Geniuses or Bad Scientists?

A book entitled Betrayers of the Truth, published in 1982 by two Science investigative journalists, William Broad and Nicholas Wade (Broad and Wade Reference Broad and Wade1982) had an extraordinary impact on the scientific community and outside it. The book dealt with news that was gaining momentum at the time: the fraudulent behaviour of some scientists. Up to that moment, scientists had been traditionally considered as lay priests surrounded by an aura of intellectual nobility, only interested in the quest for truth. In the wake of clamorous cases which had rocked the scientific community, the book soon became obligatory reading for all those dealing with the matter of scientific fraud. The book has an Appendix, which lists 34 cases of documented, or suspected, scientific fraud. The list includes a mixture of very important historical cases in the distant and less distant past, as well as cases of scientists of lesser status of that time. It also lists scientifically unlikely names, such as that of Admiral Peary, the American explorer who allegedly reached the North Pole in 1909 when he instead still was – this is the accusation – hundreds of miles away from it. Three of the 34 case have no responsible individual, thus, the real list consists of 31 names. Among the great figures of the distant past, the Appendix includes, for instance, Ptolemy, the second-century ad astronomer who lived in Alexandria. In his famous book Almagest he described astronomical observations which could not possibly be done from Alexandria. Thorough reconstructions had indeed demonstrated that they had been made four centuries before, by Hipparchus from the island of Rhodos. Ptolemy had appropriated those observations without giving credit to the scientist who had actually made them. As is well known, for 1500 years, up to the arrival of Copernicus, Ptolemy had been the reference author for any western astronomer and, until the last century, nobody had raised doubts about his observations. In addition to Ptolemy, the Appendix’s list of historical cases of suspected fraud include, among others, Galileo, Newton, Dalton, Mendel, Millikan: all of them had proposed scientific hypotheses of extreme importance that history had later proven correct.

Ptolemy aside, according to the culpability verdicts of Broad and Wade, it could paradoxically be concluded that knowledge has not advanced by the application of the scientific method, but in a significant measure thanks to the intuitions of single individuals who, at the cost of altering experimental results, have forcefully supported hypotheses that were later proven correct. Continuing the paradox, one should conclude that many of the great historical advances of scientific knowledge were eventually produced by scientists we should now consider fraudsters. The final corollary being that experimental results could be manipulated ad hoc to force them to agree with a preconceived idea. Scientists would thus be ingenious bettors, and the best among them, i.e. those who are now universally considered fathers of modern science, being those who somehow formulated hypotheses that will be demonstrated valid in future centuries, not necessarily those who succeeded in irrefutably demonstrating something. The founding fathers may have violated the canonical prescriptions of the scientific method, but what really matters is that their intuition has been proven correct.

Fortunately, things are different from the paradoxical considerations above. Provided that a correct analysis is applied, and we have tried to provide it, almost all the famous historical cases quoted by Broad and Wade in their Appendix will not fall within the category of the modern definition of scientific fraud. By using this definition, the difference between these cases and real scientific fraud is easily perceived, as the latter in no case produces advancement of knowledge: it actually puts at risk the very foundations of science.

Scientific Fraud and its Gravity: An Operational Definition and the Incremental Procedure to Judge It

From an epistemic angle, only very recently have attempts been made to define what scientific fraud actually is. As recently as 1991, David Goodstein, one of the pioneers of the topic, who has authored the first set of rules of scientific fraud, at Caltech, wrote ‘One of the reasons that nobody knows the exact extent of scientific fraud is that nobody knows exactly what scientific fraud is’ (Goodstein, Reference Goodstein1991). In particular, with respect to the historical interest of defining fraudulent behaviours, such as for instance that of Charles Babbage in the nineteenth century (Babbage, Reference Babbage1830), the very concept of scientific fraud has now been enlarged, producing guidelines and documents of general importance that are no longer linked to a single author: for instance, the Singapore Declaration of 2010 (Resnick and Shamoo Reference Resnick and Shamoo2011), the European Code for Research Integrity (All European Academies, ALLEA 2017) or even, at the level of single nations, the Guidelines for Research Integrity of the National Research Council of Italy (Caporale and Fanelli Reference Caporale and Fanelli2013). The main aim of these documents is to draw attention to behaviours that concur with the definition of scientific fraud: however, the behaviours that are considered necessary to the definition are, per se, not necessarily sufficient to talk of fraud, which is the most serious form of scientific misconduct. Other forms of misconduct, e.g., sloppy experimental work, incorrect reporting of results, failure to properly cite the work of others, are one step below fraud in a scale of seriousness.

As will be seen in this contribution, important points in the definition of fraud are the intentionality, the quantitative violation of the standard of the area (which determines the degree of seriousness with respect to standards) and even the personal advantage and/or the damage to others, which would demand the intervention of the judiciary. Operationally, it must be ascertained whether:

1. (1) specific behaviours have occurred (if yes, misconduct is found),

2. (2) whether these behaviours are intentional and, in particular, whether they are aimed at creating a fictitious support of a given hypothesis (if not, these are only cases of serious errors and/or negligence),

3. (3) whether these behaviours violate in quality and quantity the standards of the time in which they have occurred (if yes, these are cases of scientific fraud),

4. (4) whether these behaviours have damaged others and/or produced an advantage to the author of the behaviour (if yes, appeal to the judiciary power is possible).

It must be stressed that the assessment procedure of the seriousness of a finding is incremental, and goes from simple misconduct, which could be due to negligence, to cases in which appeal to the judiciary powers could be considered. It must be pointed out that, for the scientific community, points (1) to (3) above are sufficient for a definition of scientific fraud, whereas very few judiciary systems are invoked for fraud without the positive answer to point (4).

In support of the theoretical minimal framework outlined above, it is sufficient to consider the most robust, and most widely quoted, procedure for the assessment of scientific fraud: that adopted by the USA Government. Article 1 of its current version states:

This first article corresponds to points (1) and (2) of our list above, i.e., it identifies and describes the behaviours that fall within the triad which has now become universally known by the acronym FFP: Fabrication, Falsification, Plagiarism. As one of the authors of this article has extensively discussed elsewhere (Bucci Reference Bucci2015) the triad unequivocally binds scientific misconduct to behaviours connected with experimental procedures and their results, leaving aside equally questionable behaviours, e.g. ethical violations in the field of animal experimentation, or the inclusion among the authors of articles of persons who have not contributed to the work. In essence, misconduct, and thus fraud, are only linked to bogus production, counterfeiting and plagiarism of scientific data, irrespective of the conclusions that are reached.

The procedures mentioned so far do not include honest errors in the definition of scientific misconduct. This demands an additional comment on what distinguishes honest errors (or negligence) from fraudulent behaviour when one or more FFP behaviours are present. In the world of law, wilful misconduct as opposed to simple fault demands the positive intention to behave in a way in which the FFP behaviour is planned to create fictitious support for a given scientific hypothesis, or to generate the false idea that a given result has been obtained by the author, or that it is his original product (in the case of plagiarism). For instance, it is rather common to find that the results of two or more different experiments are represented in different papers with the same image, or in experimental conditions that could not possibly have produced them. In these cases, the accusation of fraud demands the proof that the double utilization of images for the description of different experiments had been wilfully made in the absence of the genuine experimental data. This is an important point, as the accused author can frequently demonstrate that the duplication of images was due to a sporadic material error, not to its serial repetition.

This brief coverage of points (1) and (2) in the list above has distinguished between honest errors and wilful or negligent misconduct, and has analysed the technical aspects of the FFP behaviours involved. The details of the procedure to decide on the transition from professional misconduct to real scientific fraud (point (3)) are described in Article 3 of the Procedure of the Federal USA Code mentioned above. It states the following:

A finding of research misconduct requires that:

• There be a significant departure from accepted practices of the relevant research community; and

• The misconduct be committed intentionally, or knowingly, or recklessly; and

• The allegation be proven by a preponderance of evidence.

Here, additional points are introduced to prove that scientific misconduct has occurred and, when verified, that they transform it in true scientific fraud. First of all, a threshold is introduced in point (1), above which an ascertained FFP behaviour reaches the level of misconduct: it must exceed the accepted standard of the scientific community. It is thus not an absolute threshold, but it varies in each sector and in each time, since the reference scientific community is the one active at the time in which a given behaviour occurs.

To better understand the importance of this new parameter, let’s consider two examples, i.e. the usefulness of thresholds in cases that are often controversial – cases of plagiarism. Plagiarism is established when text sentences, or portions of text, are reproduced more or less completely from preceding sources, which are not properly acknowledged. This, however, is not sufficient to establish misconduct because, even if the behaviour is intentional, it must be clarified whether, in the reference community, a rule permits reproducing texts by others without appropriate acknowledgment and in what quantity. Extensive recent studies have established that the reproduction of given texts is not only absolutely allowed, but it may even be useful to ease the repetition of the described results (one could think of the description of methodologies). They have also indicated that the average quantity of plagiarized texts varies with the disciplines and is always more than zero (Horbach and Halffman 2017). By checking a text under analysis against the general percentage of the discipline at that moment it is possible to find out whether the discovered plagiarism defines misconduct and, possibly, fraud. In addition, real plagiarism must be distinguished from auto-plagiarism, since the allowed thresholds standard for the two are different. One could perhaps also mention the so-called ‘subconscious plagiarism’ that should be active in the psychological condition cryptomnesia (Kellogg Reference Kellogg1994).

The other example shows clearly that the standards of given communities vary with time, implying that the misconduct judgements must be those that would have been prevailing at the time of the possible misconduct event. In many articles of biochemistry and molecular biology, results are presented of a particular type of experiment that uses the technology gel electrophoresis. This has a number of possible variants, all of which need to be documented by photographic reproduction of the result. However, sometimes researchers compose a collage, using different portions of the original photographs, as if it were a unique photographic image. Naturally, using software such as, for instance, Adobe Photoshop, dishonest researchers can put together portions of totally different experiments to ‘demonstrate’ that given results have been obtained in a single experiment. And, if they do a good job, the collage will be nearly indistinguishable from the photograph of a single experiment. For this reason, in 2004, a number of scientific journals in the area of biochemistry and molecular biology have started to prohibit, in their guidelines for authors, the construction of these collages, and are asking authors to explicitly specify if a given image corresponds to a single experiment or to a collage (Rossner and Yamada Reference Rossner and Yamada2004). In addition, after about 2008, most biochemistry and molecular biology journals have adopted this new standard. Hence, when examining a gel electrophoresis image that the analysis shows to be a collage, the date of its publication must be considered. If it is prior to 2004, or 2008, i.e., when the standards were different from today’s, retroactive judgement cannot obviously by applied. Should the same researcher use the collage procedures in papers published before, and after, 2004 (or 2008) he would be judged differently.

Points (2) and (3) of the Code insist on the intentionality, but introduce alternative scenarios for misconduct. The FFP behaviours must be known to all involved researchers, and/or they must occur ‘recklessly’, i.e. without consideration of their possible consequences. This last specification is interesting, as it leaves open the decision on whether the misconduct was wilful or simply the product of negligence.

Point (3) of the Code deserves an additional comment. It says that misconduct must be proven by a ‘preponderance of evidence’: the level of certainty that must be attained is that of civil law not that of criminal law, in which the jury must be convinced ‘beyond any reasonable doubt’. This means that, from an epistemic standpoint, the evidence of misconduct must be better founded and better integrated in a coherent picture than that favouring innocence. Therefore, it is not necessary to rule out small and/or hypothetical doubts, as is mandatory in criminal law.

As for point (4) of the procedure we have outlined above, which aims at deciding whether a given case of scientific fraud is or is not a crime, the possibilities vary with the legislative systems. There are US States which consider scientific frauds akin to any other criminal fraud, and prosecute it accordingly. Other States only do so when the scientific fraud has resulted in the misappropriation of public or private resources. And there are States which act case by case; however, only doing so when the scientific fraud has demonstrably damaged other parties or secured an advantage to the fraudster. Finally, there are States that never intervene with their judiciary powers, but leave cases in the hands of the involved institutions, which sanction the fraud internally according to rulings of a more or less general type.

Having what has been presented and discussed above in the background, we may now enter in medias res to examine in detail the cases of some of the ‘excellent defenders’ Broad and Wade have listed in the Appendix of their book.

Ptolemy, Galileo, Newton, Dalton, Medel, Millikan: Have They Been Fraudsters?

Using the incremental approach discussed above, we may now begin to dissect the evidence of FFP behaviour by some of the most important historical figures collected by Broad and Wide.

Let’s begin with Ptolemy, who was accused of plagiarism for having used the astronomical observations by Hipparchus without giving him any credit (incidentally, Hipparchus himself has been accused of having published a star catalogue taken from Babylonian sources as if it were the result of his own observations). The discussion of Ptolemy’s case is largely based on the complex reconstruction of the case made at the end of last century by Hetherington (Reference Hetherington1997). The first accusation of plagiarism is based on the systematic error of about one longitudinal grade in his catalogue of the stars. Thus, he had convinced Arab astronomers of the tenth century ad that he had not performed any observation, but had used the data of others. The suspicion was shared by Ticho Brahe in the sixteenth century ad, by Lalande in the seventeenth century, and by Delambre in 1817. The accusation was the following: to take into account the precession elapsed from the time of Hipparchus, and assuming that it would cause a shift of one longitudinal grade per century, Ptolemy had added 2 ^2/3° to the longitude of the catalogue of his predecessor. However, the shift caused by the precession in the period of time elapsed between Hipparchus and Ptolemy is in reality 3 ^2/3° which means that Ptolemy’s Tables attribute to the described stars a longitude one grade lower than the true one. This systematic error had led to the conclusion that Ptolemy had taken the data from Hipparchus, accommodating them to take the precession into account (Evans Reference Evans1987a, 1987b). However, there are alternative explanations for Ptolemy’s error: he had measured the longitude of the stars from that of a reference star, which in turn was inferred from that of the Sun. If one takes into account that the position of the Sun had been calculated by Ptolemy with an error of one longitudinal grade, the discrepancy of the Sun could easily propagate to all other stars of his catalogue. There also are other explanations for the transformation of the complex of Ptolemy’s data that would possibly rule out plagiarism; thus, the problem of the systematic errors in his catalogue is still not solved. However, there is another proof that Ptolemy’s stars were in reality Hipparchus’. In 1982, Dennis Rawlins found out that the stars described in the catalogue (magnitude, refraction, height on the horizon) could be easily observed from Rhodes, where Hipparchus lived, but not from Ptolemy’s Alexandria (90% versus 10% probability, Wlodarczyk Reference Wlodarczyk1987). It was a strong argument, but, unfortunately, Rawlins had only used 11 of the stars of Ptolemy’s catalogue. He had one additional argument against Ptolemy from the conclusions on error predictions to be expected if Ptolemy’s observations would have been made, as stated, with an armillary astrolabe: in brief, in addition to the 1° longitude error, a latitude error of a half grade with a periodical pattern would have been expected. But this was not present, reinforcing the conclusion that the observations had not been made as described (Rawlins, Reference Rawlins1982). However, alternative explanations have been proposed to explain the lack of the expected latitude error (Evans, Reference Evans1987b, Wlodarczyk, Reference Wlodarczyk1987).

The back and forth list of attacks and defences on Ptolemy’s stars catalogue seems not to have come to an accepted end. However, a dispute of a different type must also be considered, as it could lead to an accusation different from plagiarism. Ptolemy described the accurate observation of an equinox on 26 September of 139 ad, which Gingerich (Reference Gingerich1980) reluctantly writes he could not have made because of a 30 hours error. The pretended observation is easily explained with the extrapolation of data – again – of Hipparchus. Even if procedurally incorrect, the extrapolation flawlessly indicates the fabrication of an experimental result: an FPP behaviour more serious than plagiarism.

Let us move ahead by 15 centuries, to the father of the scientific method, Galileo. In his case, the accusation concerns two of the most famous mechanics experiments of all time: the measurements of the fall of different bodies from the leaning tower of Pisa, meant to deny Aristotle’s theory that bodies of different weight would reach the soil at different times, and the 1604 inclined plane experiment, which would have been then final blow to Aristotle’s theory. The experimental setup of Galileo was criticized by Koyré (Reference Koyré1953) as too inaccurate to produce the described results, and the criticism was later supported by Feyerabend in is most famous book (Feyerabend, Reference Feyerabend1974). In addition, in 1610 Benedetto Castelli had written to Galileo that, if the heliocentric theory was right, Venus, which rotated around the Sun, would have phases, similar to the Moon. Since, only six days after Castelli’s letter, Galileo had sent a series of encrypted letters to a number of people, including the Medicis, stating that he had been able to observe the phases of Venus, thus solving one of astronomy’s most important problems, Richard Westfall deduced that the observations had never been done, and that Galileo had stolen the idea from his disciple (Westfall Reference Westfall1985). The three accusation imply the fabrication of results of never-performed experiments, the last one also implying plagiarism of Castelli’s idea. It must be noted the three accusations are not based on evidence, but on deductions according to which the experiments couldn’t have been performed by Galileo. Actually none of the accusations stand up to the analysis of the facts. As for the experiment of the leaning tower of Pisa, its description was not written by Galileo, but by his first biographer, his disciple Viviani. Who, by his description, created an apocryphal legend. Koyré’s accusation on the inclined plane experiment said that the rudimentary water clock used to measure the duration of the journey of the spheres on the inclined plane could not possibly have yielded the highly precise data annotated by Galileo in the notebook which he used to derive his law. However, in 1961, an American PhD student reconstructed the setup of Galileo in all its details, obtaining exactly his results (Settle Reference Settle1961). The experiment was repeatedly confirmed (Riess et al. Reference Riess, Heering and Nawrath2005).

The last accusation is that by Westfall on the phases of Venus, which would have been impossible to observe in the six days between Castelli’s letter and the encrypted letters by Galileo. It would indeed have been impossible. However, Paolo Palmieri (Reference Palmieri2001) has convincingly shown that Galileo’s observations had probably been made months before Castelli’s letter. It is thus plausible to suggest that he had rushed to communicate his discovery not to lose the priority. In conclusion, then, the three accusations against Galileo do not stand the test of facts.

The next accusation concerns Newton’s prediction, and its experimental verification, of the speed of propagation of sound in the air. Newton was the first to realize that the speed could be deduced by assuming that the means within which the sound propagates (the air in this case) behaves like a spring, and formulated his famous equation that relates the speed of sound to the elasticity and inertia of the air. However, to measure the constant of elasticity of the air he performed a series of measurements of compression of air of isothermic type, whereas the compression of the air caused by the propagation to the sound wave is of adiabatic type (which, incidentally, was only discovered 200 years later). The constant used by Newton thus turned out to be under-evaluated, with the result that the measurement of the speed of sound turned out to deviate from that predicted by the equation by about 18%. To explain the discrepancy, Newton hypothesized that two factors, i.e., the dimension of the molecules of air, and the presence of vapours, influenced by unknown mechanisms the measured values by exactly the amount of the discrepancy. The accusation by Sheldon Lee Glashow (Nobel Laureate 1979) and, once more, Richard Westfall (in Glashow Reference Glashow2008) runs as follows:

Very strong words, indeed. However, in 1991 one of the leading scientific integrity experts (David Goodstein, mentioned above), had written that Newton’s behaviour had accommodated his theory to justify the experimental observations (Goodstein, Reference Goodstein1991). His argument for doing that was undoubtedly ad hoc, but Newton had not manipulated the data, i.e. his behaviour did not have the FFP stigma. If he had wanted to cheat, he would have altered the experimental results. Instead, as he was obsessed with the importance of precision, he kept introducing small speculative adjustments until the theory eventually agreed with the data. His mistake was to introduce erroneous adjustments (he could not do otherwise, as the appropriate knowledge advancements to avoid them were only developed centuries later) and, all considered, his errors make us smile. However, he was not intending to commit a fraud, and he was not guilty of FFP behaviour, as that behaviour implies the manipulation of data, not the ad hoc adjustment of a theory.

Dalton is the next historical defendant of Broad and Wade. As in the case of Galileo, Dalton – the founder of the atomic theory that is the base of modern chemistry – is accused of an experiment that couldn’t possibly be performed, and thus of having invented the results of that experiment. In 1803, Dalton had discovered that, if he wanted to completely consume the reagents in a reaction, a given volume of oxygen could only combine with a volume of nitric oxide (NO) 1.7 times that of its volume, or 3.4 times. The finding that the reaction of a combination of the two gases consumed the reagents only in defined proportions was an important experimental finding for the development of the atomic theory that Dalton announced for the first time in 1808. However, in the twentieth century, chemists and science historians started doubting that Dalton’s observations were real, because of theoretical thermodynamic considerations and of the inability of some scientists to reproduce Dalton’s results (Nash Reference Nash1956; Partington Reference Partington1939). However, exactly 200 years after the presentation of the atomic theory, the experiment of Dalton was repeated with great attention to all the details with which it had been performed, and it reproduced exactly his results (Usselman et al. Reference Usselman, Leaist and Watson2008). The accurate analysis of what Dalton had done has made clear that he had attempted to optimize the experimental conditions to demonstrate what he had in mind. Once he found the experimental conditions to obtain the results that agreed with his theory, even if he had been wrong in the interpretation of the chemistry of the reactions that were going on, he had correctly understood the stoichiometric combination of the gases he was utilizing. He had not falsified nor fabricated data, he had honestly reported what he had observed. And, even if he had been mistaken in interpreting the details of what was happening, he had correctly understood the element of general interest that was fundamental for the formulation of his theory.

One of the most famous historical accusations of scientific fraud is that addressed to Mendel, the discoverer of the quantitative laws of transmission of genetic characters. The accuser was Sir Roland Fisher, an extremely influential actor in the world of synthesis between Darwinism and Mendelian genetics. This is interesting, as Fisher originally was an admirer of Mendel, and actually was essential in the re-discovery and diffusion of his work. Nevertheless, in a famous paper (Fisher, Reference Fisher1936) he wrote the following statement:

This is an unusual statement: the first portion is positive, but the second is dramatically negative. Without sparing words, Fisher accused Mendel of having falsified the results of all, or nearly all, the experiments, to force them to agree with the expectations. Obviously, the claim is much more serious than the accusation, which had also been made, of the unconscious selection of data by Mendel, explainable with the bias in the classification of ambiguous phenotypes, of interrupting the counting when the results were satisfactory, of repeating it when they were not, and of repeating uncertain experiments. Fisher imagined that Mendel had falsified the results of two experiments, because, according to the reasoning of Mendel on these two experiments, the described phenotypes could only have been obtained with an extremely low probability. Fisher was actually extremely disappointed by his analysis of Mendel’s results, which he privately described as ‘abominable’ and as a ‘shocking experience’.

Naturally, given Fisher’s extremely important reputation in the area, his accusation has had ample resonance. Scores of geneticists have considered and reconsidered his accusations, publishing a multitude of contributions in support of Mendel or against him. A large number of authors have concluded that the evidence of falsification of Fisher is circumstantial and insufficient to explain the discrepancy between his calculations and the results described by Mendel, but the discussion has kept going. Perhaps an important paper by Hartl and Fairbanks (Reference Hartl and Fairbanks2007) will bring the discussions to a close. The paper summarizes all discussions on this heated matter, and concludes that Mendel could not be accused of having falsified results.

The last historically prominent ‘defendant’ chosen by Broad and Wade is Robert Millikan, the physicist who was awarded the Nobel Prize in 1923 for his very accurate measurement of the electrical charge of the electron. The crucial experiment is now described in school texts worldwide – it is the famous experiment of oil droplets nebulized in an electric field: their migration from a pole of the apparatus to the other pole had permitted demonstration that the electrical charge of the electron was a constant, which was the minimal possible measure of electrical charge, as all other were multiples of it. It had also permitted measurement of its value with an error of less than 1%, as Millikan had triumphally declared in the article that described the experimental procedure. However, Millikan became involved in a heated controversy with the Austrian physicist Felix Ehrenhaft who, using a better experimental set-up, had obtained charges of the electron lower than those of Millikan. He had questioned Millikan’s results since, he claimed, Millikan had been misled by his set-up that was too rudimentary to obtain accurate measurements. Eventually, Millikan’s results were considered correct, so he is now considered a giant of modern physics. However, in 1978, Gerald Holton, a Harvard physicist who also was a historian of science, published a study of the laboratory notebook of Millikan (Holton, Reference Holton1978). The notebook reported measurements on 175 oil droplets, whereas Millikan had only published data on 58 droplets measured during the same time. As a result, in their 1982 book, Broad and Wade claimed that Millikan, who had declared ‘not to have omitted a single oil droplet’, had falsified his experimental result, reporting only the data that agreed in the best possible way with his theory (in other words omitting charge measures much smaller than those he had reported, and which would have favoured Ehernhaft’s theory of sub-electronic charges). In essence, Millikan had not only falsified experimental data to win the controversy, he had also lied by writing that he had reported all the measured data. However, as Goodstein has demonstrated by analysing Millikan’s notebooks that were available to him at Caltech (Goodstein Reference Goodstein2000), Broad and Wade were mistaken in their analysis of Millikan’s data, as Millikan had obtained the results to be published from 63 consecutive days of observations, the last made in 1911 and 1912 with a perfected set-up. In these 63 days, he had established reliability standards (for instance the dimension of the droplets) that had led him to discard many droplets. In addition, Millikan had declared to have preferred droplets whose charge changed during the experiment, as the droplets would become charged only with multiples of the minimal charge. The crucial point, in Goodstein analysis, is that the 58 final droplets chosen for the publication were correctly selected according to experimental considerations, not to obtain better consistency with a theory. At this point, it remains to be understood why, instead of reporting faithfully what was evident in the laboratory notebook, Millikan instead decided to write (Millikan Reference Millikan1913):

And

Which of course was a gross lie. Goodstein (Reference Goodstein2000) tries to interpret Millikan’s declaration to defend him, but the defence is not convincing. The behaviour of Millikan remains highly questionable, and it consists in having falsely reported what he had actually done. Why had he done it? Perhaps because he knew he had in his hands a historical result, and decided to cut a corner in his article. This was undoubtedly arrogant: however, in his opinion, he did not ‘hide’ data that disagreed with those that were eventually published, since he felt that the selection of droplets he had made, and Goodstein agrees, was well founded. Thus, he had not falsified results, but had omitted the description of the way he had operated. The lie, however, remains.

In closing, one last case should be mentioned, which the Appendix of the book by Broad and Wade had not considered. In a book published in 1995, Gerald L. Geison used the private laboratory notebooks of one of the legendary icons of modern science, Louis Pasteur, to show that his celebrated science also had darker sides, for instance in his rush to develop the rabies vaccine he disregarded the human risks his haste entailed, and in the appropriation of anthrax’s vaccine, he used findings from a rival colleague to correct the initial defects of his own vaccine without mentioning the colleague. Geison acknowledged Pasteur’s greatness as a scientist, claiming that his aim only was to present him in the wider cultural context of the professional rivalries and the national adulation that surrounded him. His book nevertheless elicited a number of negative reactions. In particular, it was severely criticized by a number of scientists who felt that it actually was an attack on science itself (Söderqvist and Stillwell Reference Söderqvist and Stillwell1999). A particularly negative conclusion was that of Max Perutz (Reference Perutz1995) who wrote that Geison ‘burrows through Pasteur’s notebooks for scraps of supposed wrongdoing and then inflates them out of all proportion … in fact his evidence is contrived and does not survive scientific examination’.

In conclusion: Geison’s book does not question the universal fame of Pasteur’s scientific achievements, which he thinks is well deserved. But the biography also reveals that Pasteur’s personality had perhaps tracts that he exploited to successfully promote his science in the highly competitive political context of his time. Even if these may be considered somehow unpleasant they do not detract from the universal value of his achievements.