beobachtungsfehlcr" ( 18 38), in phasized that mean values alone were not enough for the science of sta-
Abhandlungen, val. z (Leipzig, 1875), pp. 372-391; also Eberhard Knobloch, "Zur Grundla-
genproblematik der Fehlertheorie," Menso Folkerts. Lindgren, eds., Festschrift (iir Helmuth '" See John Herivel, Joseph Fourier: The I\1an and the Phvsicist (Oxford 19 --) esp
ll3, 118. ' • I) • 'pp.
Gericke (Stuttgart, 1985), pp. 561-590. The curve had already been associated with atmos-
pheric retractions by Christian Kramp, Analyse des refractions astronomiques et terrestres Fo1urie~, "Notions generales sur Ia population," in Recherches statistiques sur fa
.11" djospeph
(Strasbourg, 1 799). v1 e e anset eDepartementdelaSeme vol 1iParis 18 2 ,· 2 ded r 1 8,;, 3).• pp. XXXVlll-XXXIX.
' • \ ' 1 •
... ·

97
( (
The Errors of Art-ancLNature __ _ The Errors of Ar~ and Nature
~o dire~t action except between material points extremely near, and it
tistics, and that it was necessary also to find the limits within which the
IS for th1s reason that the expression for the flow has the form which we
mean is confined. ' 2 To this end he presented with great simplicity and
assign to it. The flow thus results from an infinite multitude of acf : 1s
clarity the technique for evaluating the probability of an error of any
whose effects are added. " 1 4
given magnitude by computing a standard error of the mean and then
checking it against certain probabilities from an integral table for the er- . In additio~, Fo~rier was vastly impressed by the extraordinary diver-
ror curve. His "standard of certainty" in hypothesis testing was a bit Sity of fields m wh1ch the error law had found application. Indeed he
more stringent than the current five percent; he thought that a chance
mainta~ned that the establishment of a mathematical identity am,ong
these disparate phenomena pointed to some underlying unity that hith-
of one part in twenty thousand, corresponding to about four times our
erto had been concealed. Fourier's faith in the universality of the error
standard error of the mean, was suitable. 1 '
curve was almost mystical: ..,._
All this had been worked out fully, though never presented very
clearly, by Laplace. Fourier also used the error curve, however, as the We ought to remark, in concluding this extract, that.the principal
formula for an actual physical distribution. He did so in his landmark element of the analysis of probabilities is an e)l;ponential integral
contribution to physics, the Traite analytique de Ia chaleur, where he that has presented itself in several very different mathematical the-
showed that if all the caloric (heat) in an infinite, one-dimensional con- ories. Geometers have considered this function in an abstract man-
ductor were concentrated initially at a single point, and then allowed to ner, and as an element of general analysis .... This same function
disperse freely, the heat at any subsequent time would be distributed in is connected with general physics. It is required in order to char-
accordance with the error function. Fourier's analysis relied on exact acterize the motion oflight through gaseous milieux. We have dis-
differential equations rather than the calculus of probability, and he as- covered in recent years that it also represents the diffusion of heat
serted proudly that his mathematics was independent of any particular in the interior of solid substances. Finally, it determines the prob-
theory of the nature of heat. Still, he expressed favor for a physical ability of errors of the measures and mean results of numerous ob-
model to which Laplace's binomial derivation of the error curve could servations; it reappears in the questions of insurance and in all dif-
have been applied. "Of all modes of presenting to ourselves the action ficult applications of the science of probabilities.
of heat," he proposed, "that which seems simplest and most conform- Thus, mathematical analysis unites the most diverse effects and
able to observation, consists in comparing this action to that of light. disc~vers in the~ common properties. Its object has nothing ~f the
Molecules separated from one another reciprocally communicate, contmgent, nothmg of the fortuitous. Imprinted in all of nature it
across empty space, their rays of heat, just as shining bodies transmit is a preexistent element of universal order. This science has n~c­
their light." That is, the dispersion of heat could most plausibly be re- essary relations with all physical causes, and with most of the com-
garded as taking place in tiny, independent increments, as the rays mi- binations of spirit. It preserves only those methods that can simul-
grated along an intricate course from molecule to molecule. "There is taneously become more extensive and more simple; and its true
" Fourier, "Memoire sur Ia population de Ia ville de Paris dcpuis Ia fin du XYlic siecle," advances always recur to two fundamental points: public utility,
in ibid., vol. 2 ( 182 3), p. xx. and the study of nature. 1 ;
" Fourier, "Memoire sur les resultats moyens dcduits d'un grand nombre d'observations,"
in ibid., vol. 3 (1826). Fourier's "degree of approximation of the calculated mean to the true
' All the early applications of the error law to which Fourier alluded
mean 7"

g = j ~ (B - A>) = k .j2[(a - A)' + (b - A)' + ... ] and which have been mentioned here, could be understood in terms of
where a, b, ... were the individual measures, n the number of observations, A their mean, and
a binomial converging to an exponential, as in De Moivre's original der-
B the mean of their squares, is Vz times larger that the current "standard error of the mean,"
' 4 Fourier, The Analytical Theory of Heat (1822), Alexander Freeman, trans. (New York
and was inserted into an error curve of the form
2 t 1945), pp. 32, 460-461. '
P = • r-
V'IT
S e-x'dx,
o
.''Fourier, "E~trait d'un memoire sur la theorie analytiquc des assurances," Anna/eo~ de
chzmze et de physzque [2], 10 (1819), 177·189, pp. 188-189.
where twas measured in units of g. Sec the table in Kramp, Analyse (n. 9).
99
The-Errors of Artand Nature
The Errors of Art and Nature
ivation. All but Fourier's law of heat, which was never explicitly .tied to
malhematical probability except by analogy, were c~~11patible w1th the bodiment of terrestrial imperfection-as from the potential it held out
classical interpretation of probability. Just as probabd1ty was a measure for unifying the study of man with that of the stars. There was, it seems,
of uncertainty, this exponential function governed the chances of error. a preordained harmony between Quetelet's metaphors and the standard
It was not really an attribute of nature, but only a measure of hum.an Laplacian understanding of the error law.
ignorance--of the imperfection of measurement techniques or the m- One way to view Quetelet's average man was simply as an instrument
·accuracy of inference from phenomena that occur in finite n~mbers to of social analysis. This being "would be of only mediocre interest,"
their underlying causes. Moreover, the mathematical operations used Quetelet punned, "if I had not recognized that the average man also en-
in conjunction with it had a single purpose, to reduce the error to the joys some very remarkable properties in relation to the social system, and
narrowest bounds possible. With Quetelet, all that began to. change, and which are such as seem to me to open a vast field to a new order of social
a wider conception of statistical mathematics became poss1blc. researches. " 16 But Quetelet went immediately beyond this. As l\1aurice
Halbwachs observed in 1913, Quetelet was inspired by the philosophy
of Victor Cousin to develop the analogy between the mathematical idea
QUETELET: ERROR AND VARIATION of the average man and the moral idea of the juste milieu. 7 The juste 1

milieu was a symbol of Louis Philippe's July monarchy, but it could


When Quctelet announced in 1844 that the astronomer's error law ap- equally well have been applied to post-1830 Belgium, whose first king,
nlied also to the distribution of human features such as he1ght and gnth, Leopold I, was married to Louis Philippe's daughter. It referred simul-
he did more than add one more set of objects to the domain of this .prob- taneously to the supremacy of the bourgeois class and to the constitu-
ability function; he also began to break down its exclusive association tional monarchy, perched precariously between divine-right kingship
with error. Quetelet intended nothing of the sort, however. I-hs exten- and radical democracy. At the same time, it embodied an ideal of mod-
sion of the range of the error law seemed to him an epochal. ach1ev~ment eration, especially in political behavior, to which Quetelet already was
strongly attached.
precisely because it vindicated the presuppositions of soCial phys1cs by
proving that the concepts and formalism of astro.nomy were fully ade- The philosophical emblem of these post-1830 governments was Vic-
quate to capture the essential properties of that h1therto mystenous e,r~­ tor Cousin's eclectic philosophy, which Quetelet greatly admired.
titv man. This surprising testimony to the power of the method of cc- Cousin defended eclecticism on the ground that no particular philoso-
le~tial mechanics bolstered Quetelet's longstanding claim that the socwl phy could suffice for l' humanite tout entiere, and also because it seemed
sciences could do no better than to imitate the physical. It revealed w1th to him the embodiment of the great political achievements of his age. It
exceptional clarity the need for statisticians to familiarize themselves reflected, as Cousin proclaimed in an encomium on Charles X's resto-
with advanced mathematics if their science was ever to reahze 1ts potcn- ration of the charter, a synthesis of the old regime's exclusive monar-
ti~. . chical principle with the exclusive democratic principle of the great rev-
It is fullv characteristic ofQuetelet's scientific style that he mterpreted olution. ' 8 Cousin's taste for moderation appears clearly in the
his proud~st discovery not as an indication that the ~athematics of error concluding lecture of a series on the history of philosophy, in 1828. His
had been conceived too narrowly, but as clear evidence that hur:nan remarks on the Napoleonic wars-which, we learn, were in reality a
variation could be understood in the same terms as errors of obs~rvati_on. prolonged battle between the ideas of democracy and aristocracy--were
transcribed and published as follows:
Indeed, this interpretation was not simply the ex post facto ratJm:ahz~­
tion of an empirical discovery. Quetelet had been very close to ldenb- 6
' Quetelct, "Recherches sur lc poids de l'homme aux differcns ages," NMAB, 7 ( 1832),
fying variation and error ever since the birth of .socia.l physic~, and the separate pagination, p. 2.
'' Maurice 1-lalbwachs. La theone de l'homme moyen: Essai sur Quetelet et Ia statistique
appeal of the error law arose as much from the Impbcahon that math- morale (Paris, 1913), p. 6.
ematical order extended even to deviations from mean values--the em- ''Victor Cousin, Cours de philosophie: Introduction d l'histoire de Ia philosophie (Paris,
18z8), 13th lecture. pp. 39-.p
100
101
_, ~

The Errmiof Art andNature The Errors o(J\rt and Nature

VVhat did they bring, gentlemen? Neither the one nor the othe~ tutcd [for artists] ugliness in body as well as vice in morals and a state of
fncithcr aristocracy nor democracy]. Which was the conquero~'. sickness with regard to the constitution. " 2 " He also expressed longing for
~hich was the vanquished at Waterloo? Gentlemen, there was a mean philosophical and political position that could resolve social
none vanquished. (Applause). No, I prot:st that thosew.arshad no conflict and "conciliate the most advantageously the interests of the dif-
vanqt ;ished party; the only victors were European C!Vlhzatwn ,and ferent parties.,,'
the charter. (Unanimous and prolonged applause) Yes, messieurs, L'homme moyen, then, was not just a mathematical abstraction, but
. l ·1 b L · XVIII '. the
it is the charter g1ven vo untan Y Y outs
charter mam-
· . . ~ . a moral ideal. Quetelet held that great inequalities of wealth and vast
tained by Charles X, the charter destined to d~mmatJOn m Fran,ce, price fluctuations were responsible for crime and turmoil; he exalted the
and destined to subdue, I do not say enermes, she hasnm~~ of life of moderation, unaffected by sudden passions, and conjectured that
th b t all those who have fallen behind French CIVIlmll!On; the "higher classes" live longer than the "low people" not because of
em, u . l b ·n· t t ne
(Redoubled applause) it is the charter that IS t 1e n Ian ou cm wealth or nutrition, but because of their "habits of propriety, of tem-
of the bloody conflict of the two systems that by now have already perance, of passions excited less frequently and variations less sudden in
seen their day, to wit, absolute monarchy and the extravagances of their manner of existence. " 2 4 He listed Aristotle beside Archimedes
democracy. 19 among the discoverers of the true idea of the mean for teaching that "vir-
All this, Cousin pointed out, was evidence that Fr~nce had ~ntere~ an
tue consists in a just state of equilibrium, and all our qualities, in their
era of eclecticism' of "moderation in the philosophical o,~der . that ca~1 greatest deviations from the mean, produce only vices. "z; The average
do nothing during the days of crisis" and that becomes a neces.slty a - man embodied these mild virtues with mathematical precision; his fac-
terwards"; of the eclecticism that belongs. to no party, to no c~tene, tha~ ulties developed "in a just state of equilibrium, in a perfect harmony,
"avows itself satisfied with its century' wJth Its country and with, thl e a~ equally distant from excesses and defects of every kind, in such a way
.. I ask whether philosophy can fad to be e.c echc, that ... one must consider him as the type of all that is beautiful and
tua l or d er oftl nngs. . ·
.. all that is good. " 26 Quetclet even proposed a theory of the enlightened
when everything aroun d It Is.
?"20
. , I
Q t l t though never willing to express such unquahfied approva will to explain the greater regularity of moral statistics than those .?f
for t~~ :x~s~ing order, was enthralled by Cousin's ideal ofmoderatlOn births and deaths. It involved, in effect, a will to mediocrity, a tendency
. He quoted at length Cousin's observatiOn that the for the enlightened to resist the influences of external circumstances and
an d comprmmse. t1
.. t of a people resides in every individual, that each people mus 1ave to seek always to return to a normal and balanced state. "It is only among
;~~mmon type, and that the great man is he who most perfectly rep- men entirely abandoned to the heat of their passions that one sees those
resents this tvpe. While Cousin, however, stressed .that great ~nen must sudden transitions, faithfully reflected from all the external causes that
also preserv~ their individuality, thus uniting umformit~ with m~ltl- act on them. " 2 7
l' .t ' Quetelet proposed an identification of greatness wlt? the m~an. Hence the progress of civilization. the gradual triumph of mind, was
~~c~~;ote that "an individual who epitomized in himself, .at a gi~en equivalent to a narrowing of the limits within which the "social body"
time all the qualities of the average man, would represent at once ~1 oscillated. It must inevitably be reflected in a tendency for progress to
the ~reatness, beauty and goodness of that bei~Jg. "21 He res~)lve~ that t, e become ever more smooth and gradual, the risk of "falling into an ex-
mean alone can represent the ideals of a soCiety m pohbcs, aesthetic~,
and morals. "Deviations more or less great from the mean have constl- Quctelet, "Recherches sur lc penchant au crime aux differcns ages," ?\'MAB, 7 (r8p).
separate paginatJon, p. 6.
'' Quctclet, Sur /'homme (n. 21), vol. 2. pp. 296,297.
"'Ibid., pp. 36-37. '•ibid., vol. 1, p. 237; also Qudelet, "Penchant au crime" (n. 221. p. 44·
::~:~tel~: ~:-:hamme et le developpement de s:s facultC~s, outsa~de P~~s;~~1t:,:~ci~:1e ' 1 Quetelet, Theorie des probahilites (Brussels, 1853), p. 49·

•6 Quctdet, Sur l'homme (n. 21), vol. 2. p. 287.


( 1g
35 ; 2 vols, Brussels. 1 8-,6). vol. 2, p. 289; from Cousm, -011rs "· L J, w' ,.
'' Quctelet. Du systi!me social et des lois qui /e regissent (Paris, 1848), p. 97·
6-7.
102
103
(
The Errors of Art and Nature
The Errors of Art and Nature
lahndg the revenue his phalansteries wonld derive from in , ·d
treme" perpetually diminishing. Quetelet epitomized the moral of his pro uctlon under the n ' .. t f . crease egg
nounccd laws of s .. d, be\\ hsys ~m o soCial harmony, Quetelet an-
most influential book, On Man, as follows: . · UICI e Y angmg m Paris and of · b
sexag~nar!an twentie~~~r~a~c~u~h:~en
~;~:~~·::~;;~;~:N;~;;~w' .wc;c p~ci'elYthme th"t gomned ;he nng<~e
women and young men in their
I will finish this chapter with a final observation, which may be
seen as a consequence of all that has preceded: it is, that one of the
. principal acts of civilization is to compress more and more the lim- a perfect rule ~vera even. aw reakmg to natural law, that exercised
its within which the different elements relative to man oscillate. to discover that stat~~rerlt dJsorde~. Quetelet was particularly delighted
The more that enlightenment is propagated, the more will devia- bv the traumas of IS !Clat. aws,hsuc 1 a~ th~se of crime, were unaffected
tions from the mean diminish; moreover, as a consequence, we . revo u Jon t us md!cahng a d d. . h
realm that was scarcely affec;ed b th . I eep or er m t e social
tend to unite ourselves with what is beautiful and with what is Tl . .d . y c megu ar march of politics 3'
fi d 1e 1samefCOnS!. erations aPP ]'Ie d t o deviahons
. . . T
ra~io~alai.t\VY .an~
good. The perfectibility of the human species is derived as a nec- . from mean values
essary consequence of all our investigations. Defects and monstros- oQvartwlhton was to bring it within the domain of order
ities disappear more and more from the body; the frequency and the . · ue e e wrote:
gravity of maladies are combatted with greater effectiveness
through the progress of medical science; the moral qualities of man
!ro~1~~;~:~s:i1te tr~uble to exan:ine and to bring together observa-
wh t . . ~ care and suffiCiently numerous, one will find that
will meet with improvements no less tangible; and the more we ad-
cip~esw::1~e7~r ted :~the effect of chance, is subjected to fixed prin-
vance, the less need we fear the effects and the consequences of
f I ' a no m~ escapes the laws imposed by the all-pow-
great political upheavals and wars, the plagues of humanity. 28 er u onto orgamzed bemgs. What we call an anomaly deviates in
The implication of Quetelet's idealizations of the mean was that all ~ur ~yes from the g.enerallaw only because we are incapable of em-
deviation from it should be regarded as flawed, the product of error. This racmg e~ough thmgs in a single glance.
By placmg oneself in circumstances favorable for b , ,. .
did not imply, however, that variation must stand outside the domain of finds that ~ · d . o sen mg, one
science, for the special task of probability theory was "to establish an . , dor orgamze bemgs, all elements are subject to var:a-
t
admirable precision where one believed there were only games of flton aroun
.f a mean
. sta te, an d th at t h e variations to which the in-
'
chance." 2 9 Science, in Victor Cousin's words, was precisely "the uence o acCidental causes give birth, are ruled with SLich har
rnonv, and preclSlon,
· · t h at they can be clas~ d . d ' -
suppression of all anomaly, the ordered substituted for the arbitrary, numerically and bv order f .. . ,e m a vance
reality for appearance, reason for sense and for imagination. "3o Without which th , . , . s o magmtude, m the limits between
. . ey are compnsed. All is foreseen all is lawlike· on! .
~~~~:~~~~ .whi~~~~
general laws in the physical domain, Quetelet observed, "one can imag-
ine what dreadful chaos would be produced in the midst of these myri- leads us to suppose that all is ;ubject to the
ads of worlds circulating through space in a wholly disorganized man-
ner, and crashing against one another." He aspired to be the Newton "of
sol~'edwas
Error thatth uths b ams · h ed .f rol~lthe uni\.'erse, and Quetelet accordinglv re-
this other celestial mechanics," to find the laws that assured equilibrium e expressiOn ' aw of er .>' . . '
in the social domain, and he, like many would-be scientific thinkers of achievements of do .. " rors was mconsistent with the
seemed bett ~10 -..rn. s~Ience_. Law of accidental causes" scarcely
the early nineteenth century, derived special satisfaction from contem-
t t t. . Jerl, a. n . was cnticized. m a letter fro.m the mathematician and
plating the splendid noumenal order that he supposed to prevail beneath s. a JSt !Clan u es. Bienavme
tl . . , ' so tl I a t ()
""'"uete let eventually resolved to des-
the whir of phenomena. Just as Charles Fourier busied himself calcu- Ign a e ms curve ·simplv. as the b.mom1a . ll aw. B .
,g Quctckt, Sur /'hom me (n. 21 ). vol. 2, p. 342; also Du ;y>ti:me social (n. 27i, pp. q6-97· :: Quetekt. Du systeme social (n. 27), p. 30 1.
'9 Quctelct, Physique sociale. au essai sur le developpement des facultes de l'homme (2 vols., Jb,d., pp. 16-n.
" Ibid., P· 306; Quetclet, ''Sur que1ques. propne<cs
.. .. que prescntent les resultats d'unc serie
Brussels, 186q), vol. 1. p. v.
w Cousin. Cours (n. 18). 8th lecture, p. 19.

104
The En-ork6( Art andNuture
The Errors o{Art and j\lature
Quetclet introduced his great innovation in the use the probabilis-
tic error law, its application to real variation in nature, in 1844. He did :ow lhat height. This, Quctclet announced, was evidence offraud
so, however, without claiming that a new understanding of this math- ,lmdl lh.c was .able _to-calculate the number of men whose measurcn-Jen<
la( Jcen sl1 l tl · , j , c1 · l . · " 'L
ematical function was required. On the contrary, Quetelet, who never .. g 1 . } rcc ucc 111 ore er to gam them exemptions from their
mibtary obhgatwns. 14 -
maintained much discipline over his own penchant for metaphor,
viewed his discovery as proof that deviations from the golden mean of ~, 1\voyear~ . Quctelct pre~ented
1, later, in 1846, ~
l1I·s great d' .
IScovery m a
l'homme moyen were mere imperfections, even errors. Although the ul- ~;~l~e \opu ar fom:. As m the relatively technical paper, his expo~ition
timate effect of his innovation was primarily to broaden the possibilities . ,Js ~~~of vanatzon appealed m a discussion of mean values: ~)uetelet
of mathematical statistics, and not to promote in any significant way the wa_s ah\a}s more mteresteclm mean values than in variation for-its own
development of social science, Quetclct naturaiiy conceived his \vork as · sake. He began by distmgmshing two types of mean nllues A ·t~
a contribution to the latter. The introduction of the error law to the
m t
I: } . n an .l··
e mean, l.e wrote, may be calculated for the most incongruous sets
study of man revealed once again the homology between social and ce- ~~bjects, ~l,lt, :t reveals little or nothing about their collective character:
lestial science, confirming that the social scientist could best advance his JeiJ, ho\\cver, the v;:mation of a set of measurements follows the cus-
study by looking to the methods and even the laws of astronomy. t~71a;y law of errors, then the mean value-which is also the most prob-
Quetelet began his exposition of this new statistical law of man by pre- a,) e, (modal) value-may be regarded as a "true" mean. Thus if one
senting as clearly as possible the nature and uses of the error curve. First v.ere to compute the average height of a . ll
1 m1sce aneous assortment of
he derived the "law of possibility" from a high-order binomial, which louses on some s~reet, the result would be merely an arithmetic mean
he portrayed in terms of the expected distribution of outcomes after an ~mce the und~rlymg distribution would be altogether unsystematic If'
enormous number of repetitions of 999 drawings from a fair urn with I,nstead, a particular house were measured repeatedly and no system~ti~
equal numbers of balls of two colors. Thus 500 of one and 499 of the error w~re made,. each measurement would approximate the true value
other was most probable, 501 to 498 slightly less likely, and so on to 580- and thelf collective distribution would anproximate the ast . l
law f . . _ . · • ronom1ca
420, any result beyond which was nearly impossible. Then he observed o errors. As the measunng process was continued, the calculated
that measurements of the height of an individual, or of the position of a mean should graduall ·h · .
_
he1g t of the house. Y converge WJL mcreasmg precision to the true
h
star, would, if repeated in sufficient number, also distribute themselves
in accordance with the error curve. The expected value of the measure- Quet~let then proposed that the same reasoning ought to apply to the
ments would equal the true height or position, while each measurement ~rod~ctwn of copies, say of artistic works. If, for example, a thousand
would be subject to numerous causes of error. Finally, he queried copies were to be made of an ancient statue the G'Jadi'ator th . .
· ld d · ' , en mac-
v.:hether there might exist an homme type, a true average man, of which curaCies. wou oub~less be governed by the law of errors. Indeed, this
every real man is an imperfect replicate in the same way as a measure- could be shown empmcally, m a certain sense:
ment of height is an effort to establish the true height. Indeed there was,
he concluded. It could be seen empirically, by means of a table of the I will perhaps astonish you in saying that the experiment has b
dow' yes t I - l h ecn
distribution of chest sizes of Scottish soldiers placed beside a table of the -· . , ru y, more t 1an at ousand copies of a statue have been
astronomical error function. He followed this with a similar table of the me~sured, and though f will not assert it to be that of the Gladiator
heights of young Frenchmen who had presented themselves for con- It dJffers, many event, only slightly from it: these copies were eve~
scription. Again the conformity with the error curve seemed excellent, hvmg ones, so that the measurements were taken with all possible
except that the numbers dropped off sharply in the neighborhood im-
mediately above 1. 57 meters, and showed a corresponding surplus be-
14 Quctclet "Sur 1'·,
.
-', · d
ation des mo;cnncs ,';tc~rat!On es documents statistiqucs, ct en particulicr sur l'appn'ci-
."vlathcmatics~ of Societ . V , , t2 I r844},F,;cff. On Quctelct and the normal law, sec my "The
v. ana "'nand rror 111 Quc''l t". s· f f BJHS -
d' observations ... , " Bulletin de I'Academie Royale des Sciences et Belles-Lettres de Belgique, ~-ll!so 'p)tcttcr BduckE,- "Fro,;, Celestial Mcch~nics to Soc·i:t P~rv:~c:s res, , ,·,1 '! ,ISN(/9)8?), k5I-69d;
> '- e c · 1 t 1 · 1 . ····' ' ac~n e an
19, part z (lb52), 303-317. - . ,(T)o~;dre;!,'/~;s~f~~Pa'~~;~ocla/ Problems of the Sciences in the Early Nineteenth
J06
------- -~--- ----

The Errors of Art di!d Nature The Errors o{Art and Nature
chances of error, I will add, moreover, that the copies were subject cheerfully , bcll-sll3pcd curves to model the distribution of pro-
to deformity by a host of accidental causes. One may thus expect pensJhcs to crime, marriage, and suicide--confirrning still more im-
to find here a considerable probable error. ;s 'pressively tll;Jt crime was ;Jn attribute of the social type and not of i1
Qudclet's identification of live individuals with copies o~ statues con- vidual deviants. In his last book, Anthropomitrie, he drew curves giving
veys with unmatched clarity an impression of the way he Viewed human the development of various traits over the human life cycle, and implie~l
diversity. His supposed experiment was in fact the same as that pr~sent~d that these too would reflect the error curve provided the age axis was suit-
in 1844, the measurements of 5, 738 Scottish soldiers. Their conform1ty ably transformed to reflect real and not mere chronological age. He even
to the error curve was interpreted as implying that the distributwn was a proposed that his great law applied to entities at a higher level of organ-
genuine product of error. The soldiers had been designe~ according_ to ization, and that, for example, the area of states would be distributed ac-
a uniform pattern, that of the average man. They had faded to realize cording to the error law if there were sufficiently many to reflect accu-
rately the true state of things.l7
perfectly this imagined archetype of Scottish soldierdom on account of
a host of irregularities in their development. These were so severe that :\Tore broadly, "the same law, so simple and so elegant, applies ...
the measurements exhibited a probable error exceeding an inch. generally to all the physical laws, we add even the moral and intellectual
More particularly, the conformity of this distribution to th~ la\~ of er- laws of man. " 38 This was Quetelet's principal legacy to mathematical
rors could be explained in terms of Quetelet's customary d!sbnctiOn be- statistics. His great program of social physics, designed to form the foun-
tween constant and perturbing causes. The constant cause was the Scot- dation for an exact science of human societies, was only occasionally
tish tvpe--Scotland's average man. The perturbing causes included seen as persuasive, and those who did had little more idea how to im-
nutrit.ion, climate, and so on, all quite variable in this imperfect world, plement it than their master had. His recognition that the error distri-
so that a host of independent small errors was generated that might
bution applies not only to errors, but also to what few but Quetclet
either increase or decrease chest size. The development of soldiers was would regard as anything other than real variation was likewise taken up
thus mathematically analogous to the mean of many drawings from an at first by only a few scientists; but this idea was, if not strictly true, at
urn-which, after ;ll,
is simply the sum of a constant cause, th~ actual least accurate enough and concrete enough for his successors to do
something with it. In social statistics, it was endorsed bv a number of
ratio of black to white balls, and the inescapable but unsystematic small
errors generated by each individual drawing. Hence the binomial curve writers including the American Benjamin A. Gould, the. Italians M. L.
that regulated most games of chance should apply also t_o the sold1ers, Bodio and Luigi Perazzo, Quetelet's English admirer Samuel Brown,
and since the perturbing causes were very numerous, then effects could and even, in a more qualified wav, bv his German critic Wilhelm
be approximated with considerable accuracy by the error curve of De Lexis. ' 9 For the most important and influential developments of Quc-
Moivre and Laplace. telet's new application of the error law, however, it is necessary to look
to fields other than social science. ,
Quctelct was enormously proud of his discovery that the error law
governed human variability, and he applied it widely-:-some would say '"Surles proportions de Ia race noire."' ibid .. 21, part 1 (18qi, 96-wo; also Theorie (n 25 ),
pp. 72-78, and Anthropornetrie, ou mesure des diffirentes (acu/tes de l'homme ( 18;o; BrusscL.
indiscriminatelv-in succeeding decades. He upheld Jt as the dcfimhve !87!), p. 16.
1- Quctclct, Du systeme social (n. 27 ), p. 1 s6.
criterion of uni.ty of type, and he was able to refute polygenism during
'' Quetelct. Anthropometrie(n. 36), pp. 253-254.
the sensitive period of the American Civil War by declaring authorita-
N Sec, inter alia, B. A. Gould, Investigations m the i'vlilitar)' and Anthropoioaica/ Stallstics
tively, on the basis of a few measurements of American Indians and Ne- of American So/diers(Ncw York, 18691; Luigi Pcrozm. ··Nuo;·c applicazioni ck:J calcolo delle
groes who happened to pass through Belgium, that a single l~w of errors probal)lltta allo stud1o dci tcnomem statistici." Atti della K Accademia dei Lmcei. :\lemorie

l
delia classe di scienze morali, storiche. e ftlologiche. 10 ( 188:), 4 73-503; Samuel Brown, letter
applied to the whole of humanity. ; 6 With perhaps less ev1dence, he of Dec. 1871 1ll AQP, caiHcr 526; Wilhelm Lcxis, ""Anthropolugic und .\nthropometnc."
m Conrad et al., eels., llandworlerbuch der Stuatswissenschafi~n vols .. Jena. zd cd.;
" Quetdet. Lettres aS.A.. R. leduc regnant de Saxe Coburg et Gr;tha sur Ia theorie des pmba- vol. 1, pp. 38il-409. Quctclct hequcntlv c1tcd authors who received his ideas fa-
bilites, appliqwies aux sciences morales et poiitiques (Brussds .. '8-to). p. 1 ,6. , . . \orably; sec Physique sociale (n. 29}. passim.
,6 See Quctelet. "Sur lcs indiens 0-)tb-Bc-·Wa"s d lcs propornons de lcmscorps, Bulletur
de l"Academie Royale des Sciences et Belles-Lettres de Belgique. 1 :;. part 1 (1846), 70-;6, ond

'
108
-- - ~ - -~ ----- -

Social Law and Natural Science

- - - - - - - - - Chapter F i v e - - - - - - - - - - ity, and was n~t recognized in mathematical abstraction as a widely ap-
plicable technique until more than a decade later.

SOCIAL LAW AND NATURAL SCIENCE


MOLECULES AND SOCIAL PHYSICS

It i~ scar~ely novel to associate the statistical method in physics with the


It was 110mere accident that Quetelet's work on the error law, rather social SCience from which that phrase was derived. James Clerk Maxwell
than the much more copious and mathematically sophisticated treat- observe~ in 1873 that laws of gases could never be found by following
ments of it in the literature of observational astronomy and ~e~desy, pro- the m~twns and collisions of millions of independent particles, for in-
vided the inspiration for the most important writers on statistical math- formation about individual molecules was not available and the calcu-
ematics of the late nineteenth century. Quetelet opened up a whole new lations would, in any event, be impossibly complex. As an alternative,
perspective on the mathematical treatment of variation. lromcally,. h1s Maxwell proposed to the physicists of the British Association a new va-
perception of the wider applicability of the error law ~as made poss1~:e riety of social physics:
precisely by the depth ofh1s commitment to th~ traditional metap~ys1~s
of error but his readers were more interested 111 his results than m lus The modern atomists have therefore adopted a method which is I
opinio;s, and these results implied that much variation in nature was believe, new in the department of mathematical physics, though, it
governed by this simple and elegant law. ~h~s, while Quetele,t mte.r- has long been in use in the section of Statistics. When the working
preted his discovery as confirmation that vanabon could be neglected m members of Section F get hold of a report of the Census, ... the~
favor of the study of mean values, Maxwell and Calton, among oth~rs, begin b_r distributing the whole population into groups, accordin~
saw in it a convenient and valuable tool for analyzing with mathematical to age, mcome-tax, education, religious belief, or criminal convic-
precision the nature and effects of natural variation. . . tions. The number of individuals is far too great to allow of their
The mathematics of variation was instrumental for the Impressive tracing the history of each separately, so that, in order to reduce
achievements of the nineteenth-century kinetic theory, including Bol~z­ their labour within human limits, they concentrate their attention
mann's reduction of the second law of thermodynamics to mechamcs on a small number of artificial groups. The varying number of in-
and probability theory. It also provided the key in biology to the quan- dividuals in each group, and not the varying state of each individ-
ual, is the primary datum from which thev work
titative study of heredity, leading eventually to what IS now the most
purely statistical of the natural sciences, quantitat.ive genetics. Beyond The equations of dynamics complete!; cxpre~s· th~ laws of the
its importance for particular natural and social sciences, however, the historical method as applied to matter, but the application of these
new understanding of the error law that derived from Quetelet's work equations implies a perfect knowledge of all the data. But the small-
proved essential for mathematical statistics itself. As ~alt.on was wont to est portion of matter which we can subject to experiment consists
sav not in reference to Quetelet but to English statJshC!ans and all as- of millions of molecules, ·not one of which ever becomes individ-
tr~~omers, most scientists used the error curve in order to get nc~ of ually sensible to us. We cannot, therefore, ascertain the actual mo-
variation, whereas his own aim was to preserve and unders.tand It. New tion of any one of these molecules; so that we are obliged to aban-
techniques for handling variation were closely associated \~Ith new Ideas don the strict historical method, and to adopt the statistical method
about its nature. I show in Part Four that the mathematics of c~rrela­ of dealing with large groups of molecules. The data of the statistical
tion, the technique that, more than any other, inspired the creation of method as applied to molecular science are the sums oflarge num-
mathematical statistics, appeared first as a principle ofbwlogical hered- bers of molecular quantities. In studying the relations between
quantities of this kind, we meet with a new kind of regularity, the
110
111
/
\

Social Law and Natural Science


·Social Lawand Natw-alScience
. d , d 011 quite sufficiently under the same conditions, that we perceive wholly determinate
regularity of averages, whtch we ca~ epcn ~;ke no claim to that laws in the behavior of warm objects. For the molecules of a body
for all practical purposes: ?ut wh·~~~\c~n gs to the laws of abstract are indeed so numerous, and their movements are so rapid, that
character of absolute preCISIOn w tc 1 e on nothing ever becomes perceptible to us except average values. The
dynamics.' . . regularity of these mean values may be compared to the astonishing
. l lo y behveen statistical
constancy of the average numbers furnished by statistics, which are
By the time Maxwell began discussmg t 1e ana g . d d that this
. l l .f he had become persua e also derived from processes in which each individual occurrence is
and thermodynamtca :egu a~t ~~\ics ielded knowledge that, though
conditioned by a wholly incalculable collaboration of the most di-
new approach to questwns o p y y 'I . perfect and uncertain
l . bl · ctice was necessan Y tm verse external circumstances. The molecules are like so many in-
completeIy re ta em pr.a ' b .lt. t his approach from the be-
. · ·1
m pnnctp e. uc
s h a behef was not u1 tn
1
° l
'tl social science was not a ways
dividuals, having the most various states of notion, and the prop-
. · h and the ana ogy Wl 1 erties of gases only remain unaltered because the number of these
gmnmg, owever' . t 111axwell' s friend and corre-
. . d t · ply such uncertam Y· 1" • • . molecules which on the average have a given state of motion is
rat sed m or er o tm h h d h. : ews on the imperfection
t G thrie Tait w o s are vi b constant. The determination of averages is the task of the calculus
spond ent Pe er u
IS
, d. t hnical memoir at 1886 that e-
of probability. It would be a mistake, however, to believe that the
of statistical knowledge, observe l m ~ e~ ak up any initial distribution
cause collisions between gas mo ecu es re theory of heat involves uncertainty because the principles of prob-
of velocities and directions, ability come into application there. An incompletely proven prop-
f , h l . 1 complex question of the osition, whose correctness is for that reason problematical, is not to
~::a::~~: o~c~~~nu::~:~etna~s~~:e~~s rsolated individ~a~:~i:~ be confused with completely proven propositions of probability; the
latter represent, like the results of every other calculus, a necessary
. I . I tatistical question of the average e .
comparative y stmp e s .t 1'h·s distinction is forCJbly consequence of certain premises, and are likewise confirmed by ex-
. h · oups of a commum Y· 1
o f t e vanous gr h t. l by the extraordinary perience, whenever they are correct, if only a sufficient number of
. d on the non-mat ema tea ,
unpresse even b f h totally unpredictable, cases come under observation-which, because of the enormous
d . .th vhich the num ers o sue 1
stea mess w1 ' , uicides twin or trip c number of molecular particles, is always the case in heat theory.
h h t uncommon phenomena as s . ' d
t oug no & . pulous country, are maintaine Indeed it seems doubly necessary that they conform to conclusions
births, dead letters, c., m any po with the greatest strictness. 3
year after year. 2 ,

. . . th f t that Ludwig Boltzmann mtro- The work to which both Maxwell and Boltzmann typically referred
More impressive, perhaps, IS .e. ~c )endentlv of Maxwell, so far as when they discussed the regularities of social statistics was Buckle's His-
duced the same analogy m ~. 872-lll teti. was ~n enormously impor- tory of Civilization. Maxwell had read the relevant portions about a year
11 'fh ontext of tts presen a Jon f .
we can te .
h k f th ory d e 1.JVere d t0 the Vt.enna Academv~ o SCl-
ec before he began work on the kinetic gas theory, and expressed qualified
tant paper on t e me tc e ht to demonstrate the need admiration;4 subsequently, he referred routinely to Buckle when criti-
ence. In the intro.ductlon, Boltz~1i~n:~~~~~etic theory, and to establish cizing the idea that statistical propositions are wholly determinate.
for recourse to ratios and average l t . than those in any other Boltzmann, who enjoyed comparing the uniforril"and lawlike behavior
.. f this sort were no ess cer am
that proposi~IOns o . . h de use of an analogy with social sta- of gases with the regular profits gained by insurance companies, was one
area of phystcs. To this end, e ma
of Buckle's most enthusiastic admirers. He once remarked, in a lecture
tistics: . h in 1886: "As is well-known, Buckle demonstrated statistically that if only
It is to be ascribed exclu~i~~liht~s::e
most irregular processes yie e
c;:~~~=~t~~~~::~:.~e~;~~e~ a sufficient number of people is taken into account, then not only is the

1 Ludwig Boltzmann. "Wciterc Studien liber das Warmegleichgcwicht unter Gasmolehi·

, Maxwell, "Molecules," in SP, vol. f2,lPPK·)7r!~i~corv of Gases," in Scientific Papers (2 len," in WA, vol. 1, pp. 316·317.
, p G 'fait "On the FoundatiOns o t 1e me 1' • 4 See his letter to Lewis Campbell, Dec 1857. in Maxwell, p. 294.

vols..'Ca.mbridge, Eng., t898-t9oo), vol. 2, p. n6.


ll 3
112
Social Law and Natural-5cience
Social Law and Natural Science
number of natural events like death, illness, etc., perfectly constant, hut
The modern kinetic theorv arose soon af h. . .
also the number of so-called voluntary actions--marriages at a given ergy conservation around I8--o ·h. 't . tler t e establt~hment of en-
age, crimes, and suicides. It occurs no differently among molecules. s was some sort ()[ d . . . . ) ' w en 1 at ast became clear that heat
· vnanuc encrgv ai d ft . · l . .
Doubtless it would be too brave to argue that statistical gas theory only tablished through;ut western Eu'r~. c] ,;h:;l:~cJ~ statJsbcs .\~as well cs-
became possible after social statistics had accustomed scientific thinkers mvokcd in support of the kinet' pd II· eory of probability \vas first
to the possibility of stable laws of mass phenomena with no dependence ;d · . . Ic 1110 e Jy August Kronig in 1 8-~ H
use It to JUStify a calculation of th 1f b. ' ) /· e
on predictability of individual events. Still, the actual history of the ki- ities and gas pressure lY'scd o c rea Ion . etween. molecular veloc-
netic gas theory is fully consistent with such a claim. To be sure, the ' "· 11 a SlmplJfied mod j . h' h J
were supposed to move I·n p ll 1j· b .k e m w Jc mo ccules
idea that heat is motion has appeared intermittently throughout the pe- ara e mes ac and[; th I
of a perfectly smooth container K .. : .. . . _or )etw.een the walls
riod of modern science, while the drive to explain the properties of gases justification of this model .. . romg saJd nothmg of statistics, but his
in terms of the movements of distinct molecules or atoms can, in a gen- statistical authors had n , d~as consonalnt Witt~ the kind of rhetoric that
eral way, be traced back to the ancient atomists:"The nature of this mo- 1a c commonp ace S tf ll f
tainer are uneven in comparison with the . mce Je wa s o. the con-
tion was never settled, however, Heat might be an expansive, upward- plained, "the path of each magmt~de of an atom, he ex-
tending motion, as Francis Bacon suggested, or the vibrations of an in- calculation. In accordance ~i~l~t~~ ~:~st be so me~ular that it defies
termolecular ether, as was generally believed after the caloric theory was can suppose in lace of l . . ws of probability, however, one
discredited around 1830 until the early 185os, or the rotation of gas mol- Rudolf cJ'· p I t 1JS absolute meguiarity, complete regularitv. "7
dllSIUs w 10 published h · ·d ! '
ecules arranged in a lattice. The kinetic theory, which derived the prop- after the appearan~e of K .. . ,, IS I cas on t 1e kinetic theory soon
erties of gases from the motions and collisions of free molecules, has about c ll. . f romg s paper, made virtually the same point
o JSJons o gas molecules with th . ll f .,
been restrospeetively honored with a rich tradition of precursors, but though the angle and velocit, r'· .d , e w~ so their container. Even
none before the mid-nineteenth century had any real understanding of of reflection, he observed, } o. lflCJ ence will not generally equal those
the dynamics of their model.
The most impressive early kinetic account of thermodynamic phe- yet, according to the laws ofprobab'l't. .
1 1 Y' \vc may assume that the
nomena in gases was that in Daniel Bernoulli's Hydrodynamica (1738),
but even he did not indicate with any clarity what kind of motion his gas
C
~rge absent1any rn6o~ecudles whose an. gles fall withir; a certain inter"~e
· · ween o an 6 o tl ' '
. .d h ' I ' as 1ere are molecules whose angles of
molecules were supposed to undergo. John Herapath, the British sci- :~~~ ~F~e ave] thelsame limits, and that, on the whole, the veloc-
entific writer and railway engineer whose early nineteenth-century writ- 1C mo ecu es are not changed by th, 'd N d'f.
ings on gas theory seem to embrace certain features of what became the will be produced in th , fi I , 1 h c SI e. o . I terence
,. h ' e na resu t, t erefore, If we assume th~t f
kinetic model, was thoroughly confused and inconsistent. So nearly as CdC l110Iecule the angle and velocitv f. fl , .· ,. a or
of incidence. H • 0 Ie cxJon are equal to those
one can tell, he took heat to be the vibration of molecules in an other-
wise stationary lattice. There is no reason to suppose that any author be-
fore Waterston and Joule began writing in the 184os believed the rig- tes~~~~~;~m~:elling than thes: remarks on the stability of averages as
orous laws of gas dynamics could be derived from a model of random
molecular motions. 6 .
is the story ~f t~: :~:~~~:~~:~~·~~J~;~~~~a! 1 1~:i~~:~~cf~~ ~~: k:::eti~ g~htheory
5
m fact an achievement of tl fi t .. - , p , Sics. at was
s Boltzrmnn, "Ocr zwcitc Hauptsat7. dcr mechanischen W;innethcoric," m PS. p. 3+; also tiona] importance for ph 1~ brs tmal· gmtude, leading to work of excep-
Lectures on Gas Theory, Stephen Brush, trans. (18q6-18q8; Berkelcv, 1964), p. 4-14· ysJcs u a so mvolvmg a · t
6 See Daniel Bernoulli, Hydrodynamics, trans. T. Carmody and It Kobus (New York,
bution to the statistical approach M . ll . - hn llnpor ant contri-
. axv.e ' Its aut or, denved his un-
1968), pp. 226-229; also Eric ivlendoza, "The Kinetic Theory of Matter, J845-18)5," Archives
- Quo.tcd in lvo Schneider "R d l I c·l . , . .
intemationales d'histoire des sciences, 32 ( !<)82), 184-220; Mendoza, "A Critical Exammation
· c 1cm rc 1 'CI/st 1JCoretischcr Method,
,
wahrs'l,']']k· 110 P 1
. . 1 d., Ph,,aus1us
:. , , B't·
u Jag zur I:"unflihrung
ofHerapath's Dynamical Theorv of Gases," BJHS, 8 (HJ75). 155-165; Stephen G. Brush, The (1974-7)), 243· en l!. Ic . vstk der Case nach 1856." Archive 14
Kind of Motion We Call Heat (2 vols., Amsterdam, 1977); Robert Fox, The Caloric Theory of
, 8 Rudolf Clausius, "The Nature of the Moti h' h ,
Gases from Lavoisier to Regrwult (Oxford, 1971), passim.
Stephen Brush, ed., Kinetic Theory (3 vob.', Ne~~~)~r~c ~c ,call]Hcat" reprinted in
, 19 5;, \(). l, 1!1-134. p. 126.
114
(
Social LGwcmd Naturat Science
Social Law and Natural Science
. . . rve indirectly from Quetdct, and. his use of it
derstandmg tl.lc e. rro.. r cu } . t . II.t\' of soct.al science Ill the maintained that if gas pressure really arose from the rapid, uncon-
r
provides yet a not 1er tes Imoln: k~
l t t 1e ceu ra · · ·
He also. went bevond Quetelet, strained motion of free molecules, gases ought to interdiffuse alrnost in-
development of stabstlcal hm I mg. d. st~ibution for;nula in his the- stantaneously. Experience showed clearly that they did not. In defense
however, for the importance of.tl1e e;ror ,ltfv' g or calcHlating a mean of his theory, Clausius proposed that the molecules would collide fre-
ory was not exhausted by its ub tty wr )U> I , m . .
quently with one another, and that in this way diffusion would be greatly
. value. h· . f ll . appreciated the need to retarded. He went on to give a calculation of the mean free path of gas
Rudolf Clausius seems never to ave lu _'J . t dies of a statistical molecules, based on the simplifying assumptions of a single molecule
f d . r s from mean va ues m s u
take account o evta IOn h Cl ·us first used assumptions moving through a field of stationary particles. He then asserted without
S h .d hass own _, ausJ .
char<lcter. As Ivo c nei er . d . '!· . II. est work on the kinetic proof that multiplication by a factor of 3/4 would generalize the expres-
f th . t prescnte m liS ear
of randomness o e sor l . l blcm Meteorology was a sion to cover the case in which all particles moved with equal velocities.
. k. . 1 e~eoro og1ca. pro · ·
theory whrle wor mg on an L h . · of probability methods The entire exercise was purely theoretical, for the actual distance could
· . -t e for t e transmiSSIOn
field of frequent Impot anc l . . . t s+J.fied bv an I8so paper of not be computed so long as the volume and number density of individ-
~hich
l · and a so one as IS e ' .
to the natura sctences, ll t in consideration of variation as ual molecules was entirely unknown. The paper was important for sev-
the Dutch physiCist Buys-Ba o, . l, ·mportant." Clausius, how- eral reasons, however, not least among which is that it introduced the
, ld be seen as genume Y1
well as averages cou . . of li ht through the atmosphere, a~d ·
Ill- first probability distribution into physics. This was his formula for the
ever, worked ?~ the passage s11ow that general principles ot light distribution of path lengths, one which later became known as the Pois-
vokcd probability argume~ts to . } I: the problem of the re- son distribution. Clausius, characteristically, did no more with this
. . ld b ttamed wit 1out so vmg
transmiSSIOn cou ea t · n 36 he announced
. f 1. I b . t s 10 "orne years 1a er, 1 1 2 • curve than to calculate its mean. Once agJin, the ''laws of probability"
Aecbon o Ig .1t y a om . "T· we wish to arrive at really reliable con- served only to justify the supposition that the totality of motions in a con-
that, for the kmehc theory, ti]If II' d b. cts we must not be afraid
. · tl'sandot1era Ie su Je ., . , fined system could be adequately represented by a single mean value. ' 2
cluswns concernmg lt .d t. f the irregular mobons.
eso~·
h bl e cons1 era wn o "As soon as I became acquainted with the investigations of Clausius,"
of the somew at trou .d the irregularities into two classes. One Maxwell later wrote, "I endeavoured to ascertain" the law of equilib-
He proceeded, however, to lVI e in the individual im-
class, the "accidental inequallbes accdompalndyth~s be neglected. Only
rium velocity distribution.'' Within a few months of the publication of
" ld . 1 verage out an cou this paper in England, Maxwell had completed most of the work for his
the "normal vanatiOnsY,~ associa
pacts, wou ce_rta.m : t d' .th macroscopic temperature gra-
e WI landmark paper introducing that distribution, which he read to the Brit-
ish Association late in 1859. On May 30 of that year he had written a
clients needed to be consideredh." d. . . 1tl1c'Of" of oases as he then
b k. on t e vnam1ca · 1 " ' letter to his friend George Gabriel Stokes to inquire whether Stokes
Maxwell egan wor mg
· 1 · 8 9 after rea d'mg. an ab n.d gment of Clausius's second . h knew of any experiments that could be compared with some predictions
called tt, ear yI111 l 5 '·hich h ad recen tl Y ap peared in translation m t ,e
made from the model. ln particular, Maxwell had derived from the ki-
paper on gas t 1eory' Vv . . th responding in a significant
. lM · Clausius was ere ' netic theory the principle that gaseous friction should be independent of
Philosophzca agazzne. . . · . · ,d by Buys-Ballot, who
and creative way, to an obJection recent1y raise density. This proposition violated his intuitiort·and he \Vas accordingly
doubtful that the model could be upheld. Even the published paper,
,, Buvs-Ballot, "On the Great lmpor.~a~~~ ~ c~> J .., ( 1859), 42-49· Sec also 0. B. Shey-
f D (ations from the Mean State of the Atmos-
which appeared early in 186o, intimated that its author expected the
phcrc for the Science of Meteorology,! ~~e~h'o:fn' l1e~~o~ologv," Archive, 31 {1984-~5), 53- theory eventually to be disproved
nin, "On the History of the Stahsttca l . . , .

95· . . . the late >840S Sec lvo Schneider, "ClausiUs e:st:


w Clausius pubhshcd t!m work dunng . R l men dcr atmosphanschcn Ltcht
Evidently Maxwell took up work on the kinetic theory not because he
Anwcn d ung der Wahrscheinlichkc>tsrcdmnng
· . 8 tm a1
" Clausius, "On the Mean Lengths of the Paths Described by the Separate Molecules of
strcuung," Archive, 14(1974), 143-l5f}f· tb'C ses "PhiU'v1ag[ 3], 23 (1862),417-435, 5 12- Gaseous Bodies" (1858), in Brush. Kinetic Theory (n. 8), vol. 1, pp. '35-147-
,' Clausius, "On the Conduct1on o ea ' .,a .
534, pp. 419, 422. voL,' 2,Maxwell,
p. 427. "On the Dvnamical
. Evidence of the Molecular Constitution of Bodies," m SP,

117
( (

Social L:ciwand Natural-science ----- Social Law and Natural Science

nl~ction ~he t~; ~::tas~~~~::L~on­


thought it was true, but because he had taken an interest in working out ecules. Still, Herschel's essay points unambi uo I .
its mathematical implications. He told Stokes that although he had been beftween ideas of social statistics a;d
p tcahon
k. t' o stabsbcal
h ideas to p h ),sics
· d urmg
· the nineteenth century the
ap-
unable to determine whether the theory was valid, yet "as I found myself
able and willing to deduce the laws of motion of systems of particles act- me 1c gas t eory. •
ing on each other by impact, I have done so as an exercise in mechan- Hersch~! had not been actively involved in social statistics or insur-

ics." He also remarked: "I have been rather diffuse on gases but I have :~~:.~ but mother respe:ts his res~~rch aims were highly similar to Que-
taken to the subject for mathematical work lately and I am getting fond t . j H_e was ~he le_ad.mg practitioner of the quantitative natural-his-
of it and require to be snubbed a little by experiments. " 14 In short, Max- onca sciences
th t . m Bntam and ' in the eyes of h.IS countrymen probably
well took up the kinetic theory as an exercise in rational mechanics. He e mos en:ment man of science of his time. What Hersch~! wrote of
did not present the mathematics to Stokes, but he did offer a new for- Quetelet might equally have been applied to himself:
mula for mean free path that dearly was derived from the supposition
No_~ne has e~ert~d himself to better effect in the collection and sci-
that molecular velocities were distributed according to the error law.
enh c combm_ahon of physical data in those departments which
Moreover, the general mechanical model that made up the first half of
:~~end! for ~hen progress on the accumulation of such data in vast
his published paper consisted almost entirely of propositions that were
vo ummous masses, spreading over many succeeding years
an~ ~~t~red fr?m extensive geographical districts,-such asTer~
dependent on this distribution law. Evidently Maxwell recognized the
applicability of the law of errors to molecular velocities at the very be-
res :Ja. agnehsm, Meteorology, the influence of climates on the
ginning of his work on the kinetic theory. Probably the opportunity for
mathematical work presented by the error law had inspired him to begin ;~r~~~d:~:l :rhen~menfa of anima_! a~d vegetable life, and statistics
. anc es o that mulhfanous science political moral
work on this model in the first place. an d SOCial. •6 ' • ,
How, then, did Maxwell become convinced that the molecular ve-
locity distribution would conform to the astronomer's error law? It is ~h~se were the problems that concerned Herschel, and his essay made
now clear that the inspiration came from the ideas of Quetelet, and
more particularly from an essay review ofQuetelet's 1846 work on prob-
~e~d~~gt~:rt ~:yeomndetthhods o~ sltadtistics_ enjoyed <1,.4.ange of applicability ex-
e socta omam.
ability and its applications written by John Herschel. More than two Herschel
. . followed Lap! . ace an d D e M organ mmterpretin
. . robabili ,
decades ago Charles Gillispie pointed out the similitude of the ap- subJectively as the !ogre ofbeliefabout. d' 'd I g_p
m IVI ua events wh 1ch · t)
~•'"'
I:
proaches to probability to be found in this review and in Maxwell's pa- penectly understood. The idea of ch are tm-
per, and Stephen Brush subsequently recognized that the formal deri- ings of probability o I " h an~e, he wrote, enters the reason-
n y as t e expressiOn of our ignorance of a ents
vation of the error law given by Maxwell was in every important respect
e~ess vtew to Its exclusion
arrangements and motives, but with the . . g '
identical to the one introduced by Herschel in this essay.'' Since Max- from their result "H d
I i h' b' s. e a opted, however, a stnctly empirical line in ap-
well did not discuss or even acknowledge Herschel and may not have
~ y ~~ IS su J~~t to real phenomena. Commending John Stuart Mill
remembered where he had encountered this derivation, it is difficult to
or IS recogm~ton that true science inv~d successful rediction
know just how he decided that its argument could be extended to mol-
rath;~~han th~ I~t~itive comprehension of things in themsel~es,
hear-
,. See Joseph Larmor, ed., Memoir and Scientific Correspondence of the Late Sir George ;~: " a~ pr~ a.bt!;ty_ was the clearest example of a science upon which
Gabriel Stokes (z vols., Cambridge, 1907), vol. z, p. 10. " me ap ystca tdea of Causation" had no beart'ng The t
''C. C. Gillispie, "Intellectual Factors in the Background of Analysis by Probabilities," in
A. C. Crombie, ed., Scientific Change (New York, 1963), pp. 431-453; Brush, Heat (n. 6), caus e, " a ftera ll ',Simply
" · · or lesserm
expresses the occasion for the more fre-
esp. pp. 184-187. See'also C.W.F. Everitt, "Maxwell," in DSB, vol. 9, pp. 198-230; Eliza-
beth Wolfe Garber, "Aspects of the Introduction of Probability into Physics," Centaurus, 17
quent occurrence of a result; one cannot say that events conform them-
(197Z), 1'1-39; Theodore M. Porter, "A Statistical Survey of Gases: ·Maxwell's Social Physics," [John Herschel], "Quetelet on Probabilities ' "Edinburoh
•6 o R ev1ew,
. 92 (18 50), 1 _ 57 , p. 14 .
HSPS, n (1981), 77-116.
ll8 119
Social Law andNatural Science
Social Law and Natural Science
to explain it to his r~aders. Like Quetelet, he introduced the idea in the
selves to the laws of probability, for "the laws of probability, as acknowl-
~~~text of,a d,I,stmctl,on bet';een two kinds of mean values, arguing that
edged by us, are framed in hypothetical accordance with events. "' 7
,rue mean can exist only where deviations are sub 1·ect to the 1 f
Herschel's adherence to this positivist line did not diminish his ex- errors. Que te1et 's d.IstmctiOn
. . between true means and arithmetic meansaw o
pectations of the results that could be attained through empirical re-
became for h1m one behveen means and averages, but his main oint
search. He opened his essay with the profound observation: "experience
was the same: p
has been declared, with equal truth and poetry, to adopt occasionally the
tone, and attain to something like the certainty, of prophecy."' 8 Her- An averag~ may exist of the most different objects, as of the heights
schel identified statistics as a compelling instance of successful proph- of hous_es 111 a town, or the sizes of books in a library. It may be
ecy, and praised the flourishing insurance industry for its success in di- ?onvement, to convey a general notion of the things averaged; but
minishing the impact of chance on human affairs. Casting Bernoulli's I~volves no ~onception of a natural and recognisable central mag-
law oflarge numbers as the central truth of probability theory, he sug- ~Itude, all differences from which ought to be regarded as devia-
gested that probability in turn provided 3 model for empirical scientific tions _from a standard. The notion of a mean, on the other hand,
reasoning. Probability, he wrote, may be contemplated "as a practical docs Imply_such a conception, standing distinguished from an av-
auxiliary of the inductive philosophy . . . . Its use as such depends on erage by this very _feature, viz. the regular march of the groups, in-
that mutual destruction of accidental deviations from the regular results c~easmg to a maximum and thence again diminishing. An average
of permanent causes which always takes place when very numerous in- gives us no assurance that the future will be like the past. A mean
stances are brought into comparison. "•9 The destruction of accidental m~y be r~ckoned on with the most implicit confidence. All the
irregularities was most conspicuous in the field of social and moral sta- p~dosophi_cal_va_lue _of statistical results depends on a due appreci-
tistics, whose importance for the improvement of administration Her- ation of this dishnchon, and acceptance of its consequences. 21
schel did not neglect to mention. He was deeply impressed by statistical
Mean values, then, only t~ok_ on genuine scientific worth when they
regularities, which indicated the possibility of a science of man govern-
represent~d. a type, the dev1atwns from which were distributed in the
ing aggregates without constraining individuals. Births and marriages,
charactenshc form assumed by error. Herschel argued that statistical
he pointed out, are "free as air in individual cases," yet they
regul~nty would only prevail for these true means, and implied that this,
seem to be regulated with a precision, where masses are concerned, happrly,_ was the case for most natural and social phenomena.
clearly proving the existence of relations among the acting causes In this way Herschel communicated to Maxwell and other British
so determinate, that there is evidently nothing but the intricacy of readers Quetelet'~ belief in the universality of the error law, which he
their mode of action to prevent their being subjected to exact cal- re~ffirmcd With h1s own considerable authority. Herschel did more than
culation, and tested by appeal to fact. Taken in the mass, and in this, how~ver, fo~ his derivation of the error law-by which we are able
reference both to the physical and moral laws of his existence, the to recogmze h1s mfl~c~ce on Maxwell--was presented as an original
boasted freedom of man disappears; and hardly an action of his life one, and ~as _qUite d1fterent in character from the standard argument
can be named which usages, conventions, and the stern necessities about the hm1t of the bin~mial developed by De :\1oivre and repeated
of his being, do not appear to enjoin on him as inevitable, rather by Quetelct. Herschel pos1ted nothing about the nature of the constit-
than to leave to the free determination of his choice. 20 uent deviations, but instead made assumptions about the properties of
Most important, Herschel was fully convinced by Quetelet's appli- the er~or fu~ct10n a~ a w_hol~. He supposed first that an error of a given
magn~tude m one direction IS precisely as likely as an error of the same
cation of the error law to real variation in nature, and he took some care
magn.Jtude m another direction, and hence that the error function is
'7 Ibid., pp. 4-5, 29-30. sphencally symmetric. Next he hypothesized that the error components
'8 Ibid .. p. 1.
'9 Ibid., p. 29. "Ibid., p. 23.
' 0 Ibid., p. 42.

121
120
---~

Social Law dnaNatural Science Social Law and Natural Science


Still, Maxwell found this exceptionally abstract derivation comincing,
along perpendicular axes are strictly independent of one another-that
is the probability of an error (x,, y,) is equal to the product o~ the .prob-
and reproduced it almost exactly in his first paper on the kinetic theory.
By the time he saw the paper by Clausius that inspired his interest in the
abilities of the separate components. With these two prenllSes It fol-
lowed directly that the error function must have the form Ae-cx', where kinetic model, he had encountered Hershel's essay review of Quetelct at
A and care arbitrary, positive constants. This was the formula for the least twice. Maxwell read it when it first appeared anonymously in the
error component along a single axis which, as H.erschel rccogmzed Edinburgh Review, and forthwith dispatched a letter to Lewis Campbell
later, could easily be compounded to yield the solution for two m more explaining that probability was "the true Logic for this world.'' l'vlaxwell
remarked again on the merits of Herschel's work when he read the vol-
dimensions. 22
Herschel's derivation was described in terms of the err:rs com~utted
.
ume of his collected essays which appeared in 18 57. 2 ' His knowledge
h, b lls are dropped at a target, but there was nothing m 1t restnctmg of probability and his acquaintance with statistical reasoning, was not,
w en a dQ l ' . th t th of course, limited to this essay. He had read Laplace's Theorie analytique
it to this problem. Indeed, Herschel accept~ uete et s v1ew . a . e
error law was to be found everywhere, applymg equally to the distnb~- and had been a student ofJ. D. Forbes at Edinburgh. Forbes had been
. fh an heights and to errors of observation in astronomy, and l11S active during the late 18 30s in anthropological statistics and was a close
t lOnO um d . r .
derivation was intended to offer at once the universality a~ s1m~ !City acquaintance of Quetelet, from whom he solicited a testimonial for the
appropriate for a function found so frequently in sue~ a wide vanety of chair in natural philosophy that he subsequently occupied. ' 6 He pre-
contexts. To be sure, the derivation could not be applied dnectly to sea- pared his critique of the probability argument about double stars while
l ar quan t 1.t.1es l·ke·he . ght for which the assumption of independence of Maxwell was at Edinburgh, and read it at a meeting of the British As-
1 .. 1 , . ~
perpendicular components would have no meamng. For that matter, sociation that Maxwell attended. Maxwell was an admirer of George
the persuasiveness of Herschel's reasoning even w1th respect. to ~h~ nar- Boole's writings, and had read Mill's Logic as well as the general intro-
row domain of error did not go unchallenged. While ~he sub]echvist VY_. duction to Buckle's History of Civilization. It is the Herschel essay,
F. Donkin repeated Herschel's argument, 2 ' frequenbsts hke R. L. Ell1s however, that is imprinted most distinctly on his version of the kinetic
found in it the same lack of good sense that bolstered the m1staken fa.Ith gas theory.
in a posteriori probabilities. A matter like this is not subJect to a pnon Maxwell's introduction of the molecular velocity distribution in his
proof, Ellis maintained, and thoug~ the err?r ~aw may somehov~ rep·~ 18 59 paper was prefaced by three propmitions giving the general laws for
resent our ignorance, there is no baSIS for behevmg that abstract reason collisions between elastic particles of different mass, and show·ing that
ing "can lead us to assumptions which correspond to an,d repr~~ent out- such collisions would quickly obliterate any trace of the initial arrange-
·t ·es " In particular "there ts no shadow of reason for
war d rea l1 1 . ' · · · d ment of the system. Arguing in characteristic statistical fashion, he then
· that the occurrence of a deviation in one direction IS m e- proposed that this very confusion implies the existence of stable regu-
supposmg d. · --
pendent of that of a deviation in another, whether the two necbons dfe larities that must prevail in the mass .
. t · ht ngles or not." Herschel had been misled by a mistaken analogy
a ng a. · · h t
with the composition of forces in mechanics to assume a pnon w a If a great many equal spherical particles were in motion in a per-
could only be established from experience of actual frequencies. ' 4 fectly elastic vessel, collisions would take place among the parti-
cles, and their velocities would be altered at every collision; so that
, Ibid., pp. 19 _20 . See also Herschel, "On the Estimation of Skill in Target-Shooting." in
after a certain time the vis visa will be divided among the particles
Fan;i~r ;eg~~~i~n .~~~~~~fo~;a~~~:c~d~~:~~~~·tl~~ 6l~~ury of Probabilities and on te
Prm-
ci;le of. th~ Method of Least Square's," Quarterly Journal of Pure and Applied ~1at ematlcs,
according to some regular law, the average number of particles

I (~;"'ifli I ~lts6 7;Remarks 011 an alleged proof of the 'Method of Least Squares' containe~m
'' Sec !v1axwell, pp. 138-144, 294.
''' Sec cahier 1028, AQP; also). C. Shairp, P. G. Tait, A. Adams-Reilly, Life and Letters
. late Nu,~ber ~fthe Edinburgh Review," Phil Mag [3], 37 (r8so), 32 1 -3 2 8, PP· 3 2 5-326. ec of/ames David Forbes, F.R.S. (London, r873;, p. 123 and passim.
:lso Ellis, "On the Method of Least Squares," TPSC, 8 (1849), 204-2 19·
123
122
Social Law and Natural Science
Social Law and Natural Science
that this velocity distribution was stable, the distribution of relative ve-
whose velocity lies between certain limits being ascertainable,
locities could be found through a simple integration, which Maxwell
though the velocity of each particle changes at every collision." 7
earned out in his Proposition 5· He also calculated the joint distribution
He then gave, as Proposition 4, a derivation identical to that invented of velocities when particles of different masses were mixed together,
by Herschel. Maxwell premised first that tbe velocity along any one based on the result of his Proposition 6 that the mean energy of particles
coordinate is independent of that among each of the other two perpen- must be the same, mespective of their relative masses. In the same wav
dicular coordinates and, second, that the density of the distribution is Maxwell calculated rates of collision and mean free path distances f~~
spherically symmetric. The independence assumption implied that the systems consisting either of one or two kinds of particles. All this was ac-
total distribution was the product of three independent terms, complished with elegant combinatorial mathematics of a sort that few if
f(x)f(y){(z), and the symmetry assumption implied that this product any previous writers had found reason to use. A few years later, Maxwell
must be a function of the total magnitude of the velocity only-or, employed his distribution formula to investigate transport phenomena
equivalently since it is a scalar, to its square. That is, in Maxwell's no- m gases.
tation, . To ~axwell belongs the credit for first introducing the explicit con-
f(x)f(y)f(z) = <j:J(xl+y2+zz) sideration of probability distributions into physics. Indeed, he was ar-
The solution to this was the exponential error curve g_uably t~e first to_ use distribution formulas in a significant and produc-
F = Ae·-(x' + y' + z'). z8
tive way many SCience. The most impressive results of the kinetic theorv
The error curve was introduced to the kinetic theory, then, on the ba- during the nineteenth century were closely associated with Maxwell;s
sis of a derivation so abstract as to be applicable without modification to distrib~tion. Most notable among these was Ludwig Boltzmann's dem-
telescopic observations, balls dropped at a target, and the distribution of onstration that the second law of thermodynamics could be understood
velocities in a system of rigid, elastic molecules. This is not the kind of m terms of mechanics and probability theory. At the time Maxwell be-
argument that would emerge from a detailed study of the collisions that gan ~ork on _the kinetic theory, the second law was a purely ther~ody­
a molecule in a gaseous system might undergo. It was instead the result namic pnnc1ple, expressed entirely in thermodynamic terms-specifi-
of an intuition. Possibly Maxwell had been prepared to recognize the cally heat, temperature, and entropy. The second law exJ?.ressed the
significance of a distribution formula for the kinetic model by his pre- observed tendency for heat to flow from warm to cold bodie;· or more
precisely, the amount of mechanical work that could be ~erformed
vious work on the stability of Saturn's rings, which in certain respects
had ended unsatisfactorily. 2 9 In any event, it seems probable that the ve-
when this passage of heat was harnessed to an idealized, reversible en-
locity distribution formula was his very first result in the kinetic theory,
gine. Boltzmann felt that this formulation by Camot and Clausius
and the cause of the sudden interest he took in it.
lacked conceptual clarity, which could only be achieved through con-
Maxwell's work shows, rather more compellingly than Quetelet's,
sideration of the mechanics of heat. Interestingly, his reformulation of
that the same mathematics can lead to radically different results when
the second law was closely tied to his search for a satisfying proof that the
applied to a new object. Quetelet had been able to do no more with the
error curve as a law of natural variation than to assert its applicability to velocity d1stnbution of molecules must be governed by l'vlaxwell's for-
heights, penchants for crime and suicide, and so on. When incorpo- mula.
rated into the kinetic theory, however, the error curve could be put to Alre~dy i~ 1866 Maxwell had recognized that there might be prob-
work to yield powerful new conclusions and methods. Assuming always lems ~1th h1s first derivation of the velocity distribution. Since the "as-
sumptiOn that the probability of a molecule having a velocitv resolved
'7 James Clerk Maxwell, "Illustrations of the Dynamical Theory of Gases," in SP, vol. 1,
parallel to x lying within given limits is not in any way affected by the
p, 380 knowledge that the molecule has a given velocity resolved parallel to y"
,s Ibid.
'9 See Stephen Brush, C.W F. Everitt, and Elizabeth Garber, Maxwell on Saturn's Rings now appeared "precarious," he sought to "determine the form of the
(Carnbridge, Mass., 1983l.
125
124
(

Social Law cmd·Natural-Seience Social_ Law and Natural Science....


g~ed from symmetry to show that the derivative of E must always be neg-
function in a different manner.">o His new derivation, which formed
ative unless the error-curve distribution prevails when it will be z"r
the starting point for much of Boltzmann's work, presupposed that the
tremendou~_frequency of collisions would produce and maintain an
That is, he demonstrated not only that Maxwell's distribution was ~t~~
b.lc' but also tha~ any other distribution must converge to it. Even more,
equilibriunVstate, in which both the velocity distribution and the rate of
smce the quantity E was proportional to the negative of his expression
collisions of every specified type would be stable. Since, Maxwell ar-
for entropy, the tendency for entropy to reach a maximum was equiva-
. gued, there is no reason for transformation cycles to run preferentially
lent to the tendency for a system to approach the Maxwell distribution. B
in one direction, the number of collisions involving two molecules with
The relation between probability theory and the second law was
initial velocities v, and v, and final velocities v/ and v; must equal the
brou~ht out more clearly in 1877, when Boltzmann applied a combi-
number of collisions of the opposite sort. Maxwell showed that this
natonal model to the distribution of molecular energies. He had first in-
equality would prevail only if the molecular velocities were distributed
troduced combinatorics to the kinetic theory in 1868, because he was
in accordance with the error law. concerned that for a system with a finite number of molecules and a
Around the time Maxwell published this second paper on the kinetic
fixed total energy, the energy of the last molecule must be completely
theory, the young Austrian Ludwig Boltzmann became familiar with
dete.r~ined by that of the other n-1 molecules. Hence the energies of
Maxwell's striking new approach to thermodynamics. In 1866 he had colhdmg molecules would not be fully independent, and Maxwell's
written a paper on the mechanical meaning of the second law in which 18.67 derivation could not be fully rigorous. 34 The problem of distrib-
he modestly claimed "to give a purely analytical, completely general utmg a fixed amount of energy among a finite number of molecules was
proof of the second law of thermodynamics, as well as to discover the similar to a problem solved by Laplace to support his calculation of the
theorem in mechanics that corresponds to it";'' but he quickly recog- probability that the sums of the angles made by the planes of the orbits
nized both the power and the necessity of Maxwell's new statistical ap- of the planets with the ecliptic would be as small as it is, had the orbits
proach for studies of thermal phenomena in mechanical terms. Soon he been produced by chance. Posed abstractly, in terms of the urn model
had extended the Maxwell distribution law to complex, polyatomic the query was: "An urn being supposed to contain n + l balls marked
molecules subject to external forces, and then in 1871 he succeeded in ~ith th~ numbers o, 1, 2, 3, ... n, one draws a ball from it a~d places
formulating Clausius's thermodynamic concept of entropy in terms of It ba~k 111 the urn after drawing. One asks the probability that after i
distributions of molecular velocities." drawmgs the sum of the numbers drawn will be equal to s. "35 Boltz-
Armed with the generalized distribution law, Boltzmann returned to m~~n constraine? the system to a fixed total value and sought the prob-
the problem of reducing the second law to mechanics in the remarkable ability that any given molecule would have a certain energy.
paper of 1872 whose preface, on the relations of probability mathemat- To do this, he was obliged to divide the total energy into a finite num-
ics, statistics, and scientific certainty was quoted above. In this paper, ber .of m.tervals of width £, and to suppose every energy interval to be
Boltzmann defined a quantity E (subsequently immortalized as H), a eqmpoSSible, a priori, for a given molecule, up to the total energy avail-
function of the distribution of molecular velocities, which he showed to able. ~he actual pro.bability that an arbitrary molecule has energy n£ is
be minimized when the Maxwell-Boltzmann distribution prevailed. As- then given by a fraction whose denominator is the total number of com-
suming, as Maxwell had, that the collision rate between molecules binations satisfying the energy constraint and whose numerator is the
within any specified energy range varies in proportion to the product of
their frequencies in the entire population of molecules, Boltzmann ar- 33 Boltzmann, "Weitere Studien" (n. 3).
34 Boltzmann, "Studicn iiber das Gleichgewicht der lcbendigen Kraft zwischen bewegten
matenellcn Punkten" (1868), in WA, vol. 1, pp. 8o-81.
3° Maxwell "The Dvnamical Theory of Gases" (1867), in Papers, vol. 2, p. 43· "Laplace, Theorie analy:!que des probabilites (Paris, zd ed., 182o), in Oeuvres, vol. 10, p.
'' Quoted in Marti~ Klein, "The Development of Boltzmann's Statistical Ideas," Acta
257. Sec also Oettmger,. Untersuchungen iiber die Wahrscheinlichkcitsrcchnung," in
Physica Austriaca, Suppl. X (1973), 53-106, P· 57·. . Crclles, Tournai fur dze reme und angewandte Mathematik, 2 6 ( 1 843 ), ( 1 8 ) and 6
'' Edward E. Daub, "Probability and ThermodynamiCS: The Reduction of the Second (1848), esp. vol. 26, 311 . 332 . 34 47' 3
Law," Isis, 6o (1969), 318-330.
126 127
( (
Social Law and Natural Science
----social Lihirand Naturai-science
Galton was the first to use the statistical methods of error analysis in or-
number of these for which the given molecule occupies the prescribed
der _to_ analyze real variation, and his achievements as a biological and
energy level. Already in 1868 Boltzmann was able to derive the Maxwell
stahshcal theorist were due in no small measure to the interest he took
velocity distribution following this procedure. In 1877 he approached in variation for its own sake. As Victor Hilts and Ruth Cowan have
the problem with a broader perspective, and sought not just the proba- s~own, Gal~on seems always to have been more interested in the excep-
bility tha~ny given molecule will have some particular energy, but that
tJ~nal than 111 the average: "Some thorough-going democrats may look
the entire system will be characterized by a given energy distribution. with complacency on a mob of mediocrities, but to most other persons
Once again, Boltzmann assumed that every permissible "complex- they are the reverse of attractive. "3 8 The statistical techniques of astron-
ion"--that is, every assignment of energy levels to the individual mol-
omer~, whic~ Qu"etelet had sought to imitate, seemed to Galton wholly
ecules consistent with a fixed total energy-was equipossible. The prob- unsUitable, smce they were designed to eliminate errors and deviations
lem was to calculate the probabilities of the various possible "state the very objects that he most wished to study and preserve. Galton ad~
distributions" (wa. w,, W 2 , ••• wp), where each wi designates the number of
mired Q~etelet's use of the error law, perhaps precisely because he paid
molecules with energy level i. Using some very sophisticated combi-
no attention to Quetelet's interpretation of it. Galton understood the er-
natorial mathematics along with the familiar Stirling approximation,
ror law as an invaluable means for taking account of natural variation,
Boltzmann found that this probability could be given in terms of a "per-
an~ h~ was accordingly critical of contemporary statists who, because of
mutability measure," which was a special form of the H-function he had
their mfatua~ion with averages, failed to make use of it. He castigated
discussed five years earlier. Since he had already shown that the H-func-
th~m f?r their neglect of the natural diversity that makes society inter-
tion decreases continuously until the system reache~ equilibrium at estmg m a noteworthy passage in Natural Inheritance entitled "The
Maxwell's distribution, he announced the problem solved. "Thus we
Charms of Statistics": "It is difficult to understand why statisticians com-
can see that it can justifiably be said: that distribution which is the most monly limit their inquiries to Averages, and do not revel in more com-
probable of all, corresponds also to the state of thermal equilibrium. For prehensiv~ views. Their souls seem as dull to the charm of variety as that
if an urn is filled with slips in the manner indicated above, then it will of ~he nahve of one of our flat English counties, whose retrospect of
be most probable that the distribution corresponding to thermal equilib- Sw_Itzerland was that, if its mountains could be thrown into its lakes, two
rium will be written on the slips that arc drawn."3 6 The second law of nUisances would be got rid of at once. "39
thermodynamics was equivalent to a tendency for a system of molecular Galton's most important work, as well as his sustain,~d commitment
velocities to approach its most probable state, the Maxwell-Boltzmann to the study of human variety, derived from an ideal that differed in one
distribution, and entropy was proportional to the logarithm of probabil- other important respect from the aims and outlook of nineteenth-cen-
ity.'7 tury s_tatisticians. Galton was the founder of eugenics, the evolutionary
doctrme that ~he ~ondition of mankind can most effectively be improved
through a s~Ienbfically directed process of controlled breeding. Al-
GALTON AND THE REALITY OF VARIATION
though public health reform movements had led men like William Fan
to give attention to the biological determinants of health and even of
Like Maxwell and Boltzmann, Francis Galton was not content simply achievement, 40 the usual aim of statistics as social science had been to
to assert, with Quetelet, the universal applicability of the error law, and
" ' 8 Galton, "Preside~t's Addr.:s.s," JAI, 18 (1889), 401-419, p. 407. See also Victor L. Hilts,
he forthrightly opposed the social physicist's interpretation of this curve Stahshcs and Soc1al SCience, m Ronald N. Giere. RichardS. Westfall, eds .• Foundations
as proof that everything exceptional could best be interpreted as a flaw. of Sczent},fic Method: Th~ l'fmeteenth Century (Bloomington, 1973), pp. 206-233; Ruth S.
~ o;an, Francts Galton s Stahshcal Ideas: The Influence of Eugenics," Isis, 63 ( 197 2), 509-
' 6 Boltzmann. "Uber die Beziehung zwischen dem zweitcn Hauptsatze dcr mcchanischen
2
Warmegleichgewicht und der Wahrschcinlichkcitsrcchnung rcspektive den Satzen tiber das Galton, Natural Inheritance (New York, 1889), p. 62.
39
Warmcgleichgewicht," in WA, vol. 2, p. 193. . See Victor L. Hilts, "William Farr{ t8o7-1883) and the 'Human Unit,' "Victorian Stud-
40

n See Klein, "Boltzmann's Statistical ideas" (n. 31), and Thomas S. Kuhn, Black-Body zes, 14 ( 1970), 143-150.
Theory and the Quantum Discontinuity, 1894-1912 (Oxford, 1978), chap. 2, esp. pp. 46-54.
129
128
-------- ~ ---- -- -·- --
(

· ·Social Law and Natmal Science


Social Law and Natural Science
improve men by providing them with education and with homes and
views underwent any s.ignificant shift during the years preceding his first
workplaces free of physical and moral contamination. Calton rejected
publication on "hereditary genius" in 1865. That his views were un-
the principle that human character was determined by environmen~.
timely is strongly suggested by his failure to \Vin any real support for eu-
In this sense, at least, Galton can accurately be called a conservative.
genics until the 189os. Calton's ideology was too unorthodox, its im-
Although he felt nothing but disdain for those members of the aristoc-
plications too revolutionary, for it to be seen as an ordinarv contribution
racY who had neither been "nobly educated" nor had "any emment
to political discourse. To be sure, his object was not to ;vcrturn social
kin~men. within three degrees," he firmly believed that men are not "of institutions, but to reinvigorate them, by replacing blind faith in a mor-
equal v~e, as social units, equally capable of voting, and the rest."
ibund tradition with reliance on the growing powers of science. But to
Calton glibly assumed that "social worth" among all but the intel~ectual
m~ke talent rather than property the basis of nobility, or to replace the
and scientific classes was distributed roughly in the same proportions as
ex~s~mg House of Lords with an upper legislative chamber of hereditary
income, and he rated professionals above tradesmen, shopkeepers above
abrhty, would be to transform the entire basis of aristocracy and not
artisans, and the latter above unskilled workers, in innate merit. These
merely to change its personnel. Galton was not really a conse;;ative but
gradations, he insisted, reflected differences of biology, not merely of
a deeply committed reformer whose tastes ran to c~nservatism only in
upbringing-at least in liberal England. 'There can be no doubt but certain respects.
that the upper classes of a nation like our own, which are largely and
Galton's biography suggests that his dedication to eugenics derived at
continually recruited by selections from below, are by far the most pro-
least as much from Quaker idealism and the Victorian intellectual ar-
ductive of natural ability. The lower classes arc, in truth, the 'resid-
istocracy's reverence for science as from social or political conserva-
uum'. "4' It was Galton who coined the terminology of the nature-nur-
t~sm. 44 He was descended from a line of Birmingham Quakers whose
ture debate, and he held steadfastly to the position that inherited
sizable fortune was derived from banking and the manufacture of arms,
qualities provided by nature were by much the most important in deter-
among other things. The family was becoming increasingly genteel dur-
mining the achievement of individuals. He sought to support th1s opm-
ing the early nineteenth century; his father had joined the Anglican
ion with a variety of studies, and felt it to be confirmed most decJsJvcly
church, and his t:'o older brothers became landed gentlemen, devoting
bv a comparison of identical and fraternal twins. The uncanny resem-
their lives to huntmg and other forms of refined idleness. Francis Calton
blance between identical twins in matters of behavior as well as appear-
seems to have feared a similar destiny, which he was not content to fol-
ance, as revealed by a litany of anecdotes, seemed to imply that even
low. tie was the product of some remarkable educational efforts on the
complex abilities were largely determined by heredity. 42 . .
part of his elder sisters, leading to the acquisition of certain skills at a very
Galton cannot, however, be construed as simply a conservative, or h1s
tender age, which inspired Lewis Terman to assign him an unparalleled
eugenic creed as a response to any resurgence of conservatism em~rging
intelligence quotient. His mother aspired for him to become, like her
in opposition to the reform movements leading up to the franch1se ex-
father Erasmus Darwin, an eminent physician, and although he found
tension of 1867. To begin, there is little evidence that he ever took an
the medical profession no more appealing than had his older cousin,
active interest in ordinary political affairs, 4 3 and still less that his political
Charles Darwin, he gave evidence already during his youth of great per-
·" Galton, English Men of Science: Their Nature and Nurture (1874; New York, 18;sl, P· sonal ambition. In particular, Calton was driven during adolescence
17. Sec also Galton, "Hereditary Improvement," Fraser's ,\1agazme, n.s. 7 ( 1873i. 116-130,
p. 127; Galton, Hereditary Genius: An Inquiry into Its Laws and Consequences (London, and early adulthood by an overpowering desire "to do good."
1869), pp. 86-87. . . . , dN At Cambridge, Galton suffered two debilitating mental breakdowns,
4' Galton, "The History of Twins as a Cntenon of the Relative Powers of Nature .an ', ur·
ture," JAI, 5 ( 1876), 391-406; see also Stephen). Gould, Thetv1ismeas.ure o{Man (New). ork,.
which prevented him from taking honors, and he seems then to have
198 1 ); Daniel J. Kevles, In the Name of Eugenics: Genetics and the Uses of Human Heredzty begun to feel doubts about his own abilities. What was worse the ideal-
(New York, 198;). d f 'l d. · istic and ambitious young man was unable to find a purpos: worthy of
-n See however Ruth Schwartz Cowan, "Sir Francis Galton and the Stu 'o r ere 1ty m
the Nindteenth cdntury" (Ph.D. Dissertation, Johns Hopkins University, 1969), p. ;6. who
observes that Galton already was voting Conservative as a student at Cambndge. ,, .... See MacKenzie; also Noel Annan, "The lntcllcctual Aristocracy," in J. H. Plumb, cd.,
S.udzes m Socza/ History (London, 1955), pp. 2cp·267.
Social Law and Natural Scierzc:e
Social Law and Natural Science
his energies, and when his father died, leaving him a con~idera~le in-
heritance that freed him from the obligation to study medtcme, hts own I always think of you in the same way as converts from barbarism
aimlessness became an intolerable burden. The period that followed think of the teacher who first relieved them from the intolerable
was one of "wanderings," in the fullest sense. He traveled through the burden of ~heir superstition. I used to be wretched under the weight
Middle East, without any particular goal or destination, and discovered, of the old-fashwned "argument from design" of which I felt though
as he later observed, that experience of new cultures tends to upset the I was unable to prove to myself the \Vorthlessness. Consequently
assumptions of the traveler. In particuiar, he began to question his re- the appearance of your "Origin of the Species'' formed a real crisis
ligious faith. While at Cambridge, he had written ~oetrylamentmg the in my life; your book drove away the constraint of my old supersti-
failure of his materialistic society to live up to Chnsban tdeals, and ex- tion as if it had been a nightmare and was the first to give me Free-
pressing a desire to be a citizen "of some state, modelled after Plato's dom of thought. 48
scheme, and overruled by Christianity."45 By this time he had become
This, clearly, was not the reluctant concession of an admirer of Chris-
persuaded that the era of Christianity was past, that the. Christian faith
tianity to the weight of scientific counterevidence to his faith. Galton's
lacked the power to guide a life. In contrast, he found m Egy~t reason
disenchantment with the Christian religion derived not from doubts
to be "much impressed by the nobler aspects ofMussulman ClvthsatlOn,
concerning the truth of its dogmas, or from scruples about the humane-
especially, I may say, with the manly conformity of their every-day prac-
ness of its ethical code, but from a conviction of its ineffectuality in the
tice to their creed, which contrasts sharply with what we see among most
industrial society that was Victorian England. The argument from de-
Europeans, who profess extreme unworldliness and humiliation on one
sign, this intellectual justification for Christianity, was for Galton the
day of the week, and act in a worldly and masterful manner dunng the
last support of the old faith to crumble. By the t86os Galton had come
remaining six. "46 .
to regard science as a robust and manly replacement for an enfeebled
Science did not cause Galton's crisis of faith, but it did make Itself
Anglicanism, and his eagerness to interpret the theory of evolution as a
available as an alternative to Christianity. In Egypt, Galton met a
refutation of Christianity was born of his belief that this ancient religion
French St. Simonian, Arnaud Bey, who proposed a scientific object for
must be demolished to make way for the future. Darwinism was a de-
his travels, and from 1845 to 186o his life was increasingly shaped by his
cisive defeat of Christianity by science, a conquest of the old order bv
activities in the Royal Geographical Society-Galton was among the the new_,.....-~·/ ~
most intrepid African explorers 4 7-and by his meteorologic~! ~e­
Thus Calton, like many of the intellectuals who came forward to at-
searches. Science, having blessed Galton's wanderlust w1th the d1gmty
tack the old religious order during the decades after 1859, saw in evo-
of purpose, at last gave him an object fully worthy of his talents and ~hil­
lution a rallying cry for the reconstruction of society in the image of sci-
anthropic longings in the form of the biological principle. of evolution.
ence. He stated plainly that his own distinctive contribution to this
Galton's reaction to the Origin of Species was correspondmgly mflated ·
movement, eugenics, was a substitute religion--or rather, as he vvould
He wrote to his cousin a decade later:
have it, the true religion. Calton rhapsodized about the jov and serenity
~; Quoted in Galton, vol. 1, p. 177. Sec also Galton, Men of Science (n. 41 ), p. 218; Ruth comparable to that of religious contemplation, to be de;ived from ob~
Cowan, "Nature and Nurture: The lntcrplav of B1ology a1;d Pohbcs m the Work of FranciS
Galton "Studies in History of Biology, 1 ( 1974), 133-208; Kcvks, Eugemcs (n. 42); Ravmond
serving the solidarity of the fabric of nature, and from furthering the
E. Fan~hcr, "Biographical Origins of Francis Galton's Psychologv," !sis, 74! 1983), 22_7.-2 33- processes of race improvement revealed by the theory of evolution to be
~6 Galton, Inquiries into Human Faculty and 1ts Development \ 1g83; London .• d cd ..
190 7), p. '54- Sec also Galton. tv1emories of My Life (London, ;d cd .. 1909), p. 88: wh~:c
the goal of nature itself. He also called for "the establishment of a sort
Galton noted a sharp contrast between his own carousmg and \!oslcm p1etv. Galt?n s nLcl- of scientific priesthood throughout the kingdom, whose high duties
lectual development during his years of travel is insightfully presented by Derek W. F arrest, would have reference to the health and well-being of the nation in its
Francis Galton: The Life and Work of a Victorian Genws (London, '97 4).
4 7 On the relation of Galton's African travels to his later work, sec Ravmond E. Fancher,', broadest sense, and whose emoluments and social position would be
"Francis Galton's African Ethnography and Its Role in the Development of his Psvchology, made commensurate with the importance and variety of their func-
flJHS, 16 (1983), 67-79.
48 See facsimile of this letter in Galton, vol. plate
1, 2.

1 33
(

·Social Lawarrrl Natural Science Social Law and Natmal Science


theory nor Gregor Mendel's combinational model of hereditarv trans-
tion. "4" The need for this priesthood he justified with an argument
nmswn can be regarded as statistical in the sense used here-that is, em-
about the increasing complexity of society that had already endeared it-
ploymg a mode of reasoning based on stable numerical frequencies. ,z
self to several generations of social scientists and would-be technocrats.
The statistical approach was introduced to biology by Galton, who
The average culture of mankind is bec0me so much higher than it was Darwin's cousin. When, as an old man, Galton looked back on his
was, and the branches of knowledge and history so various and ex- career, his tendency to approach problems statistically seemed so thor-
tended, that few are capable even of comprehending the exigencies oughly ingrained in his character that he attributed it to a genetic in-
of our modern civilisation; much less of fulfilling them. We arc liv- hentance-from the Galton side of the family, not the Darwins. Ava-
ing in a sortof intellectual anarchy, for the want of master minds. riety of anecdotes survive telling of how young Galton employed a
T'he general intellectual capacity of our leaders requires to be sextant to ascertain the dimensions of some attractiYe African women
raised and also to be differentiated. We want abler commanders, and of how he invented a series of concealed devices for registering fidg~
states1~en, thinkers, inventors, and artists. The natural qualifica- ets and other unconscious acts as an index of the dullness of meetings.
tions of our race are no greater than they used to be in semi-bar- Perhaps, however, it is unnecessary to assume an unusually vigorous
barous times, though the conditions amid which we are born arc stabstJcal gemmule in the Galton ancestry to account for this aspect of
vastly more complex than of old. The foremost minds of the pres· lm.character. Galton reached maturity at the peak of the Europe::m sta-
cnt day seem to stagger and halt under an intellectual load too tistical movement, and his early numerical studies of meteorological
heavy for their powers. ;o and geographical statistics were, in an extended sense, part of it. It is per-
haps also relevant that his father, a successful businessman and banker
The theorv of evolution bv natural selection provided the context in
which statistical biology was- introduced, and within which it has since
inc~lcated in Galton a~d his brothers the most fastidious accountin~
habits-as the regular financial reports in the son's letters from Cam-
been most fruitfully developed. Darwin's own ideas, though rightly re-
bridge plainly reveal. >3
garded as cmcial for the development of what Ernst Mayr calls "popu-
Although Galton wrote a paper in 1873 for the Roval Statistical So-
lation thinking" in the biological sciences, can only in retrospect be con-
ciety (successor to the Statistical Society of London), l~e was skeptical of
strued as statistical. Perhaps it only begs the question to point out that
the scientific worth of the society's activity. In 1877 he proposed that
Darwin claimed to be unstatistical by disposition, and that there is little
Section F, statistics and economic science, be expelled from the British
evidence of any influence on his work from social statistics beyond the
Association. "Usage has drawn a strong distinction," he wrote, "be-
wcll-lnown inspiration he received from a completely different aspect
tween knowledge in its generality and science, confining the latter in its
ofT. R. Mal thus's theory of population. But even though Darwin's the-
strictest sense to precise measurements and definite laws , which lead bv"
ory involved the production of variation by causes that were poorly
~~--~---~~--------------­
understood in general and wholly ignored in detail, and the differential Bwmetrik~, an cff~rt was made by Pearson to establish that statistics was the natural fulfillment
survival over the long run of certain variants, Darwin never developed of Darwm s evolutiOnary approach, based on the supposition that since the studv of evolution
was propcrlv a statistical affair, Darwin's thought must have been dccplv statistic.al. Galton re-
anything like a quantitative model of evolutionary change.'' Neither his ported to Pearson a negative conclusion, in which both Frank and Lca'nard Darwin had con-
curred, that their father had a "non-statistical" mind. '·! fear vou must take it as a fact that
49 Galton, English Men of Science (n. 41), p. 195. Galton often raised the possibility of a
Da:Wm had no li~:ng fo.r statistics." See letters ~rom Galton to P~arson, 4 Feb. 1901 and 8 Feb.
scientific priesthood, in whose work statistics would have a prominent place. On the rehgwus ',901,111 FGP, fi,c 245118 E;alsoPcarsontoCalton, 3)ulv 1901 and w)uly l')Ol in file 293
character of eugenics, sec Galton, "The Part of Religion in Human Evolution," National Re- [',. On the other hand, George Darwin had written to Galton in a letter some 2 5 vcars earlier
view, 23 (1894), 755-763 that he and Calton shared "a common family weakness fm statistics" (4 Jan. 18; 5: file 19o A).
'0 Galton, "Hereditary 'Talent and Character,'' J\lacmillan's Magazine, 12 (1865), 157-
'' It rs clear, however, that chance was an important issue in Darwin's thcorv of evolution
166, 31 8- 3 27, p. 166. from the beginnmg; see Silvan S. Schweber, "The Origin of the Origin Revisit~d," Journal o{
s• On Da;win and probabilistic thinking, see M.J.S. Hodge, "Law, Cause. Chance, :~d- the Hzstoryof Bwlogy, 10 (1977), 229-316: Schwcber, "Aspects of Probabilistic Thought in
aptation and Species in Darwinian Theory in the 183os, with a Postscrip\ on the 1930s, .'!.' Great Bntarn Dunng the 19th Century: Darwm and 1vlaxwcll," in Heidelberger, Probabilitv
Michael Heidelberger et al., cds., Probability since 18oo: lnterdrscrplmary Stud res o{Screntzfzc (n. 51), pp. 41·96. · "
Development, Report Wissenschaftsforschung, 25 (Biclcfdd, HJ83), z8;-3~9· In H)OI. when "Sec Cowan, "Nature and Nurture" (n. 45); also Galton, vol. 1, p. 107.
Karl Pearson and W.F.R. \Veldon were setting up, in collaboration wrth Ce1lton, tire JOUrnal

134
Social Lawand Natum!Science
Social Law and Natural Science
such exact processes of reasoning to their results, that all minds are
obiiged to accept the latter as true. "54 Calton believed in the need for Calton also used the methods of the social statisticians to debunk
those conventional religious belief~ whose unscientific character
higher mathematics in statistics, and saw little evidence that members
seemed to him .to render them untenable in the modern age. The most
of the statistical societies were interested in the mathematical aspects of
notonous contnbution to this crusade was his remarkable "Statistical In-
their discipline. For that reason, he thought the old Section F could
quiries into. the Efficacy of Prayer," a paper in which he sought to re-
most appropriately be transferred to the Social Science Congre~s.
solve set~nt.If1cally whether piety or prayer brought any objective advan-
. Nevertheless, Calton was not untouched by the preconceptions and
tages to tts mtende? beneficiaries. Its effect was to crush mystical piety
beliefs of the social fact gatherers. It is significant that his program for
under a heap of miscellaneous statistical facts. Sovereigns, whose lives
the cuantitative study of heredity was originally founded on a "simile"
whic h explicated the process of hereditary transmission in terms of :o~ial
1 were the object of regular prayerful appeal of whole nations, proved to
live no longer than other members of the prosperous classes. Galton
processes. ss Galton took as his starting point the principle of ~tahsbcal
supposed that clergymen could be expected to plead on occasion for
regularity developed by Quetelet and the British statists. In hts :vork tt
appeared in a slightly more abstract form as the "axiom of stabshcs that
t~~ir own health, yet found that their life span was similar to that of phy-
SICians or attorneys. The final blow was struck by the practice of insur-
large samples taken out of the same populatio~ at r~nd?m are stahst~­
cally similar. "s6 Galton did not characterize th1s s1mt!anty mathen:ab-
a~ce companies, which, it seems, did not distinguish between the lives
of the pious and the worldly. They even offered slave vessels the same
cally, or derive it from probability theory. Instead, he explamed tt m
advantageous rates as missionary ships. "If prayerful habits had influ-
terms of an easily understood analogy. He proposed, for example, t~at
ence. on temporal success, it is very probable, as we must again repeat,
the selection of genetic material to be passed on from parents to offspnng
was similar to a process of "indiscriminate conscription: thus, if a large
t~at msurance offices, of at least some descriptions, would long ago have
army be drawn from the provinces of a country by a gen.eral conscr~p­
d1s~overed ~nd made allowance for it. It would be most unwise, from a
tion, its constitution, according to the laws of chance, wtll reflect wtth
bus~ness pomt of view, to allow the devout, supposing their greater lon-
gevity even probable, to obtain annuities at the same low rates as the pro-
surprising precision, the qualities of the population whence it was t~k~n; fane."sY'
each village will be found to furnish a contingent, and the compos1bon
The statistical studies that Calton performed to promote and facilitate
of the army will be sensibly the same as if it had been due to a system of
eugenic intrusion into human heredity were fundamentally dependent
immediate representation from the several villages. " 57
on Quetelet's error law. There are some respects in which his dedication
Galton introduced this important simile between the phenomena of
to this and other statistical Jaws revealed concerns similar to that of the
biological inheritance and the statistical behavior of a free society in
1 86 9 , only a year or two before Maxwell and Boltzmann proposed .the
~ocial st~tisticians. Calton was a man of order, and he displayed a special
mterest 111 subj~cting the irrational and inexplicable to scientific prin-
~
~Iple. He was hnnselfburdened with an unstable psyche, reflected in his
same analogy for physics. The nearly perfect coincidence is perhaps JUSt
luck, but their independence in proposing this analogy seems clear, and
Idea that "the tableland of sanity upon which most of us dwell, is small
is at least highly suggestive. Seemingly without exception, those who ap-
m area, with unfenced precipices on every side, over any one of which
plied statistical thinking to any of the sciences dur~ng t~e second ~1alf of
we ma~ fall." He once performed some introspective experiments on
the nineteenth century thought in terms of analog1es w1th the soCial sci-
para.n~1a and fetishism whose success was so great that he had difficulty
ence of statistics.
regammg a normal perspective. 59 He was fascinated by the phenomena
<4 Francis Galton, "Considerations Adverse to the Maintenance of Section F .. · , "JRSS. of mental imagery, to which he believed a certain portion of the popu-
40(1877) 468-473, p. 471.
ss See the discussion of Galton in chapter 9· 58Calton. "Statistical Inquiries into the Efficacy of Prayer," Fortnightlv Review n.s. 12
s6 Galton, "Family Likeness in Stature," PRSL, 40 (1886), 42-73, P· 43· (187~),125-,135, p. 134. Galton's statistical argument on prayer was criticiz~d bv Ccdrge John
<" Calton, "On Blood-relationship," PRSL, 20 (1871-72), 394-402, pp. 397-398. Ro:nancs, Chnstran Prayer and General Laws (London, 1874 ), pp. 253 _2 68. -
5
Galton, Memorzes (n. 46), pp. 38, 276-277.
q6
137
Social Law and Netural Science
Social Ltnvand Natural Science
errors. Spottiswoode was genuinely interested in his Asian mountains
Jation was congenitally susceptible. He reported that whenever the cli-
but be regarded his brief memoir primarily as a methodological contri:
mate of opinion becomes favorable to supernaturalism, these .!1ormal
butwn to natural science. He saw Quetelct's use of the error law as the
mental processes are misinterpreted, and seers of vision appear m every
basis for a general test which could be applied to identify those phenom-
communitv. Galton took great pains to establish a quantitative study of
ena for which a common cause could be sought with reasonable expec-
phenomer;a like these, and thereby to show how. a multitude of subcon-
tatiOn of success. Spottiswoode was, consequently, pleased to explain to
scious mental processes "admit of being dragged mto ltght, recorded and
Galton "the far-reaching application of that extraordinarily b~autiful
. treated c:tatistically, and how the obscurity that attends the initial steps
law." 6 >~- -
of our thoughts can thus be pierced and dissipated. "6c ./
Galton was deeply impressed by Spottiswoode's exposition of the error
Galton was rarely satisfied with counting alone, and he bronght even
law, of which he first learned in 186o or 1861 at about the time Quctelct
to th~se psychological studies the aims of_the hereditary ;uantifier,. as
Journeyed to London for the International Statistical Congress there. He
Galton would have it, as well as the quantther ofhered1ty. fhough ITlln-
met Quetelet at that time, and consulted his book of 1846 on the meth-
imally proficient at algebraic operations, Galton h~d received trai2ing
ods of statistics before he began using the error curve, in 1869. Galton
in mathematics at Cambridge and had acqmrcd, uke so many Cam-
remai~ed one of the most loyal partisans of the error law throughout his
bridge graduates, a sound geometrical intuition as well as an ability to
hfe. Even though he was among the first to propose an alternative dis-
set up problems correctly even when he was unable to solve them. Smce,
tribution function, the so-called log-normal, in connection with a cer-
the statistical studv of heredity required techniques for takmg account or
tain class of data, that formula involved no rejection of the conventional
variation, Galton. was pleased to have available the astronomical error
error law. Galton simply believed that sensations governed by Fechner's
law. He insisted, however, that in biological studies it was absurd to use
law-according to which the perceived force of a given alteration in a
such expressions as "probable error," since variation in this domain was
stimulus is inversely proportional to the total level of that stimulus-
genuine and not the mere product of error. 6 ' For eugenic purposes, 1t
could be expected to reflect the same structure of deviations in the error
;vas nat:nal to look upon the exceptional as more interesting and im-
distribution function. Accordingly, a geometric rather than an arith-
portant than the average. metic mean should be used, and the treatment of variation similarly
Galton mav have encountered the error function in his meteorolog-
modified. 64
ical researche-s at the Kew Observatory during the 185os, but he became
Despite his enthusiasm for the exceptional, Galton often used the er-
interested in learning about its use only after he became aware of its ap-
ror curve precisely in the fashion developed by Quetelet and Spottis-
plicability as a law of distribution 62 In 1861 his friend and fellow geog-
woode, as a definition of type. He sought, for example, to ascertain
rapher William Spottiswoode published a paper that made use of the e.r-
whether the basic forms of fingerprints represented genuine differences
ror law as a definition type. His object was to ascertain whether a certam
of type by determining if the variation within each were governed bv
group of Asian mountain ranges had been produced by a single cause-
subject, as ever, to numerous incidental sources of vanatwn. lnvokmg
Quetelet's law_ 6 ' He also based his use of the remarkable technique ;f
composite photography on the principle that deviations from any given
Ouetelet on the principles of type and of natural variation, he proposed
type must be distributed in the same wav as errors of observation. So
that the discrepancies in the orientation of the principal axes of these
long as deviations from type were accidental, flaws and eccentricities
ranges from their mean ought to be governed by the astronomical law of
6' Galton, Memories (n. 46), p. 304. Sec William Spottiswoodc, ''On ~lountain
oo Galton, Human Faculty (n. 4 6), p. 145; alsop. 1 21 and "The Visions of Sane Persons,"
Rang?s~ An Apphcati?n of,~f~~ Ca}culus of Probabih,ties to Physi':al Geography, fournal of the
Fortnightly Review. n s 29ir881), 729·740. Roya. Ceograpillcal Soczet,: 31 ,1861), 149-154- Spothswoode s source for a geometrical ap-
6< Galton. 1\:atural Inheritance (n. 39), p. 58. . . . proach to :-1ountam fonnatwn was the Petersburg academician O.H.vV. Abich.
6, After learning of Spottiswoode's usc of the error law, drsn:s:ed below: he pro;11ptl:: l~okcd
, 1' 4. See~. rancrs Galton and Donald MacAlistcr. "The Geometric i\.kan. in Vital and Social
1
m George Biddcll Airv's On the Algebraical and Numencal I heory of Errors or Obsenatwn Statistics, PRSL, 29 ( 1879), 365-376.
(London, 1 86 1 ). Ther~ is no evidence that he consulted such works to support tus meteoro- 6' Galton, Finger Prints (London, 1892), p. 19
logical researches.
139
Social Law and Natural Science
SocialLaw and Natural Science
be prepared to concede that the average height of a large, isolated pop-
would average out, and the result of superimposing photo~raphic im-
ulation will be constant from year to year. Indeed, "the experience of
a es would be a picture very similar to the type itself. That ts, the tech-
modern statistics" shows that not only the mean height, but also the
n7que could be used to discover Quctelct's average man, .although Gal-
ton did not want to take the average over a whole population, but rather fraction of the population in any given range of heights, must likewise
over the individuals belonging to a distinctive human type. By t~kmg
be very nearly uniform over time. The variation of stature, he went on,
numerous, rapid photographs of perfectly aligned individuals on a SI~gle will be governed by a particular law, the "law of deviation from an av-
plate of film, Galton believed he had uncovered the. pure typ~ or the erage." This law is "perfectly general in its application," as all of statis-
criminal, the consumptive, and so on. Notwithstandmg the b.eauty of tics demonstrates. "Wherever there is a large number of similar events,
the exceptional, Galton's composites were always "better lookmg than each due to the resultant influences of the same variable conditions, two
their components, because the averaged portrait of many persons IS free effects will follow. First, the average value of those events will be con-
from the irregularities that variously blemish the looks of each of stant, and, secondly, the deviations of the sevctal events from the aver~
age, will be governed by this law." 69 The error law was not confined to
them. " 6~/ . . .. .
Interestingly, Galton held to this defimtiOn of type even when 1t pre- the physical domain, for if the error curve expresses the distribution of
sented severe-and, as Pearson believed, illusory-obstacles to t~e at- stature, "then it will be true as regards every other physical feature-as
tainment of eugenic objectives. Thus Galton began in the 188os to Iden- circumference of head, size of brain, weight of grey matter, number of
tify the concept of race with a distribution governed by the error Ia:"' and, brain fibres, &c.; and thence, by a step on which no physiologist will
for that reason, tending always to remain. stable. ·:It
is the e:senba~ no- hesitate, as regards mental capacity. "7o ,..,../ ·
tion of a race that there should be some 1deal typical form uom \'/hich Hereditary Genius was intended t;~establish that exceptional ability
the individuals may deviate in all directions, and towards which thw in a variety of pursuits from music, justice, and statesmanship to wres-
descendants will continue to cluster. " 67 Since the type was defined by a tling and rowing was inherited. Galton argued that whereas a moderate
stable point, and not merely as the mean of an arbitrary distnbutiOn, de- talent might on occasion be held back by deprived circumstances, and
viations were only imperfectly perpetuated. Galton became pers~aded a mediocre one elevated to high office by education or background, ex-
that evolution was based on discontinuous variation, or ''sports, . and ceptional hereditary ability was both necessary and sufficient for the at-
that no significant progress towards a eugenic society could be ach1eved tainment of real eminence in a given field. Hence he decided that his
using natural or continuous variation. Sports, ho.wever, represented case would be adequately established if he could show, which he did,
new types, or new centers of stability.' towards which the o!~spnng of that the most distinguished judges, statesmen, and wrestlers were far
. their own exceptional progeny would 111 turn tend to regress. . more likely to be near relatives of other eminences in these activities
Galton began using the error law in his first book on biOlogical he- than was an ordinary member of the population. His argument required
redity, the ;otorious Hereditary Genius of 1869. Like Que.telet, he no use of mathematical formalism more elaborate than the most
stressed its universality. Since "so much has been published m recent straightforward calculation of numerical probabilities.
years about statistical deductions," Galton began, every reader should Galton, however, saw in the error curve both the possibility of in-
creasing the impressiveness of his study and a technique for quantifying
66 Galton, "Composite Portraits," Nature, 18 (1878), 97-100, P· 98; see also Human Fac-
a range of attributes that previously had resisted exact investigation.
ulty (n. 46), P· 23°· l R · t d Mcd·ocritv i•1
67 Galton, Human Faculty (n. 4 6), p. 10. See a so " egresswn owar s ' I .. : : "Whenever we have grounds for believing the law of frequency of error
Hereditarv Stature," JAl, 15 (1886), 246-263; "Pedigree Moth-breed;~g, ~~a means of 'e~­
fying cert;in important constants in the General Theory of Heredity, m I ransactwns oft e
to apply, we may work backwards, and, from the relative frequency of
occurrence of various magnitudes, derive a knowledge of the true rcla-
Entomological Society of London (1887), PP· 1 9- 2 8. . . . ,
:t
. .
6s Galton believed that the different forms of fingerpnnts were mstances of var~lh lh~cs:.
"The Patterns in Thumb and Finger Marks," Phd Trans •. B, 182 (1892), 1-23 .. )n t c ".' er
0

60 Galton, Hereditary Genius (n. 41), pp. 26-29.


issue, see William B. Provine, The Origins of Theoretzcal Populatwn Genetzcs (Ch1cago, Ibid., pp. 31-32.
1971).
Social Law and Natural Science
SocialLaw and Natural Science -
Statistical units proved indispensable for Galton's suhsequent quantita-
tivc values of those magnitudes, expressed in units of probable error," 7 '
tive study of inheritance and his method of intercomparison.
he wrote in 1875. In Hereditary Genius, Galton assumed that 250 of
Calton's use of statistical units in connection with the analvsis of ob-
every million men are properly eminent. This group corresponded to a
jects that could be ordered by rank but not directly measure~] became
tiny area at the right end of the error curve. Calton then divided the ax1s
more sophisticated during the early 187os. ln 1874 he published his
between this point and the mean into five equal distances and, usmg this
study of English IV! en of Science, an effort to identify those traits which
small interval as his unit, marked one more class above minimal emi-
nence and seven below mediocrity. These fourteen intervals represented contribute to great achievement in various scientific fields. His material
the grades of intelligence or statesmanship, as defined by mathematics. consisted, for the most part, not of direct measurements, but of re-
According to this scheme, the classes immediately above and below sponses to a survey that he distributed among his scientific associates.
the mean--which Calton labeled A and a, respectiveh---cach con- Galton had intended to arrange the replies to his survey of these men in
tained about 2 57,000 men per million. The next higher grade of intel- order according to degree of religious piety, energy, business sense, and
ligence, B, and the next lower grade of stupidity, b, numbered 161,000 so on, much as he later did with reports of ability to form mental im-
each. The density declined at an accelerated rate so that the first emi- ages, 73 and from this ordering by rank to place each man in the appro-
nent category, F .(and the corresponding state of abject idiocy, {), con- priate class of ability as defined by relative rarity and the error law. He
tained only 2 3 3 men per million. Beyond the seventh division at each hoped to supplement this information with the judgments of biogra-
end of the curve, and extending to infinity, were X and x, each with a phers of scientists and other great men, to whom he appealed to rank
mere one man per million//' their subjects on his statistical scale, "in respect to every qualitv that is
discussed. "7 4 From this information, Calton planned to find relations
It will, I trust, be clearly understood that the numbers of men in
between each of these qualities and scientific achievement bv field. It is
the several classes in my table depend on no uncertain hypothesis.
significant that Calton's first "correlation diagram," a ch~rt plotting
Thev are determined by the assured law of deviations from an av-
erag~. It is an absolute fact that if we pick out of each million the head size against weight, which Calton never published, was prepared
one man who is naturally the ablest, and also the one man who ts in conjunction with this study of English scientists. 7;
the most stupid, and divide the remaining 999,998 men into four- In the event, the data proved too unreliable to be handled by rigorous
teen classes, the average ability in each being separated from that techniques. "It had been my wish to work up the materials I possess with
of its neighbours by equal grades, then the numbers in each of these much minuteness; but some months of careful labor made it clear to me
classes will, on the average of many millions, be as stated in the that they were not sufficient to bear a more strict or elaborate treatment
table . . . . Thus the rarity of commanding ability, and the vast than I have now given to them. "~ 6 Calton found it impossible to attach
abundance of mediocrity, is no accident, but follows of necessity, reliable ratings to the replies, or to distinguish between the F's and G's
from the very nature of these things. 7 ' of scientific ability. As a consequence, the most important statistical re-
sult of this inquiry was not the analysis of scientific character, but the
Although this introduction of the error law yielded no concrete results
statistical technique that Calton had devised to enable him to draw exact
in Hereditary Genius, and added little if anything to its persuasiveness,
conclusions from this data as reliably and efficiently as possible. This
the mathematics of error analysis was itself enriched by its application to
was his method of "statistics by intercomparison," which he was able to
elusive objects such as intelligence. Because he \vas unable to meas~ue
these traits directlv, he was obliged to have recourse to statlstlcal umts, "Calton. "Mental Imagery," Fortnightly Review. n.s. rS (r88o). pp. 312-324
'"Calton, "On Men of Science, Their \faturc and Nurture," PRJ, 7! r874l, 227-236. p.
the great ancestor~ of which were these fourteen classes of intelligence. 23+
" 1 Sec Victor L. Hilts. "A Guide to Francis Calton's English tvfen a( Science.·· Transactions
"' Galton, "Statistics by lntercomparison, with Remarks on the Law of Frequency of Error," o( the American Philosophical Society, 65. part 5 ( 1975), z s-26.
Phil ;vfag [4], 49 ( 1875), 33-46, p. 37 ' 6 Calton, English !vlen a( Science (n. 41), p. vii.
7' Calton, Hereditary Genius (n. 41), PP· 34-35·

143
Sodal Law ancLNaturai Science Social LaH' and .Natural Science
apply to his data on men of science and which he published separately term "ogive" to designate the curve generated by this procedure. He fre-
the next vear. quently described its shape by imagining a series of men whose heights
The ~ethod of intercomparison was expressly designed to bring attri- are normally distributed arranged in a row in order of stature. The curve
butes that could be ranked but not measured within the purview of sta- that touches the top of each head will be an ogive; it will slope sharply
tistical analysis. It was "a method for obtaining simple statistical results near the ends, and very gradually at the middle.
which has the merit of being applicable to a multitude of objects lying Although the method of intercornparison proved difficult to use fruit-
outside the present limits of statistical inquiry." 77 Having recently read fully in connection with self-characterizations given by prominent men
Fechner's Elemente der Psychophysik, Galton recognized that only a of science, it still had certain noteworthy advantages over ordinarv sta-
slight ; crease in inaccuracy would be introduced if the median were tistical procedures. It was far more convenient for rank data tha;1 the
substituted for the mean-and it was, for his purposes, far more con- technique in Hereditary Genius. Since the ogive is completely deter-
venient. 78 He also abandoned the whimsical notion of defining a statis- mined as soon as two points on it are known, an anthropologi~t could
tical unit as one-fifth the distance between the mean and the threshold mfer the complete distribution of heights in a tribe using only two meas-
of genius. Instead he adopted the customary measure of the width of dis- urements, if only the barbarian chief could be persuaded to assemble his
tributions used bv astronomers, the "probable error" (or median error). men in order of stature. Again, if the anthropologist wished to learn the
This was, in any-study based on ordering by rank, much easier to learn "strength of pull" of these same aborigines, he was no longer obliged to
carry around an elaborate device capable of registering continuous mag-
than the "dispersion" (Vz times the modern standard deviation),
nitudes. Instead, the tribesmen need only be asked to try their strength
which also was occasionally used by contemporaries to measure the
against two bows ofknown stiffness. When the percentage that can draw
width of a distribution.
each bow is known, the corresponding ogive will give the distribution of
Using the method of intercomparison, it no longer was necessary to
strengths in the entire tribe. So
rneasure directly every individual in a group, and record all pieces of in-
These anthropological procedures, recommended by Galton in 1874
formation separately. TI1e presumptive applicability of the error law im-
in the British Association handbook .Notes and Queries in Anthropology,
plied that only two pieces of information need be known in order to
did not exhaust the practical uses of analysis by intercomparison. Galton
characterize the entire distribution. Hence everything necessary could
later conceived a yet more ingenious use of his method of fitting
be learned simply by arranging the group in order, beginning with those
ogives-this one a contribution to one of the most time-honored bene-
possessing the lmvest degree of the attribute in question and proceeding
ficiaries of probabilistic analysis, the theory of elections. If a legislative
to the highest. The middlemost individual in the series will then rep-
body is seeking to reach a decision on some matter subject to a contin-
resent its mean-"in at least one of the many senses in which that term
uous range of choices, such as the allocation of rnonev the reorescnt-
atives need not haggle endlessly in order to locate the t;~c point of con-
mav be used. "79 Galton later adopted Cournot's term "median" to de-
scri.be this middle term, but he was aware from the beginning that ex- sensus, which is the median. Instead they can simply take a recorded
pected mean and median were identical so long as the normal law pre- vote on any two amounts that fall within the range of variation. Thus if
vailed. He went on to point out that the individuals one-fourth and 6o percent think that at least A pounds should be allocated, while 2::;
three-fourths of the way along the curve would represent the probable percent favor at least B pounds, the ogive passing through these points
error of the series These two values, mean and quartile, were sufficient need only be constructed and the median can be ascertained directlv. s,
to characterize or compare populations. Galton used the architectural The technique of intercomparison reinforced Galton's awarenes~ of
the utilitv of units of probable error, which were indispensable to him
--Calton, "lntercomparison" (n. 71 ). p. 33·
-8 Gustav Theodor Fechner, Elemente der Psychophysik (2 vols., Leipzig, zd unaltered ed., when he began the work that led him to the statistical method of cor-
1907), voL 1, pp. 120·129 and passim. Calton referred very favorably to this book in a letter
Galton, "President's Address" (n. 38.1, p ..p J.
of 1875; see Galton, vo!. 3B, p. 464.
-q Galton, "lntcrcomparison" (n. 71 ), p. 34· '' Galton. "The \1cdi~n Estimate," BAAS, 59 ( 1899), 6-,8-640.

144 145
Social Law and Natural Sdence
relation. Still, the ogive was an outgrowth of the error curve. The in-
formation it conveyed could equally be expressed in terms of the error
distribution, and Galton's belief in its universality was in this sense de-
rived directly from Quctelet. That faith, sha1cd by Maxwell and Francis
Edgeworth as well, was instrumental in the most important work in sta-
tistical mathematics of the nineteenth century. Quctelet's idea was thus
of signal importance. His influence was not wholly due to the weight of
his logic. Galton and Maxwell did not simply go beyond him; they also
differed with him in some important matters of interpretation. Still, the
submergence of local disorder before general principle-Quctclet' s ver-
PART THREE
sion of the invisible hand--gained wide acceptance because the idea of
"Order in Apparent Chaos" \vas as welcome to scientific thinkers in
Creat Britain and Germany as in Belgium and France. The nature of
this conviction was brought out most clearly in Francis Galton's book
Natural fnheritance.
I know of scarcely anything so apt to impress the imagination as
the wonderful form of cosmic order expressed by the "Law of Fre-
quency of Error." The law would have been personified by the
Greeks and deified, if they had known of it. It reigns with serenity
and in complete self-effacement amidst the wildest confusion. The
huger the mob, and the greater the apparent anarchy, the more
perfect is its sway. It is the supreme la\v of Unreason. Whenever a
large sample of chaotic elements are taken in hand and marshalled
in the order of their magnitude, an unsuspected and most beautiful
form of regularity proves to have been latent all along. 'T'he tops of
the marshalled row form a flowing curve of invariable proportions;
and each element, as it is sorted into place, finds, as it were, a pre-
ordained niche, accurately adapted to fit it. 8 "
'' Galton, Natural Inheritance (n. 39), p. 66, adapted from "President's Address," /AI, 15
(!886 ), 489-499· '·

146
THE SCIENCE OF UNCERTAINTY

The acceptance of indeterminism constitutes one of the most striking


changes of modern scientific thought. With few exceptions, scientists
and philosophers previous to the late nineteenth century would have
agreed with\Augustus De Morgan that to sav an event occ~rs bv chance
is to say that'it occurs for no reason at all. lt to speak nonsens~, to con--
is
fuse a mere word with a cause. Mathematical probability, "far from pro-
ceeding out of any admission of events happening bv chance, is a con-
Through the great beneficence of Providence, what is sequence of the directly opposite belief; for, if preceding occurrences
given to be foreseen in the general sphere of masses, es- had been perfectly fortuitous, the arrival of one event would furnish no
capes us in the confined sphere of individuals; what sta- probability whatever for the repetition of the same under similar circum-
tistics indicates to be a definite law for the country, or stances."' The view of Quctclet and Buckle that the regularities of sta-
even for the town, cannot be discerned in hearth and tistics constituted laws pointed in the same direction. The inabilitv of
home. -VALENTIN SMITH (1853) the social scientist to predict in detail the behavior of individuals w~s of
no more consequence than minute puffs of wind that can alter the out-
What, statistics prove that there arc laws of history? Laws? come of individual coin tosses.
Yes, it proves how mean and disgustingly uniform t~1c Yet Quetclet's statistical laws seemed to some not wholly satisfactory,
masses are: is one to call laws the effect of inertia, stupid- even at the time. Statistical regularities, it was argued, prove nothing
ity, aping, love and hunger? Well, we will admit it, but about the causes of things, or they embrace acts so diverse as to be almost
with that the following proposition is also sure: so far as uninterpretablc. Most significantly, it was often pointed out, especially
there are laws in history, laws are worth nothing and his- by physicians, that statistics is quite incapable of justifying conclusions
tory is worth nothing. -FRIEDRICH NIETZSCHE ( 187 4) about individuals. Quetelet sought to escape the force of these objec-
tions by treating social physics as a science of societv, and not mcrelv of
individuals, but only the most dedicated partisans of statistical social ,sci-
Number is but another name for diversity.
-W. S. }EVONS (l874l ence were willing to concede that a mere mass regularity could consti-
tute a natural law. Those less inclined to grant this were roused from
their lethargy by the claim that statistical la\vs had shown human free
will to be illusory. The defense of human freedom inspired a v.idt:-rang-
ing revaluation of statistical thought during the late nineteenth ccnturv.
Statistics came to be seen not as the method of physical science, appli~d
to society, but as a new scientific strategy, more problematical in many
respects than the old, but also one with great promise. If the analogies
between statistics and physics survived, it was not because of the suc~ess
' iAugustus de Morgan]. "Quctdd on Probabilities," Quarterly Journal of Education. 4
i1X32), 101-110. p. 102.

\ 149
---- The Sci{!nc-€ of Uncertainty
of Quetelet's social physics, but because certain crucial areas of physics
Chapter Six
soon came to be seen by many as merely statistical. Maxwell's radical
new argument that the second law of thermodynamics could be no more
than a statement of probability adumbrated a new understanding of
physics. Significantly, it was put forth as a defense of metaphysical free-
STATISTICAL LAW AND HUMAN FREEDOM
dom, in opposition both to the alleged statistical disproof of free will and
to other contemporary arguments against the possibility of freedom
based on energy conservation and brain physiology.
Maxwell's line of thinking led eventually to C. S. Peirce's depiction s~;r~riticis_m of the pretensions of statistics began with the science it-
of a universe of chance and, less directly, to the discovery by quantum s~t . k ~eady ~l the early nineteenth century, the statistical approach was
physicists of the 1920s that the most fundamental particles of nature ex- a ac I~ on t e ground that mere statistical tables cannot demonstrate
hibit irreducible chance in their movements and interactions. In the causa lty, or that mathematical probability presupposes (what is un-
aftermath of this stunning new development in physics, the indetermin- t~mkable) th~ _occurrence of events wholly by chance. The intent of
ism of nature has become almost a part of common knowledge. Its in- t. ese ~arly cnhcs was _not to suggest the inadequacy of causal laws in so-
terest and significance can hardly be gainsaid. Clearly, chance is rec- C!al'~~~~nce, but to reJect the scientific validity of statistics. But statistics
ognized as a fundamental aspect of the world in a way that it was not s~n I've ' and when the same arguments were rehearsed later in the
nmete~n_th ccntur~, t~e~ were meant to imply that these shortcomin s
before.
Still, this story is not simply one of a new appreciation of the empire of statistics were mtnns!c to its subject matter and not fl . _g
method. · ' aws m 1ts
of chance. Randomness first attained real standing in scientific thought
not as a source of massive uncertainty, but as a small-scale component !he new _interpret~tion of statistics that emerged during the I86os and
of an overarching order. The recognition of chance stemmed not from l8;os ":as hed t~ a_ View of society in which variation was seen as much
the weakness of science, but from its strength--or rather, its aggressive more Vital. Statistical determinism became untenabl . 1. h
social think h d b . . e preCise y w .en
imperialism, the drive to extend scientific determinism into a domain . f hers w o use num ers became mwlllhng to overlook the di-
that had previously been seen by most as the realm of inscrutable · d.lVI·d ua 1s m
vemtv, o. . t e. compo nen t m · society, and hence denied that
whimsy. Even Quetelet was initially surprised and shocked at the reg- r~gulanhes m the collective society could justify anv particular co I
S!ons ab t ·t b • · nc u-
ularities of crime and suicide. In many ways, the drive to find reliable ou ' s ~em ers. The connection between this interest in hu-
laws of social phenomena has not succeeded. But its failure has been man heterogene,_ty a,nd statistical uncertainty is exemplified bv the re-
marks of a_ certam french statistician, Bourdin, who stood ~ at the
'~eetmgs ~he annou~ce
most pronounced not where phenomena are truly random, but where
unknown or unmeasurable causes are perpetually shifting the probabil- 1869 of International Statistical Congress to that
ities of events in ways that cannot be predicted. The indeterminism of ~~n oe~ not extst as a umtary being." Statistical results might be cer-
probability is so reliable and highly structured that randomness seems to tam, positive, ~nd absolute in the domain of the physical sciences he
conceded, but m respect to society that science could attain to n '
mo~e
disappear from the end result. However great the metaphysicai interest
than averages. These results he went on " d, . bl h o
of a universe of chance, it is at least equally significant from the prosaic v 1 "b ". ' . ' un ema Y ave a certam
standpoint of the historian that chance has proved no obstacle to sci- ,. a ue, _ut they a:e, like the science itself, essentially variable."' Que-
'~let, emznence gnse of the Congress and chairman of this particular ses-
entific prediction and control. 2
SIOn, protested that statistical research could yield "ad c . bl e laws ,' cwr
mna
' On these matters, see Ian Hacking, The Taming of Change (Cambridge, Eng., 1986); also
Lorenz Kruger, "Philosophical Arguments for and against Probabilism," in Prob Rev. r .. ·
~eptieme ague, 1 8 rOh 33-165, pp. 52-
session du Congre~~n~e:,;:t:~~~~~e s~~~i~;;qn~: (s~a,~~~~lucs," in ~o1mPte Rendu de Ia
' "Theorie de la statistiq t d , . .
>3·

1)1
( (
Statistietfl Law andlluman Freed!Jm Statistical La wand Human Freedom
society as for nature, but Bourdin spoke for an emerging consensus descriptions in terms of which statistics was defined. Statistics was de-
against which Quetelet's protests were unavailing. . . rived from the Latin status ("etat, situation"); it was a highly detailed
The influence of these social discussions on natural science and phi- geography that could do no more than "reveal the state of production
losophy will be seen in chapter 7, where we find them debated again i~ and of consumption of one or several nations at a designated point in
the more structured context of gas physics. This line of thought culmi- time (epoque, or for several successive points in time. "• This might
nated in a sense in the work of Charles Sanders Peirce, who made the have some interest, but it was subject to instant obsolescence. Sinclair's
spont~neous production of chance variation an indispensable ~ondition great statistical compilation for Scotland, Say remarked, probably
of evolutionary change, whether physical, biological, or soc1al. They ceased to be accurate within a year of its publication. Statistics was the
also bore fruit in the growing interest in the analysis of variation,. and not most transitory kind of knowledge, and for that reason Say thought it
merely of mean values, evinced by the late-century mathematical stat- preposterous to speak of "laws" of mortality or of population derived
isticians. To be sure, Galton gave little attention to the debates on hu- from statistics. He wrote: "Variable relations are not laws: they
man freedom, but Francis Edgeworth was closely familiar with them, change."3
and Wilhelm Lexis's important work on dispersion can only be under- Long before Quetelet invented his doctrine of statistical uniformity,
stood in the context of this tradition. which may be seen in part as an answer to objections like this, the prom-
inent Italian statistical writer Melchiorre Gioja had countered Say with
the observation that most statistical objects vary slowly, if at all, and
THE OPPONENTS OF STATISTICS hence that its results remain approximately valid for an appreciable pe-
riod. 4 Evidently the defense of statistics moved Say slightly, for Quetelet
Champions of statistics of Quetelet's generation preferred not to view it was able to procure a letter from him indicating that statistics could con-
as different from other sciences, but they did see in the strategy of mass tribute to knowledge of society if pursued competently.; Fundamen-
observation a singular virtue that permitted the methods of natural sci- tally, however, Say's conception of political economy was ill-suited to
ence to be applied to social matters. The great merit of stati~tics wa.s that offer statistics a place of any significance. Say argued that there are two
it eliminated perturbations by ignoring individuals and lettmg th~lf un- kinds of facts: general or constant facts, arising from the action of laws
predictable activities average out. That th~s was gen.ui~ely novel1s con- of nature in well-defined cases, and particular, or variable, facts, also
firmed by the writings of the leading critics of statistic~, most. of them the result of inviolable laws, but modified by special circumstauces that
positivistically oriented Frenchmen, who rejected preCJSel~ this feature conceal the true underlying relations. Facts of the second sort, which
of it. The method of averaging, they argued, does not clanfy, but con- constitute the subject matter of statistics, are like the jets of water in a
fuses by mixing together things that are fundamentally different. Any so- fountain that can counter the force of gravity and suspend a heavy object
cial science that views the differences among individuals as random, in equilibrium. They may be curious and amusing, but they make a
they argued, is irremediably flawed. Auguste Comte, the so~ial scien- poor basis for science. Say viewed Adam Smith's Wealth of Nations as
tist, argued that one must analyze carefully in order to establish causes a confused mixture of political economy and statistics; his declared in-
and recognize their heterogeneous effects on different parts of t~e ~op­
' Jean-Baptiste Say, Traite d'economie politique, ou simple exposition de Ia maniere doni se
ulation. Medical opponents held that because statistical generalizations fonnent, se distribuent, et .1e consomment les richesses (2 vols., Paris, 1803), vol. 1, p. v.
could not be applied to individuals, they were useless and even immo- 3 Say, "De !'objet et de l'utilite des statistiques," Revue encyclopedique, 35 ( 1827), 529-5 53,

pp. 547-548; also Cours complet d'economie po/itique pratique (2 vols., 2d ed., 1852; repr. Os-
ral. nabrtick, 1966), vol. 2, p. 486; re\'iew of joseph Lowe, "The Present State of England," in
As early as 1803, before statistics had even become associated spe~if- Revue encyclopedique, 18 (1823), 312-324.
• Melchiorre Gioja, Filosofia della statistica (1826; Mendrisio, 1839), pp. 7-16; also Indole,
ically with numbers, Jean-Baptiste Say argued vigorously that nothmg estensione, vantaggi della statistica (Milan, 1809).
like a reliable science could be formed from it. It would be absurd, he ' See Quctelet, Physique sociale, ou essai sur le developpement des facultes de l'homme (2
explained, to base political economy on the miscellaneous numbers and vols., Brussels, 1869), \·ol. 1, p. 104; vol. 2, p. 447· The letter was written in 1832.

152 153
( (

Statistical ba-w and Human Freedom Statistical Law and Human Freedom
tention was to secure the foundations for a science of economy based ferent characters. Tables of mortality, for example, may be valid for an
solely on general or reliable facts. 6 average, but are false for both indigent and leisured classes, as every in-
Naturally, Say denied that his was a purely deductive science. An en- surance company must soon discover.'' In short. the numerical results
lightened ideologue, he eschewed, in the customary and acceptable obtained by amassing information about a variety of individuals lack the
manner, that spirit of system which had given rise to scholasticism and purity and clarity needed for anv genuinely scientific theory of political
to the Cartesian vortices.7 Political economy could not, however, he economy. Statistical results, collected regularly and presented coher-
based on the grab-bag of facts presented by statistics, but only on selected ently, convey an impression of the condition of a people whose value
experimental facts chosen especially for the pristine clarity that arises must not be gainsaid, but which cannot provide the basis of a science.
from the absence of modifying factors and for the availability of com- Since Auguste Comte was deeply impressed by the interpretation of
plete knowledge about them. 8 Say conceded that economic ?henomena science of the ideologue physiologists Bichat, Broussais, and Cabanis,
are often formidably complex, and he disavowed on behalt of pohbcal and particularly by their reasons for rejecting mathematics in the study
economy any pretense to explain them completely. For that reason, it of life, it is not entirely fortuitous that his view of statistics should bear a
seemed all the more absurd to suppose that they explain themselves. certain resemblance to that of Say. Comte rejected the use of mathe-
The truth must be discovered through exact analysis, and facts only take matics in social science, for it was a central truth of positivism that each
on meaning when elucidated by reason according to economic princi- science must have its own distinctive method and that reduction of sci-
ples. Hence political economy is logically prior to statistics, and is es- ences is impossible. Beyond that, he was wholly unsympathetic to math-
sential to raise understanding above the "dangerous empiricism" that ematical probability. Like Destutt de Tracy, he thought Condorcet un-
"applies the same methods to opposite cases which it believes similar." justified, and Laplace and his imitators positively ridiculous, to seek to
The truths of political economy need not be tested against undigested apply probability to judicial decisions and testimonies. This he regarded
facts. For example, it is a law that interest is proportional to risk, and as an "abuse ... of the mathematical spirit," and accordingly he ap-
though a particular lender may not know the risk, or some other circurn- plauded Poinsot's refutation of Poisson which, he thought, relieved him
stance may confuse an individual case, the general law remains, ~·and of the obligation to present his own full critique. Comte went so far as
must recover its whole empire at the moment when the causes oi per- to reject the mathematics of probability itself-he was the most promi··
turbation, themselves the effect of some other general law, cease to nent scientific writer since d' A! em bert to do so. He wrote: "As to the
act. "9 philosophical conception upon which such a doctrine rests, I believe it
At the same time, and for identical reasons, Say denied that political radically false and tending to lead to the most absurd consequences ....
economy could attain greater exactitude through mathematics. The law It would habitually lead us in practice to reject as numerically improb-
of supply and demand is universal and invariable, but the price of a bot- able (invraisemblables) events which will nonetheless take place."'z
tle of wine cannot be predicted from it, for it will fluctuate with availa- Probability had led geometers into "utter delusion," Comte thought, for
bility of capital, changing expectations and tastes, weather, and a host it "presupposes that the phenomena considered are not subject to
of other causes. For this reason, experience furnishes no fixed value to law."ll
which calculation can be applied.' 0 Even if the problem of fluctuations Comte's social pl1ilosophy was not so incompatible with statistics as
could be overcome, most of the numbers presented by statistics are se- '"""''' '

'' Sav. "Del' objet" in. 3), p. 548. On Sav. sec Claude :vienard, '"1 'hrcc Forms of Resistance
riously misleading, for they lump together facts of fundamentally dif- to Statistics: Sav, Cournot, Walras," History of Polztical Economy, 12 ( 1980), 524-541.
"Auguste Comtc, Cours de philow-phie positive (6 vols., Paris, 1S3o-1842), vol. 2 (r8)5),
6 Sav Traite (n. 2), pp. iv-ix. p. 371; also vol. 4 ( 1839). pp. s13-51 ). See also "'Appendice, XIV" ( 1819), in Revue occiden-
Say: Traite(6thed., 1841; rcpr. Osnabrtick, 19661. voL 1. pp. 15-16. tale philoso-phique, social, et -politique, 8 i 1882), -'}00-409. For l'oinsot rebuttal to Poisson, see
B Say, Traite(zded., Paris, 1814), vol. 1, pp. xxv-xx,·i1;6thed., vaL 1, pp. 6-;. Comptes rendus hebdomadaires des seances de /'Academie des sciences, 2 (1836), 380, 389.
<>Ibid., 1st ed., vol. 1, pp. viii-ix. '' Comte, A General View of Positivism ( 1848i. J H. Bndges, trans. (New York, 1975), p.
' 0 Ibid., zd ed., vol. 1, pp. xxx-xxxii. 28.

154
i
I
(
Statistical Law and Humon~Freedom
Statistical Law and Human Freedom
to prevent Quddet from plagiarizing his general historical conception.
domain standing above that of individuals. Indeed, the regularities of
Comte had argued from the early 1820s that the object of social science
statistics had helped to define that domain.
was to learn the natural course of civilization, \Vhich, he propmed,
The statistical method was much more difficult to defend in relation
could not be redirected by politics, but only smoothed or made more
:o medicine, particularly diagnostics and therapeutics, and opposition to
turbulent. Comte thought the secular course of progress to be largely a
1t \vas more in evidence there than in any other field. For phvsicians,
consequence of advancing knowledge. Quetclet adopted these ideas
the direct object of concern was quite naturally the sick individ.ual, and
wholesale, and with them the title of Comte's science, physique sociale.
it was far from clear that the mean result from some large number of
Comte was greatly annoyed, observing that "a Belgian savant'' had ex-
assorted trials in a hospital provided an appropriate basis for treating that
propriated this designation and misconstrued it as mere statistics, thus
individual. Blind use of averages threatened to supplant that fe ne sais
obliging him to coin a new title for the same science, sociology. q This
quai, medical tact, upon which physicians prided themselves. Accord-
theft was all the more offensive because Comte was unimpressed by the
mgly, thcso-called numerical method in medicine was viewed by many
achievements, actual or potential, of statistics. He meant by social sci- w1th suspiCIOn, and subjected on occasion to vitriolic attacks.
ence a study of beliefs and of institutions in history, not the construction
Medical statistics, under other names, was at least as old as political
of a mythi~al average man. Comte held that the current state of society arithmetic, and the numerical method was applied in a variety of in-
involved a confusion of the three fundamental stages--theological, s_tances_ even to therapeutics during the late eighteenth century. 1 7 Phi-
metaphysical, and positive-so that mass statistics could only smear to- lippe Pmel, who admired Condorcet's work, championed the numerical
gether what ought to be kept distinct. A "genuine and precise statistics method during the first decades of the nineteenth centurv ~but the col- 1

of the social body," he thought, was impossible unless based on study of lection of numbers in order to guide the physician in his. ;1ractice on in-
the historical development of human institutions. 15 dividual patients was largely a product of the 182os and 183os. The
There were a few other social scientists during the nineteenth century name with which that method became most strongly associated was that
who entirely rejected this new science of statistics. Paul von Lilienfeld of Pierre Charles Alexandre Louis. Louis' most famous and controver-
argued in 1873 that the law of large numbers, the foundation of statis- sial work was a quantitative study of the effects of bloodletting on pneu-
tical science, was no law at all, since it did not act with necessity, and monia, published in 1828. His results were equivocal-early bleeding
hence that even the most interesting statistical tables presented not "nec- seemed to lessen the duration of the illness slightly, hut to have far less
essary natural laws" but only "accidental circumstances. " 16 This argu- effect than had generally been thought-but his work became noto-
ment, hcm;ever, was less compelling against statistics as social science rious. This was due partly to the circumstance that his results were in-
than it was against other applications of statistical method, for it could terpreted as a challenge to conventional practice. J'\Iore generally, how-
be answered by pointing out that the object of statistics was not the in- ever, the numerical method was put forward as a s~'stematic technique
dividual, but society. Statistics maintained much of its persuasiveness so for improving medical treatment at a time of diagnostic triumph and
long as observers were willing to admit the existence of a coherent social therapeutic crisis, and hence much more was at stake than anv partic-
ular remedy. '
,. Comte, Cours (n. 12), vaL 4, pp. 7, 252. See Julien Freund, "Quctclct et Auguste
Corntc,"' Academic royale de Belgique, Adolphe Quetelet, 1796-1874 (l'vlemoria[) (4 vols , Louis exalted his numerical method as "simultaneously natural and
Brussels, 1974), vol. 4, pp. 46-64. rigorous. "' 9 It offered an unprecedented level of precision, .and was fully
'' Comtc,' "Plan des travaux scientifique necessaires pour n'organiscr Ia societe"' ( 1822), in
Opuscules de philosophie sociale, r819-1828 (Paris, r883), 6o-18o, p. 130. Comte's disciples
were less hostile to statistics, though thev did not usc them. See C \\\rouboff, "De Ia me- ~ ,.. Sec Ulrich Trohler, Quantification in British Medicine and Surgerv, 17 ;-rl-l8JO. with
thode dans Ia statistiquc," <1 review of Quctdefs Physique sociale, in La philosophic positi\'e. 6 Specw/ Reference to 1ts Introduction into Fherapeutics (Ph. D. Dissertation. ·of Lon-
don, '9791-
z 'l-43 .
Paul von Li!icnfdd. Gedanken i.iber die Socialwisser>schaft der Zukunft (187'3; 3 wds., ,, Sec Erwin Ackcrknecht, :'vfedicine at the Paris Hospital,
pp. 47-48. I Baltimore, 1967),
Berlin, 1901 ), vol. 3, p. to; also vol. 1, pp. 387-388. Lilienfdd conceded the usefulness of
averages in politicJ! economy and medicine 1v·ol. 1, pp. !Ol-IO:). "' P. C. A. Louis. Recherches sur les effets de Ia dans quelques maladies inf!amma-
to ires et sur/' action de el. des vesicatoires la pneumonic !Paris, 1835), p. 70.

1 57
·Statistical Law and Humdtl Freedom Statistical Law and Human Freedom
consistent with accepted medical standards. To those who objected that ability is to appeal to chance and to give up the possibility of certitude.
individual cases were too diverse to be grouped together statistically, Amador further claimed that the repetition of events in the past proves
Louis responded that diseases can be classified by type, even though no nothmg about the future. When probability is applied "to real facts in
two cases are identical, and that if his opponents chose to deny all n> the physical and moral world, it becomes either useless or illusory."''
semblance among separate cases, then nothing would be left in h- If this inflexible and mechanical calculus were n1adc the basis of
eine but pure individuality. Then, there would be no prospect of attain- medicine, Amador announced, healing would cease to be an art and be-
ing general knowledge by any method. In fact, "it is precisely on come a lottery. The numerical method denies the variability of medical
account of the impossibility of comprehending each case with, as it facts, which can only be fully appreciated through induction and med-
were, mathematical precision, that it becomes necessary to count; since ical intuition. Since the enumerators lump together disparate cases.
the errors, the inevitable errors, are the same for two groups of diseases their aim is clearly "not to cure this or that disease, but to cure the most
treated by different procedures, these errors arc compensated and can be possible out of a certain number. This problem is essentially anti-med-
neglected, without sensibly altering the exactitude of the results. " 20 The ical." After all, since "the law of the maioritv has no autho~itv over re-
numerical method captures what every physician does when he learns fractory cases," the physician must eithe,r ig~ore the results of statistics
from experience, Louis wrote, and he had been led to it naturally, even for these variant individuals or condemn them to death. 2 4 Even strong
involuntarily. Still, the difference between those who count and those advocates of the numerical method, !ike the Englishman William Guv
who are content saying more or less "is the difference of truth from error, admitted that its application to individual cases \vas problematical. z; Ri~
of something clear and genuinely scientific from something vague and suefio d'Amador argued, somewhat oddly, that it was the task of nature
without value. " 2 ' Louis acknowledged that his numbers were too small, to .consen.'e th~ species, and that since medicine had no desire to usurp
and appealed for a more systematic collection of them. His student Dan- tlm function, It could not be judged by results en masse. z6
vin argued that the only impediment to the attainment of certainty with Louis pleaded in defense of his method that medical statistics could
the numerical method was the unavailability of sufficiently large num- demonstrate variability as well as uniformity and that its object was not
bers, and maintained that once the number of cases reaches 500 the re- necessarily "Ia determination d\m hornme moyen ou irnaginaire. " 2 7
sulting proportions are thereafter fixed and exact. 22 The numerical method was by no means banished, although the level
In the view of opponents of the numerical method in medicine, of of enthusiasm had begun to wane by 18 50; the cumulative results were
whom there was no shortage, Louis' strategy was based on a false analogy s~~ficient by 1865 to fill a massive German handbook compiled by
between medicine, whose facts are complex, variable, and often hid- Fnednch Oesterlen. 28 Most writers on the numerical method had little
den, and physics, where they are always simple and uniform. Risueflo understanding of probability, and tinv samples were common. For this
d'Amador, who launched a debate in the Academic Royale de Mede- they were sharply criticized. In addition, resistance to the verv idea of
cine in 1836 with a memoir on numbers and medicine, denounced the basing medicine on numbers, however large, remained strong. ~Auguste
use of probability in therapeutics as anti-scientific. Among his argu- Comte contended that the numerical method \Yas merelv "absolute em-
ments was the assertion made familiar by Say that statistics is mere de- pir~cism, masquerading under frivolous mathematicat appearances,"
scription, valid only for the fleeting moment. His respect for the math- and could only lead to a "profound, direct degeneration of the medical
ematics of probability was on the same level as Comtc's; he called it " Ris~CllO d?.mador. "MenH;~rc sur lc calcul des probabilites appli4UC aIa mcdccinc," Bul-
skepticism embracing empiricism, and remarked that to resort to prob- letm de I Academte Ruyale de tvleuecme, 1 1.1 ~-,6). 6n-68o, p. 6 24 . This was followed bv "Dis-
cusstOI>s sur Ia stahstiquc mcdicale," pp. 68 4 . So6. '
' 4 Ibid , pp. 6·>4-63 5·
'" Ibid , p. 76 '' William Augustus Cuv, "On the Value of the Numerical :\lethod as applied to Scicnct:,
" Ibid., p. 85; Recherches anatomiques, pathologiques, et therapeutiques sur Ia phthisie
but e~pcCiallyto \?YSlologv and ,\lcclicmc," )RSS, 2 1183 0 !, 25-.+7, p. 4 o.
(Paris, zd ed., 1843), p. xx. ' 6 Dtscuss1ons (n. 23), p. 8os.
"B. Danvin, De La methode numerique et de ses avantages dans fetude de Ia medecine ' 7 Ibid., p. 741. .
(Paris, 1831), pp. 30-31. ' 8 Fr. Ocsterlcn, Ilandbuch der medicinischen Statistik (T(ibingcn, 1 865).
T

Statistical Law and Human Freedom Statistical Law and Human Freedom
art"-to the trial of therapeutic procedures at random. '9 During the Bernard's point was not, as the physician Double argued, that statis-
1840s and 185os, an alternative to statistics for improving therapeutics tics "strip man of his individuality, "Band he certainly did not agree with
began to gain increasing support. This was experimental physiology, a Cruveilhier that general results in therapeutics are unattainable "be-
field in which it was and b.<> remained customary to emphasize control cause in medicine there is nothing but individuals. "34 Nor did he favor
over all aspects of the experiment rather than the averaging out of errors. the doctrine that the imperfect knowledge gleaned through observation
The reductionist Emil Du Bois-Reymond was forthright and vocal in should be supplemented with some obscure medical tact. Bernard's
opposing numerical tables and stressing complete determinism. 30 The modest aim was to find through completely controlled experimental
most influential critique of medical statistics, however, came from the manipulations the general laws that governed vital phenomena deter-
positivist Claude Bernard. ministically, without exception. He thought it absurd to label statistical
Bernard did not, like Comte, reject the applicability of mathematics results laws, when statistics had been applied precisely because the con-
to biological phenomena. The precision of mathematics was always de- stituent facts were not fully comparable. The numerical method gives
sirable, he wrote, for the discovery of exact relations was the proper aim only probabilities, he insisted, whereas the only proper aim of scientific
of all science. But the profitable employment of mathematics presup- medicine is certainty. It cannot rest content with relative proportions,
posed, in his view, that all relevant phenomena were completely under- but must know why, when eighty are cured, twenty still die, so that
stood, and on this basis he maintained that the study of vital phenomena knowledge will be complete and medicine reliable. Bernard proclaimed
was not yet ready for mathematics. The complexity of vital phenomena again and again the need for experimental determinism; "the absolute
constituted the central problem of physiology, but no solution to it could principle of experimental science is conscious and necessary determin-
be attained by averaging variety out. Instead, the physician must learn ism in the conditions of phenomena. "35 Physicians, he wrote, "have
to understand in detail.
nothing to do with what is called the law oflarge numbers, a law which,
Bernard asserted repeatedly that physiological functions are not ho-
according to a great mathematician's expression, is always true in gen-
mogeneous, either over diverse individuals or over time for a single in-
eral and false in particular. This amounts to saying that the law of large
dividual. There is, he held, no average pulse; there is only a pulse when
numbers never teaches us anything about any particular case. What a
resting, or exercising, or eating. It was the height of folly to seek, as some
physician needs to know is whether his patient will recover, and only the
unnamed physiologist allegedly had, the "average European urine" by
search for scientific determinism can lead to this knowledge. "16
collecting from a railroad station urinal in a great city. Averages "con-
Bernard doubtless underestimated what could be accomplished
fuse while aiming to unify and distort while aiming to simplify," and
through careful controlled trials performed on large numbers of patients
consequently, the physiologist "must never make average descriptions of
using drugs and procedures whose mode of action was imperfectly
experiments, because the true relations of phenomena disappear in the
understood, if at all, and his denunciation of statistics has become a con-
average. "3' This was most especially the case in medical therapeutics.
siderable embarrassment to his latter-day medical admirers. But phys-
"True enough, statistics can tell you if an illness is more serious than
iologists, in fact, still have little use for statistics, which they rarely pres-
another; you can tell your patient that, of every hundred such cases,
ent except out of a sense of duty, and then, quite often, badly. Those
eighty are cured ... but that will scarcely move him. What he wants to
know is whether he is numbered among those who are cured. " 32 proach. "in Michacllleidelbergcr et al., eds., Probability since 18oo: Interdisciplinary Studies
of8Scientlf1C Development. Report Wissenschaftsforschung, 25 (Bielefeld, 1 9 83 ), 275 _2 86, p.
2 2.
"' Comtc, Cours (n. 12), vol. 3 (1838), p. 420.
;o See Brigitte Lohff. "Emil Du Bois-Rcymonds Thcoric des Experiments," in Gun;7er , '-'Quoted in Joseph Schiller, "Claude Bernard et la statistique," Archives internationa/es
;..1ann, ed., l\aturwissen und Erkenntnis 1m 19. Jahrhundert: Emzl Du B01s-Reymond \ r.d- d h1st01re des sciences, 17 ( 1963), 405-418. p. 408.
desheim, 1981), llj-128, p. 122. '
4
Quoted in Paul Delaunay. "Les doctrines medicales au debut du X!Xe siecle: Louis et Ia
'' Claude Bernard, An Introduction to the Study of Experimenta/1\.1edicine ( 1865), H. C. methode numcriquc," in E. A. Under.vood, ed., Science. Medicine, and History ( 2 vois .•
Greene, trans. (New York, 1957), p. 135. London, H)5'l), vol. 2, pp. 321-330, p. 326.
''Bernard, Principes de medecine experimentale, L. Delhoume, ed. (Paris, 1947), p. 6;. "Bernard, Introduction (n. 31), p. 53·
quoted in William Coleman. "Neither Empiricism nor Probability: The Experimental Ap- ' 6 Ibid., p. q8.

160
Sta-tistical Lawand Human Freedom Statistical Law and Hurnan Freedom

who, like Bernard, seek universally valid principles but not, in the first some time whether statistics might be inconsistent with hmnan free-
instance, exact quantitative ones, can simply ignore statistics. Bernard's dom. Although the great champions of statistics. including Quetelet
critique should be recognized as among the most insightful commen- and Buckle, were aware that its success depended on a strategy of ignor-
taries on statistical reasoning up to his time, impressive for its clear for- ing individual phenomena, it was almost universallv associated with an
mulation of epistemological issues if not for its tolerance of alternative impressive extension of the domain of knowledge, and not with its lim-
approaches or its understanding of probability mathematics. Bernard itations. The more skeptical view of statistics th;t gained influence dur-
understood perfectly that adopting a statistical approach implied ac- ing the 186os and 187os was largely a reaction to the extravagance of its
ceptance of a considerable domain of ignorance-the lumping together pretensions.
of heterogeneous material so as to establish generalizations that could The possibility that statistical regularity might be seen as inconsistent
not be applied with certainty to individuals. with human freedom troubled Quetelet almost from the beginning of
Bernard thought this expedient entirely unnecessary, as did Say and his career. Even before he invented his mricanique sociale, it appears, he
Comte. Certainly none of these positivistically inclined French oppo- was subjected to sharp criticism for his attempt to quantify the human
nents of statistics believed that the imperfections of that science had any sciences. Already in 1829 he complained of accusations that his re-
implications for a proper understanding of nature or society. Thus, al- searches were materialistic and that he viewed states as cadavers, or that
though their role in the development of statistical thinking was signifi- his observations on the constancy of crime led to a dangerous fatalism.
cant,· it is still more important to look at those critics of statistics who Yillerme, evidently, had been similarly reproached. In a letter of 1829
thought detailed knowledge of individuals was inaccessible and who to Quetelet, he observed that natality is an economic phenomenon-
thus accepted the statistical study of mass phenomena, with all its lim- the production of people (births) being determined by their expenditure
itations, as the best available approach to certain sciences. For these (dcaths)-and then remarked: "You sec that if you are reproached for
writers, the reinterpretation of statistics required a new understanding of viewing societies as cadavers, I can equally be reproached for viewing
scientific law, either as an expression of the possibilities of human them as merchandise. "38
knowledge, or perhaps even as a characterization of the real world. Quetclet's 1848 paper seeking to reconcile statistical regularity with
free will was criticized in the Belgian Academy by two conservative
scholars. P. De Decker argued that statistical regularity must be attrib-
STATISTICS AND FREE WILL uted to divine will, not laws of probability, while M. Van Mecnen
maintained that moral statistics is a contradiction in terms, since noth-
ing can be known by men about the human soul. 39 Tom Cradgrind's
As early as 1819, Thomas Young remarked that the surprising regularity
remarks in Hard Times, quoted at the head of Part One, reveal Charles
of phenomena like the dead letters in the Paris postal system implied no
Dickens's view of the effects of such teachings on young minds. From
"mysterious fatality" and that Laplace among others had interpreted
William Whewell, Quetelet's ideas brought implied criticism in the
them "as implying something approaching more nearly to constancy in
form of displaced congratulations. No subject, wrote Whewell, could be
the original causes of the events, than there is any just reason for infer-
better suited to re\eal the relation between moral statistics and free will
ring from them. "37 The standard interpretation of statistical regularity,
than the one Quetelet had chosen for his 1848 paper: marriage. Whew-
however, remained precisely the one that Young opposed. Quetelet's
ell clearly was more skeptical of the conclusions that might be drawn
statistical laws were not onlv seen as indicating some constancy in un-
derlying causes, but came ~ven to be associated with an ill-defined de- '"Letter, 25 April 1829, Villerme to Qucte!ct, cahicr 2560, AQP; also Quetekt. "Re-
terminism, if not mysterious fatality, and it was seriously debated for cherches statJstiqucs sur lc Ro\'aumc des Pays-Bas." NJv!AB. s {1 pp. v, 33·
19 "De !'influence de libre arbitre de l'homrnc sur Ies faits sociaux. ~.v1AB, 21 ( 1 R4K), by

17 Thomas Young, "Remarks on the probabilities of error in phvsical observations. and on


De Decker, pp. 69-92. and Van :V1eenen. pp. 93-112. See also a review bv Moritz Wilhelm
the density of the earth, considered especiallv with regard to the reduction of expenments on Drobisch, "Moralischc Statistik .. , Leipziger Hepertorium der deutschen uniausli:indischen Lit-
eratur, )g. 1 849. voL 2. pp. 28- 'l9·
the pendL;lums," Phi/Trans, 109 (1819), '70-95. p. 71
(
Statistical La wand H umanFreedom
Statistical Law and lluman Freedom
the statistics of crime. Only external influences, not the internal
process of the mind, can be given an exact numerical form through sta- 186o session of the International Statistical Congress held · I d
h · ·· ' m "on on
tistics, wrote Whewell. "Your statistical results are highly valuable to the t) ~se ~ccusabons were summarized in their most extreme form by th~
legislator, but they cannot guide the moralist. A crime is no less a crime, l nn;c ~omort, a former pup!~ <:me! continuing correspondent of Que-
because it is committed at the age of greater criminality, or in the month telet s, and a strong advocate ot statistics, \Vho served as president of the
of more frequent transgressions. "4o Cong~ess. It had been alleged, he explained, that the numerical studv
of SOCiety '
-t- Objections like these, however, were not often heard. Until 1857,
most readers found the regularity of moral statistics somewhat surprising
le~ds ne~essar!ly to Pantheism and the destruction of true religion,
in view of the evident freedom of the human will, but only slightlv, if at as It depnves, m man's estimation the Almighty of His power offree
all, threatening. Thus George Boole, who followed Scottish Common self-de~ermmation, making His world a mere machine, working
Sense in regarding metaphysical freedom as a self-evident fact of con- accordmg to a general, pre-arranged scheme, the parts of which are
sciousness, wrote: "The consideration of human free-agency would capabl~ of mathematical measurement, and the scheme itself of
seem at first sight to preclude the idea that the movements of the social m~mencal expression, that it leads to fatalism, and therefore de-
system should ever manifest that character of orderly evolution which pnves ~1an of his dignity, of his virtue and morality, as it would
we are prepared to expect under the reign of a physical necessity. Yet P:~ve him to be a mere wheel in this machine, incapable of exer-
already do the researches of the statist reveal to us facts at variance with CISing a free choice of action, but predestined to fulfil a given task
such an anticipation. "·P and to run a prescribed course, whether for good or for evil. 43
The forceful arguments in Buckle's 184 7 History of Civilization de-
stroyed this spirit of statistical moderation. The reception of that book In. the wake ~f.this challenge to accepted views, it became necessary
should suffice to dissolve a modern reader's residual inclination to sup- to dun~ more cnhcally about the character and implications of statistic;]
pose that the expansion of statistics was associated in the first imtance
!
laws. l.Ie urgency of the task was depicted with unmatched melodrama
with an increasing tendency to philosophical probabilism. On the con-
~y Wilham Cyples, author of an essay in Cornhill Magazine on the
Morality of the Doctrine of Averages." Statistics "brings the human
trary, statistics was intended to expand the domain of exact scientific cer- heart to a standstill," he wrote.
tainty to include man and society as well as physical and organic nature.
Buckle's book was an enormous success, reaching a popular as well as When the choice lies betwixt this wholesale ruin of the human
an intellectual audience. The fear he provoked that a new and all-em- worl?, and concluding that some fev.· persons, led awav bv an en-
bracing determinism had at last succeeded in excluding the possibility t~usJasm for statistics, have applied logic to a matter ~utside the
of divine or human freedom extended from America and Britain to Ger- lrm1ts of proof, an ap~eallie: to common sense; and a protest may
many and even to Dostoevsky in Russia, whose underground man com- safely b~ entered agamst tlm modern superstition of arithmetic,
plains about statistics and then about Buckle. 4" It is far from clear that which, If acquiesced in, would seem to threaten mankind with a
Darwin or Comte was discussed with greater urgency during the 186os later and worse blight than any it has yet suffered,-that not so
and 187os. much of a fixed destiny, as of a fate expressive in decimal fractions
£ ll" .. ,
Friends of statistics were obliged to take note of the controversy in a _mg upon us, not personally, but in averages.
which their science was suddenly engulfed, for the charges directed at ~a:ious possibilities were put forward to deflect this argument from sta-
them threatened to undermine the legitimacy of their science. At the tJ~bcs agamst human freedorn and responsibility. One of the more cre-
40 Whewe!l to Quetdet, 7 Oct. 1847, cahier z644, AQP. ~tJve was suggested b~ the same William Cyples. "If a prisoner standing
"'' Boole, An Investigation of the Laws of Thought (1854; New York rcpr, 1958), p. 10.
"See Fyodor Dostoevsky, Notes from Underground (186.4), :\1irra Gimburg, trans. (New
m the dock pleaded m bar of punishment, that the commission of his
York, 1981), pp. 23, 2). l 143 "The A1d?ress of the Prince Consort on opening as President the Fourth Session of the
n em a bona Stahshcal Congress,'" JRSS. 2 3 ( 1 86o), 2 80 .
('

Statistical Law and Human Freedom


Statistical Law and Human Freedom
crime was necessary for maintenance of the statistics, it would be per-
general fa:t of statistical regularity could in no way justify claims about
fectly logical . . for the judge to urge that, in the same way, it was req- the behaviOr of any constituent individual. He wrote:
uisite for his ten years' penal servitude to be ordered, to prevent an ir-
regularity in the annual tables of sentences. "44 Several of Buckle's If free :Vill :xists at all, it cannot on any hypothesis introduce more
reviewers challenged the supposition that free will ought to express itself confuswn mto sta~istical calculations than any other cause of ac-
in the form of disorder. One, at least, maintained that the lawlike reg- ~Ion, of the o~eratwn and nature of which we are ignorant, but it
ularities of statistics do not contradict freedom, but involve merely "an IS the very object of the science to which Mr. Buckle refers to en-
observed uniformity of succession" in the phenomena whose cause re- able _u~ to make general assertions about the effects of such causes
mains unexplained. 45 and It Is the strangest perversion of its doctrines to infer from the~
British and American critics of Buckle also found it attractive to stress that unknown causes do not exist. If the question whether one man
the negative side of what had appeared in Quetclet and Buckle as the should ~r should not murder another had to be decided by a throw
principal justification for adopting a statistical approach. Statistics em- o~ the dice, the uncertainty whether the murder would take place,
bodied the long view; it was the expedient through which the confusion \\-ould be qu_Ite as great as It could be if the question depended on
and unpredictability of individual phenomena gave way to splendid reg- free WilL With respect to the dice, we can forctel to a nicetv how
ularity and order in the mass. But a mass regularity composed wholly of many Sixes and aces will be thrown in ten thousand throws but we
individual idiosyncrasies could hardly stand as proof of some iron ne- arc absolutely unable to forctcl what any particular throw '\vill b,
cessity regulating the events whose diversity had been evaded by the very nor docs our certaii:ty as to the general result help us in the lea~~
act of averaging. Lord Acton, who opposed Buckle with the claim that degree to a conclusiOn as to the ])articular c.J 11 c CI'I11· 5 · . . 1
· , IS SUIC V an
history was really about individuals, and had nothing to do with statis- ex~ct parallel to the case of human action. We can forctel it~ ag-
tics, argued that "it is only to men as persons that free-will belongs: look gregate, but we cannot foretel its individual results. 49
at them in masses, and they become machines; with their personality
you abstract their freedom. "46 T. C. Sanders, more sympathetically, ob- . ~~ckle's counterposition of free will and statistics was less controver-
served that Buckle's statistical viewpoint was fully appropriate for social ~Ia m_ France and Belgiu,m, perhaps because his book attracted far less
history, but could not justify any conclusions about the freedom of in- ttenbon there. Quetelet s grand pronouncements on l'homme moyen
dividuals. 4 7 and physzque socwle had also been received more coollv in Fra , , tl .
m G . t B ·t · Q l . _ - nee 1an
These issues were treated more clearly and comprehensively by Fitz- . rea n am. ucte et hnnsclt was pleased at the publicity he re-
james Stephen, writing in the Edinburgh Review in 1958. According ceived from_ Buckle's book, and he quoted at length from the ~elevant
to Stephen, Buckle had erred in assuming that free will must act irreg- c~apter m lm book Physique sociale of I869. The statistical determinism
ularly, or that the role of divine providence must be to cause deviations did r_wt offend hu_n._ When John Herschel complained to him of the dis-
from natural laws. At the same time he asserted that laws of nature per- crtdit cast on stahsh~s by Buckl~'s unjustified remarks on free will, Que-.
tain only to the minds which conceive them, not to the "facts which :,e ct resp~nded placidly that lm good friend had analvzed these ideas
trop de rzgueur." 50 ' '
they are supposed to govern. "4 8 Most pertinently, he insisted that the
... [William Cyplcs], "Morality of the Doctrine of Averages," Com hill Magazine, to ( 1864),
, By bco~tr~st, in Germany, where Quetclet had been little noticed ex-
218-224, p. 224- cept . y specialist~ bc~ore 186o, Buckle's book elevated both men to
45 Anon., "Statistical Averages and Human Actions," Temple Bar, 1 5 ( t86s), 495-504.
prommence and mspircd a controversy over free will and statistics at
4 6 Lord Acton, "Mr. Buckle's Thesis and Method" ( •858), in William McNeill, cd., Essays
in the Liberal Interpretation of History (Chicago, 1967), 3-21, p. 8. least as vehement as that in Britain. Oddly enough, Buckle was tra,ns-
47 T. C. Sanders, "Buckle's History of Civilization in England," Fraser"s Magazine, 56 latcd by Arnold Ruge, the aging Young Hegelian, by then an exile i~1
(1857), 409-424; Sanders, Saturday Review, 4 (185;), 38-40.
4 8 [James Fitzjames Stephen], "Buckle's History of Civilization in England," Edinburgh Re-
;: Stcph,ct~, ',',Buckle's llis,tory" (n. ~8), p. 4 73-
view. 107 (1858), 465-512, p. 483. Sec also his "The Studv of Historv,'' Cornhi/IJVIagazine,
3 (t86t), 66G-68o; 4 (t86Ii, 2541, p. 27- Q8 uetdet, Nollcc sur Su John Frederic \Villiam Herschel ., AOB (8 )
pp. ' 5-186; also cahicr 12 89 , AOP. • ' '39 1 72, 153-197,
166
Statistical Law and Human Freedom
Statistical Law and Human Freedom
England, who had decided to make the translation and exposition of that no law can applv to the mass wl . h d
Buckle to his countrymen his life work. This implied no rejection of his dividuals. Wagner ~as nonetileles. 11~ ocs not also pertain to the in-
former commitments, he explained; Buckle's history captured the true dicated the existence of . l s persuaded that mass regularity in-
genume aws actmg 0 · d. ·d
spirit of Hegelian philosophy. Its author, he went on, should not be too accounted for their fai·] , t 1 , . n every m IV! ual, and he
ure o express t 1ems ] . ·f, l
harshly judged for his inability to comprehend the true concept of free- to the disturbing forces '"hi'cll fit
. · "
l le :es
o en cancc t 1e ff, ·t H b .
unJ orm y by pointing
dom, and it would be a mistake to speak "as if Buckle were a materialist, statisticians should not t . If e ec s. e eheved that
. res content With the de t . -
when he is really only an Englishman. "st Buckle's remarkable success ties, but decompose arld , . b. d . mons ration ot regulari-
. recom me ata d I
in Germany was likely due to the fortunate appearance of his book dur- likeness by finding actual . Tl fl m or cr to c emonstrate law-
d causes. ne uctu"lbon of .
ing the ascendancy of liberalism there-for which, according to one an marriages by season or w·th . . c . cnmcs, suicides,
· . 1 gram pnces seemed to vV • I
critic, it helped set the tone. Buckle was received with less enthusiasm statishcal argument against f d . f- I · : · agner tne Jest
by academic historians, like Droysen, who tended to find his book most ree om o t 1e w1ll tl J } 1
their annual regularity bv it· ·If .. , fifi.. , 10.ug 1 Jc t 10ught that
.!. , se was su Cient to re , ·t "th ! . l
noteworthy for its Dilettantism us and unphilosophical empiricism. s: a bI Ity the idea that thes ,, t ld . }ec WI ltg 1 prob-
Nevertheless, it was principally he who made statistics a philosophical 1 . e e\ en s cou be produced b f .11
t le will is inherently diverse h d. . y ree WI ' since
problem, in Germany as in England, and his history inspired a number . . . w en uncon J!IOncd by th
actmgmsociety. Wagneratt t·d I . econstantcauses
of scholars to go back and investigate the writings of Quctelct. a land where an autocrat d rae ed t 1e tmost a_ttcntwn with his fable of
ecrce at tne begm · . f
Among the earliest and most prominent of these was Adolph Wagner, number of marriages ii1 th, d·f:r nmg o every year the
e 1 1crent age gr , · ·d b
who was also among the few to express an unreservedly favorable opin- profession, and weapon aile·] c . . f ]] oups, SUlCI e y age, sex,
' • nrncs o a sort. tl ·
ion of Quetelet's and Buckle's contributions. Wagner was the son of a aspect of life was con troll d It . . !· . • s, so Jat VIrtually everv
holistically inclined Giittingen biologist, but though he dedicated his e . Is
h as t l1e power to accomplish . ll t! . o wwus he pointe 1 t tl ,
, c ou ' lat no state
book on statistics to his father, he preferred to interpret science in terms artificially through h . · ·~ll 115-yet what "could never be effected
uman WI anc1authorit · · , d1 .
of rigid natural laws. His field was economics, and as a young man he remarkable manner due to th , . . 1 y IS e~ecute JY Itself in a
embraced the deductive approach to that science along with free-market ety. "54 e natura orgamzatwn of human soci-
liberalism. Similarly, he took Mill's doctrine of universal causation as Dcspite his enthusiasm for Buckle and Quetcl '
the point of departure for statistics, observing that while this postulate firm conclusion as to the ·J. f . b. . et, \-v agner reached no
might seem self-evident "its significance immediately becomes apparent acknowledged the existenc~e afJonls ctween statistics and free will. He
. . o su JjectJve fcelmgs 0 f l
once we recollect that we so often explain our own actions and events in 1ty, and while he felt that Q t l t d · mora rcsponsibil-
our lives in a diametrically opposite manner."» . l ue c e an others ha I t 1
m t Jeir efforts to get beneath this a . , t no Jeen successful
However great his admiration of statistics, Wagner was unwilling to regularity of phenomcm W· 'ppal rei.1tlcontradJcbon with the lawlike
. '' agncr t lOtio Jt ., sue , , ·t- l ..
allow !hat the regularities revealed by statistics were themselves natural ITHgl.lt yet be possibl, Otl , . I . ~ ' , cc~s u reconcdJation
. lCIS
C · Fischer published a tra 't. . . . f.
1.:. WCJe CSS ClfClllllSj) t 'f'/
laws. He challenged the coherence of the law oflarge numbers, insisting . . · L.'C · Jc materialist].
.
bme-honored disproo£~ of h
< c agamst rcc wdlm 1 s~
. f- I
. 1 d'
' )9, lllc u mg all the
I' Arnold Rugc, "Ueber Heinrich Thomas Buckle, und zur zwcitcn Auf!agc" ( 1B64), in !I. .fi · · uman rcec om ·mel .1tt ·1 t.
T. Buckle, Ceschichte der Civilisation in England (2 vols., Leipzig, 7th cd., 1901), pp. xvi- m cance to the results of t· t. t. , . 'fl . ' ' .n JU mg special sig-
s a Is Jcs le arownc t tl t .
xvii. On Rugc's ambition to translate Buckle, sec the letters of another young Hegelian in Eng- ru led by laws even ·Js ,·t, , b . . . "' . • .n . Ja society can be
lish exile, Karl Marx to Friedrich Engels, 18 June 11i62, in Briefwechsel, ,\Jarx-Engels Geo- '· s mem ers rehm .1 .., t· ·
amtausgabe (MEGA), vol. 3, sec. Ill ("Ciashiittcn in Taunus." 1970). p. 78. he dismissed with conteinpt 0 . . !' , cer am measure of freedom
1 ' job. Drovscn, "Die Erhebung der Ceschichte zum Rang einer Wissenschaft," Histo-
· ne nug 1t as pia 1"bl
gregation of blind men , ld f, .' u_s Y suppose that a con-
rische Zeitschrift, 9 (1863), 1-22. On German liberalism, see James J. Sheehan, German Lib- ,4 . cou orrn a SOCiety With vision, or that atoms
eralism in the Nineteenth Century (Chicago, 1978); on Buckle and the German liberals, sec
Wagner, DleGesetzmiissigkeit in den scheinb . .. , '
G. F. Knapp, A us der /ugend eines deutschen Celehrten (Stuttgart, 1927), p. 155.
~om Standpunkt der Statistik (I lam burg, 1864), ar ~~l~k,~hrhcnen menschlichen Handlungen
"Adolph Wagner, "Statistik," in J. C. Bluntschli and K. Bratcr, cds., Deutsches Staats-
n pp. 63-80, and two rCVICWS he wrote for ZG$~ 4 Ce also his note 22 on p, 54. [he essav
Worterbuch ( 11 vols., Stuttgart and Leipzig, 1867), vol. 1 o, 400-481, p. 457· !89-203, p. 192. m the latter of which he ret . t , 21 ( 18165), 273-291, p. 28o, and 36 {I88o ,·
~ rca s sonlew lat fronl this position. , ,,
Statistical Law and Human Freedom (
Statistical Law and Human Freedom
free will, implying that a thief, for example, "is not free in reference to
arc unaffected by force even though macroscopic bodies follow rigidly his decision to steal, but only in whether to do so on horseback or 011
the laws of gravitational attraction." foot. " 57 Lotze held, .however, that crimes of the same legal category arc
No argument against free will was likely to win unreserved popular ?f very d1ffcrcnt eth1cal worth, so that the regularity obtained by lurnp-
acclaim, and the one based 011 statistics was no exception. Its influence mg together hundreds of such acts can indicate nothing about the meas-
was predominantly negative, for it was routinely considered, and then ure of evil in a society, and hence cannot be explained in terms of anv
rejected, in late nineteenth-century German discussions of free will and constancy of causes, but remains a great mystery. .
determinism. The argument was, to say the least, not without its Raws, This view, which would seem to deny that enumeration of moral acts
but the very imprecision of its logic served to hinder the emergence of l:a~ any value or legitimacy, was not widely adopted by German statis-
an accepted, sharply defined refutation. The most common opposing ticians. They did, however, insist on a fundamental distinction between
argument had already been implied in an odd passage by Quetelct, those who committed and those who abstained from crime. That is thev
when he noted that the enlightened will tended to combat all perturbing rejected Quetelct's ploy of converting crime from a series of discr:te in--
forces and thus to maintain a state of equilibrium more perfect even dividual acts into a continuous social phenomenon through the assign-
than that of nature. Wagner's title highlighted the inconsistency of sta- ment of penchants. The Austrian Leopold Neumann reflected that the
tistics with caprice (Willkiihr) rather than will (Wille). In the Kantian wide diversity of responses to reasonably uniform circumstances-the
tradition, the enlightened will was unambiguously a source of order, fact that some placed under given conditions commit crimes, and others
and the Leipzig philosopher Moritz Wilhelm Drobisch placed great em- do not-confirmed the existence of a large measure of personal auton-
phasis on this aspect in his influential book on moral statistics and hu- omy. 58 Drobisch held that criminal statistics do not reveal tendencies
man freedom. Gustav Schmoller maintained that the stability of the will
shared by a whole population, but only those of a smallminoritv so that
no ~n.ference is possible from an all-embracing average to th~ ~raits of
provided the only possible explanation for statistical regularity, si nee
material factors could never exert a causative influence in the domain mdividuals. He wrote: "~nly through a great failure of understanding
of spirit. Carl Giiring argued that the systematic variation of marriages
can the mathematical fiction of an average man ... be elaborated as if
with grain prices was no argmnent for the powerlessness of will, but the
all individuals ... possess a real part of whatever obtains for this average
opposite; it shows that human rationality can sometimes overcome blind person."5<J
impulse. Hermann Siebeck made the point that willful acts do not can-
cel to vield the results of moral statistics, but sum to them. 16
Th~t mass regularity need not imply causal necessity was also a wide- THE SCIENCE OF DIVERSITY
spread argument among German statistical critics, who developed it
much more fully than any other group One variant of this approach was
introduced by the philosopher Hermann Lotze, who sought to refute D.robisch': <:rgument against determinism, however, opens up a much
what he took to be the core of Buckle's position, that a fixed quantity of w1dcr stahstJcaiissue, the extent to vvhich statistics presupposes homo-
evil is present in society, and must somehow be expressed. If this con- . '" Hermann Lotze, Mikrokosmus: ldeen zur Naturgesclzichte und Geschicllte der 1\fen-
stai1CY of evil were genuine, Lotze thought, it would indeed contradict :chhezl: Versuch emer Anthropologie (3 vols .. Leipzig. t8s6-5S-tq), \'Ol. 3, p. -;8: also Grund-
;~ge der praktlschen Ph.llosophie: Diktate aus den .'iorlwmgen (Leipzig, 1 8S 4 i, pp. 24_2 S
11 J. C. Fischer, Die Freiheit des mensch lichen Willens und die Einheit der Naturgesetze l hts arg~m1cnt \vas repeated appwvmgly by Fncdnch Albert Lange, Geschichte des Materi-
!Leipzig, zd cd., 1871), pp. 235-236. ah~IT)us ,2 vols., lscrlolm. 1866), vol. 2, pp. 479-480. It was challenged on the ground that
<6 SeeM. W. Drobisch, Die moralische Statistik und die menschliche Willensfreiheit: Eine ~~v tatt~ns ofmo~al worth would ~~·<;rage out by Johannes Wahn, "Kritik dcr Lchrc Lotzcs von
Untersuchung (Leipzig. 1867); Gustav Schmollcr, "Die ncucrcn Ansichtcn tiber llcvolkc- cr ,mcnsd1hchcn Wahlfrcthclt, Zeztschnft /!ir l'hilosophie und philosophische Kritik
rungs- und Moralstatistik" (1869) in Zur Litteraturgeschichte der Staats- und Sozialwissen- (1888), 88-q1, pp. lllff. . 94
schaften (Leipzig, 1888), pp. 172-203; Carl Giiring, Ueber die menschliche Freiheit ~nd Zu- sH L. N. [Leopold Neumann/, "Zur Moralstatistik," Preussische Jahrbiicher 27 ( 1 871 · , "··
2
rechnungsfiihigkeit. Eine kritische l!ntersuchung (Leipzig, 18;6), p. 128; IIermann Stcbcck, 2 47, P: 245; also Rtchard Wahle, "Einc Vcrtcidigung dcr Willcnsfr .. 'l1c 1' t "ze' 1'ts' h. ·rt' /' -p 1'·
"Das Vcrhalh;iss des Einzclwillens zur Gesammthcit im Lichte dcr Moralstatistik," Jbb, 33 ·ph
/os~J dh'l h'hK'' "' .cn,,,url!-
ze un P z omp 1sc e rzt1k, 92 (1888), 1-64, pp. 35-36.
(1879), 347-370. Sec also C. Schaarschmidt, "Zur Widcrlcgung des Dcterminismns," Phi- ' 9 Droi)ISch, D1e moralische Statistik (n. s6), p. 18; also "Moralischc Statistik" (tl. 3<)).
losophische Monatsscllrift, 20 (1884), 193-218, pp. 213-214.
( Statistical Law and Human Freedom (
Statistical Law and Human Freedom
geneity in the phenomena counted. This issue was widely discussed, statisticians received a shock in 1828 when A. Balbi, A.M. Guerry and
and not exclusively in response to Buckle's claims or even in reference Benoiston de Chateauneuf showed that crime and education we;; not
to the matter of human freedom. As early as 1831, an appreciative re- at least on their face, inversely correlated, but that, on the contrary:
viewer of Quetelet's first papers on mecanique sociale cautioned that a areas where education was most widely distributed tended also to have
mean value taken by itself could not provide an adequate representation more crimes. Unwilling to forsake their campaign for public education,
of any given trait over a whole society. One must also know the ex- statisticians were obliged to reinterpret their science as a subtle and com-
tremes, the critic observed, pointing to the great moral and intellectual plex affair. One could no longer suppose, as had the devout educational
benefits bestowed on society as a result of the accumulation of wealth by reformer Lucas, that the influence of moral factors on crimes against
a few privileged individuals. The mean might suffice for most problems persons could be established directly by comparing a table giving the
in physics, but the most important questions of social mechanics re- state of civilization based on instruction and leisure with a table of the
quired more detailed knowledge. 60 pertinent crimes. Balbi, after learning of the preliminary results, de-
The coherence of the average man concept was also sharply chal- cided it was essential to consider a multitude of factors-extremes of
lenged during its early childhood. In 1843 Cournot pointed out that the wealth and poverty, density of population, the proximity ofborders or of
mean taken for each side from a great number of right triangles could in seacoast, cities, and the presence or absence of released criminals as
no way represent the type of a right triangle, since it almost certainly well as ignorance and superstition-before the numbers could be in~er­
would not be a right triangle a.t all. For the same reason, the mean of all ~reted. Guerry, similarly, examined a wide range of factors that might
organs and limbs for a particular species of animal would likely fail even mfluence the level of crime, and concluded somewhat vaguely that dif-
to be a viable organism. Hence l'homme moyen might well prove to be ferences of culture and mores were the most important factors. 64
simply l'homtne impossible. 6 • The Prussian pastor and astronomer The issue of crime and education inspired recognition that in statistics
J.W.H. Lehmann showed that the Schuss, or spurt of growth accom- the effects of given causes arc often masked, and can easily be confused
panying adolescence, disappeared completely when a curve of mean with other factors not explicitly considered. Alphonse De Candolle ar-
height according to age was plotted, since the Schuss occurred at differ- gued that the positive relationship between crime and instruction de-
ent ages in different men. 62 Quetelet did not accede to these objections. rived from no direct effect of the latter, but only reflected the tendency
He held up his subsequent demonstration of the applicability of the error for education to be most widespread in prosperous districts where diffe;-
law to humanity as proof that the average man was a genuine type, not- entials of wealth were most pronounced. He emphasized that in social
withstanding Cournot, and he answered Lehmann by asserting that the science, even more than in physics, it was essential to sort out the effects
growth spurt was no normal characteristic of human growth, but merely of perturbing causes before drawing conclusions. 6 ; In England, Charles
an artifact of contemporary civilization. 6 > That these rebuttals per- Morgan and others found reason to dismiss Guerry's results as a kind of
suaded anyone, however, is unclear. false correlation, inadequate to justify a firm conciusion on the relation
The value and usefulness of averages was challenged in a more prac- of crime to education. 66 Quetelet, too, accepted De Candolle's argu-
tical way by the uncongenial results that too frequently presented them- 64 A. Balbi and A. M. Guerry, Statistique comparee de l'etat de /'instruction et du nombre
selves as the outcomes of direct and simple-minded studies. Reformist des crzmes dans les divers Arrondissements des Academies et des Cours de France (Paris, 18 28 ),
~large pnntedtablc; M. Lucas, "Influence morale de !'instruction ct de Ia civilisation en gen-
6o Sec Bulletin universe/ des sciences et de l'industrie, 6th sec., Bulletin des sciences geo- eral sur Ia dnmnutwn des delits ct des crimes," Bulletin universe/ des sciences et de l'industrie
graphiques, 28 (1831), 113-136. 6th sec.: B,ulletm des sc~ences geographiques, 15 ( 1828), 105-1 17; also E. Jlcreau's review of·
6• Cournot, p. 214. the ~alb1-Guerry chart, rbrd., 16 (1829), 6-10; A. Balbi, "Rapport du nombre des crimes id'ctat
6' Jacob Wilhelm Heinrich Lehmann, "Bemerkungen bei Gelegenheit der Abhandlung de I.mstruchon pubhque en France," ibid., 20 (1829), 252-264; A.M. Guerry, Essai sur Ia
von Quetclet: Uber den Menschen und die Gesetze seiner Entwicklung, in diesem Jahrbuchc, statrstlque morale de Ia France (Paris, 1833), p. 4o.
65 Alphonse De Candolle, "Considerations sur Ia statistique des dclits," Bibliotheque uni-
Jahrgang 1839," in H. C. Schumacher, ed., fahrbuch {iir 1841, 137-219, p. 139; fahrbuch fiir
ve~;lle des scrences, belles-lettres et arts, 104 ( 1830), Litterature, pp. 159 -186.
!843. 146-230, pp. 146-147·
6, Quetelet, Du systeme social et des lois qui le regissent (Paris, 1848), pp. 23ff. Charles Morgan, review of Guerry in Athenaeum, no. 303 (1833).

173
(
Statistical Law and Human Freedom
( Statistical Law and Human Freedom (
each separate event comprised in the statistics has its invariable and un-
ment, though he added with his customary retrospective prescience that
conditional antecedents. "7a
he had never expected the alleviation of crime to result from mere in-
~ enn, we may. re~a~l, held that probability statements apply only to
struction, but only from moral education. 6~ The bewildering complex-
senes and not to md1v1dual events. The use of probability presupposes
ity of statistics was invariably rediscovered whenever a result arose that
that the eve~ts can. be reg.ard~d as homogeneous for certain purposes,
contradicted expectations or prejudices. The relatively high suicide rate
and Venn reJected 1ts apphcatron to testimonies, judicial decisions and
some aspects of medic.in~ .on the ground that too much is gen;rally
among Protestants troubled German scholars later in the nineteenth
century, and writers like the Gottingen economist Helferich found rea-
known about the pecuhanhes of the individual cases for an assumption
son to suppose that Protestantism could not itself be a cause of suicide,
of ~?m.ogen.eity to be acceptable. Venn understood that the use of prob-
but rather was entangled with other factors. 68 Problems of heterogeneity
~bd1ty 1mph~s ~Is~ a measure of ignorance about individuals, proceed-
were invoked by William Lucas Sargant to criticize Buckle's conclu-
~~g from theu mtncate and complex differences. He held that in statis-
sions on crime and free will, and by others to point out the deficiencies
tics-though not, for practical purposes, in games of chance-there is
of sanitary statistics. 69 also irregularity. over the very long run, reflecting the gradual evolution
A coherent tradition of critical thinking about the character of statis-
of customs, behefs, and laws. Probability, then, applies to that limited
tics only developed in the wake of Buckle's discomfitting pronounce- extent of order that characterizes in an imperfect way the middle term. 7'
ments. It was predominantly German; the French showed a minimum These ~~ss regularities constituted, for Venn, the fundamental data
of interest, and the British, being less professionalized, tended to be o~ pro.ba~1hty .. He opposed Quetelet's view that they provided a suffi-
more impulsive and to consign their contributions to the relatively Cien.t Jushficatwn for inferring the existence of an average man, repre-
ephemeral critical reviews. This was, however, by no means upiversally sentmg the type of a race or people. This error, he thought embodied
the case. The remarks of W. S. Jevons and, especially, John Venn on a deep subjectivism, and was analogous to the defective arg~mentation
this issue were studied and cited for years afterward. o~ ~aplace a~d ~e ~?rgan, for it assumed a fixed, pre-existent proba-
Venn's Logic of Chance, which first appeared in 1866, had received b1hty ~egulatmg md1v1dual trials, from which laws of the mass were to
a considerable impetus from Buckle's History. Venn was no militant on ~e denved. He wrote: "We are concerned only with averages, or with the
free will. He confessed himself sympathetic to a limited doctrine of ne- smgl~ event as deduced from an average and conceived to form one of
cessity, which allowed a prominent role for the will but insisted that a senes. VY_e start with the assumption, grounded on experience, that
identical antecedents must produce the very same consequents. On the there IS umformity in this average, and, so long as this is secured to us,
other hand, he rejected as a dangerous fallacy the doctrine, which he we can. aff?r? to be perfectly indifferent to the fate, as regards causation,
labeled fatalism, that events occur quite independently of men's mo- of .the md1V1duals which compose the average. "7• That is, probability
tives, and he categorically denied that even his less extreme version of ne1th.er. presupposes nor demonstrates causality.
philosophical necessity received the slightest measure of support from Wdl~am Stanl~y Jevons stood diametrically opposite Venn on the
the existence of regularities in statistics. Universal causality could bees- pr?pe~ mterpretabon of probability, but concurred with him on the im-
tablished only from systematic covariation, not from broad uniformities; phcah~ns. of mass regularity for the understanding of individual events
"no amount of regularity seems to me to bring us nearer to proving that or pred1cho~ of the future. Jevons is best known to posterity as a pioneer
o~ ~he margm~l approach to economics and, to a lesser extent, as a lo-
67 Quetelet, "Recherches sur le penchant au crime aux differens ages," NMAB, 7 (1832),

separate pagination, pp. 26ff., esp. pp. 43-44. gician .an? philosopher of science, but he began his career as a mete-
6b Helferich, review of Wagner in Gottingische gelehrte Anzeigen, 1865, vol. 1, 486-506, orologist m Australia and he took a leading role in the Manchester Sta-
p. 500.
"" William Lucas Sargant, "Lies of Statistics," in Essays of a Birmingham Manufacturer (4
7~ John Venn, The Logic of Chance (London and Cambridge, Eng., 1866) p. 33 .
vols., London, 1869), vol. 2, 56-182, pp. 166-167; Anon., "The Fallacies and Shortcomings 7 lbzd., pp. 20-22, 2 34- 23 6. ' 5
of our Sanitary Statistics," The Social Science Review and fournal of the Sciences, n.s. 4 (t86s), 7' Ibid., p. 330; sec alsop. 43 .
234-250, 358-363, 403-414, 481-495; 5 (1866), 21-43, 310-321, 436-447; 7 (t866), 98-110.

174 175
Statistical Law and Human Freedom Statistical Law and 11 uman Freedom

tistical Society during the 186os and 187os. Like Venn, he received a and of society, are always complicated, so that "laws and explanations
considerable impetus from Buckle, and he discussed the nature of sta- are in a certain sense hypothetical, and apply exactly to nothing which
tistics at some length in the context of his wide-ranging remarks on prob- we can know to exist. "79 Science proceeds through abstraction, by sub-
ability. Jevons set out from a classical view of probability derived. from stituting imaginary objects for those subsisting in the real world. Hence
the instruction of De Morgan and the writings of Laplace and Po~s.so~. "those who speak of the uniformity of nature, and the reign of law, mis-
Our knowledge is always uncertain, he explained, so that. probab1hty IS interpret the meaning involved in those expressions. Law is not incon-
essential for scientific investigation. "Events come out hke balls from sistent with extreme diversity. "8o
the vast ballot-box of nature,"?> he remarked, and indeed the world Like Laplace, Quctelet, and De Morgan, Jevons thought the concept
might have been governed by chance. "Happily the Universe in whic~ of pure, irreducible chance unthinkable. He was, however, opposed to
we dwell is not the result of chance, and where chance seems to work It the idea that the events of the world are regulated in all their details by
is our own deficient faculties which prevents us from recognizing the blind, purposeless laws of the mechanical sort. He noted that by varying
operation of Law and Design. "74 Jevons marveled that natural order ex- the initial conditions, the same laws could be made to yield quite dif-
tended even to errors and deviations, a result he attnbuted to Quetelet. ferent effects, observing that on this account one could still think in
This remarkable order provided assurance that the same principles of terms of purpose and design even after the triumph of Darwinism. Je-
scientific method which had proved so fruitful in the study of nature vons ridiculed Comte and Buckle, arguing that the inaccessibility of
could be applied with like success to man. "Litt~e allus.ion need be made laws of individuals ruled out the discovery of determinate social laws,
in this work to the fact that man in his economic, samtary, mtellectual, and he most emphatically rejected the possibility of a science of history,
aesthetic or moral relations may become the subject of sciences, the maintaining that society was unstable, and that the slightest disturban~e
highest a;1d most useful of all sciences~ Every one who is engaged in sta- could magnify itself until everything was altered. With regard to free
tistical inquiry must acknowledge the possibility of natural laws govern- will, moral responsibility, and the efficacy of prayer, Jevons denied they
ing such statistical facts. "75 . . .
could be contradicted by statistics. "Laws of nature are uniformities ob-
Despite his unqualified acceptance of the subjective log1~ of pro~a- served to exist in the action of certain material agents, but it is logically
bility, and not obviously reconcilable with it, Jevons empl:~SJZed them- impossible to show that all other agents must behave as these do. The
terest of natural variety and the appropriateness of probability theory for too exclusive study of particular branches of physical science seems to
c d. . "76 h generate an over-confident and dogmatic spirit. " 8 • Science should re-
dealing with it. "Number is but another name wr Iversity, e an-
nounced, and he rejected the "fallacious impression . : . that the th~~ry spect the teeming diversity of the world, and recognize its own imper-
of probabilities necessitates uniformity in the happemng of events, as fection. Jevons wished to make clear "that atheism and materialism are
Buckle had "superficially" believed. nOn the contrary, he argued, the no necessary results of scientific method.'%
theory of probability positively mandates runs of luck of a limited exten.t.
Jevons remarked that in statistics "quantities which are called errors ~n
one case, may really be most important and interesting phenomena 111 Statistik: BETWEEN NATURE AND H!STOHY

another investigation. When we speak of eliminating error we really


mean disentangling the complicated phenomena of nature." 78 He e.m- For practitioners of the academicized German version of statistics, the
phasized repeatedly that real events, especially the phenomena of mmd idea of a comprehensive mean value uniting the whir of appearances
into a lawlike totality had very limited appeal. The only prominent ex-
7' Jcvons, Principles (part 3, n. 1), p. 239; see alsop. 149.
74 Ibid., p. 2
o; Ibid., p. 3 34· See also preface and pp. 3 59ff., csp. P· 374·
79 Ibid., p. 458.
So Ibid., p. 750.
o6 Ibid., p. 1 56.
,, Ibid., p. 65 5·
8' Ibid., p. 737; sec also pp. 759-761.764-765.
8' Ibid., p. 766.
' 0 !bid., p. 3 39-

1 77
Statistical Law and Human Freedom Statistical Law and Human Freedom
ception to this generalization is Alexander von Oe~tingen, _the conser-
moyen) of a nation. " 88 Wappaus, however, was a statistician of the old
vative Lutheran theologian and statistician, whose mflue_nhal and ;uc-
school, writing in 1859 when a new tradition of university statistics
cessful tome on moral statistics and social ethics appeared m 1866. Even
based on numbers and dedicated to the solution of social problems was
Oettingen did not subscribe to this vi~w unreservedly, :~r _h~ w~s deep~y
just beginning to take hold. The new generation, educated principally
concerned aIJou t the alleged
inconsistency between statistics and free
· ·. . . . . fl, Oct- in history though partly in economics, was motivated by an entirely dif-
will, and hence was reluctant to speak wtthout quahfi~ahon o a_w .. ,.d-
ferent set of concerns. They were far more closely attuned to issues of
tingen sought to find the middle way between co~lcctiVIsm and ~~dnl
methodology, and had definite preconceptions about the character of
ualism. On the one hand, he opposed what he viewed as the soctal de-
social science. Buckle's book appeared at just the right time to form a
. · . 0 f H'dolnh
tef!11111!Snl r
Wagner and Quetelet, .
and even· ·
persuaded
} .
key part of their collective education. Still more important, he directed
Wagner during the latter's brief tenure at Dorpat th~t lm Ideas on tm
them to Quetelet, whose interpretation of statistical science seemed to
matter were " exaggera· ted . "83 At the ·same time he rqected the mdl\!19- I
ualism of his critic, the criminologist Emil Wilhe_lm ~ahl~erg, _and ~e them at once inspiring and deeply flawed. Quetclet's exaltation of mean
dismissed Drobisch as a Pelagian and ''proud phansee . for his claim t~1at values and dismissal of variation appeared to them as a central deficiency
of his approach.
most individuals were uninfluenced by the urge to crune. 84 The ~mon
of free individuals into a higher collective whole seemed to Oetbnge~1 The idea of a science built on the regularities of mean values stirred
essential to the very concept of a society; he even 1mputed Germany s opposition for a variety of reasons. One objection, whose appeal was
victorv in the Franco-Prussian War to the contrast betw~en her deep wide both within and outside of Germany, simply took note of the cir-
sense .of community and French atomism. S; Octtmgen'~ am~ ':as to ex- cumstance that the regularities of statistics were not causal. In Ger-
emplify this· social dimension, not to find the causes of mcltvtdual h,c~ many, the Tiibingcn statistician Fallati and the prominent Berlin pub-
havior and for this purpose there was no need to break t~e nu1~1bus lic-health writer J. L. Casper both remarked as early as the 184os that
down. ',rhe regularity of mean values was sufficiently inspmng by 1tself; the observed constancy of statistical aggregates could not by itself justify
"if all individuals were free to conduct their lives as autonomous and un- the term "law," since there was no basis for ccrtaintv that the same reg-
fettered selves, how could this constancy in the ethical act1v~ty of the ularity would persist into the future. ''1 The most compelling evidence of
whole have arisen and how could it be explained?" 86 Still, C~ct~mgen d~­ this deficiency was the impossibility of applying those so-called laws to
nied that this regularity was "so undeviating, that we can mfer from It individuals. A critic of Adolph Wagner's hook observed that a law must
the existence of a necessary law of nature. "87 provide a determinate and necessary relation between cause and effect,
The Gottingen professor]. E. Wappaus is perhaps also to be excluded and hence must apply not only to an abstract whole, but also to the parts.
from the general claim that German statisticians were comparatively Since statistical regularities do not prevail for small numbers, they can-
unimpressed by the virtues of wide mean values. He annom~ced that the not apply with necessity to large numbers either. Hence the law of large
truths of statistics "arc only valid for the totality of a population, .consid- numbers is a misnomer, and "statistical laws" arc at best numerically ex-
ered as a whole, or, as Quetelet puts it, for the average man (l homme pressed properties of lands or nations. ()o
/From the 185os, German numerical statisticians stressed the search
''On Wagner, sec Oettingcn, Die Moralstat~s!ik, in ihrer Bedeutung {iir eine Socialethik
(Er'·-n en ,dcd 188o) p. 19·andWagncr,ZCS\V, 36(188o), 159-203,p. 192. 1
for systematic covariation, and not mere regularities, as the primary aim
',~" ATcx~ndcr \"~;n Octti,ngcn, lJie Moralstatistik und die Christliche Siit/nle~;ed\ e~­
.
e;suc?·
ner Soci~lethik aufempirischer Crundlage, vol. 2, Die Christliche S1tten e Jre: e u~ 1ve '- 11 - ""j. E. Wappiius, Allgemeine Beviilkerungsstatistik: Vorlesungen (z vols., Leipzig, 1859),
wicklun der Cesetze Christ/ichen lleilslebens im Orgamsnws d.er Menschhe1t (Erl.angen, vol. 1, p. 17. Franz Vorlandcr also argued that statistical principles apply only to mean values;
s~ ) g \Vahlberg pp. 8-9 on Drobisch p. 33: see also Wahlberg s reVIeW of Oettmgtn, Dw see "Die moralische Statistik und die sittlichc Freiheit,·· ZCS\V, 22 ( 1866), 477-511, p. 4b3.
9
.. ~ Fal!ati, Einleitung in die Wissenschaft der Statistik (Tiibingcn, 1843), p. 54;). L. Casper,
.~~:als~~~istik in ihr~r Bedeu,tung (n 83), (1866]_1st ed .. in Z~SW, 26 (dl}o), 567-576.
\n.
'' Octtingen, l'VIoralstatistik und Christliche Slttenlehre 84), p. >· Uber die wahrscheinliche Lebensdauer des Menschen (Berlin, 184 3), pp. 29-30. I treat the Ger-
man discussion of statistical heterogeneity in greater detail in ""Lawless Society: Social Science
s6 Octtingcn, Die 1\lloralstatistik in ihrer Bedeutung (n. 82). P· 37·. .
88
sc Octtingcn, Ueberakuten und chronischen Selbstmord: E111 Ze1tb1/d (Dorpat, 1 1I. P· 12 · and the Reinterpretation of Statistics in Germany, 185o-188o," in Prob Rev.
''
0
Anonymous review in Jbb, 4 ( 1R6sl. 286--;o1.
178
179
Statistical Law and Human Freedom Statistical Law and Human Freedom
of their science. Ernst Engel, founder of the Berlin statistical seminar in cialist and barbaric communist doctrines and intrigues"94 of the time
which so many German statisticians received their training, and .to a~l was forceably brought to public attention by the events of 1848, and
appearances one of the most fervent admirer.s .of Quetelet, argued m lm then purged of its radical implications by the Heidelberg political sci-
first essay on statistics, in 1851, that regulantles amounted only to em-
entist Robert von Mohl, who substituted for "the false social doctrine"
pirical l~ws, not causal ones. 9' Twenty-five years later: in .his iloge for
of class conflict a "true science of society" based on organic diversity and
Quctelet for the International Statistical Congress meetmg 1!1 Budapest,
"communities of interest. "95 The explosive growth of industry begin-
Engel announced that their leader's infatuation with mean values had
ning in the 185os produced disruptions that made the worker question
led the congress in the wrong direction, and that "not the largest, but
seem all the more pressing. Numerical study of society was undertaken
relatively local averages constitute what is really worth knowing. " 92
in large measure to facilitate the discovery and implementation of a
Wagner. aimed his book at the discovery of partial causes through .the moderate solution to it.
presentation of tables giving suicide rates in relation to every p~ss.lble
Virtually all of the influential German statisticians of the 186os and
variable-climate, weather, time of day and year, sex, age, religiOn,
profession, education, political and economic conditions, and so forth.
187os were or became members of the Verein fur Sozialpolitik, which
was formed in 1871. Most, including Engel and Schmoller, favoreda
He felt that suicide would be explained, or reduced to law, when all
variation over time and place was in this way accounted for. That the solution to the problems of workers through the organization of volun-
search for causes of variation constituted the highest aim of statistics was tary cooperatives, 96 although Adolph Wagner, who renounced his free-
similarly endorsed by Bruno Hildebrand, Adolf Held, Leopold Neu- market orientation in a "Damascus experience" about 1870, champi-
mann, and Gustav Schmoller, among others. 93 oned state socialism and became increasingly influential after Bismarck's
These practical reasons for stressing the study of variation, however, change of economic policy in 1878.97 In any event, all these "academic
were supported by, and ofter! subordinate to, an organic conception of socialists" (Kathedersozialisten) were opposed to rigid laissez-faire eco-
state and society widely shared among academic liberal reformers who nomics, and especially to the supposition that economic phenomena
opposed both socialism and laissez-faire liberalism. Statisticians of th.e are wholly governed by natural laws, before which governments are
generation that came to maturity during the 186os derived much of thelf powerless. Schmoller, Lujo Brentano, and G. F. Knapp, among others,
philosophical outlook from the German idealist tradition; ~ut to the ex- were greatly vexed by the so-called "iron law of wages," which held that
tent they adopted the viewpoint of their predecessors ~hey ~1d so ~ecause the "wage fund" in any economy at a given time was fixed, so that col-
it seemed to answer the needs of their own day. Social science m Ger- lective bargaining could do no more than shift wages from one class of
many was closely bound to the so-called worker question. The idea that workers to another. 98 The Sozialpolitikers were persuaded that this so-
society was a dynamic and powerful entity, requiring a science of its own 94
Anon., "Ncuere deutsche Leistungcn auf dern Gebiete dcr Staatswisscnschaften ··
which the state could ignore at its peril, had been introduced to Ger- Deutsche Vierteliahrsschrift (1854), no. 3. 1-78, p. 12. '
"'Robert von Mohl, "Gescllschafts-Wisscnschaftcn und Staats-Wissenschaften ·· ZGSW
many by the radical Lorenz von Stein in an 1843 study of socialism and 7 (1851), p. 25 and passim. See also Erich Angcrmann, Robert von Moh/, 1799 - 1 7 ,. . Lebe~
8
communism in France. This one beneficial result of the "ill-formed so- und Werk eJ~esalthberalen Staatsgelehrten (Ncuwicd, 1962), and Eckart Pankokc. Sociale
Bewegung-Socia/e Frage-Sociale Politik: Grund/ragen der deutschen "Socialwissenscha(t"
zm 19 Jahrhundert (Stuttgart, 1970).
9'Ernst Engel, "Mein Standpunkt der F'rage gcgeniiber ~b die ~tatistik cine selbstandige 96
Sec Gustav Schmoller, "Die Arbeiterfragc.'" Preussische fahrbiJcher, 14 (1864), 393- 4 2 4 •
Wisscnschaft oder nur cine Methode sci" ( 18 51), repr. 111 l~ngel. Das stal!shsche Se~nmar 523-547; 15 (1865), 32-63.
und das Studium der Statistik tiberhaupt," Zeitschrift des kiinzgl. preusszschen statzstzsc.?en 97
Sec Wagner reviews inZG~W, 34 (1878), 199-233, p. 211, and Lujo Brcntano, Mein
Bureaus, 11 (1871). 188-194, p. 189. On Engel sec Ian llacktng. "PmSS!an Numbers. m
Le~e~ 1m Kampf um.?Ie sozzale En.~wicklung Deutschlands (Jcna. 1931 ). pp. 63-76.
Sec ~chmollcr, Arbe1tcrfrage (n. 96), p. 413; )ames Sheehan. The Career ofLujo Bren-
Prob Rev.
9' Engel, "L.A.). Quctelet: Ein Gcdachtnisrede," ibid .. 16 (1876), 207-220, p. 217.
tano: A Study o(Lzberahsm and Social Reform in Imperial Germany (Chicago, 1 966), esp. p.
o' Wagner. Gesetzmiissigkeit (n. 4), part z; Brun,o Hildebrand. "D~c W!sscnschafthchc Auf-
gabe der Statistik." Jbb. 6 (1866), 1-11; Adolf Held srev1ew of Queklcl 111 fbb, 14 ( 1870), 81-
~ 1. Even Jevons, Enghsh founder of the mathematical and deductive approach to econom-
lcs-:-margmallsm-callcd in 1876 for a new branch of political and statistical science to in-
95; Leopold Neumann, "Zur Moralstatistik" (n. 58); Gustav Schmoller, "D1e ncucren Anslch-
vestigate the limitations of laissez faire. See T. W. Hutchison, "Economists and Economic
ten" (n. s6).
Pohcy m Bntam after 187o." History of Political Economy, 1 (1969), 231-255, p. 2 36.
(
Statistical Law and Human Freedom Statistical Law and Human FreedJ1 ..
called law was inconsistent with empirical evidence of an actu~l rise in head of the statistical-topographical office primarily because the post be-
wages, that it had survived because it supported the presumed .mterests came open at a convenient time. He was two decades older than most
of myopic manufacturers, and that in the end adher~nce to 1t would German statisticians of this new generation, and was virtually the only
only promote a more radical and violent process of soc1~l ch~nge .. one without formal academic training in statistics or some other branch
Classical economics, in fact, was rejected almost entlr~ly m Wt!hel- of Staatswissenschaft (Ernst Engel, a graduate of the mining school at
mian Germany. It seemed to sanctify greed as the expressiOn of n~tural Freiberg, was also an exception on both counts). Nevertheless, he was
law and to deny the possibility of conscious reform. It made soctety a a notable success as a statistician, and wrote two highly influential pa-
machine rather than an organism, an aggregation rather than a .com- pers on the theory of statistics while an administrator that justified his
munity, and though it was conceded to have introduced economic dy- assumption of a university post at Tiibingen in 1867. His writings, even
namism by liberating the energy of individuals, t~e free ma.rket was seen more than Wagner's, formed the point of departure for the "younger
as having since degenerated to financial speculatw~ and mmdless Man- historical school" in statistics, much as the "older historical school" of
chestertum. According to the historical view, the ltberal hypertrophy of Wilhelm Roscher, Bruno Hildebrand, and Carl Knies defined their ori-
atomistic individualism and rigid deduction embodied traits character- entation on economic matters. 100
istic of England and France. The time had come to subordinate these In his first paper on statistics, written in 1863, Riimelin addressed
to the essentially German ideas of community and history. Ger~an eco- himself to the longstanding debate in Germany as to whether statistics
nomics in the late nineteenth century meant historical econom1cs, and was properly a science or a method. His answer, conveniently, was that
the new enthusiasm for numerical statistics reflected a commitment to it was both, although he made the methodological definition primary.
empiricism rather than abstract deduction as well as a heighte~~d con- Statistics, Rumelin argued, is intrinsically a technique for the observa-
cern about social unrest. The German reinterpretation of stahshcs was tion and study of "mass phenomena." As such, it is applicable precisely
parallel to that of economics. 99 The historical ec?nomists ?pposed the to those composite phenomena made up of thoroughly heterogeneous
idea of statistical law for the same reasons they obJected to timeless ~at­ individuals. Invoking a distinction set out by the French statistician Du-
urallaws of economics. The physics of the average man defined ~oc1e~ fau in 1840, 101 Ri.imelin explained that in the sciences of nature, the in-
in terms of the similarities of individuals, whereas the key to soctal SCI- dividual is wholly or largely typical, so that a single well-recorded fact is
ence was seen by historicist Germans as the harmonious interaction~ of sufficient to justify an induction. Society, by contrast, is the domain of
diverse social groups. Quetelet's statistical regularities offered a ~ub]~ct diversity, and each person, though governed by law, is yet subject to so
matter which could justify the establishment of statistics as a umveiSity many perturbing causes that the action of particular laws cannot be in-
discipline, but only if statistics were redefined to accord with acceptable ferred from single cases. This dichotomy, he explained, is really a matter
German social and political ideas. . .. . of degree; animals are less uniform than plants, people than apes, mod-
That redefinition began with the work of Gustav Rumeltn. Rumeltn erns than ancients, adults than children, Caucasians than Negroes,
was trained at Tubingen in theology, and served as a pastor and then men than women, and the educated than the unschooled. The gulf be-
teacher before his political activity led to his election as a Frankfurt par- tween man and nature, however, was so wide that the human sciences
liamentarian. Subsequently he became director of the department. of "could never have raised themselves above infancy, where they have
churches and schools in the Wurttemberg ministry of culture, dunng been for some generations and where they in part remain, if there were
which time he presided over a controversy regarding government trea~­ no means of observation through which the inadequacy of individual-
ment of Catholic schools that amounted to a minor Kulturkampf. H1s
' 00 See the biography by Ri.imelin's much younger brother-in-law Gustav Schmoller in
position had already become untenable by 1861, and he was appointed Charakterbilder (Munich and Leipzig, 1913), pp. 140-188. On the older historical school 'see
Gottfried Eisermann, Die Grundlagen des Historismus in der deutschen Nationaltikon~mie
oo On the historical school, see Pankoke, Sociale Bewegung(n. 95); Sheehan, Brentano (~ 1 · (Stuttgart, 1956).
00 ' P. A. Dufau, Traite de statistique, ou theorie de !'etude des lois d'apres /esquelles se de-
9 8); Ulla G. Schafer, Historische Nationaltikonomie und Sozwlstatzstzk als GesellschaftswJs-
senschaften (Cologne and Vienna, 1971). veloppent les faits sociaux (Paris, 1840), p. 24.
D!l!l_.lllii!.Ti!ll._ _.....,_,.___ .__ , ______ .. _ __

Statistical Law and Human Freedom


Statistical Law and Human Freedom
"law of large numbers" reflect a fundamental misunderstanding. Hear-
ized and idiosyncratic experience can be alleviated, and our experience gu?d m 1867 that the term "law" should be reserved for expressing are-
graspe d as a who '! e. "w·~ . . . latton of causeand effect that is elementary, constant, and recognizable
Although statistics by its very nature dealt w1th diVerse objects,. and m every case. The peculiar function of statistics, in contrast, is to facil-
hence could not reduce the phenomena to a few simple laws, neither Itate the acquisition of knowledge about domains like society where a
should the statistician rest content with a general impression of the complex multitude of forces acts simultaneously. Statistic;! analysis
whole. The lawlikeness of statistical aggregates, shown by Quetclet, was m1ght lead to laws, but once the law is found, the need for statistics must
merely a starting point. "In the end," Riimelin wrote, "the intere~t .of v~nish, since every case can then be explained. A genuine law of sex at
moral statistics lies not at all in the demonstration of these regulantles bnth, for exa~ple, would not belong to statistics but to physiology, and
in willful acts of man, but rather in the perpetual movement and alter-
would apply w1th certainty to every individual case. ' 0 7
ation that these numbers undergo. "' 0 3 Indeed, statistical uniformity is
Althougl~ Rum~lin r~mained a wholehearted proponent of statistics,
mostly superficial, attained only when one considers a great ~ass and he became mcreasmgly 1m pressed in subsequent years by its limitations.
smears together the variety of the phenomena. An average by 1tself, he In a lecture on the laws of history, delivered ip 1878, he reflected that
observed, is completely inadequate, for a given mean of wealth, hOL~smg nearly two decades of intensive statistical research had yielded him noth-
space, age of death, or price of grain could ari~e fro.m completely differ- mg. that could properly be called a social law. For this reason, he ex-
ent distributions. 104 More significantly, the dlspos!hon to emphasize a plamed, he had been led to consider how different in character physical
single aggregate figure from a large, miscellaneous population mt~st in- phenomena were from psychical, and how unreasonable to suppose that
hibit rather than promote the search for causes. Hence, accordmg to
~he same concept of law applied to each. He had decided that where the
Ri..imelin, the proper statistical procedure is to fracture the populat1on
1dea of freedom is involved, the method and ideal of knowledge must
into tiny pieces, and then regroup these in various ~v~ys .. For example, change. 108 In view of these considerations, the uncertainty of statistical
by-finding the conditions under which crime and su1c1dc mcrcase or d.e-
knowledge ceased to be a defect, as it had seemed to the F;ench positiv-
crease, and the subgroups for which they are most commo~, the statl~­
~sts: and was transformed into a virtue, an accurate reflection of the real-
tician goes beyond the mere assertion of regularity and attams proposi-
Ity 1t. was meant to describe. Riimelin became persuaded that statistics
tions of real worth and interest. prov1dcd the appropriate method for dealing with the collective behavior
Even so the tendencies unveiled through this procedure would per-
of high!~ diverse. individuals-social science-precisely because it did
tain not t~ any particular individual, but to "the collective whole of
not requne the d1scovery of fixed and timeless laws .
greater or lesser groups of persons or processes. "•o> Like. Drobisch, he
. Ri.imelin's definition of statistics as a method of mass observation and
lm characterization of its principal object, society, in terms of f~nda­
denied that statistical conclusions could be used as a baSIS to mfer psy-
chological states; "What I say of the forest docs not hold ~or the se~aratc
mental d~versity won wide and almost immediate acceptance in the Ger-
trces."106 Because statistical generalizations do not perm1t pred1ctton of
man statistical community. Georg Mayr of the Universitv of Munich
and Bavarian stat.istical office held that statistics was the ~nly suitable
individual cases, Rumelin held that phrases like "statistical law" and
'"' Gustav Rcimelin, "Zur Thcoric dcr Statistik, I" (1863), in !\eden und Aufsdtze (Frci- meth~d for studymg human communities, and Etienne Laspeyres ex-
burg, 1875), 208-264, pp. 218-219. . ., . , . . . _ alted 1t as a new category of scientific inference, the key to research in
•c> Rcimelin, "Moralstatlsttk und WillcnsfrClOC1!, mtbtd., 370-37;. P· 375· ,
.a• Rumelin, "Zahl und Arten der Haushaltungcn in Wlirtten:bcrg nach dcm Stand der tho~e .areas where "all. other things" cannot be set equal. :oq In this guise,
Zahlun vom 3 Dec. 1864," in Wiirttembergische Jahrbiicher fur Stattsttk und Lande.skunde, statistics won a promment place in logic textbooks, such as Christoph
Jg. 86~. 16 2 - 217 , pp. 173 , 185; "Ergebnisse der Zahlung der ortsanwesendc1; Bevolkerung
1
nach dern Stande vom 3 December 1867," in ibid., Jg. 1867, 174-n6;P· 192. 1 he sarne,pomt '"7 Rumclin, "Ueber den Begriff cines so.cialen Gcsetzes" (186-.) ibid _3 8
was made by Fr. J. Neumann, "Unsere Kenntniss von den soCialen Zustandc um uns, Jbb, w8 Riii j' "U b G d G l h I , ., 1 l, pp. -10,
, ... ne m, e cr cscze er esc uc tc," Reden und Au{sdt;::e. Neue Folge (Frcibur
18 (1872), 279-341, p. 288. . ·· h · k · and I ubmgen, 1881), 11 8-'48, p. 139 . g
.o, Rumelin, "Statistik," in Gustav Schonberg, cd., Handbuch der pull lise en Oe ononl!e, wo Georg Mayr,. Die Geset~mdss;¥keit im Gesellschaftsleben: Statistische Studien (Munich,
1
Finanzwissenschaft und Verwaltungslehre (3 vols., 1\ibingcn, 3d ed., 189 ), vol. 3, 8o3-8n, 18
77), part 1, E!Jcnne Laspcyres, D1e KathcdcrsocJa!Jstcn und die statistische Con rcssc "
Deutsche Ze1t- und Streitfragen, Jg. 4 (1875), Heft 51, 1 n- 1 84 , p. 1 64 . g . '
p. ~~'tumelin, "Zur Thcoric dcr Statistik, ll" (1874), in Reden (n. 102), 265-284. P· 270.
( ( (
\ Statistical Law and Human Freedon,
Statistical Law and Human Freedom
.
S Igwar t' 110 On the matter of the actual existence of social !aws.' irre- studies in 1861 at Munich, where he studied political economy with the
s. . . h R.. 1 d1d not head of the Bavarian statistical office, Hermann. He records that in 1862
spective of the fitness of statistics to d1scover t em, ume 10 1" l"k,
lead but followed younger statisticians. Reform~rs an~ mora !sts I e he was deeply impressed by the "Enlightenment" viewpoint in David
Sch:noller, Knapp, and Wilhelm Lexis invested m the ,de_a of the Ge- Friedrich Strauss's Leben Jesu, but his politics and social philosophy be-
. h ~t or community their hopes for the preservation of order gan to change at about that time when he moved to the University of
memsc a,, , ' b ·d d b
through shared values and mutual assistance-the lat_te_r to e gm e .y Berlin. There the Prussian parliament was deadlocked in its struggles
professors of economy. As Knapp indicated most exphe~tly, to argue that with Bismarck, at a time when a revolution seemed imminent, and he
humanity was subject to uniform and unvarying laws was to er~se the began to feel that liberalism was sterile and ineffectual. The sermon of
essential distinction between man and nature and to deny the lmp?r- a Straussian liberal preacher at this time impressed him as vapid, even
tance of unique cultural values. The societies that subsumed human m- contemptible, and though he still regarded himself as a "nominal lib-
dividuals were not mechanical aggregates ruled by necessa:y laws, but eral" he found the argument about the powerlessness of liberalism in
diverse communities drawing strength from common feelmg and de- Ferdinand Lassalle's Was Nun striking and persuasive. He took his doc-
fined by a unique history rooted in freedom. torate from the classical economist Helferich at Gottingen, whom he
Georg Friedrich Knapp was one of the most outspoken oppone~ts_of held in high esteem, but by the time he finished he was persuaded that
the idea of statistical law, and more generally of the idea that soCieties "dogmatic" liberal political economy was a useless Gymnastik, and that
are subject to fixed natural laws of any sort. He wa~'. f?r the same reason, the solution to real problems could never be discovered through the ab-
a sharp critic of Quetelet, although his was a cntle~sm born of ~ deep stractions of formal theory. He sought out a more practical education at
. ss of intellectual debts and the consequent need to sift out Engel's statistical seminar in Berlin, and his first real position was as a
conSCIOUSne ' l d" t ble
what was valuable and true from what was mi~ ea mg or un en~ . statistical administrator. He set up the city statistical office in Leipzig in
Knapp was among the few German statisticians m the schoo~ of hlstor~ 1867, and was called to a post at the University of Leipzig two years later.
ical economics and Sozialpolitik who knew_ much ~athematlcs, and he From 1874 until the end of the First Worlc:l War he held a professorship
took his mathematics seriously. He had no mterest m ~ere works of sta- at Strasbourg, although by 1875 his interests had moved away from sta-
tistical compilation, and was disgusted while_ at Go~u:g~n by the ~ld­ tistics. " z
fashioned and uninspired writings of the agmg statistician. Wappaus. In Leipzig, Knapp was able to seek out Drobisch, whose remarks on
Hence he fully approved ofQuetelet's devotion to mathematics, tho~gh moral statistics he found highly admirable. He also benefited from a pa-
Quetelet's actual work impressed him less than the mathe~abcal per on "Adam Smith and Quetelet" by his close friend Adolf Held,
models of population and mortality published by Joseph Fo~ner_ a_nd published in 1867. Held expressed measured admiration for both these
Ludwig Moser. Knapp's intellectual ori~ntation was devoutly ~lstoncl~t, authors, but he dismissed Quetelet's idea of a social physics and criti-
but this did not stifle his lively interest m abstract and deductive models cized the Belgian astronomer for his lack of interest in the autonomy and
of population change. ''' . freedom of individuals. Quetelet's shallow treatment of mankind as a
Knapp, the son ofJustus Liebig's sister an_d her h~sband, one of Lie- "homogeneous mass," according to the historicist Held, was obviously
big's early students, was born in Giessen. His autobiography r~cords an the outcome of his experience of advancing democracy and leveling.
early infatuation with French culture, but it also reveal~ that_ his _moth~r Held also denied that the statistical regularities uncovered by Quetelet
placed flowers on the grave of the_ nationa~ist economist F~ledn~h L~st could properly be regarded as laws, or that they contradicted free will. ••1
while the family was on vacation m the Tuol. He began his umvemty
" 2 Knapp, Aus der fugend (n. 52); also Knapp, "Ernst Engel, Erinnerungen a us den jahren

11o ~hri~oph S~~drt, {:J~:~~ v~~~·!t:~~~~~gke~tdi~ds~:~s~~\~~~ ::n~~i~h~~ Quellen dar-


1 I86s-1866," in Einfiihrung in einige Hauptgebiete der Nationaliikonomie (Munich and Leip-
zig, 1925). pp. 322-327.
ge:;~llt (~i;~~: 1~~9)~1~nd especially, Theorie des Beviilkerungs-Wechsels: Abhandlungen zur "'Adolf Held, "Adam Smith und Quetelet," Jbb, 9 (1867), 249-279, p. 271; see also his
Angewandten Mathematik (Brunsw1ck, 1874). review of Quctclct's Physique sociale in fbb, 1..1 (1870), 81-95.

186
I
Statistical Law and Human Freedom ·
Statistical Law and Human Freedom
larity really did provide the key to social science, but not if it was ex-
Knapp included in his first book on demographic r~odels a chapter plained in ~en~s. of some mysterious and diffuse force that negated the
"Are there Laws of Mortality?" in which he detenmned that unless freedo_m of mdiVIduals, or of properties held in common by all members
death was a fixed function of age, independent of time and place, which of ~oc1ety. SocJCty was not a mere aggregation of like individuals, but a
it was not, it could not be justified to speak oflaws in this regard. ~~also umon of free ~ers~ns imbedded in a common culture that depended as
argued that there was no reason to be astonishe? by the regulanbes of much on then d1fferences as on their similarities. For this reason,
statistics.'14 During 1871 and 1872 Knapp published a senes of papers Knapp thought Quetelet's ideas on the error law was justified for the an-
on Quetelet and other modern statistical writers in which he developed thropological domain but useless to social science, and he went on to
more fully his interpretation of statistics. He attributed to Quetelet two ~e~y that probabilistic error theory could contribute significantly to sta-
interconnected sets of errors, the social determinism of physics and the t~stics. A statistician cannot use the methods of error analysis for the verY
atomistic individualism of anthropology, both inconsistent with true so- s1mple reason that every individual is genuinely different. There are 11~
cial science. The first was the work of Quetelet the physicist, and was errors, but only variation, and hence there is no "true" value beneath
the source of the doctrine of "natural laws of society," that, when ex- the diversity of phenomena awaiting discoven·. 117 ,·
aggerated by Buckle and Wagner, had inspired the suppos~d conflict _be- Kn_app's emphasis o_n the intrinsic importa1;ce of human diversity was
tween statistics and human freedom. Knapp defined the astronmmcal ass~cwted w1th IllS behef that the observed regularity of various statistical
conception of society" as the view "that forces act on socie_ty which, as senes co~~tituted the most secure knowledge that was accessible through
we recognize from the regularity of their effects, seem to be mdependent the empmcal study of society. Unlike Wagner and the early Ri..imelin
of those affecting individual events and actions, and that therefore mu~t Knapp refused to believe that true laws of society could be uncovered
be conceived as external forces. ""5 This, incidentally, was the defim- through the study of variation. According to Knapp, it was the great
tion that guided Wilhelm Lexis's work on dispersion, which was de- shortcomu~,g of Quetcle_t's French contemporary A. M. Guerry to be-
signed as a refutation of Quetelet and his followers. Quetelet's oth~r lieve tha_t soe1ai-I11Stoncal constants" such as a timeless relation be-
great error, according to Knapp, arose from his confu~ion o; socJal ~CI­ ~een ~nrne and education could be discovered through statistical ma-
ence with the natural science of man, anthropology. Knapp s objection n~pulatJon. In this respect, Quetelet's viewpoint on crime seemed to
here was to the theory of the average man, which explained social phe- hn~ altogether more satisfactory, for he read Quetelet as being content
nomena in terms of a set of penchants that were, at least to a first order to mtcrpret crune wholly 111 terms of the society in which it occurred.
of approximation, common to all members of society. This view made Soc1ety, K1~app believed, is so complex an entity that a negative relation
society a mere sum of individuals. Both the anthropologiCal and _the as- betwee~ ~nme and education in one culture could in no way be taken
tronomical errors, Knapp thought, were characteristic of Enlighten- as defimhve for another, and he implied that statisticians could do little
ment liberalism and of French thought generally, and indeed were re- mo:c th<~n affin~1 the close relationship betv,:ecn individual actions and
sponsible for the habitual instability of French political life, particularly soCJal m1heux. Smce he was equally persuaded that deductive principles
"the frightful catastrophe of the present," the Paris commune. 1,(, . ~f.populabon or economy could not be applied directly to real social en-
The proper understanding of statistics, according to Knapp, was m- tities, his viewpoint amounted to a denial of the possibility of social laws
consistent with both these conceptions, yet also derived from the "un- of any sort. "s · ,
philosophical head" of Adolphe Quetelet, a~~ even err:b~died the Knapp:s wide-ranging critique of Quetelet was widely read and ex-
proper lesson of "true Queteletismus." Recogmhon of statistical regu- erted an mfluei:ce comparable to those of Wagner, Ri.imelin, Oettin-
gen, and Drob1sch. His positive ideas about statistical science, along
"• Knapp, Sterblichkeit (n. 111), pp. 95-101. . . . . ,
"' Knapp, "Bcricht tiber die Schriftcn Quctelet's zur Sozialstahshk und Ant~ropolog1c,
Jbb, , 7 ( ,g 71 ), 16 7- 174 , 34 2-358, 427-445, pp. 438-439. Quctclct almost ccrtam1ywould not tcl~~;s ~~~~~·P"A._Qt uetcl/cbtbals Th(e~rctikcr," Jbb, !8 (!872), 89-124; also his review of Que-
have endorsed this statement, though Buckle might and Oethngcn dtd (sec n. 86). See also . o orne ne m , 17 1o71), 16o-16~.
1' 8 Knapp, "Quetelet als Theoretiker .. (n. 1 , 7), pp. 9 g_ 100.
review by Knapp in fbb, 16 (1871), 182-186. . . , .
""Knapp, "Die ncuercn Ansichten uber Moralstahshk, Jbb, 16 (1871), 237-250, P· 250.
-1WIIII•I!iilillllllilll-~~m-•""'"""'"""'"'"_. ____ ~---~-----

(
Statistical Law and Human Freedom
Statistical Law and Human Freedom
social heterogeneity, and insisted that a strategy of mass observations was
with those of Rumelin, were enlisted in the campaign to make statistics
indispensable if reliable knowledge about society was to be gained. Their
the basis for the emerging science of sociology. That campaign was con-
aim was not to find laws of society, nor to pursue international compar-
ducted less effectively in Germany, where the academic social science
ison. Following Knapp, lnama-Sternegg maintained that since each na-
of statistics was increasingly giving way to bureaucratic compilation by
tion is an intricate, highly differentiated community, possessing a dis-
188o, 119 than in Austria, where academic statisticians maintained the:r tinctive individual character, the search for cross-cultural constants was
theoretical and ideological interests for at least a decade longer. In Vt-
likely to be less fruitful than the alli;mce of statistics with history. His
enna as in Berlin, a statistical seminar had been instituted under the
joint direction of the university and the government statistical offic~ dur-
aim was not to dissolve sociology in traditional political history, but to
create a social history founded on the quantitative study of mass phe-
ing the early 186os. The head of Austrian statistics, Baron Czoermg,_ re-
nomena. That statistics lacked the power to give strict proof of causal
spected the organization of Belgian statistics sufficiently to adopt a stm-
connections did not vitiate, but confirmed, its appropriateness as the
ilar model, and his successor, Adolf Ficker, strongly approved of
method of social science. 121
Quetelet's attempt to transform this descriptive discipline into an induc-
Notwithstanding Durkheim's celebrated study of suicide, the battle to
tive science, even if he was disturbed by the Belgian's ostensible mate-
make sociology statistical was lost, at least in the short term. Naum
rialism and his tendency to dissolve the individual in society. The oc-
Reichesberg, a Swiss academic and biographer of Quetclet, maintained
cupant of the statistics chair at the University of Vienna, Leopold
Neumann, had already in 1865 expressed skepticism that statistical reg- the faith during the 189os, and his book of 1893 on statistics and social
science is perhaps the most comprehensive summary of statistical ideas
ularities could amount to laws of society. 120 The Austrians, evidently,
developed in Germany and Austria during the preceding three decades,
were sensitive to the same issues as their German neighbors.
but academic sociology was unmoved. 122 The Danish social mathe-
During the late 187os, leadership of Viennese statistics passed to F.
matician Harald Westergaard tried without luck to revive the old science
X. von Neumann-Spallart and K. T. von lnama-Sternegg. These wnt-
of statistics and to infuse it with greater mathematical sophistication. 12 >
ers developed the interpretation of statistics as a science of mass phe-
nomena, derived from Ri.imelin and Wilhelm Lexis, into an argument To be sure, the statistical study of society did not by any means disap-
for the importance and appropriateness of statistical sociology. They pear. Ferdinand Tonnies wrote on the methods of statistical social anal-
ysis into the twentieth century. 124 The historian Karl Lamprecht sought
held that an alternative was needed to the deductive sociology of Herbert
to revive Buckle's approach while purifying it of atomistic materialism
Spencer, the American H. C. Carey, the Russian P. F. Lilien-
feld and the German Albert Schaffle. They criticized the tendency to
"' Sec F. X. v. Ncuma~~1-Spallart, "Sociologic und Statistik," Statistische Monatschrift, 4
relv' 011 biological or physical analogies and argued that neither Carey's (,Ii)i)), l-IS, 57-7~; Karl Ihcodor von lnama-Stcrnegg, "Vom Wcscn und den Wcgcn der
in;istcnce on analyzing society down to its most elementary parts nor Sonalwtssenschaft, 1b1d., 7 (1881), -t81-488; idem, "Geschichte und Statistik" ibid. 8
(1882), 3-1 5; Idem, "Zur Kritik dcr Moralstatistik," Jbb, N, F. 7 ( 1883 ), 50)·>;25; idem. "l)ie
Schaffk's unwillingness to view society as anything but a seamless web Quellen de.: lustonschen Bcvi.ilkerungsstatistik," Statistische Ivlonatschrift, 12 (1886), ,s7•
40.8; tdcm, DIC Qucllcn dcr historischen Prcisstatistik," ibid., pp. 579-594; idem, "Neue Bei-
represented a viable approach to social science. trage zur allg~memen,Methodcnlehre der Statistik," ibid., 16 ( 1890), JOI-llO; idem, "Geo-
One error was common to both these writers, according to the Aus- graphic und Stahshk, 1b1d., 17 (i89i), 375-385; Gustav Adolf Schimmcr, "Die Statistik in
trians-an insistence on perfect certainty in social science. Neumann- 1hren Beziehungen zur Anthropologie und Ethnographic ., ibid. iO ( 1 88 4 ) 2 62 -,6-
. "' ~;e Na(nn Reichesberg,Die Statistik und dieGesell;chafts~issenschaf/(Stutt~a:t, 1893 );
Spallart and Inama-Sternegg deemed this unattainable. Along with ~?em, Was 1st Stattstik," ZeaschnfUii~ sch"'·eizerische Statistik, 33 ( 189;), 269 - 2 ;5; idem,
their colleague Gustav Adolf Schimmer they stressed the importance of Adolf Quctelct als Moralstahsttker, Ibid., 29 ( 18cn), 490-498; idem, ''Ocr bcruhmtc Statis-
hker Adolf Quetclct, sein Leben und scin Wirken; Ein biographische Skizze." ibid., 32
(1896), 418-460. .
"9 Sec Anthony Obcrschall, Empirical Social Research in Germany, 1848-19I4 (Paris,
::: Harald We,s~c.rgaard,,;·zur Theorie der Statistik," Jbb, N. F. iO ( I88s), 1-23 _
19 ~.~)Sce Adolf Ficker, Quetelet Nekrolog, Statistische Monatschrift, 1 (1875), 6-14; Leopold Ferdmaml I onnics,, Emc neue Methode dcr Verglc1chung stahstischcr Rei hen," Jahr-
buch fur Gesetzgebung, \erwaltung, und Volkswirtschaft im Deutschen Reiche, 33 (Iqo 9 ),
Neumann, "Ubcr 'Theoric der Statistik," Oesterreichische V1ertel1ahresschnft fur Rechts- und
699-720.
Staatswissenschaft, 16 (1865), 40-62.
~;.;,m;- - - - - -

Statistical Law and Human Freedom


bv ado ting the statistical viewpmnt . 0 f R un:e.ll:,
. "I' Knapp. ' and Lexis r and
psychrc~l
125

: dd' p .t t Wilhelm Wundt's nondetermnmhc causa.rty. - - - - - - - - - Chapter Seven - - - - - - - - -


we mg I o . . t d . nathemahcal socJal sta-
The Italians became increasmgly mteres e m I . t' I kin
tistics near the en d f h° , t .
t e cen ury' 126 and the concrete stabs rca wor
h' t. t d than
1 des ~hat statistics de-
d 1 0 ears enormously more sop Is Ica e TIME'S ARROW AND STATISTICAL UNCERTAINTY
Germany aroun 9° app ' .
hat of Quetelet's generation. Hemz Maus cone u ' . d
t the credit for the establishment of a bond between sociology an IN PHYSICS AND PHILOSOPHY
~lr
serves. I cr· .· esearch at the end of the nineteenth century. 127 f
empmca so · " · b t th ture o
The most important effects of these discussions a _ou. I' e na R b t
statistical knowledge, however, were felt by other drscip mes: o .er
Campbell and Wilhelm Lexis, the first writers to compare the ~~pe;sr?~
German economists and statisticians of the historical school viewed the
idea of social or statistical law as the product of confusion between spirit
of actual distributions with expected values from ~ ~om ma or:: and matter or, equivalently, between history and nature. Their sense of
model, were both inspired by their opposition to certam ~dea~ of Qt .~
inconsistency between mechanical law and progressive history was in
t I t and Buckle. 'fheir important mathematic~! contnbubons, dis
most cases uncorrupted by any actual knowledge of physics, but the con-
e e d . ·h· t 8 belonged to the traditions discussed here. Norton
cusse Ill c ap er • d f 1 J an idea of tradiction they identified proved more pervasive than they tcalizt:d.
Wise has suggested that these German ideas, an par Icu ar y 'b
"statistical causality" developed by Karl Lamprecht, may~ have cont~I -
That the laws of Newtonian mechanics arc fully time-symmetric and
hence can be equally run backwards or forwards could not easily be rec-
uted to the radical new views on causality that arose m ~en~an qt;an-
onciled with the commonplace observation that heat always flows from
1 . d . er1suing decades. 128 Even
tum p 1YSICS unng · .
before the nse o quan urn
·, 1 .. t' f... d warmer to cooler bodies. This discrepancy became for a time one of the
h •sics, a close and significant relationship between socia. sta IS I~s, ~~1
~1~ origins of probabilism in physics is apparent. It pertams to t e ast
deepest theoretical problems of the dynan1ical-or, as it came to be re-
garded, statistical-gas theory. James Clerk Maxwell, responding to the
third of the nineteenth century.
apparent threat to the doctrine of free will posed by thermodynamics and
"' Sec M. Norton Wise, "l!cJ"'
n [)cJ Sums Count? On the Cultural Origins of Statistical statistics, pointed out that the second law of thermodynamics was only
Causalitv,". in· ·Prob . .
p Rev. "N OVC app 1lCJZIOilC dCl ca. 1C·olodelle probabilit:'! allo studio
d' ·dci fcno- probable, and that heat could be made to flow from a cold body to a
u6 Sec Lmg1 crozzo,
. · · · u
" Atti della R Acca d.emw. del. L.mce~,. . Memorie della classe 1 SC!enze mo-
· · 1 d 11 warm one by a being sufficiently quick and perceptive. Ludwig Boltz-
mcmstoriche,
rali, stattshcte.. ·, .
filologJche. ·
10 { t88z), 473-5°3·
. Antonio Cabaglio. Storia e teorra genera e e a
mann resisted this incursion of probabilism into physics but in the end
statistica (Milan, 188o). .. I I ... I en Sozialforschung " in Rene Kiinig.
he was obliged, largely as a result of difficulties presented by the issue of
"' Heinz Maus, "Zur \forgcsch.'c Jtc ~c cr ~~~m~co; J Geschichte u;d Grundprobleme
ed., J/andbuch der empmschen Sozwltorsct / ' inc;cas~d sophistication in Germany one mechanical reversibility, to admit at least the theoretical possibility of
(Stuttgart, 1967), 21-56. P· 33· To ~p6wa_c b~n und Statistik, edited by Heinrich Braun chance effects in thermodynamics. The American philosopher and
need only look at the Archn• fur sozw ed. e:t~g~ g Sombart Max Weber, and Edgar Jaffe,
from J888-t903, or its contmuahm~, c ttc . v erncr ' physicist C. S. Peirce determined that progress-the production of het-
Archiv fiir Sozialwissenschaft und So;,ratpohllk.
,~Wise, "How Do Sums Count? (n. 125). erogeneity out of homogeneity-could never Bow from rigid mechani-
cal laws, but demanded the existence of objective chance throughout
the universe. His position can by no means be represented as the con-
sensus view at the end of the nineteenth century, but neither was it a
radical departure from all contemporary discussion.

1 93
( Statistical Uncertainty (
Statistical Uncertainty
siasm for Hamilton waned after he left Edinburgh for Cambridge in
18 50, he always took ca~e _to avoid implicating the scientific tempera-
~ent as an ob~tacle to ~ehgwus belief. "I have endeavoured to show that
BUCKLE'S LAWS AND MAXWELL'S DEMON
It IS th~ peculiar function of physical science to lead us to the confines
of the ~ncomprehensible, and to bid us behold and receive it in faith, till
Statistical law had been presented to the world by Quetelet and Buckle
such time as the mystery shall open. , 3
as proof that disorder and chance were epiphenomenal. The application
Max~ell was not persuad~d that the bounds of the incomprehensible
. by Maxwell of that most remarkable law of unreason, Quetelet's error
1 curve, to the otherwise intractable problem of molecular velocities im- were bemg ~ushed back qmte as rapidly as some of his contemporaries
plied no vindication of what was still almost universally regarded as an would_ have It, an~ he was troubled by the extravagant pronouncements
unthinkable and self-contradictory illusion of untrained minds, objec- sometimes m~de_m the name of science. Prominent among the writers
tive chance, but an impressive extension of the domain of scientific or- :ose necessJtanan ~!aim~ bothered him was Henry Thomas Buckle.
der. The combinatorial operations it made possible were both rigorous xwell had bee~ highly 101pressed by certain aspects of Buckle's first
and elegant, and it was no more to be anticipated that the reduction of volu~e and menti~ne.d Buckle to his friend and subsequent biographer
thermodynamic propositions to mechanical ones would introduce a Lewis Campbell withm a few months of its appearance: "One ni ht I
new element of uncertainty into the subject than that the application of read 16o ~a.g~sof Buckle's History of Civilization-a bumptious !ok
number would have such an effect on social science. The idea that mac- s~?ng positivism, emancipation from exploded notions and that style of
roscopic regularities such as the second law of thermodynamics are only t mg, but a great deal of actually original matter, the true result of fertile
probable was manifestly a development of certain ideas associated with study, and not mere brainspinning."4 Evidently Buckle set hi. l'k
the statistical approach, but it involved at the same time a repudiation · fh'·f,ll . . m,JeSO
m,my o Is_ ~ ows, to thmkmg about the relation between the remark-
of the statistical viewpoint presented by Quetelct and Buckle. able_ reg~lanbes of statistics and free will, for three months later he wrote
That Maxwell should play a major role in the reinterpretation of sta- to h1s fnend R. B. Litchfield:
tistical reasoning was fully consistent with his character and commit-
ments. He was, of course, a physicist of exceptional creativity, revealed No~ I am going to put down something on my own authority
in work on electricity and magnetism that began in 1854 no less clearly w?Ich you must not take for more than it is worth. There are cer-
than in his papers on gas physics. He was always eager to look at old tam men w?o write books, who assume that whatever things are or-
truths from new perspectives, and in his inaugural lectures at Aberdeen derly,_ certam, and capable of being accurately predicted by men of
in 1856 and King's College, London, in 186o he warned his students expenence, belong to one category; and whatever things are there-
against "assuming that the higher laws which we do not know are ca- sult of conscious action, whatever are capricious, contingent, and
pable of being stated in the same forms as the lower ones which we do cannot be f?reseen, belong to another category.
know. "• A deeply religious man, Maxwell was sensitive to the limita- All the bme I_ have lived and thought, I have seen more and
tions of natural science. His philosophy professor at Edinburgh, Sir more reason t? disagree with this opinion, and to hold that all want
William Hamilton, had observed that since the Deity could not be sub- o~ ord~r, capnc~, and unaccountableness results from interference
ject to necessity in the material universe, the study of nature could never with liberty, w~Ich would, if unimpeded, result in order, certainty,
attain more than "probable certainty,''• and although Maxwell's enthu-
~nd trustworthmess (~ertainty of success of predicting). Remember
' Maxwell, "Inaugural Lecture at Aberdeen, 2 Nov. 1856," Notes and Records of the Royal do not say that capnce and disorder are not the result of free will
Society of London, 28 (1973), 69-81, p. 78; also "James Clerk Maxwell's Inaugural Lecture at
King's College, London, 186o," American Journal a{ Physics, 47 (1979), 928-933, p. 930.
on his life generally, sec c. W.F. Evc;i~ •..t~x'~ II~ S a~we~ ~ rclationshi,p_to Hamilton, and
3 Maxwell, "Aberdeen Lecture" (n ) 8 0 M ,
' Sir William Hamilton, Discussions on Philosophy and Literature (New York, 1856), p. 41;
also pp. 275, 297· On Maxwell and Scottish Common Sense, sec George Elder Davie, The et al.' eds., Springs of Scientific Creativity (M. c s ~~~nh Be CreatiVIty, m Rutherford Aris
Democratic Intellect: Scotland and her Universities in the Nineteenth Century (Edinburgh, 4 Maxwell, pp. 294-295· mneapo IS, 19 3), pp. 71-141.
1961); Richard Olson, Scottish Philosophy and British Physics, 1750-t88o (Princeton, 1975).
195
194
Statistical Uncertainty (
Statistical Uncertainty
(so called), only I say that there is a liberty which is not disorder,
and that this is by no means less free than the other, but more. 5 scar~e scie.ntific posts when dedicated dissenting scientists had difficultv
findm~ suitable positions. The appeal to physics embodied a drive t~
No less objectionable to devout and conservative men like Maxwell sever ties w~th natural theology, to remove biology from the empire of
were the public lectures and popular writings of the Victorian scientific teleology. 1 he obJect, as Huxley put it, was to "reduce all scientific
naturalists. Maxwell seems to have gotten along well with T. H. Hux- problems, except t~ose which are purely mathematical, to questions of
ley, but there is ample evidence that he found John Tyndall's ideas m~lecular phys~cs. 8 ~ uxley was appalled by the audacity of those who
wrongheaded and disagreeable. In a commentary inspired by some re- clai.med authonty to judge scientific theories such as Darwin's on the
marks of P. C. Tait, Maxwell implied that Tyndall had "martyred his basi~ of what arno~1~t~d to "sacerdotal pretensions,'"9 and Tyndall. dis-
scientific authority," adding: "If he writes it in a dry manner it is bad playmg, equal sensitivity to the purity of science and the sanctity of the
enough, but the harm is confined to students. But if he seasons it for the scholar s tt:~f, boldly announced the "impregnable position of science"
public and the public swallows it, then it is a sad misuse of words to say as follows: We claim, and we shall wrest from theology the entire do-
that this is useful work. " 6 More playful, but perhaps equally pointed, n: am of cosmological theory. All schemes and systems which thus in-
was the "Tyndallic Ode" that Maxwell composed to parody Tyndall's fn~ge upon the domain of science must, in so far as they do this submit
very successful lecture style. Maxwell's Tyndallic lecturer began with to Its control, and relinquish all thought of controlling it. "'o '
some showy demonstration experiments, then resolved that "These If natural science was to implement these ambitious claims, it could
transient facts I These fugitive impressions I Must be transformed by a~ford to !~ave no gaps, and the idea of chance in science was viewed
mental acts I To permanent possessions," and finally proceeded to con- With ~k:phcism by Huxley and Tyndall. One reason for opposition to
struct a metaphysical tower of Babel upon his experimental sand: Da:w~n s theory was that it replaced teleological purpose by chance
Co to! prepare your mental bricks,
v~nat1on~what John Herschel was reputed to have called "the law of
hJgglcdy-piggledy· "" B " ,h " D · · f
Fetch them from every quarter, · Y c ance, arwm o course meant undi-
rected, not uncaused, but his defenders were nonetheless disturbed bv
Firm on the sand your basement fix
the abse.nce of a "law of biological variation. Thus Huxley wrote t~
With best sensation mortar.
Ho?ker m 1861: Because no law has yet been made out, Darwin is
The top shall rise to heaven on high-
obliged to speak of variation as if it were spontaneous or a matter of
Or such an elevation,
cha1~ce, so tha~ the bishops and superior clergy generally (the only real
That the swift whirl with which we fly
atheists and believers m chance left in the world) gird at him as if he were
Shall conquer gravitation. 7 another Lucretius. " 1 1

Scientific naturalism sprang up in Britain in defense of Darwin's the- T~is was by no means the last mention in late Victorian scientific dis-
ory of evolution, but its prominent spokesmen included the physicist cussi~n of the a~cient atomists or of issues involving chance and me-
Tyndall and the mathematician W. K. Clifford as well as biologists like chamcal determ11mm With which they were habitually associated. John
Huxley, and its rhetoric was saturated with concepts and terminology . ~ T(.LH. ldluxley8, "The Scientific Aspects of Positivism.,. in Lav Sermom 4.ddresses andRe
v1ews on on. 1 95 ), p. 144 . · ·,' , -
from mechanics. Naturalism was in part an instrument of profession- '' Huxlcv. • "On th c H ypo tt1es1s· t1wt niH
• · 1
lecte dE'ssays (9 vols New y k 68) Ilia s arc Automata and 1ts J Iiston·'" ' 1 o_ ) · C' 1
alism, and biologists felt far less threatened by the scientific imperialism l . · · . ' "A . Ill o-
torian Conflict B ~' S . or '. 19 d '~o: I, p. 249· Sec also Frank :\1 Turner, 'The Vic-
of physics than by the Anglican clergymen who insisted on bringing bi- 356-376. c cen C!cncc an RcligtmL A Profcsswnal Dimension," Isis, 69 ( 1978),

en;; {~:~v·V~~~al~,97";')he B~~fast8Addrcss" ( '874), in George Bassala ct al.' cds.' Victorian Sci-
ology within the domain of natural theology and, worse, who occupied
D . ' '43 47 , pp. 474-475.
6
'Ibid., pp. 305-306.
See Maxwell comment on galleys ofTait rebuttal to Tyndall in JCMP 7655 lll d/5. by 1
't~e ~;;:ntfi! ~~:rz~~:;~ (~~~~~:irgr~ti~a;'he Recep). tion of D
6 arwin's Theory of Evolution
"L dH 1 ·~ ' s., 1 973. pp. 7, 1.
1 Maxwell, pp. 63 5-636. eonar ux ey, L1,e and Letters of Thomas H. Huxley (2 vols. New York 190 ,) \'CJJ
1, p. 245· , • • .

1 97
Jii"iWFk ·w

(
Statistical Uncertainty
Statistical Uncertainty
t:1otes which chase each other in all directions througl·lt.t ']"h'
Tyndall found it possible to enlist ancient atomism in the naturalists' t f th · ·bl · 1s mo-
wn o . c VIS! ~ motes, he tells us, is but a result of the far more
cause by sidestepping Lucretius, whose De rerum natura contained the complicate~ motion of the invisible atoms, which knock the motes
troubling doctrine of the swerve, and invoking directly his predecessor
about. In ht~'~ream of nature, as Tennyson tells us he
Dcmocritus, who had thoughtfully refrained from burdening the pop-
Saw the flaring atom streams
ular historian with any surviving works. In Tyndall's controversial1874
And torrents of her myriad universe
address to the British Association in Belfast, he listed six fundamental
Running along the illimitable inane
Fly on to clash together again, and ~1akc
scientific propositions of Democritus and pronounced the first five of
them "a fair general statement of the atomic philosophy as now held."
~nother and another frame of things
These were the following: For ever."
1. From nothing comes nothing. Nothing that exists can be de- Ad. ·
b n d tt I; no wonder t~at h~ should have attempted to burst the
stroyed. All changes are due to the combination and separation of o~ s o Fat: b~ makmg hts atoms deviate from their courses at
molecules. 2. Nothing happens by chance. Every occurrence has ~tmt~ unccrtam tnnes and places, thus attributin t0 th . k' d
its cause from which it follows by necessity. 3· The only existing mahonalf. h. h ·n. g ema m of
l· . . • ree wt ' w IC on ll!S materialistic theory is the only ex-
things are the atoms and empty space, all else is mere opinion. 4· p anahon of that power of which we ourselves arc conscious.''
The atoms are infinite in number and infinitely various in form;
they strike together, and the lateral motions and whirlings which , Maxwell
, . certainly· did no t wts ·11 to Imply
· that natural science should
b
thus arise arc the beginnings of worlds. 5· The varieties of all things c constramcd . to .su ppor t ccr t am ·. rc 1..
Igious doctrines. "Th , t, f
depend upon the varieties of thcir.atoms, in number, size, and ag- c h ange
'bl. of .sctcntifi
. c h ypo tl Jcscs ts
. naturally much more ra!)id ethan ra ctl o. t
fB 1
gregation." o. I teal mterpretat wns, " h e to ld a clergyman in 1876 "so tl t 'f a
mterpr t f · £ d d ' 1a t an
Tyndall's line of thinking was more troubling to Maxwell even than h l c ~ ton ts oun e on such an hypothesis, it may help to keep the
Buckle's since it was precisely one of Maxwell's specialties, atomism, t::?.:;e~s :~ove hgr?und Ion~ after it ought to be buried and forgot-
that was dressed up to support this deterministic and materialistic doc- : . e o~g t It appropnate for Christians to seek to harmont.ze
th
trine. Maxwell had long held that dynamical explanation represented a etr sc1ence
were valid w1th
I · [;~~'th : b. ut msisted
th etr · . that the results of this effort
scientific ideal, and he had written just before the storm broke, in 1856, . '7 on y for the mdtvtdual involved, and only for a limited
that "if we know what is at any assigned point of space at any assigned
:~~~· reli~:~~~n::~~~~c~b~~thtil: !:c~i~~;~~~ntcl~ odf. ctontcmpor~ry science
instant of time, we may be said to know all the events in Nature. We f. d p . liS IC um provmg as his
cannot conceive any other thing which it would be necessary to ~t.cn ~ . G. Tat~ ~nd Balfour Stewart explained, that, the oppo~c:t~ts of
know." 14 It was not a little discomfiting to sec these ideas applied vig-
orously to life and mind, in explicit opposition to religion. Maxwell pre-
dfatt~1,
" d
and not rehgw.us meu,_ ar.c the true dognntists '"' "N
ehghted · . ' · ··
more m speculation,,. observed Lewis Campbell of~ ·t·
. d
o nun ever
II
an yet no , . . 1v axwc ,
ferred to cast his lot with Lucretius: popular diss~~~a·st ever ~w~c Jealous of the practical application or the
When Lucretius wishes us to form a mental representation of the I . ,. . na t~n o w at appeared to him as crude and half-baked
t leones about the htghest subjects· · "tq Kar II").. , came away from 3
carson
motion of atoms, he tells us to look at a sunbeam shining through
a darkened room (the same instrument of research by which Dr ''Maxwell, "Molecules," in SP vol 2
' 6 Maxwell, p. 394 . ' · · P· 1Tl·
Tyndall makes visible to us the dust we breathe), and to observe the
'7 Ibid., p. 405
'' Tvnda\1, "Belfast Address" (n. 10), p. 443· Sec also Frank M. Turner, "Lucretius among
the Victorians." Victorian Studies, 16 ( 1972-73), pq-341:\; and Turner, The Greek I leritage in
on'~,s l•uture
\eler Guthrie Tail and Balfour Stewart Th
State (London, 1 s75 ), p. v. '
, .
e Unseen Umverse, or Ph)'sicai Speculations

Victorian Britain (New Haveu, 1981). Maxwell, p. 322 .

" .Maxwell, p. 2 38.


199
_ _ _illlllll_ _.,..,.._ _ _ ..,.,_.,_
.. - - -

Statistical Uncertainty 1
Statistical Uncertainty
the correlative term ~rder, is_ not a property of material things in
Cambridge examination by Maxwell for the Smith's prize with a less
themselves, but only m relation to the mind that perceives them.
charitable view: A memorandum-book does not, provided it is neatly written ap-
The conversation turned on Darwinian evolution; I can't say how ~ear confused to an illiterate person, or the owner wh~ underst~nds
it came about, but I spoke disrespectfully of Noah's flood. Clerk ~t tho~oughly, but to any other person able to read it appears to be
Maxwell was instantly aroused to the highest pitch of anger, re- mextncably confused. Similarly the notion of dissipated energy
proving me for want of faith in the Bible. I had no idea at the time could not oc~ur to a being who could not turn any of the energies
that he had retained the rigid faith of his childhood, and was, if pos- of nature to his own account, or to one who could trace the motion
sible, a firmer believer than Gladstone in the accuracy of Gene- of _eve~y mol~cule an~ seize it at the right moment. It is only to a
sis. 20 bemg m t~e mtermed1ate stage, who can lay hold of some forms of
~nergy whde oth:rs elude his grasp, that energy appears to be pass-
It was in this context that Maxwell developed his ideas about the sta-
mg from the avmlable to the dissipated state. 22
tistical character of the second law of thermodynamics, and more gen-
erally about the inescapable imperfection of human knowledge. Max- M~xwcll first adum~rat~d a_ connection between the indeterminacy of
well first pointed out that his dynamical theory of gases implied the certam thermodynamic pnnCiples and their statistical character in 1868
possibility of violating the second law in 1867, in a playful letter to P. w~en he com~ar~d t~le tendency for gas molecules to assume the nor~
C. Tait. There he introduced the "very observant and neat-figured rna,! ve~oCity distnbubon to the mixing of black and white balls in a box.
being," later dubbed by William Thomson Maxwell's "demon," whose T\\oo yea~~ later he wrote to the young physicist Rayleigh that the second
mission was to "pick a hole" in the second law. The demon needed law had the same degree of truth as the statement that if you throw a
merely to be set to guard a small hole in the clastic wall between gases tumblerful of_wa.~er into the sea you cannot get the same tumblerful of
at different temperatures and it could-by allowing only the most en- w~ter ~ut ~gai~1. 23 Maxwell began to develop an argument about the
ergetic molecules to pass from the cold to the hot side, and only the least wider Implications of the statistical method in physics in the inaugural
energetic in the opposite direction-cause heat to flow from a cold gas lecture he delivered ~~ 1871 upon becoming head of the new Cavendish
to a warm one. Notwithstanding Thomson's name for this fictional crea- L~borato~y ~t Cambndge. Th_er~ he remarked, explicitly and publicly,
ture, which Maxwell found objectionable, no supernatural powers were th_at pred:chons base? on statistical knowledge were inherently uncer-
required, but only an exaggerated level of ordinary ones. 21 t~m. ~he_ gas la~s c~Jdently were of a different character from clvnami-
Maxwell's invention implied that some physical principles, among cal pnnciples,. yie~dmg a form of knowledge whose implicati~ns he
them the second law, were really as much attributes of human percep- thou~ht had wide mterest. Maxwell pointed immediately to the vexed
tions as of nature itself. He wrote a decade later in the Encyclopaedia queshon of human freedom, arguing:
Britannica under the heading "Diffusion":
... the sta~istical method ... ' which in the present state of our
'The idea of dissipation of energy depends on the extent of our knowledge .'s tl:e only available method of studying the properties
knowledge. Available energy is energy which we can direct into any o~ real bodies, mvol~es an abandonment of strict dynamical prin-
desired channel. Dissipated energy is energy which we cannot lay ciples, and an adoptwn of the mathematical methods belonging to
hold of and direct at pleasure, such as the energy of the confused
agitation of molecules which we call heat. Now, confusion, like :: Maxwell, "Diffusion," in SP, vol. 2, p. 646 _
(Mad~~~~'~~~:, fa~~:::~~:J~{a~le~~1 • Life of ~ohn Willia.m Strutt, Thi:;d Baron Rayleigh
w Karl Pearson, "Old Tripos Days at Cambridge, as seen from another Viewpoint," Math- J· t J?
Irreversibility," in The Kind of Mot' 9
. l 1
·1 See(~lso Stcphcn Brush, Randomness and
•on e ,a 11eat • vo s. Amsterdam 19 ~6) 54 -6
ematical Gazette, 20 ( 1936),
27-36. 59~, ~as~m; s!at .' A
a s?Brush, "lrrcvcrsihility and Indeterminism:'· chap. 2 of h'is t. 31 4: P·.
" The original letter is printed in C. G. Knott, Life and Scientific Work of Peter Guthrie
Tait(Cambridge, 1911), pp. 213-214. ScealsothcundatcdlettcrofMaxwell toTaitin ibid., ~~S ,\ e tom!c Ilwory of Matter from Boyle and Newton to Landau and Onsag;~ (~~inc!t~~~~
pp. 214-215.
201
200
Statistical Uncertainty ,
Statistical Uncertainty
the theory of probability. It is probable that important results will
~~'~/~'~:0~'~
M 1 1
dofcmc of f<cc will
{ ~s evzew
published anonym~usly in the
w,;
868 by Fleeming Jenkin, who evide~tlv sent
ll1 1
1
be obtained by the application of this method, which is as yet little l axwe t de pr~ofs m advance. 27 His argument, cast as a defense ~f the
known and is not familiar to our minds. If the actual history of Sci- "ucr.ehan octnne of the swerve, was virtually the same as Maxwell's·
ence had been different, and if the scientific doctrines most famil- Jenkm wrote that "if mind or will deflects matter as it moves it ' ,,
iar to us had been those which must be expressed in this vvay, it is duce a~! t~e
consequences claimed by the Wilful school, ~;ld ~;;t ~~~
possible that we might have considered the existence of a certain add" neither energy not matter to the universe. "28
kind of contingency a self-evident truth, and treated the doctrine of tic~lhk~~~:t~incentiv~ for .Maxwell's reflections on the limits of statis-
philosophical necessity as a mere sophism. 24 .th tl I e gefwas his desne to show that freedom was not inconsistent
fWI d 11e aws o nature prop,er1y known b. y contemporary science His
By 1873 Maxwell had developed this line of reasoning into a full ar- un amental argument was that the statistical method the l . ,
gument against the increasingly threatening doctrine of mechanical de- b 1· r h . , on y means
~ w liC l umans can attam general knowledge of a molecul . .
terminism. His aim was not to use physics to demonstrate the existence y1elds 0 ] , 1· . ar umverse
. £ n. y genera IZahons about the mass of molecules and provides ,
of human freedom, for he adhered to the Common Sense tenet that be-
lief in free will arises naturally from reflection on the mind's own activ-
m ormahon about individuals. He wrote in his 1871 textbook on he~~~
ity. He carried out his examination of physical knowledge precisely in Ilt is th~refore pos.·sible that we may arrive at results which t} l
order to show that known scientific principles in fact proved nothing on t 1ey fanly
. , t th e .cJacts as Iong as we are supposed to deal
re prescn • 10ug
with1
this vital issue. Maxwell did not, however, deny natural science all rel- atgas m mass, would cease to be applicable if our faculties and in-
evance to metaphysical issues. He suggested that by showing what free- I were so ·sh arpe~1e d t11at we could detect and lay hold of
s,. ruments
h
dom could not be, physics offered valuable guidance in identifying what eac mo ecule and trace tt through all its course.
it was. The inherent interest of this project, as well as the need to dem- f For the ~a me reason, a theory of the effects of education deduced
onstrate the possibility of free action that did not violate any laws of phys- ro~ a stu y o~ the returns of registrars, in which no names of in-
ics, inspired Maxwell to devote considerable attention to the physical dlv~duals are given, might be found not to be applicable to the ex-
interpretation of free will. ~edr~endce ol.b schoolmaster who is able to trace the progress of each
Already in 1862 Maxwell had observed to Lewis Campbell that the m lVI ua pupJI. 29
conservation of energy permitted the soul to be switchman, but not
mover of the body. "There is action and reaction between body and Hence
. , k the need to tesor
-, . t to statistics
· · guaranteed not only that nothin
soul," he wrote, "but it is not of a kind in which energy passes from one ~~:~ e~~:~~ <~~the p~rhcular CI.rcumstances of individual molecules
le aws o then motion might bear no determi . , I .
bu~
to the other." The direction which the soul could give to the energy of
the body was comparable, he thought, to the relation between trigger the observable regularities of the mass. nate re ahon to
and gun, or pointsman and train. Maxwell disclaimed any pretense to earlMaxwell
.I 8 developed
£ , thes,. .
e arguments . . fully ll1
most . a paper that he read
having solved this issue-"lt is well that it will go, and that we remain
in possession, though we do not understand it," >-but he was not long
2
"D y \~ l/3or Eranus, a club of past Cambridge Apostles titled
. oes 1e rogress of Physical Science tend to give any advanta~~ to the
willing to rest secure in ignorance. The frequent invocation of conser-
Uniformity" ( 1903), in Charles Hartshorne et . 1 d ·
vation laws against human freedom convinced the defenders of meta- Peirce 18 vols., Cambridge Mass 19 ,
,
.u)a · , c s. , Collected Papers of Charles Sanders
· ., ,1-19,o ,.o 1 6 6- R·
physical freedom of their obligation to demonstrate the possibility that '~ ce Jenkin's letter to Maxwell 10 )ulv 8'68 111
S · · ' r '' p. ;o.
' 8 [Fleemino Jenkin] '"I'!Je At(ln' · 1 .1 ' L , JC'v!P, 765 5/ll,
h Box
. . 1.
,l,I!-128, p. I IS. This paper was widclv cir~u
" ' uc 1corv of L f . ·· N'
the will could operate without any expenditure of energy. 26 The most- ucre IUS, ort BntJsh Review, 48(!868).
1 homson, 'The Structure of Matter aJJd. tile Ulatetd, faSI:d quote?. by, among
Maxwell, "Introductory Lecture on Experimental Physics," in SP. vol. z, p. 253. m Yo Clencc ' 1 B J \'others,
· · William
, .
' 4 ( n. IO, ) 101-128, p. 109, and Tait and St U . ' n assa a. ·lctonan Science
'' .Maxwell, p. 336. '9 Maxwell, T'heoryo(Heat(I87J· Lonedwart, n~een Umverse(n I8), pp. J8J-18z.
'6 As C. S. Peirce noted of Simon Newcomb, defenders of free will were strangely indiffer- ' on, 19041, pp. 315-316.
ent to violations of Newlon's third law of motion. See Charles Sanders Peirce, "Variety and

202
'Statistical Uncertainty
Statisticai Uncertainty
opinion of Necessity (or Determinism) over that of the Contingency of
viation is sufficient to determine into which of l:vvo vallers we shall de-
Events and the Freedom of the Will?" Here atomism, thermodynamics,
scend. The doctrine of free will asserts that in some cases. the Ego alone
and statistics, a trio of sources for those deterministic arguments seen as
IS the determini1~g cause. "3' The statistical character of knowledge in-
most compelling in mid-Victorian Britain, were turned on their heads
sured that no fin1te observer could be in a position to refute this possi-
and reinterpreted as evidence for the possibility of human freedom. bility.
Maxwell began by noting that the statistical method, which "has La,~
A few years later, Maxwell encountered a refinement to this solution
place for its most scientific and Buckle for its most popular expounder,
to the free-will problem that seemed to him more satisfactory. His
was by its nature an imperfect one, applicable precisely when the course
source was a group of French and Belgian Catholic scientists who were
of individual events cannot be charted or explained. Its use, he observed
similarly repelled by the deterministic materialism of Tyndall, and also
depends on an assumption "that the effects of widespread causes, though
of Comte's disciple Emile Littre and the American John Draper. As
very different in each individual, will produce an average result on the
~1ary J~ Nye has shown, Ignace Carbonelle, Joseph Delsaulx, and Ju-
whole nation, from a study of which we may estimate the character and
lie~ Thmon had, hke Maxwell, found in the kinetic gas theory justifi-
propensities of an imaginary being called the Mean Man." Far from
cat.wn for the belief that some measure of uncertainty and unpredicta-
being unique to social science, however, this form of reasoning consti- bility characterized scientific laws. They were especially interested in the
tuted the basis for "all our knowledge of matter," assuming that the mo-
?roblem. of Brownian motion, the irregular movements of tiny but vis-
lecular hypothesis was true. "A constituent molecule of a body has prop- Ible particles suspended in fluids, which was first linked to the kinetic
erties very different from those of the body to which it belongs," he theory by Delsaulx. Maxwell found his inspiration in the work on fluid
continued, and "those uniformities which we observe in our experi- mechan!cs of another member of this group, Joseph Boussinesq, who
ments with quantities of matter containing millions of millions of mol- showe~ 111 1878 that ~nder certain conditions the differential equations
ecules are uniformities of the same kind as those explained by Laplace re~ulat~ng a m.echamcal system should have multiple solutions at cer-
and wondered at by Buckle arising from the slumping together of mul- tam pomts ~f smgularity, and hence that determinate forces might pro-
titudes of causes each of which is by no means uniform with the duce no umquely determined motion. In that event, it would be possi-
others. "w ble for a "directing principle" such as the will to determine which
This, evidently, left open a certain space of ignorance, within which possible solution actually occurred. 32
the will could operate without its direct effects being perceived. In this Oddly enoug~, Maxwell developed this line of thinking most fully in
respect "our free will is at best like that of Lucretius's atoms," extending a letter to Franc1s Galton, with whom he did not regularly correspond
onlv over an infinitesimal range. Under certain circumstances, how- and whom he once characterized as a man "whose mission it seems to
eve~, these limitations might be transcended. Maxwell wrote: "It has be to ride other men's hobbies to death. ">J Maxwell's comments, which
been well pointed out by Professor Balfour Stewart that physical stability appeared ~sa postscript to an othenvise brief and routine letter renewing
is the characteristic of those systems from the contemplation of which IllS subscnphon to the Philosophical Club, were as follows:
determinists draw their arguments and physical instability that of those
I?o you have any interest in Fixt Fate, Free Will, &c. If so Bous-
living bodies, and moral instability that of those developable souls,
smesq [of hydrodynamic reputation] "Conciliation du veritable de-
which furnish to consciousness the conviction of free will." The mind,
terminisme mecanique avec l' existence de Ia vie et de Ia liberte mo-
he indicated, was a system for generating and regulating instabilities,
,, Ibid., pp. 440-441
through which human freedom and moral responsibility are expressed.
"In the course of this our mortal life, we more or less frequently find
C ''
1
sr Mary )o Nyc: 'The Moral Freedom of Man and the Determinism of Nature: The
atl~o tc Sy_nthcsts of Sctcncc a,nd Htstory in the Revue des questions scientifiques," BJHS, 9
ourselves on a physical or moral watershed, where an imperceptible de- ( 97 ), 2 74292, pp. 277:281, fhe tdea of a dtrcctmg pnnctplc denvcs from A. A. Cournot
a1nd was revtved by Boussmesq s mentor, St. Vcnant. Nye also mentions Claude Bernard in
t 11s connection.
' 0 Maxwell, 434-444, pp. 438-439.
" Maxwell, p. 390.
204
205
( ( Statistical Uncertainty (
Statistical Uncertainty general,ization~, if.not in the laws of nature themselves. U;in
~:tt~~s f.,m~buuiles, he rema<kcd in •n unpublished "";usc,ipt
one of
rale" (Paris, 1878) does the whole business by the theory of the
. . e popu abon of a watering-place, considered as a mere number
singular solutions of the differential equations of motion. Two
other Frenchmen have been working on the same or similar tracks. vanes m the same way whether its visitors return to it season after seaso~
Cournot (now dead) and de St. Venant [of elastic reputation Tor- oHr wheth~r the annual flock consists each year of fresh individuals ";;

~ew, th•t the imperlection of statistic•! knowledge had rec;i::~


e was nght , of co urse, b u t Its
· h ou ld not be forgotten that these 'd·
sion of Prism &c.). were
Another, also in the engineering line of research. Philippe Bre-
ton seems to me to be somewhat like minded with these. scarce y any attention before 1857 except by those wh . t d.
tirely and th t M ll b o reJeC e 1t en-
There are certain cases in which a material system, when it h ' a axwe egan pursuing this line of thought only when
comes to a phase in whiGh the particular path which it is describing . e ~;s pro~o~ed by the vehement expression of statistical and mechan-
coincides with the envelope of all such paths may either continue ICa etermmlsm by Buckle and Tyndall.
in the particular path or take to the envelope (which in these cases d Th~ per~uasivfe force of Maxwell's conclusions on statistical knowl

P~~::i unsympathetic. to pwbabilism in thcmlodyn~.;,~"~ :.: ~~::


is also a possible path) and which course it takes is not determined e ge IS evl ent rom the circumstance that Boltzman h -
by the forces of the system (which are the same for both cases) but
when the bifurcation of path occurs, the system, ipso facto, invokes p . by problems ansmg Wlthm statistical gas theory to move a I
' . . . ~ no reqwre t e embrace of
some determining principle which is extra physical (but not extra way 111 the same direction. Still statistics d'd t · h ong
non-mechanical ca .
natural) to determine which of the two paths it is to follow. 'bl b lusahon, much less of mdctcrminism for it was a!
When it is on the enveloping path it may at any instant, at its . 1e to e leve
ways poss1 · th e un d er1ymg
· phenomena to be ' mechanical!)'-
own sweet will, without exerting any force or spending any energy, d t
t' e ermmec
F and only the Iarge-sca Ic numencal
. 'c·l . regularities probabilis-
go off along that one of the particular paths which happens to co- IC. rane~s a ton saw no reason t f
account of the stat· t' . I I . o move away rom determinism on
incide with the actual condition of the system at that instany IS Ica c 1aracter of know led · · .
dicated b h · . ge, as IS m some way m-
In most of the former methods Dr. Balfour Stewart's, &c. there where h/ IS unres~ons.Ive res~onse to Maxwell's unsolicited missive,
he had be:~t7,~~~:~:~r~~~~wn~l~tr~sp~c~ive p~yc.hological experiments
was a certain small but finite amount of travail decrochant or trig-
ger-work for the will to do. Boussinesq has managed to reduce this
seem to o . ?.ne o n ow distinctly cause and effect
to mathematical zero, but at the expense of having to restrict cer- g vern everythmg. ~6 Boltzmann eventually accepted the im-
tain of the arbitrary constants of the motion to mathematically def-
inite values, and this I think will be found in the long run very ex- :: Undated manuscript, JCMP, 7655 /V fli J.
replyGalton to Maxwell ' 27 Feb · 1379111
is as follows: · JCMP '7 6 55/II (Box 1). The relevant text of Galton's
pensive. But I think Boussinesq's method is a very powerful one
against metaphysical arguments about cause and effect and much Very many thanks for the free will &c.-After all . . .
pnnciple with . that of unstable equilJ'b num
. 7- I am• doels thebusv
questiOn not comcidc in final
better than the insinuation that there is something loose about the rat 1cr 1· t ·1
on tl1c workmgs of my own mind • a d .am ·.a Imost fnghtcned
. 't<> us
fi dnow
h witd. 1 experiments
· l
11
laws of nature, not of sensible magnitude but enough to bring her & eftiec t seem to govern everything -If l n ow JShnct y cause
Century" (March) & care to look t. h you 1appen to sec. the forthcoming "Nineteenth
. a as ort paper m it by m "P h . ,
round in time. ;4 as you WI 11 sec describe somethin of what 1 , c syc _omctnc .acts"-it will
That Maxwell's argument about the imperfection of physical knowl- are almost out of the reach of con~ciousness i~~~~~ll~i lj~t I contmue to drag ideas that

edge and even the possible incompleteness of mechanical laws should


. I look forward With mfinite interest to--moreover g h .
SJ!Jon of your sorting demon at the R II . . -to c?r Sir W. 1 homson's expo-
,
to see art illustrations of the medievalbcv:titutwn. I have httlc doubt that some will go
have developed out of his thoughts on the implications of statistics for
physics seems completely plausible-perhaps too plausible-in retro- In "Free Will-Observations and Inferences" M. d 88
eluded even more decisively that "man is littie 11 m il9 (1 4), 4?6-413, P· 412, Galton con-
spect. To be sure, Maxwell himself came to sec this connection as a nat- must ~nderstandthe word 'spontanci 'in the JO~c.
Jan a conscwusmachine," and that "we
ural one. He often noted how much looseness there was in statistical word chance.' He thereby affir I .ty same sense that a scJcntJfic man understands the
. ms 11s Ignorance of the pr ·c · f
not m any way deny the possibility of d t . . I c Jse causes o an event, but he does
e enmnmg t 1cm."
,. Maxwell to Galton, 2.6 Feb. 1879. FGP, folder 191. l have silently closed a parenthesis
in the first paragraph. The material inside brackets is Maxwell's. 207
Statistical Uncertainty Statistical Uncertainty

plications of his statistical approach, but always ~mphasized the positive economists from Germany and Austria who sought to reinterpret statis-
content and power of its accomplishments, not Its hm1tabons. tical reasoning so as to purge it of all connection with atomism, me-
chanical determinism, and natural law. Boltzmann grew up under the
ascendancy of liberalism, and such evidence as we have suggests that he
BOLTZMANN, STATISTICS, AND IRREVERSIBILITY was a steadfast liberal throughout his life. That his political ideas did not
change seriously under the pressure of the "worker question," the Ger-
The kinetic gas theory constituted Boltzmann's life work, .and the la~­
man unification, and the depression that began in 1873 may perhaps be
guage and concepts of probability theory were central to lns researc~.m
attributed in part to the greater remove from politics that is possible for
this field from the beginning. He was never comfortable, howev~r, w1:h physicists than for economists, although it may be significant that in
the idea that the form of his mathematics implied any ind~tern:macy ll1 Austria, deductive, liberal economics held sway even when economists
the resulting laws. Boltzmann would have derived .no satisfaction from in the new Reich had moved almost unanimously to historicism. Boltz-
"picking a hole" in the second law of thermodynamics, as Maxwell had. mann, it may be observed, applauded the precision and efficiency of a
He always stressed instead the certainty of science: free economy, defending capitalism against those who derided it a; wor-
ship of Mammon. w His loyalty to a culture of Recht und Wissenschaft
A precondition of all scientific knowledge is the principle of the can be seen in his opposition to the radical pan-Germans and anti-Sem-
complete (eindeutig) determination of all natural processes, or, as ites who became increasingly troublesome for the University of Vienna,
applied to mechanics, the complete determination of all move- and indeed the whole Austrian empire, during the 188os and thereafter.
ments. This principle declares, that the movements of a bod_Y do Boltzmann's love for Austria was shown in his dedication to a cosmo-
not occur purely accidentally, going sometimes here, sorn.etimes politan empire, and his lament about 1866 as the "year of misfortune"
there, but that they are completely determined by the cncum- can be seen as reaffirmation of his liberal faith. 4o It is also worth noting
stances to which the body is subject. 37 his persistent fascination with the United States, which he visited three
As we have seen, Boltzmann invoked the analogy of gas theory wi~h s~­ time~ . . .
cial statistics in order to bolster, not undermine, the certainty ofh1s sci- In' science, Boltzmann applauded mechamcal explanation m every
entific conclusions. Nevertheless, he did not glide over diffic~l~ies-at domain. He called Darwin's theory of evolution a mechanical one, and
least not for long-but faced them with resilience and creativity. ~Is he argued that human thought, whether expressed in scientific theory
faith in molecular models was so deep as to be virtually a matter of pnn- or in fondness for music, was "mechanically necessary." As a young
ciple-he thought continuity was meaningful even in ma.thematics only physicist, he worked for a time in the lab of Hermann von Helmholtz,
as the limit of finite differentials> 8-and he held to atomism even whe.n the great champion of the ideas of the Naturwissenschaften against the
he found himself compelled by it to concede that certain macroscopic Geisteswissenschaften. Boltzmann's consistent adherence to atomism
laws, such as the second law of thermodynamics, were only statements transgressed the guiding metaphysics not only of historical economists,
of high probability. . . but also of late-century energeticists and Machian positivists like Wil-
Boltzmann was an Austrian, the son of a V1ennese tax offi~1al.. He ~as helm Ostwald and Ernst Zermclo, who held that the transformations of
born in 1 844 , about a decade later than most of the holistic lmtoncal energy perceptible in experiment and observation must be accepted as
" Boltzmann, "Uber die Grundprinzipicn und Grundglcichungen dcr Mcchanik" ( 1 t\c)9), 19 Boltzmann, "Ubcr die Principicn der Mcchanik" (1900), in PS. p. 322.
40 See Engelbert Broda, Ludwig Boltzmann: Mensch, Physiker, Philosoph (Vienna, 1955),
in PS, pp. 276-277. . h. f · .• d y , II ·ague p. 7; also Boltzmann, "josef Stefan" (1895), in PS, p. 102. On fin-de-sieclc Austrian liberalism
'" It mav be noted iu this connection that, accordmg to IS nenu an 1enna co c
Franz Clc~cns Brcntano, Boltzmann thought Joseph Bertrand's paradox of probab1hty (see ~1. and the revolts against it, see Allen Janik and Stephen Toulmin, Wittgenstein's Vienna (1\:cw
, chap. 3, above) would be cleared up once the fallacy of ~ssunm,;g a conhm1um, was .r~c­ York, 1973); William J. McGrath, Dionysian Art and Populist Politics in Austria (New Ha-
;;;nizcd. Sec Brcntano, "Von dcr Unmoglichke1t absoluten Zufalls (1916) Ill Versuch u er ven, 1974), and, especially, Carl Schorskc, Fin-de-Siecle Vienna: Politics and Culture (New
York, 1980).
die Erkenntnis (Hamburg, 1970), p. 141.
209
I
Statistical Uncertainty Statistical Uncertainty
primary, not explained as interactions of purely hypothetical molecules. tisti~al parameters, such as the product of frequencies of molecules of
During his last years-before he fulfilled his statistical destiny by taking specified typ~s. He made no allowance for variation around some mean
his own life-Boltzmann came to feel scientifically isolated. Faced with :alue, a~d h1s conclusion admitted no exceptions. He wrote in 1 8 72 : "It
a wide resistance to the reduction of communities or of perceptible phe- 1 ~ acc_ordmgly ngorously proved that, whatever the initial distribution of
nomena to individuals, whether human or atomic, he began to call kmetic energy may have been, it must always necessarily approach the
himself a "reactionary," one who has been left behind, "the last onere- Maxwellian form after a very long time has elapsed. "43
maining of those who embraced the old [mechanical scientific picture] . -~o be sure, Boltzmann had acknowledged as early as 1 868 that some
with their full souls." But he held to his principles, vowing "so much as mitial_ configurations, among them the one in which all molecules
it lies in my power, to work out in as clear and logically-ordered a way move~~ the s~me direction with the same velocity, perpendicular to the
as possible the results of the old classical theory in order that the many w~lls :V1th which the~ collide, would not converge to the Maxwell dis-
good and eternally useful doctrines which I am persuaded are yet con- ~nbutw~. 44 Th~ possibility of exceptions to his H-theorem only became
tained in it will not have to be discovered a second time." 4' ~,nteresti~g to h1m, ho,~ever, when he was confronted with the so-called
Boltzmann's earliest paper on the kinetic theory made frequent ref- reversibility paradox by his colleague and former teacher at Vienna
erence to probability, and used probability mathematics extensively, yet Jose~ Loschmidt That the second law could be violated simply bv re~
it contained, as Lorenz Kruger has pointed out, not a trace of indeter- versmg the veloCities of all particles in a closed system, causing tim~, in
minacy. Boltzmann there treated probabilities as perfectly interchange- effect, to nm_ backwards, had been discussed privately in Britain since
able with frequencies. That is, he moved from probabilities expressed as 1 867, when It was noted by William Thomson on the letter in which

the fraction of time which an arbitrary molecule would, in the long run, Maxwell ~ad first mentioned his sorting demon to their mutual friend
spend in any given state to the actual proportion of molecules in that P. G. Ta1_t. 45 Loschmidt, an atomist like Boltzmann, was inspired to
state at a given time without evincing the least awareness that this might present this argument, which he invented independently of Thomson
be problematical. Like Clausius, he invoked probability largely in order an~ Maxwel~, b~ his concern over the "heat death" that would ne~es­
to justify the introduction of simplifying assumptions so as to rise above sanly ensue m ~1me if, a~ the second law indicated, entropy must in-
the chaos of molecular motions and to attain simple, analytic expres- crease n:onotomcally untd all the energy in the universe has been con-
sions. Boltzmann believed then that he had given a rigorous analytical verted . mto the disordered motion of heat. Boltzmann quickly
proof of the second law, reducing it to a proposition in mechanicsY Al- recogn~zed that he was confronted with a deep paradox, the problem of
though his methods had become much more refined by 1872, when he rec?ncdm~ the flow of heat, which manifestly depends on the direction
introduced his H-theorem, that derivation also flowed from the as- of hme, ~1th the laws of mechanics, which do not. Elastic collisions in
sumption that frequencies were interchangeable with probabilities. a mechamcal system are perfectly reversible, but heat is alwavs observed
to flow from warm to cold bodies. '
Boltzmann presupposed, as indeed had Maxwell in his technical papers,
that collisions of every sort occurred in exact proportion to certain sta- Boltzmann responded to this objection with some of his best technical
work, b~t he rc~uired more than a decade to formulate a satisf1ctorv an-
4' Boltzmann, "'Ubcr die Entwicklung dcr Methodcn der thcoretischcn Physik in ncucrer
s:vcr to It. Dun'g the interim he offered several contradictory re~olu­
Zeit" (1899). in PS, p. 205. More generally, see Erwin N. Hiebert, "'The Energetics Contro-
versy and the New Thermodynamics,'" in Duane H. D. Roller, cd .• Perspectives in the History hons of the paradox, without ever admitting that his opinion~ had al-
of Science and Technology (Norman, Oklahoma, 1971), pp. 67-86; also Daniel Gasman. The tered m the least. He conceded immediately what he now held up as
Scientific Origins of National Socialism (New York, 1971); and, for a French perspective.
chap. 1 of Mary Jo Nyc, l\llolecular Reality (New York, 1972). W" Boltzmann, "Weitere Studien Ober das Wiirmcglcichgewicht untcr Gasmokkuler' " in
4 ' Boltzmann. "Ubcr die rnechanische Bedeutung des zwciten llauptsatzes dcr Warmcthco·

ric," in WA, vol. 1, pp. 9-35; see also Lorenz Kruger. "Reduction as a Problem: Some Re- I ~o:~/:. ~c~d~~ranslaAhont
tic;
, .. , yszca
fromSMartlin ~lcin,
"The Development of Boltzmann's Statis·
~s rzca, upp. X (1 973 ), 53 . 106, p. 73 .
44
marks on the History of Statistical Mechanics from a Philosophical Point of View,·· in Jaakko Boltzmann, Studrcn uber das Gleichgcw· ht d 1,b d. r ·
Hintikka et al., eds., Probabilistic Thinking. Thermodynamics and the Interaction of the His· materiellen Punkten"' ( 868) · WA I rc cr c en rgen Kra,t ZI\!Schcn bewegtcn
. I , ll1 , VO . I, p. 96.
tory and Philosophy of Science (Dordrecht, 1981), 147-174, pp. 152-154· 45 See Klem. "Boltzmann's Statistical Ideas"' (n. 43 ), p. 75 .

210 211
( ( Statistical Uncertainty (
Statistical Uncertainty
theoretically, only the cha yn:micfs IS one~~ thos.e principles that "have,
The second law of thermod . .
wrote, it is, for an arbitrary system at a given time "extremely probable " . rae er o propositions m prob bTty "b
that H is very near to its minimum value; if it is greater, it may increase practically equivalent to natural laws. "5> a I I ' ut are
or decrease, but the probability that it decreases is al~ys greater. "53 Al-
though any given deviation from the Maxwell distribution has a finite
By the standard of his insuran
be a highly improbable place
need for caution in a I .
· ·1 h
in~:es~m~:d
. '.
k .
el real umverse appeared to
o tzmann began to feel the
·

probability, that probability declines precipitously as the extent of the . . pp ymg statistical thermodynamics to th ,I
ld H
wor . IS attitude m d t d h e rea
ginning of the new ce~~~ry ~w~~ tAat e~unicated most clearly at the be-
deviation increases. The decline is so steep, in fact, that if one choose
some arbitrary H-value significantly above the minimum, and observes
introduced statistical m h Y. e mencan Josiah Willard Gibbs, who
all cases in which the system reaches that deviation in a very great time ec ames as a purely d d t' .
served that "th b . e uc Ive science and ob-
reg~~
interval, the chosen level will almost always be a local maximum. The ere can e no mistake in d h ·
hypothesis with the facts of n t £ to. t e agreement of the
spect. The only error into h .a ~re, or not mg IS assumed in that re-
system will exceed that level by only an insigniijcant amount-Boltz-

between the premises and t~e IC olne .can fall, is the want of agreement
mann, always disdainful of differentials, said no't at all-and then de-
cline back to its minimum. . cone uswns, and this 'th
h ope, m the main to avoid "56 B It . , WI care, one may
The observation that perceptible fluctuations from the minimum are
he did disclaim all obi' f. t o zmann did not go quite this far, but
exceedingly rare provided Boltzmann with his answer to another so- kinetic gas theory was aig; IOn fo ac.count for the actual universe. The
called paradox that arose in the mid-189os, this one called the recur- wrm o ratiOnal me h · h
not the actual universe but th 1 .c ames, w ose object was
rence paradox. It was formulated by Ernst Zermelo, an assistant in Max rations of an arbitrary e aws go:ernmg the molecular configu-
~~~assume
Planck's laboratory, who argued from a recurrence theorem by Poincare gaseous system with cert. . d
Nevertheless, Boltzmann did offer sorn properties.
that, given enough time, any system of molecules must eventually re-
the second law of thcrm d . . . e opmwns as to the place of
sibilities immediately p~es~~:~~~~~ 111 th~ observable universe. Two pas-
turn to a configuration arbitrarily close to its initial state. Since no such
may simply have begun in a h. hl e~se ves. One was that the universe
recurrences have been observed, Zermelo argued, the molecular hy-
pothesis must be false. Boltzmann answered straightforwardly that these
evolving gradually towards th~r:a;:p~~~~b.le state, from which it was
recurrences must be quite rare, and that, for example, a small volume
able, however that the
,
. f q
regiOn o space ne I b .
~mm. It was also conceiv-
h
of gas could be expected to separate spontaneously into nitrogen and ox-
sen ted a monumental fl t t' f h Ig 1 onng t e earth repre-
ygen only once in years. He wrote: . uc ua IOn o t e entropy h
10 ,o' 0 bon by man was not wholl • t . . curve, w ose observa-
y IOr mtous smce w'th 1
t th' fl
One may