11
On Types of Scientiﬁc Inquiry:
The Role of Qualitative Reasoning
David A. Freedman
One type of scientiﬁc inquiry involves the analysis of large data sets, often
using statistical models and formal tests of hypotheses. A moment’s
thought, however, shows that there must be other types of scientiﬁc
inquiry. For instance, something has to be done to answer questions like
the following. How should a study be designed? What sorts of data should
be collected? What kind of a model is needed? Which hypotheses should
be formulated in terms of the model and then tested against the data?
The answers to these questions frequently turn on observations, qualitative
or quantitative, that give crucial insights into the causal processes of
interest. Such observations generate a line of scientiﬁc inquiry, or markedly
shift the direction of the inquiry by overturning prior hypotheses, or provide
striking evidence to conﬁrm hypotheses. They may well stand on their
own rather than being subsumed under the systematic data collection and
modeling activities mentioned above.
Such observations have come to be called ‘‘Causal Process Observations’’
(CPOs). These are contrasted with the ‘‘Data Set Observations’’ (DSOs) that
are grist for statistical modeling (Collier, Brady, and Seawright, chap. 9, this
volume). My object in this essay is to illustrate the role played by CPOs,
and qualitative reasoning more generally, in a series of well-known episodes
drawn from the history of medicine.
Why is the history of medicine relevant to us today? For one thing, medical
researchers frequently confront observational data that present familiar
challenges to causal inference. For another, distance lends perspective,
221
222 David A. Freedman
allowing gains and losses to be more sharply delineated. The examples
show that an impressive degree of rigor can be obtained by combining
qualitative reasoning, quantitative analysis, and experiments when those
are feasible. The examples also show that great work can be done by spotting
anomalies, and trying to understand them.
QUALITATIVE REASONING: CASE STUDIES
FROM EPIDEMIOLOGY
Jenner and Vaccination
The setting is the English countryside in the 1790s. Cowpox, as will be
clear from the name, is a disease of cows. The symptoms include sores on
the teats. Those who milk the cows often became infected, with sores on
their hands; by the standards of the time, the illness is rarely serious. In
contrast, smallpox is one of the great killers of the 18th century.
In 1796, Edward Jenner took some matter from a cowpox sore on the
hand of dairymaid Sarah Nelmes, and inserted it into the arm of an eightyear-old
boy, ‘‘by means of two superﬁcial incisions, barely penetrating the
cutis, each about half an inch long.’’ The boy was ‘‘perceptibly indisposed’’
on the ninth day, but recovered the following day. Six weeks later, Jenner
inoculated him with matter taken from a smallpox pustule, ‘‘but no disease
followed’’ (Jenner 1798, Case XVII).
Jenner published 23 case studies to demonstrate the safety and efﬁcacy
of ‘‘vaccination,’’ as his procedure came to be called: vacca is the Latin term
for cow, and vaccinia is another term for cowpox. Despite initial opposition,
vaccination became standard practice within a few years, and Jenner
achieved international fame. By 1978, smallpox had been eradicated.
What led Jenner to try his experiment? The 18th century view of disease
was quite different from ours. The great Scottish doctor of the time, William
Cullen, taught that most diseases were ‘‘caused by external inﬂuences—climate,
foodstuffs, efﬂuvia, humidity, and so on—and . . . the
same external factors could cause different diseases in different individuals,
depending on the state of the nervous system’’ (Porter 1997, 262).
Despite such misconceptions, it was known that smallpox could somehow
be communicated from one person to another; moreover a person
who contracted smallpox and survived was generally immune to the disease
from that point on. As a preventive measure, patients could be deliberately
infected (through scratches on the skin) with minute quantities of
material taken from smallpox pustules, the idea being to induce a mild case
of the disease that would confer immunity later.
This procedure was called ‘‘inoculation’’ or ‘‘variolation.’’ It was not free
of risk: serious disease was sometimes caused in the patient, and in people
On Types of Scientiﬁc Inquiry 223
who came into contact with the patient (smallpox is highly contagious).
On the other hand, failure to inoculate could easily lead to death from
smallpox.
By the early part of the 18th century, variolation had reached England.
Jenner was a country doctor who performed variolations. He paid attention
to two crucial facts—although these facts were not explicable in terms of
the medical knowledge of his time. (i) People who had the cowpox never
seemed to contract smallpox afterwards, whether they had been inoculated
or not. (ii) Some of his patients who had been ill with cowpox in the past
still wanted to be inoculated; such patients reacted very little to inocula-
tion—
What renders the Cox-pox virus so extremely singular, is, that the person who
has been thus affected is for ever after secure from the infection of the Small
Pox; neither exposure to the variolous efﬂuvia, nor the insertion of the matter
into the skin, producing this distemper. (Jenner 1798, 6)
These two facts led him to a hypothesis: cowpox created immunity
against smallpox. That is the hypothesis he tested, observationally and
experimentally, as described above. In our terminology, Jenner vaccinated
a boy (Case XVII) who showed no response to subsequent inoculation.
Immunity to smallpox had been induced by the vaccination.
By ‘‘virus,’’ Jenner probably meant ‘‘contagious matter,’’ that being a standard
usage in his time. Viruses in the modern sense were not to be discovered
for another century. By a curious twist, smallpox and cowpox are viral
diseases in our sense too.
Semmelweis and Puerperal Fever
The time is 1844 and the place is Vienna. The discovery of microbes as
the cause of infectious disease will not be made for some decades. Ignac
Semmelweis is an obstetrician in the First Division of the Lying-in Hospital,
where medical students are trained. (Midwives are trained in the Second
Division.) Pregnant women are admitted to one division or the other,
according to the day of the week that they come to the hospital, in strict
alternation. Mortality from ‘‘puerperal fever’’ is much higher in the First
Division (Semmelweis 1981 [1860]: 356).
Eventually, Semmelweis discovers the cause. The medical students are
doing autopsies, and then examining the ‘‘puerperae’’ (women who are giving
birth, or who have just given birth). ‘‘Cadaveric particles’’ are thus
transferred to the women, entering the bloodstream and causing infection.
In 1847, Semmelweis institutes the practice of disinfection, and mortality
plummets (Semmelweis 1981 [1860]: 393–4).
224 David A. Freedman
But how did Semmelweis make his discovery? To begin with, he has to
reject conventional explanations, including ‘‘epidemic inﬂuences,’’ which
meant something different then:
Epidemic inﬂuences . . . are to be understood [as] certain hitherto inexplicable,
atmospheric, cosmic, telluric changes, which sometimes disseminate themselves
over whole countrysides, and produce childbed fever in individuals predisposed
thereto by the puerperal state. [‘‘Telluric’’ means earthly.] Now, if the
atmospheric-cosmic-telluric conditions of the City of Vienna are so disposed
that they cause puerperal fever in individuals susceptible thereto as puerperae,
how does it happen that these atmospheric-cosmic-telluric conditions over
such a long period of years have carried off individuals disposed thereto as
puerperae in the First Clinic, while they have so strikingly spared others also
in Vienna, even in the same building in the Second Division and similarly vulnerable
as puerperae?’’ (Semmelweis 1981 [1860]: 357).
The reasoning is qualitative; and similar qualitative arguments dispose of
other theories—diet, ventilation, use of hospital linens, and so forth.
Now he has to discover the real cause. In 1847, his revered colleague Professor
Kolletschka is accidentally cut with a knife used in a medico-legal
autopsy. Kolletschka becomes ill, with symptoms remarkably similar to
puerperal fever; then he dies. Again, qualitative analysis is crucial. Close
attention to symptoms and their progression is used to identify Kolletschka’s
illness with puerperal fever (Semmelweis 1981 [1860]: 391). Tracing
of causal processes comes into play as well:
Day and night this picture of Kolletschka’s disease pursued me. . . . I was
obliged to acknowledge the identity of the disease, from which Kolletschka
died, with that disease of which I saw so many puerperae die. . . . I must
acknowledge, if Kolletschka’s disease and the disease from which I saw so
many puerperae die, are identical, then in the puerperae it must be produced
by the self-same engendering cause, which produced it in Kolletschka. In Kolletschka,
the speciﬁc agent was cadaveric particles, which were introduced into
his vascular system [the bloodstream]. I must ask myself the question: Did the
cadaveric particles make their way into the vascular systems of the individuals,
whom I had seen die of an identical disease? This question I answer in the
afﬁrmative. (Semmelweis 1981 [1860]: 391–2)
The source of the infectious agent could also be a wound in a living person
(Semmelweis 1981 [1860]: 396). Once the cause is discovered, the
remedy is not far away: eliminate the infectious particles from the hands
that will examine the puerperae. Washing with soap and water is insufﬁcient,
but disinfection with chlorine compounds is sufﬁcient (Semmelweis
1981 [1860]: 392–96).
Semmelweis’ work was accepted by few of his contemporaries, due in
On Types of Scientiﬁc Inquiry 225
part to his troubled and disputatious personality, although his picture of
the disease was essentially correct. Puerperal fever is a generalized infection,
typically caused by bacteria in the group Streptococcus pyogenes. These bacteria
enter the blood-stream through wounds suffered during childbirth (for
instance, at the site where the placenta was attached). Puerperal fever can
be—and today it generally is—avoided by proper hygiene.
Snow and Cholera
John Snow was a physician in Victorian London. In 1854, he demonstrated
that cholera was an infectious disease, which could be prevented by
cleaning up the water supply. The demonstration took advantage of a natural
experiment. A large area of London was served by two water companies.
The Southwark and Vauxhall company distributed contaminated water, and
house-holds served by it had a death rate ‘‘between eight and nine times as
great as in the houses supplied by the Lambeth company,’’ which supplied
relatively pure water (Snow 1965 [1855]: 86, data in Table IX).
What led Snow to design the study and undertake the arduous task of
data collection? To begin with, he had to reject the explanations of cholera
epidemics that were conventional in his time. The predominant theory
attributed cholera to ‘‘miasmas,’’ that is, noxious odors—especially odors
generated by decaying organic material. Snow makes qualitative arguments
against such explanations:
[Cholera] travels along the great tracks of human intercourse, never going
faster than people travel, and generally much more slowly. In extending to a
fresh island or continent, it always appears ﬁrst at a sea-port. It never attacks
the crews of ships going from a country free from cholera, to one where the
disease is prevailing, till they have entered a port, or had intercourse with the
shore. Its exact progress from town to town cannot always be traced; but it has
never appeared except where there has been ample opportunity for it to be
conveyed by human intercourse. (Snow 1965 [1855]: 2)
These phenomena are easily understood if cholera is an infectious disease,
but hard to explain on the miasma theory. Similarly,
The ﬁrst case of decided Asiatic cholera in London, in the autumn of 1848,
was that of a seaman named John Harnold, who had newly arrived by the Elbe
steamer from Hamburgh, where the disease was prevailing. . . . Now the next
case of cholera, in London, occurred in the very room in which the above
patient died. (Snow 1965 [1855]: 3)
The ﬁrst case was infected in Hamburgh; the second case was infected by
contact with dejecta from the ﬁrst case, on the bedding or other furnishings
226 David A. Freedman
in that fatal room. The miasma theory, on the other hand, does not provide
good explanations.
Careful observation of the disease led to the conclusion ‘‘that cholera
invariably commences with the affection of the alimentary canal’’ (Snow
1965, 10). A living organism enters the body, as a contaminant of water or
food, multiplies in the body, and creates the symptoms of the disease.
Many copies of the organism are expelled from the body with the dejecta,
contaminate water or food, then infect other victims. The task is now to
prove this hypothesis.
According to Sir Benjamin Ward Richardson, who wrote the introduction
to Snow’s book, the decisive proof came during the Broad Street epidemic
of 1854:
[Snow] had ﬁxed his attention on the Broad Street pump as the source and
centre of the calamity. He advised the removal of the pump-handle as the
grand prescription. The vestry [in charge of the pump] was incredulous, but
had the good sense to carry out the advice. The pump-handle was removed
and the plague was stayed. (Snow 1965 [1855]: xxxvi)
The pump-handle as the decisive test is a wonderful fable, which has
beguiled many a commentator.
What are the facts? Contamination at the pump did cause the epidemic,
Snow recommended closing the pump, his advice was followed, and the
epidemic stopped. However, the epidemic was stopping anyway. Closing
the pump had no discernible effect: the episode proves little. Snow explains
this with great clarity (Snow 1965 [1855]: 40–55, see esp. Table I on p. 49
and the conclusory paragraph on pp. 51–2). Richardson’s account is therefore
a classic instance of post hoc, ergo propter hoc.
The reality is more interesting than the fable. Snow was intimately familiar
with the Broad Street area, because of his medical practice. He says,
As soon as I became acquainted with the situation and extent of this irruption
of cholera, I suspected some contamination of the water of the much-frequented
street-pump in Broad Street. . . . but on examining the water, on the
evening of 3rd September, I found so little impurity in it of an organic nature,
that I hesitated to come to a conclusion. (Snow 1965 [1855]: 38–39)
Snow had access to the death certiﬁcates at the General Register Ofﬁce,
and drew up a list of the cholera fatalities registered shortly before his
inspection of the pump. He then made a house-to-house canvass (the death
certiﬁcate shows the address of the deceased), and discovered that the cases
clustered around the pump, conﬁrming his suspicion. Later, he made a
more complete tally of cholera deaths in the area. His ‘‘spot map’’ displays
the locations of cholera fatalities during the epidemic, and the clustering is
On Types of Scientiﬁc Inquiry 227
apparent from the map (Snow 1965 [1855]: 44–45; Cholera Inquiry Committee
1855: 106–9).
However, there were a number of exceptions that had to be explained.
For example, there was a brewery near the pump; none of the workers contracted
the disease: why not? First, the workers drank beer; second, if water
was desired, there was a pump on the premises (Snow 1965 [1855]: 10).
For another example, a lady in Hampstead contracted cholera: why? As it
turned out, she liked the taste of the water from the Broad Street pump, and
had it brought to her house (Snow 1965 [1855]: 44). Snow gives many
other such examples.
Snow’s work on the Broad Street epidemic illustrates the power of case
studies. His refutation of the usual explanations for cholera, and the development
of his own explanation, are other indicators of the power of qualitative
reasoning. The analysis of his natural experiment, referred to above,
shows the power of simple quantitative methods and good research design.
This was the great quantitative test of his theory that cholera was a waterborne
infectious disease.
In designing the quantitative study, however, Snow made some key qualitative
steps: (i) seeing that conventional theories were wrong, (ii) formulating
the water hypothesis, and (iii) noticing that in 1852, the Lambeth
company moved its intake pipe to obtain relatively pure water, while
Southwark and Vauxhall continued to draw heavily contaminated water. It
took real insight to see—a priori rather than a posteriori—that this difference
between the companies allowed the crucial study to be done.
Snow’s ideas gained some circulation, especially in England. However,
widespread acceptance was achieved only when Robert Koch isolated the
causal agent (Vibrio cholerae, the comma-shaped bacillus) during the Indian
epidemic of 1883. Even then, there were dissenters, with catastrophic
results in the Hamburg epidemic of 1892 (Evans 1987).
Inspired by Koch and Louis Pasteur, there was a great burst of activity in
microbiology during the 1870s and 1880s. The idea that microscopic lifeforms
could arise by spontaneous generation was cast aside, and the germ
theory of disease was given solid experimental proof. Besides the cholera
vibrio, the bacteria responsible for anthrax (Bacillus anthracis) and for tuberculosis
(Mycobacterium tuberculosis) were isolated, and a vaccine was developed
against rabies. However, as we shall see in a moment, these triumphs
made it harder to solve the riddle of beriberi. Beriberi is a deﬁciency disease,
but the prestige of the new microbiology made investigators suspicious of
any explanation that did not involve microorganisms.
Eijkman and Beriberi
Beriberi was endemic in Asia, from about 1750 until 1930 or so. Today,
the cause is known. People need minute amounts (about one part per mil-
228 David A. Freedman
lion in the diet) of a vitamin called ‘‘thiamin.’’ Many Asians eat a diet based
on rice, and white rice is preferred to brown.
Thiamin in rice is concentrated in the bran—the skin that gives rice its
color. White rice is obtained by polishing away the skin, and with it most
of the thiamin; what is left is further degraded by cooking. The diet is then
deﬁcient in thiamin, unless supplemented by other foods rich in that substance.
Beriberi is the sequel.
In 1888, knowledge about vitamins and deﬁciency diseases lay decades
in the future. That year, Christiaan Eijkman—after studying microbiology
with Koch in Berlin—was appointed director of the Dutch Laboratory for
Bacteriology and Pathology in the colony of Java, near the city now called
Jakarta. His research plan was to show that beriberi was an infectious disease,
with Koch’s methods for the proof.
Eijkman tried to infect rabbits and then monkeys with blood drawn from
beriberi patients. This was unsuccessful. He then turned to chickens. He
tried to infect some of the birds, leaving others as controls. After a time,
many of his chickens came down with polyneuritis, which he judged to be
very similar to beriberi in humans. (‘‘Polyneuritis’’ means inﬂammation of
multiple nerves.)
However, the treated chickens and the controls were equally affected.
Perhaps the infection spread from the treated chickens to the controls? To
minimize cross infection, he housed the treated chickens and the controls
separately. That had no effect. Perhaps his whole establishment had
become infected? To eliminate this possibility, he started work on another,
remote experimental station—at which point, the chickens began recovering
from the disease.
[Eijkman] wrote ‘‘something struck us that had escaped our attention so far.’’
The chickens had been fed a different diet during the ﬁve months in which the
disease had been developing. In that period (July through November 1889),
the man in charge of the chickens had persuaded the cook at the military hospital,
without Eijkman being aware of it, to provide him with leftover cooked
[white] rice from the previous day, for feeding to the birds. A new cook, who
started duty on 21 November, had refused to continue the practice. Thirty
years later, Eijkman was to say that ‘‘[the new cook] had seen no reason to give
military rice to civilian hens.’’ (Carpenter 2000, 38)
In short, the chickens became ill when fed cooked, polished rice; they
recovered when fed uncooked, unpolished rice. This was an accidental
experiment, arranged by the cooks. One of Eijkman’s great insights was
paying attention to the results, because the cooks’ experiment eventually
changed the understanding of beriberi.
Eijkman’s colleague Adolphe Vorderman undertook an observational
study of prisons, to conﬁrm the relevance to humans. Where prisoners were
On Types of Scientiﬁc Inquiry 229
fed polished rice, beriberi was common; with a diet of unpolished rice,
beriberi was uncommon. Beriberi is a deﬁciency disease, not an infectious
disease.
The evidence may seem compelling, but that is because we know the
answer. At the time, the picture was far from clear. Eijkman himself thought
that white rice was poisonous, the bran containing the antidote. Later, he
was to reverse himself: beriberi is an infectious disease, although a poor
diet makes people (and chickens) more vulnerable to infection.
In 1896, Gerrit Grijns took over Eijkman’s lab (Eijkman suffered from
malaria, and had to return to Holland). Among other contributions, after
a long series of careful experiments, Grijns concluded that beriberi was a
deﬁciency disease, the missing element in the diet being concentrated in
rice bran—and in other foods like mung beans.
In 1901, Grijn’s colleague Hulshoff Pol ran a controlled experiment at a
mental hospital, showing that mung beans prevented or cured beriberi. In
three pavilions out of twelve, the patients were fed mung beans; in three
pavilions, other green vegetables. In three pavilions, there was intensive
disinfection, and three pavilions were used as controls. The incidence of
beriberi was dramatically lower in the pavilions with mung beans.
Still, medical opinion remained divided. Some public health professionals
accepted the deﬁciency hypothesis. Others continued to favor the germ
theory, and still others thought the cause was an inanimate poison. It took
another ten years or so to reach consensus that beriberi was a deﬁciency
disease, which could be prevented by eating unpolished rice, or enriching
the diet in other ways. From a public health perspective, the problem of
beriberi might be solved, but the research effort turned to extracting the
critical active ingredient in rice bran—no mean challenge, since there is
about one teaspoon of thiamin in a ton of bran.
Around 1912, Casimir Funk coined the term ‘‘vitamines,’’ later contracted
to vitamins, as shorthand for ‘‘vital amines.’’ The claim that he succeeded
in purifying thiamin may be questionable. But he did guess that
beriberi and pellagra were deﬁciency diseases, which could be prevented by
supplying trace amounts of organic nutrients.
By 1926, B. C. P. Jansen and W. F. Donath had succeeded in extracting
thiamin (vitamin B1) in pure crystal form. Ten years later, Robert R. Williams
and his associates managed to synthesize the compound in the lab.
In the 1930s, there were still beriberi cases in the East—and these could be
cured by injecting a few milligrams of the new vitamin B1.
Goldberger and Pellagra
Pellagra was ﬁrst observed in Europe in the eighteenth century by a Spanish
physician, Gaspar Casal, who found that it was an important cause of ill-
230 David A. Freedman
health, disability, and premature death among the very poor inhabitants of
the Asturias. In the ensuing years, numerous . . . authors described the same
condition in northern Italian peasants, particularly those from the plain of
Lombardy. By the beginning of the nineteenth century, pellagra had spread
across Europe, like a belt, causing the progressive physical and mental deterioration
of thousands of people in southwestern France, in Austria, in Rumania,
and in the domains of the Turkish Empire. Outside Europe, pellagra was recognized
in Egypt and South Africa, and by the ﬁrst decade of the twentieth century
it was rampant in the United States, especially in the south. . . . (Roe
1973: 1)
Pellagra seemed to hit some villages much harder than others. Even
within affected villages, many households were spared, but some had pellagra
cases year after year. Sanitary conditions in diseased households were
primitive: ﬂies were everywhere. One blood-sucking ﬂy (Simulium) had the
same geographical range as pellagra, at least in Europe; and the ﬂy was most
active in the spring, just when most pellagra cases developed. Many epidemiologists
concluded the disease was infectious, and—like malaria or yellow
fever—was transmitted from one person to another by insects.
Joseph Goldberger was an epidemiologist working for the U. S. Public
Health Service. In 1914, he was assigned to work on pellagra. Despite the
climate of opinion described above, he designed a series of observational
studies and experiments showing that pellagra was caused by a bad diet,
and is not infectious. The disease could be prevented or cured by foods rich
in what Goldberger called the P-P (pellagra-preventive) factor.
By 1926, he and his associates had tentatively identiﬁed the P-P factor as
part of the vitamin B complex. By 1937, C. A. Elvehjem and his associates
had identiﬁed the P-P factor as niacin, also called vitamin B3 (this compound
had been discovered by C. Huber around 1870, but its signiﬁcance
had not been recognized). Since 1940, most of the ﬂour sold in the United
States has been enriched with niacin, among other vitamins.
Niacin occurs naturally in meat, milk, eggs, some vegetables, and certain
grains. Corn, however, contains relatively little niacin. In the pellagra areas,
the poor ate corn—and not much else. Some villages and some households
were poorer than others, and had even more restricted diets. That is why
they were harder hit by the disease. The ﬂies were a marker of poverty, not
a cause of pellagra.
What prompted Goldberger to think that pellagra was a deﬁciency disease
rather than an infectious disease? In hospitals and asylums, the
inmates frequently developed pellagra, the attendants almost never—which
is unlikely if the disease is infectious, because the inmates could infect the
attendants. This observation, although far from deﬁnitive, set Goldberger
on the path to discovering the cause of pellagra and methods for prevention
or cure. The qualitative thinking precedes the quantitative investigation.
On Types of Scientiﬁc Inquiry 231
Pellaga is virtually unknown in the developed world today, although it
remains prevalent in some particularly poor countries.
Fleming and Penicillin
Alexander Fleming was working at St. Mary’s Hospital in London, under
the direction of Sir Almroth Wright, studying the life cycle of staphylococcus
(bacteria that grow in clusters, looking under the microscope like clusters
of grapes). Fleming had a number of plates on which he was growing
staphylococcus colonies. He left the plates in a corner of his ofﬁce for some
weeks while he was on holiday. When he returned, one of the plates had
been contaminated by mold. So far, this is unremarkable. He noticed, however,
‘‘that around a large colony of a contaminating mould the staphylococcus
colonies became transparent and were obviously undergoing lysis’’
(Fleming 1929: 226).
Bacteria ‘‘lyse’’ when their cell walls collapse. What caused the lysis?
Rather than discarding the plate—the normal thing to do—Fleming
thought that the lysis was worth investigating. He did so by growing the
mold in broth, watching its behavior, and trying ﬁltered broth on various
kinds of bacteria. The mold, a species of Penicillium, generated a substance
that ‘‘to avoid the repetition of the rather cumbersome phrase ‘mould broth
ﬁltrate’ [will be named] ‘penicillin’ ’’ (Fleming 1929: 227). It was the penicillin
that caused the bacteria to lyse. Fleming showed that penicillin
destroyed—or at least inhibited the growth of—many kinds of bacteria
besides staphylococcus.
Penicillin’s therapeutic potential went unrealized until Howard Florey
and his associates at Oxford took up the research in 1938 and found processes
for puriﬁcation and larger-scale production. Due to the exigencies of
World War II, much of the work was done in the U. S., where a strain of
Penicillium that gave high yields was found on a moldy cantaloupe at a market
in Peoria. (Industrial-scale development was being done at a nearby
Department of Agriculture laboratory under the direction of Kenneth
Raper, and people were encouraged to bring in moldy fruit for analysis.)
Penicillin was widely used to treat battleﬁeld injuries, largely preventing
gangrene, for example. Along with the sulfa drugs (prontosil was discovered
by Gerhard Domagk in 1932) and streptomycin (discovered by Selman
Waksman in 1944), penicillin was among the ﬁrst of the modern antibi-
otics.
CONCLUSIONS
In the health sciences, there have been enormous gains since the time of
Jenner, many of which are due to statistics. Snow’s analysis of his natural
232 David A. Freedman
experiment shows the power of quantitative methods and good research
design. Semmelweis’ argument depends on statistics; so too with Goldberger.
On the other hand, as the examples demonstrate, substantial progress
also derives from informal reasoning and qualitative insights.
Recognizing anomalies is important; so is the ability to capitalize on accidents.
Progress depends on refuting conventional ideas if they are wrong,
developing new ideas that are better, and testing the new ideas as well as
the old ones. The examples show that qualitative methods can play a key
role in all three tasks.
In Fleming’s lab, chance circumstances generated an anomalous observation.
Fleming resolved the anomaly and discovered penicillin. Semmelweis
used qualitative reasoning to reject older theories about the cause of puerperal
fever, to develop a new theory from observations on a tragic accident,
and to design an intervention that would prevent the disease. The other
examples lead to similar conclusions.
What are the lessons for methodologists in the 21st century? Causal
inference from observational data presents many difﬁculties, especially
when underlying mechanisms are poorly understood. There is a natural
desire to substitute intellectual capital for labor, and an equally-natural
preference for system and rigor over methods that seem more haphazard.
These are possible explanations for the current popularity of statistical
models.
Indeed, far-reaching claims have been made for the superiority of a quantitative
template that depends on modeling—by those who manage to
ignore the far-reaching assumptions behind the models. However, the
assumptions often turn out to be unsupported by the data (Duncan 1984a;
Berk 2004; Freedman 2005; chaps. 1 and 9, this volume). If so, the rigor of
advanced quantitative methods is a matter of appearance rather than sub-
stance.
The historical examples therefore have another important lesson to teach
us. Scientiﬁc inquiry is a long and tortuous process, with many false starts
and blind alleys. Combining qualitative insights and quantitative analysis—and
a healthy dose of skepticism—may provide the most secure
results.
FURTHER READING
Brady, Collier, and Seawright (chaps. 1 and 9, this volume) compare qualitative
and quantitative methods for causal inference in the social sciences.
As they point out,
it is difﬁcult to make causal inferences from observational data, especially
when research focuses on complex political processes. Behind the apparent
On Types of Scientiﬁc Inquiry 233
precision of quantitative ﬁndings lie many potential problems concerning
equivalence of cases, conceptualization and measurement, assumptions about
the data, and choices about model speciﬁcation. (22 this volume)
These authors recommend using a diverse mix of qualitative and quantitative
techniques in order to exploit the available information; no particular
set of tools is universally best. Causal process observations (including
anomalies and results of accidental experiments, even experiments with N
‫ס‬ 1) can be extremely helpful, as they were in the epidemiological examples
discussed here.
The role of anomalies in political science is also discussed by Rogowski
(chap. 5, this volume). He suggests that scholars in that ﬁeld may be excessively
concerned with hypothesis testing based on statistical models. Scholars
may underestimate the degree to which the discovery of anomalies can
overturn prior hypotheses and open new avenues of investigation. Anomalies
that matter have been discovered in case studies—even when the cases
have been selected in ways that do considerable violence to large-N canons
for case selection. He also suggests that failure to search for anomalies can
lead to a kind of sterility in research programs.
Scientiﬁc progress often begins with inspired guesswork. On the other
hand, if guesses cannot be veriﬁed, progress may be illusory. For example,
Snow (1965 [1855]: 125–33) theorized that—by analogy with cholera—
plague, yellow fever, dysentery, typhoid fever, and malaria (which he calls
‘‘ague’’ or ‘‘intermittent fever’’) were waterbone infectious diseases. His supporting
arguments were thin. As it turns out, these diseases are infectious;
however, only dysentery and typhoid fever are waterborne.
Proof for dysentery and typhoid fever, and disproof for the other diseases,
was not to come in Snow’s lifetime. Although William Budd (1873)
made a strong case on typhoid fever, reputable authors of the late 19th century
still denied that such diseases were infectious (Bristowe and Hutchinson
1876: 211, 629; Bristowe et al. 1879: 102–3). In the following decades,
evidence from epidemiology and microbiology settled the issue.
Plague is mainly spread by ﬂeas, although transmission by coughing is
also possible. The causal agent is the bacterium Yersinia pestis. Yellow fever
and malaria are spread by mosquitoes. Yellow fever is caused by a virus.
Malaria is caused by several species of Plasmodium, one-celled organisms
with a nucleus and an extravagantly complicated life-cycle spent partly in
humans and partly in mosquitoes. The medieval Black Death is usually
identiﬁed with modern plague, but this is still contested by some scholars
(Nutton 2008).
Buck et al. (1989) reprints many of the classic papers in epidemiology;
some classic errors are included too. Porter (1997) is a standard reference
on history of medicine. Jenner’s papers are reprinted in Eliot (1910
234 David A. Freedman
[1897]). Bazin (2000) discusses the history of smallpox, Jenner’s work, and
later developments, including the eradication of smallpox; the last recorded
cases were in 1977–78. There is a wealth of additional information on the
disease and its history in Fenner et al. (1988).
Inoculation was recorded in England by 1721 (Bazin 2000: 13; Fenner et
al. 1988: 214–6). However, the practice was described in the journals some
years before that (Timonius and Woodward 1714). It was a common opinion
in Jenner’s time that cowpox created immunity to smallpox (Jenner
1801; Baron 1838: 122). Over the period 1798–1978, techniques for producing
and administering the vaccine were elaborated. As life spans became
longer, it became clear that—contrary to Jenner’s teachings—the efﬁcacy of
vaccination gradually wore off. Revaccination was introduced. By 1939, the
virus in the vaccines was a little different from naturally-occurring cowpox
virus. The virus in the vaccines is called ‘‘vaccinia’’ (Bazin 2000: chap. 11;
Fenner et al. 1988: chaps. 6–7, esp. 278).
Bulloch (1938) reviews the history of bacteriology. Bacteria were
observed by Hooke and Leeuwenhoek before 1700. Otto Friderich Mu¨ller
in Denmark developed a workable classiﬁcation before 1800, improved
about 50 years later by Ferdinand Cohn in Germany.
Some of Koch’s work on anthrax was anticipated by Pierre Franc¸ois Rayer
and Casimir-Joseph Davaine in France. Likewise, Pasteur’s experiments disproving
spontaneous generation built on previous work by others, including
Lazzaro Spallanzani; contemporaneous research by John Tyndall
should also be mentioned.
Freedman (2005: 6–9) reports on Snow and cholera. For detailed information
on Snow’s work, see Vinten-Johansen et al. (2003). Evans (1987)
gives a historical analysis of the cholera years in Europe. Koch’s discovery
of the vibrio was anticipated by Filippo Pacini in 1854, but the implications
of Pacini’s work were not recognized by his contemporaries.
Henry Whitehead was a clergyman in the Soho area. He did not believe
that the Broad Street pump—famous for the purity of its water—was
responsible for the epidemic. He saw a gap in Snow’s argument: the fatalities
cluster around the pump, but what about the population in general?
Whitehead made his own house-to-house canvass to determine attack rates
among those who drank water from the pump and those who did not. Then
he drew up a 2 ‫ן‬ 2 table to summarize the results. The data convinced him
that Snow was correct (Cholera Inquiry Committee 1855: 121–33). Snow
made this kind of analysis only for his natural experiment.
William Farr, statistical superintendent of the General Register Ofﬁce,
was a leading medical statistician in Victorian England and a ‘‘sanitarian,’’
committed to eliminating air pollution and its sources. He claimed that the
force of mortality from cholera in an area was inversely related to its eleva-
On Types of Scientiﬁc Inquiry 235
tion. More speciﬁcally, if y is the death rate rate from cholera in an area and
x is its elevation, Farr proposed the equation
y ‫ס‬
a
b ‫ם‬ x
The constants a and b were estimated from the data. For 1848–49, the ﬁt
was excellent.
Farr held the relationship to be causal, explained by atmospheric
changes, including attenuation of noxious exhalations from the Thames,
changes in vegetation, and changes in the soil. After the London epidemic
of 1866, however, he came to accept substantial parts of Snow’s theory—
without abandoning his own views about miasmas and elevation (Humphreys
1885: 341–84; Eyler 1979: 114–22; Vinten-Johansen et al. 2003:
394).
For better or worse, Farr’s belief in mathematical symbolism had considerable
inﬂuence on the development of research methods in medicine and
social science. Furthermore, the tension between the pursuit of social
reform and the pursuit of truth, so evident in the work of the sanitarians, is
still with us.
There are two informative web sites on Snow, Whitehead, and other
major ﬁgures of the era (these sites were active as of January 8, 2010):
http://www.ph.ucla.edu/epi/snow.html
http://johnsnow.matrix.msu.edu/index.php
Loudon (2000) is highly recommended on puerperal fever; but also see
Nuland (1979) for a more sympathetic account of Semmelweis’ life. Hare
(1970: chap. 7) discusses efforts to control puerperal fever in a London
maternity hospital in the 1930s. The strain of Staphylococcus pyogenes causing
the disease turned out to be a common inhabitant of the human nose
and throat (Loudon 2000: 201–4).
A deﬁnitive source on beriberi, only paraphrased here, is Carpenter
(2000). He gives a vivid picture of a major scientiﬁc advance, including discussion
of work done before Eijkman arrived in Java.
The discussion of pellagra is based on Freedman, Pisani, and Purves
(2007: 15–16). Goldberger’s papers are collected in Terris (1964). Goldberger
(1914) explains the reasoning that led him to the deﬁciency-disease
hypothesis; Goldberger et al. (1926) identiﬁes the P-P factor as part of the
vitamin B complex. Carpenter (1981) reprints papers by many pellagra
researchers, with invaluable commentary. He explains why in Mexico a
corn-based diet does not lead to pellagra, discusses the role of tryptophan
(an amino acid that can be converted to niacin in the body), and points
236 David A. Freedman
out the gaps in our knowledge of the disease and the reasons for its disap-
pearance.
An excellent source on Fleming is Hare (1970), with Goldsmith (1946)
adding useful background. Today, ‘‘penicillin’’ refers to the active ingredient
in Fleming’s mold broth ﬁltrate. What is the cell-killing mechanism? In
brief, cell walls of most bacteria include a scaffolding constructed from sugars
and amino acids. Components of the scaffolding have to be manufactured
and assembled when the cells are dividing to form daughter cells. In
many species of bacteria, penicillin interferes with the assembly process,
eventually causing the cell wall to collapse (Walsh 2003).
Some species of bacteria manufacture an enzyme (‘‘penicillinase’’) that
disables penicillin—before the penicillin can disable the cell. There are
other bacterial defense systems too, which explain the limits to the efﬁcacy
of penicillin. Penicillin inhibits cell wall synthesis by a process that is reasonably
well understood, but how does inhibition cause lysis? That is still
something of a mystery, although much has been learned (Walsh 2003: 41;
Bayles 2000; Giesbrecht et al. 1998).
Penicillin only causes lysis when bacteria are dividing. For this reason
among others, a rather unusual combination of circumstances was needed
to produce the effect that Fleming noticed on his Petri dish (Hare 1970:
chap. 3). Was Fleming merely lucky? Pasteur’s epigram is worth remembering:
‘‘Dans les champs de l’observation, le hasard ne favorise que les esprits
pre´pare´s.’’1
Almroth Wright, Fleming’s mentor, was one of the founders of modern
immunology (Dunnill 2001). Among other accomplishments, he developed
a vaccine that prevented typhoid fever. Wright was a close friend of
George Bernard Shaw’s, and was the basis for one of the characters in The
Doctor’s Dilemma.
1. This may be liberally translated as, ‘‘In the practice of observation, chance
favors only the prepared mind.’’