The Bernoullis and The Origin of Probability Theory: Looking Back After 300 Years
The Bernoullis and The Origin of Probability Theory: Looking Back After 300 Years
The Bernoullis and The Origin of Probability Theory: Looking Back After 300 Years
--------~--------
26 RESONANCE I August 2000
GENERAL I ARTICLE
-R-ES-O-N-A-N--C-E-I--AU-9-U-s-t-2-0-0-0------------~-------------------------------2-7
GENERAL I ARTICLE
Structuring work of what we call 'statistics' today. This term was invented
knowledge in the around 1780, i.e., about 100 years after Ars Conjectandi was
right framework conceived by Jacob Bernoulli. Also, the term 'distribution of a
and posing random variable' was not known. Applying the Chebyshev in-
solvable classes of equality on distributions, the law of large numbers can be
problems, can proved very easily. This shows that general progress in science,
make originally i.e., structuring knowledge in the right framework and posing
complicated solvable classes of problems, can make originally complicated
abstract results abstract results more understandable and practically useful. In
more the following sections, we briefly describe the contributions of
understandable the Bernoulli scientists to the development of probability theory
and practically in the 17th and 18th centuries and their relationships to other
useful. mathematicians at their time.
2. Jacob Bernoulli
--------~--------
28 RESONANCE I August 2000
GENERAL I ARTICLE
1685. He found his brother Johann Bernoulli to be a good Jacob and Johann
student and a discussion partner, as Johann had studied arts and Bernoulli were the
medicine at his father's orders. In 1685 Johann finished his first to master
master of arts degree. Leibniz's method,
and it was through
The brothers studied the new infinitesimal calculus of G W
the brilliant papers
Leibniz (1646-1716), which had just been published in 1684. In
that they and
the words of G Shaffer (1982, p. 215): "They were the first to
Leibniz published
master Leibniz's method, and it was through the brilliant papers
from 1690 onwards
that they and Leibniz published from 1690 onwards that calcu-
that calculus
lus became known to the learned world". In 1687 Jacob Ber-
became known to
noulli became a professor at the University of Basel and, despite
the learned world.
his discontent with university affairs, he worked on the theory
of series, the law of large numbers and complete induction.
According to the comments of van der Waerden [5], the law of
large numbers, which was simply called the 'main theorem', was
proved almost certainly in the years 1687-1689, just after Jacob
Bernoulli was appointed professor at the University of Basel.
The work on his 'main theorem' started'around 1684, which can
be documented by written letters. The name 'law of large
numbers' was given much later by Poisson in 1837. The dating
of the discovery of the theorem can be traced by Jacob's own
testimony just before his death: 'A problem in which I've en-
gaged myself for 20 years'.
-R-ES-O-N-A-N-C-E--I-A-U-9-US-t--2-00-0------------~------------------------------2-9
GENERAL I ARTICLE
--------~--------
30 RESONANCE I August 2000
GENERAL , ARTICLE
When did Jacob start writing his book? Johann Bernoulli wrote
in a letter to Leibniz on February 16,1697: My brother has been
working for many years on a book which he calls Ars Conjectandi.
From the Meditationes it is known that all preliminary work was
-R-ES-O-N-A-N-C-E--I-A-U-9-U-st--2-00-0------------~-------------------------------~
GENERAL I ARTICLE
In games which finished in 1690. Why did Jacob leave the book unfinished?
are ruled by Many people thought that Jacob was not convinced by his
chance the events examples. Did he delay his work because he looked for convinc-
are uncertain; but it ing examples? In the last years before his death he worked again
is possible to on his book. On October 3, 1703 he wrote to Leibniz: '1 have
determine how completed the larger part of my book, but the most important
much closer part is missing, where I show how the basics of Ars Conjectandi
someone is to win can be applied to civil, moral and economic matters'.
than to lose. The publication process after his death was delayed because the
family was suspicious about Johann and the fact that in the last
years the brothers were involved in fights over mathematical
originality and the relationship turned bitter. Jacob's only son,
a painter, started to bring together all the related material with
the help of Varignon, the faithful disciple of the Bernoullis.
Nicolaus was the editor of the whole project.
-32------------------------------~-------------------------------
RESONANCE I August 2000
GENERAL I ARTICLE
ability theory was no formal theory at all at that time. People had The new calculus
the notion of determining chances by mathematical tools, like merits being
combinatorics, and called them 'a priori probabilities' since named after the
they are determined without looking at empirical facts. The law two Bernoullis with
of large numbers was seen to be an instrument for determining the same right as
probabilities 'a posteriori', i.e., from empirical observations. If it being named after
is not possible to determine probabilities a priori then one can himself.
find them by observing an experiment long enough and Jacob
Bernoulli's theorem (the law of large numbers) makes sure that
you can measure probabilities with increasing precision, even if
it takes a very, very ('eternally') long time. The history of
statistics has shown that much time had to elapse before it was
possible to implement this idea in practice, as for example we use
it today. The notion of precise reasoning is an ideal, difficult to
rationalize and very complicated to implement in everyday life.
Jacob Bernoulli published about 60 papers but no books. Be-
cause of his scientific success he became a member of the Paris
Academy of Science in 1699. The bookArs Conjectandi appeared
in 1713, eight years after his death on August 16, 1705, with the
help of his cousin Nicolaus and his brother Johann, who suc-
ceeded him in 1705 as a professor at the University of Basel. The
epitaph of Jacob Bernoulli can be found in the cloister of the
'Muenster'in Basel and in the bottom part of the frame one can
see the logarithmic spiral with the words: Eadem mutata resurgo
(I resurrect changed in the same way).
--------~-------
RESONANCE I August 2000 33
GENERAL I ARTICLE
Here are some common caveats: By the law oflarge numbers you
can't improve the uncertainty of the basic random experiment.
You only know that you are measuring the basic probability
with more and more precision. Even this creates many confu-
sions. The law only states the convergence of the empirical
proportion if the observations are generated from a constant
true process. But in absolute terms the deviations can become
larger and larger and the improvement in precision does not
always hold for absolute values. Many lotteries publish tables
which show the frequency of the numbers drawn. The implicit
appeal is that numbers which are drawn less might have a better
chance to be drawn next or numbers which have occurred too
often are not likely to show up in the near future. It is very
difficult to explain to laymen that this is not a consequence of
the law of large numbers. The basic random drawing mecha-
nism which is independent of the past stays the same. The
-34-------------------------------~------------R-E-S-O-N-A-N-C-E---1-A-u-gu-s-t-2-0-0-0
GENERAL I ARTICLE
chance to draw the number which occurred least (or most) is The law of large
exactly the same as it was for the first time or any other time in numbers just
the past. The law of large numbers just states that the measure- states that the
ment of the underlying parameters of the random experiment measurement of
can be made more precise - in relative terms. It could be that the underlying
Jacob Bernoulli was the first one to see this new road of science parameters of the
which led to a probabilistic revolution in empirical research random
work in the last century. Loosely speaking, the law of large experiment can be
numbers has a learning interpretation from a philosophy of made more
science point of view: Working with observations which are precise - in
exposed to uncertainty and measurement errors will not create relative terms.
knowledge accumulation with absolute certainty. In the words
of Jacob Bernoulli, empirical knowledge has to be judged with
'moral certainty'. Like the definition of convergence of series,
moral certainty is a concept which implies a thought experiment
that probabilities could possibly be measured with increasing
precision. But the price to pay is that the number of experiments
which have to be repeated under the same conditions could
become unrealistically large to be useful as a practical concept.
This shows that the frequentistic notion of probability has its
limits: How should we measure the probability of an event
where a repeated empirical measurement is not possible? It was
about 100 years later that Thomas Bayes (1702 -1761) and Simon
Laplace (1749-1827) found other notions of probability which
gave a totally different interpretation for 'a priori' and 'a poste-
riori' probabilities.
3. Johann Bernoulli
--------~--------
RESONANCE I August 2000 35
GENERAL I ARTICLE
4. Daniel Bernoulli
-36------------------------------~------------R-E-S-O-N-A-N-C-E--1-A-U9-U-s-t-2-0-0-0
GENERAL I ARTICLE
Strasbourg and wrote a dissertation de respiratione on the me- 'I feel robbed of my
chanics of breathing in 1721. He loved to study mathematical total
problems and he published his exercitationes mathematicae in Hydrodynamica for
Venice in 1724, a result of his studies in Italy, which consists of which lowe not an
four rather independent topics. One topic was the integration of iota to my father
a differential equation which was later named after Riccati (1676 and I have lost in a
-1754). Nevertheless the work attracted so much.attention that single moment the
in 1725 he was offered a position in St. Petersburg (for mechan- fruits of a 10-year
ics) at the newly founded Academy of Science. The story goes long work.
that it was not clear which Bernoulli got the offer from the
Academy. It could also have been Nicolaus II (born in 1695)
who could speak 4 languages at the age of 8 and who was a
professor of law in Bern. So, their father Johann I asked for
clarifications and as a result both the Bernoulli brothers got
offers from the Academy. Daniel travelled with his elder brother
Nicolaus II to work in St.Petersburg in 1725. Unfortunately,
Nicolaus II died after one year at the age of 31 from fever and this
led Daniel Bernoulli to look for other appointments. In 1733 he
returned to Basel after eight very fruitful years at the Imperial
Academy in St. Petersburg. At the University of Basel, he first
got the chair for anatomy and botany and the chair for physics
in 1750. His main work is the book Hydrodynamica which was
completed by 1734 but not published until 1738. In this book
the basic equation, Bernoulli equation of fluid dynamics is
given. His father Johann I Bernoulli wrote about the same time
the book Hydraulica which appeared in 1742 and he predated it
to 1732 - just to ensure priority for himself. To Daniel's big
surprise he found in this book several results related to his own
work. Daniel got the impression that his father wanted him to
feel that he had known these results much earlier. Since Daniel
saw himself in the position of a plagiarist, he wrote in a letter to
L Euler: 'I feel robbed of my total Hydrodynamica for which I
owe not an iota to my father and I have lost in a single moment
the fruits of a 10-year long work'. After Daniel left St.Petersburg,
he began an intensive and fruitful scientific collaboration with
L Euler another citizen of Basel) between 1726 and 1767. The
other two main scientific letter exchanges were with Christian
-R-ES-O-N-A-N-C-E--I-A-U-9-US-t--2-00-0------------~-------------------------------37
GENERAL I ARTICLE
What is the fair Goldbach (1690-1764) and his nephew Johann III (1744-1807).
price of a game Daniel Bernoulli had three rather different scientific working
with infinite fields: medicine, mathematics including probability theory,
mathematical and physics, especially mechanics. Here we concentrate on his
expectation? contributions to probability theory and statistics. His most
important work in this area was the book 'Specimen Theoriae
Novae De Mensura Sortis' (literally: Exposition of a new theory
for the measurement of chance; but often just called Probability
Theory) published in 1738 which was started during his stay in
St. Petersburg. An English translation of 'De Mensura Sortis'
appeared in Econometrica (1954), Vo1.22, 23-36. Here he started
the theory of utility called 'emolumentum' in Latin (see Vol. 2
of the collected work of D Bernoulli) but also called 'moral
value', 'moral expectation' (or in older versions 'hope'). The idea
for this paper was from his cousin Nicolaus: What is the fair
price of a game with infinite mathematical expectation? (Recall
the 'St.Petersburg Paradox' which appeared in Resonance, Vol.
1, No.9, 1996 and Vol. 2, No.1, 1997.) In 1760 he treated a
medical problem (today: epidemiology) in a similar way: What
is the rate of mortality of smallpox across age groups? He
touched also the realm of demography by determining the
average length of a marriage for every age. He also wrote a paper
on the question: Are the angles of the planes of the planets
around the sun randomly distributed? The solution he offered
was obtained by what we call today a geometrical probability
approach. The maximum observed difference was an angle of 6
degrees and 54'. What is the probability that the other planes fall
in this interval 'randomly'? Daniel Bernoulli also won 10 prizes
(some of them shared) for the competition of the Paris Academy
which shows the popularity of his research work.
5. Nicolaus Bernoulli I
Are the angles of
Nicolaus Bernoulli I (October 21, 1687 - November 29, 1759)
the planes of the
was the son of Nicolaus Bernoulli and studied mathematics with
planets around the
his two uncles, Jacob and Johann. He started his dissertation De
sun randomly
usu artis conjectandi in jure (the use of probability theory in law)
distributed?
in 1704 with Jacob Bernoulli. He finished his 'licenciat' at the
-38------------------------------~------------R-E-S-O-N-A-N-C-E--1-A-U9-U-s-t-Z-o-o-o
GENERAL I ARTICLE
-R-ES-O-N-A-N-C-E--I-A-U-9-U-st--2-00-0------------~~-----------------------------3-9
GENERAL I ARTICLE
black pebble is drawn. The question is, can you do this so often
that it becomes 10 times, 100 times, 1000 times, etc. more pro-
bable (that is morally certain) that the number of white and
black pebbles chosen are in the same 3:2 ratio as the pebbles in
the urn, rather than any other different ratio?'
--------~--------
40 RESONANCE I August 2000
GENERAL I ARTICLE
-R-ES-O-N-A-N--CE--I-A-U-9-U-s-t-2-0-0-0------------~-------------------------------4-1
GENERAL I ARTICLE
Suggested Reading
[1] Daniel Bernoulli, De mensura sort is, Econometrica, 22, 23-36, 1954.
[2] C C Gillispie (ed.), Dictionary 0/ Scientific Biography, Vol. 2 (H Berger-
Ch. Buys Ballot), New York: Charles Scribner's Sons, N ew York, 1970.
[3] G Shaffer, The Bernoullis in Encyclopedia 0/ Statistical Sciences,
S Kotz and N L johnson (eds.)Wiley, New York, 1982.
[4] G Shaffer, The significance of jacob Bernoulli'sArs Conjectandi for the
philosophy of probability today, Journal 0/ Econometrics, 75, 15-32,
Address for Correspondence 1996.
Wolfgang Polasek
[5] B L van der Waerden, (ed.) Die Werke von jacob Bernoulli, Vol. 3,
Wirtschaftswissenschaftliches
Wahrscheinlichkeitsrechnung, Birkhauser Verlag, Basel, 1975.
Zentrum
[6] S M Stigler, The history 0/ statistics: The measurement 0/ uncertainty
Statistik und Okonometrie
be/ore 1900, Harvard University Press, Cambridge, 1986.
University of Basel,
[7] S M Stigler, The Bernoullis of Basel, Journal o/Econometrics, 75, 7-14,
Holbeinstr. 12, CH-40S1 Basel,
1996.
Switzerland.
Email:
[email protected]
--------~--------
42 RESONANCE I August 2000