Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Institute of Mathematical Statistics is collaborating with JSTOR to digitize, preserve and extend access to
Statistical Science.
http://www.jstor.org
Abstract.R. A. Fisher'saccountofthedeclineofinverseprobabilitymeth-
ods duringthe latterhalfof the 19thcenturyidentifiesBoole,Venn and
Chrystalas the keyfiguresin this change.Carefulexaminationof these
and otherwritingsof the period,however,revealsa different and much
morecomplexpicture.Contrary to Fisher'saccount,inversemethods-at
leastin modifiedform-remained theoretically untilthe1920's,
respectable
whentheworkofFisherand thenNeymancausedtheireclipseforthenext
quartercentury.
Keywordsandphrases:R. A. Fisher,inverseprobability, of
history
statistics.
247
A bag contains threeballs, each ofwhichis either The conclusion to this episode in the historyof the
whiteor black, all possible numbersofwhitebeing historyof statisticsis somewhatbizarre.Of his trinity
equally likely.Two at once are drawn at random of authorities-Boole, Venn and Chrystal-Fisher
and prove to be white;what is the chance that all thoughtBoole was an opponent of inverse methods,
the balls are white? but Boole was not; Venn was an opponent,but only
Any one who knows the definitionof mathemat- in part; and Chrystal was an unqualified opponent,
ical probability,and who considersthis question but on grounds Fisher would have found repugnant,
apart fromthe Inverse Rule, will not hesitate for had he known of them.
a momentto say that the chance is 1/2; that is to
say,that the thirdball is just as likelyto be white
as black. For thereare fourpossible constitutions 5. INVERSE PROBABILITY FROM 1880 TO 1930
of the bag:
What was the actual impact of these critics? Con-
10 20 30 40 traryto what Fisher suggests,they did not eliminate
W 3 2 1 0
inversemethods.Edgeworthand Pearson, perhaps the
B 0 1 2 3 two most prominentEnglish statisticiansof the gen-
eration immediately preceding Fisher's, both re-
each of which,we are told, occurs equally often mained sympatheticto Bayesian methods.Moreover,
in the long run,and among those cases there are we have the testimonyof Fisher himselfthat he had
two (10 and 20) in whichthereare two whiteballs, "learned it at school as an integralpart ofthe subject,
and amongthese the case in whichthereare three and for some years saw no reason to question its
white occurs in the long-runjust as oftenas the validity"(Fisher, 1936a, page 248). Indeed, he had to
case in whichthere are onlytwo. "plead guiltyin my originalstatementof the Method
of Maximum Likelihood [Fisher, 1912] to having
Chrystalthen goes on to correctlycalculate that,in
based my argument upon the principle of inverse
contrast,the "application of the Inverse Rules" leads
probability.. ." (Fisher, 1922, page 326).
to posteriorodds of 3 to 1 in favorof the third ball
The real effectof Boole, Venn, and Chrystal and
being white,and concludes:
other critics appears rather to have been to cause
No one would say that ifyou simplyput twowhite the exponents of inverse methods to hedge their
balls intoa bag containingone ofunknowncolour, claims for the theory. For example, William Allen
equally likelyto be black or white,that this action Whitworth,the author of a popular 19th century
raised the odds that the unknown ball is white textbookChoice and Chance, dealt with objections to
fromeven to 3 to 1. It appears, however,fromthe the rule of succession by conceding that expressions
InverseRule that ifwe findout that the twowhite such as "entirelyunknown" in its formulationwere
balls are in the bag, not by puttingthem in, but "vague." He proposedthat theybe replaced in the rule
by takingthem out, it makes all the difference. by the explicithypothesisthat "all possible probabil-
ities [are] equally likely,"and noted that:
Indeed it does. Chrystal'serroris exactlythe point of
the closely related Bertrand box paradox (Bertrand, Though the cases are very rare in which the
1907, pages 2-3). radical assumption of the Rule of Succession is
In the lightof this fundamentalmisunderstanding, strictlyjustified,the rule may be taken to afford
Chrystal's objections to inverse probability can a rough and ready estimate in many cases in
scarcelybe describedas intellectuallydevastating.He which the assumption is approximatelyjustified.
was merelyone of many (e.g., D'Alembert and Mill) [Whitworth,1901, page 193]
whose intellectualattainmentsin otherareas led him
to uncriticallyaccept his own untutoredprobabilistic This defenseessentiallyoriginateswithEdgeworth,
intuitions.As Jevons once noted, "It is curious how who was an importantdefenderof inverse methods
oftenthe most acute and powerfulintellectshave gone throughoutthis period (see Stigler, 1978, page 296;
astray in the calculation of probabilities" (Jevons, 1986,page 310). In 1884,at the beginningofhis career,
1877, page 213). (In 1893, shortlyafterChrystalread Edgeworthwrote a review of Venn's Logic, entitled
his paper beforethe Actuarial Society of Edinburgh, "The Philosophy of Chance," which appeared in the
JohnGovan read a paper beforethe same body,point- English philosophicaljournal Mind. (Nearly 40 years
ing out the errorsand confusionsin Chrystal'spaper. later, in the twilightof his career, Edgeworthwould
It went unpublished,however,until 1920, when the returnto the same subject withan articleof the same
eminentmathematicianE. T. Whittakerread a simi- titlein the same journal, this time reviewingKeynes's
lar expose before the London Faculty of Actuaries Treatise.) Edgeworthtook an empiricaland pragmatic
(Whittaker,1920).) view of the subject, and, as noted earlier, may well
Fisher was, in fact,being too modest when he as- ArthurKoestler's The Sleepwalkers,1959) and so is
cribedthe demiseofinverseprobabilityto Boole, Venn repeated fromone misinformedsource to another. It
and Chrystal.The two mostimportantpersons in that does not occur to someone to check the authenticity
undertakingwere none otherthan Fisher himselfand of such a story,any more than it would occur to him
Neyman. (Thus forEgon Pearson, the inverseproba- to check whether Einstein was responsible for the
bilityapproach "had been foreverdiscreditedbyFisher special theoryof relativity,or whetherWatson and
in his 1922 paper .. ." (Reid, 1982, page 79).) Human Crick discoveredthe structureof DNA.
nature being what it is, no matter how cogent or Even when a person has first-handknowledgeof
convincingthe argumentsofthe opponentsof inverse the events about which he is writing,the passage of
probabilitywere, until a credible alternativeto the time may lead to a subtle erosionin the accuracywith
Bayesian methodologywas provided,any attemptto which those events are remembered.A notable ex-
demolishthe edificeofinverseprobabilitywas doomed ample is Karl Pearson's historicalaccount of correla-
to failure(see, e.g., Pearson, 1920, page 3). tion (Seal, 1967; Plackett, 1983). As Stigler notes,
The HarvardmathematicianJulianLowell Coolidge Pearson's commentary"reflects well neither upon
was perhaps merely being more candid than most Pearson nor the generaltrustworthiness of the latter
when he wrote (1925, page 100): recollections of great scientists" (Stigler, 1986,
page 344, n. 11).
Why not, then, reject the formula outright? Under the rubric of sins of commission may be
Because, defectiveas it is, Bayes' formulais the placed an interrelatedcomplex of causes including
only thingwe have to answer certain important subconsciousbias, dogmatism,sensationalismand de-
questions which do arise in the calculus of prob- liberate distortion.Everyone "knows," for example,
ability.... Thereforewe use Bayes' formulawith that the nightbeforehe was fatallywoundedin a duel,
a sigh, as the only thing available under the the unfortunateEvariste Galois stayed up feverishly
circumstances: writingdown a sketch of his theoryof equations so
'Steyning tuk him for the reason the thief tuk that it would not be lost to posterity.In realityGalois
the hot stove-bekaze there was nothing else had publishedan outlineof his resultsmonthsearlier,
that season.' [Kipling, Captains Courageous, and although he did write furtherdetails down the
Chapter 6] nightbeforethe fatal duel, therewas not the urgency
oftendepicted. Reality does not make nearlyas good
a story as the piquant version in circulation. As
6. DISCUSSION Rothman (1982) discusses, this is not an isolated
Paradoxically,the historyof science when written incident in Galois's biography: several of the best
byscientiststhemselvesis sometimesseriouslyflawed. known accounts of Galois's life (those of Bell, Hoyle
A typologyof possible reasons forthis suggeststwo and Infeld) are marredby serious inaccuracies which
generalcategories,involvingsins of omissionand sins occur because of-rather than in spite of-the ability
of commission. of their authors to appreciate the technical achieve-
First and foremost,there may be simplya lack of ments of Galois; "the misfortuneis that the biogra-
interest,resources,time or training.A common man- phers have been scientists" (Rothman, 1982,
ifestationof this is the uncriticalcopyingof earlier, page 104). Similarly,Stigler (1982) argues that many
secondary,oftenhighlyflawedaccounts withoutcon- accounts of Bayes's originalpaper are seriouslyinac-
sulting original sources. Everyone "knows," for ex- curate;here foundationalbiases oftenled statisticians
ample, that during the Middle Ages the Ptolemaic of the statureof Pearson, Fisher and Jeffreys to mis-
modelofthe solar systemwas modifiedbythe addition read into Bayes their own viewpoints.
ofepicycleupon epicycleto artificiallyforceagreement Fisher's account ofthe historyofinverseprobability
with increasinglyaccurate experimentaldata. But in is marredforreasons fallingintoboth ofthese general
reality, nothing of the kind occurred: the original categories. Due perhaps in part to poor eyesight,
Ptolemaic model of one deferentand one epicycle Fisher was never veryscholarlyin documentingpre-
provided a remarkablygood fit to the observational vious work; this was to prove vexatious years later
data available prior to the time of Tycho Brahe; when Neyman and otherswould criticizehim fornot
indeed, given the mathematical sophistication of adequately acknowledgingEdgeworth'searliercontri-
Ptolemy's original system, more simplified models butions to maximum likelihood (Savage, 1976,
weretypicallyemployedthroughoutthe Middle Ages, pages 447-448; Pratt, 1976).
not more complex ones (see, e.g., Gingerich,1973, Nevertheless,throughouthis lifeFisher had a seri-
page 95). But this misconceptionfitspopular preju- ous interest in historical matters. Leafing through
dices about the science of the Middle Ages (see, e.g., Todhunter, he was quick to note the Bernoulli-
FISHER, R. A. (1971-74). Collected Papers of R. A. Fisher 1-5 efficiencyof maximum likelihood estimation.Ann. Statist. 4
(J. H. Bennett,ed.). Univ. Adelaide. 501-514.
FRY, T. C. (1928). Probabilityand Its EngineeringApplications.van RAMSEY, F. P. (1926). Truth and Probability.In The Foundations
Nostrand,New York. of Mathematics and Other Logical Essays (R. B. Braithwaite,
GINGERICH, 0. (1973). Copernicus and Tycho. ScientificAmerican ed.) 156-198. Routledge and Kegan Paul, London (1931).
229 86-101. REID, C. (1982). Neyman-From Life. Springer,New York.
GosSET, W. S. (1908). Probable errorof a correlationcoefficient. RoOT-BERNSTEIN, R. S. (1983). Mendel and methodology.History
Biometrika6 302-310. ofScience 21 275-295.
HAILPERIN, T. (1976). Boole's Logic and Probability.North-Holland, ROTHMAN, T. (1982). Genius and biographers:the fictionalization
Amsterdam. of Evariste Galois. Amer.Math. Monthly89 84-106.
HARDY, G. F. (1889). Letter. Insurance Record 457. (Reprinted, SALMON,W. C. (1981). JohnVenn's LogicofChance. In Probabilistic
Trans. FacultyActuaries8 180-181, 1920.) Thinking,Thermodynamicsand the Interactionof the History
JEFFREYS, H. (1939). Theoryof Probability.Clarendon Press, Ox- and Philosophy(J. Hintikka,D. Gruenderand E. Agazzi, eds.)
ford.(2nd ed., 1948; 3rd ed., 1967.) 2 125-138. Reidel, Dordrecht.
JEVONS, W. S. (1877). The PrinciplesofScience, 2nd ed. Macmillan, SAVAGE, L. J. (1976). On re-readingR. A. Fisher (withdiscussion).
London. Ann. Statist. 3 441-500.
KEYNES, J. M. (1921). A Treatise on Probability.Macmillan, SEAL, H. L. (1967). The historicaldevelopmentof the Gauss linear
London. model. Biometrika54 1-24.
KOESTLER, A. (1959). The Sleepwalkers.Macmillan, New York. SHAFER, G. (1976). A Mathematical TheoryofEvidence. Princeton
MILL, J. S. (1843). A SystemofLogic, Ratiocinativeand Inductive, Univ. Press, Princeton,N.J.
Being a Connected View of the Principlesof Evidence and the STIGLER, S. M. (1978). Francis Ysidro Edgeworth,statistician(with
Methods of ScientificInvestigation.John W. Parker, London. discussion). J. Roy. Statist. Soc. Ser. A 141 287-322.
(Many later editions.) STIGLER, S. M. (1982). Thomas Bayes's Bayesian inference.J. Roy.
MONTMORT, P. R. (1713). Essai d'analyse sur les jeux de hazards, Statist. Soc. Ser. A 145 250-258.
2nd ed. Jacques Quillan, Paris. (lst ed., 1708.) STIGLER, S. M. (1986). The HistoryofStatistics:The Measurement
NEYMAN, J. (1929). Contribution to the theory of certain test of UncertaintyBefore 1900. Harvard Univ. Press, Cambridge,
criteria.Bull. Internat.Statist.Inst. 24 3-48. Mass.
NEYMAN, J. (1934). On the two differentaspects of the represent- TODHUNTER, I. (1865). A Historyof the Mathematical Theoryof
ative method: The method of stratifiedsampling and the Probability.Macmillan, London. (Reprinted by Chelsea, New
method of purposive selection. J. Roy. Statist. Soc. 97 York, 1949.)
558-625. VENN, J. (1866). The Logic of Chance. Macmillan, London. (2nd
NEYMAN, J. and PEARSON, E. S. (1928). On theuse ofinterpretation ed., 1876; 3rd ed., 1888; reprintedby Chelsea, New York,
of certain test criteria for purposes of statistical inference. 1962.)
Biometrika20 175-240, 263-294. VON KRIES, J. (1886). Die Prinzipien der Wahrscheinlichkeits-
PASSMORE, J. (1968). A Hundred Years of Philosophy,2nd ed. rechnung.Eine Logische Untersuchung.Freiburg. (2nd ed.,
Penguin, New York. Tubingen, 1927.)
PEARSON, K. (1892). The Grammarof Science. Walter Scott, Lon- VON WRIGHT, G. H. (1941). The Logical Problem of Induction.
don. (2nd ed., 1900; 3rd ed., 1911.) Finnish LiterarySoc., Helsinki. (2nd rev. ed. Macmillan, New
PEARSON, K. (1907). On the influenceof past experienceon future York, 1957.)
expectation.Philos. Mag. (6) 13 365-378. WHITTAKER, E. T. (1920). On some disputedquestions ofprobabil-
PEARSON, K. (1920). The fundamentalproblemof practical statis- ity (with discussion). Trans. FacultyActuaries77 163-206.
tics. Biometrika13 1-16. WHITWORTH, W. A. (1897). DCC Exercises in Choice and Chance.
PLACKETT, R. L. (1983). Karl Pearson and the chi-squared test. (Reprintedby Hafner,New York, 1965.)
Internat.Statist. Rev. 51 59-72. WHITWORTH, W. A. (1901). Choice and Chance, 5th ed. GeorgeBell
PORTER, T. M. (1986). The Rise ofStatisticalThinking:1820-1900. and Sons, London.
PrincetonUniv. Press, Princeton,N.J. WINSOR, C. P. (1947). Probability and listerism.Human Biology
PRATT, J. W. (1976). F. Y. Edgeworth and R. A. Fisher on the 19 161-169.
Corment
Robin L. Plackett
Sandy Zabell deservesour thanksfordiscovering ing why Fisher could not have relied on them to
furtherdetails of what Boole, Venn and Chrystal provide consistent arguments against this form of
forexplain-
wroteonthesubjectofinverseprobability, statistical inferenceand for an analysis of how far
Fisher's claims concerningthe eclipse of inverseprob-
ability are justified. Like everythingelse connected
Robin L. Plackett is Emeritus Professorof Statistics, withFisher,mattersare indeed complex,and Zabell's
Universityof Newcastle upon Tyne. His mailing ad- paper providesa good topic fordiscussion.
dressis: 57 Highbury,Newcastleupon TyneNE2 3LN, At the height of his career, Fisher was certainly
UnitedKingdom. familiar with what mattered in developments of