Beyond the fringe: when science moves from innovative to nonsense


  • Simon Silver

    Corresponding author
    1. Department of Microbiology and Immunology, University of Illinois, Chicago, IL, USA
    • Correspondence: Simon Silver, Department of Microbiology and Immunology, University of Illinois, Chicago, IL 60612, USA. Tel.: +1 312 996 9608; fax: +1 312 996 6415; e-mail:

    Search for more papers by this author


Microbiology has experienced examples of highly productive researchers who have gone beyond just interpreting their experimental results with hypotheses and published nonsense that was readily recognized as such by readers. Although the most discussed cases of this pathology come from physics, studies of single-celled microorganisms, virology, and immunology have provided many examples. Five cases are described here along with some generalizations. These are the Lamarckian inheritance of acquired characteristics reported by distinguished and experienced researchers, vectorless DNA transfer and incorporation of bacterial DNA into chromosomes of plants years before vector construction of genetically modified plants was invented, water with memory of immunoglobulin IgE, a new electromagnetic radiation method for identifying bacterial and viral pathogens by the discoverer of human immunodeficiency virus, and the claim of isolation of a new bacterial isolate with arsenic replacing phosphorus in DNA. These examples represent very dissimilar areas, and the only common factor is hubris on the part of experienced researchers. Secondarily, failure of peer review sometimes happens, and journal editors do not step in, sometimes even when alerted before publication. These failures of the publishing process teach us that unnecessary mistakes occur and should warn us all to watch our own enthusiasms.

The general picture

This is a commentary on the publishing of science, beyond the fringe from what is recognized as the innovative results and hypotheses leading from them (Kuhn, 1962), and not on the scientific results themselves. In this time of open-access online publishing, sometimes reports are altered after publication online, at the option of the editor (sometimes without or sometimes with authors’ agreement). This new process is also open to beyond the fringe problems concerning what publication now means. The topic here is that creative and experienced experimentalists frequently overly interpret their results, going from far more than mere hypothesis to what is quickly recognized by the peer community as snake oil. This phenomenon is not new.

Two useful monographs cover the processes by which one can judge innovative real science from beyond the fringe ideas, with examples mostly from physics. Park (2000) has a long interest in this problem, especially with regard to flying saucers and claims of governmental cover-up of beyond the fringe physical science. Friedlander's (1995) book is titled ‘At the fringe ….’, so we move here to ‘Beyond the fringe’, recognizing that this phrase was used 50 years ago for a British stage comedy that had strong academic roots.

Irving Langmuir (a Nobel laureate physical chemist) perhaps started modern consideration of these problems, when he called this ‘pathological science’ in an unpublished 1953 lecture at General Electric Company (where he worked). That lecture was recorded and later transcribed and published (Langmuir & Hall, 1989). Langmuir considered it pathological when the excess enthusiasm by scientists (often distinguished and experienced) ran beyond reason. Langmuir himself, however, was victim to this situation in his unwarranted defense of a model for protein structure. The model (Senechal, 2012) might be described as heterocyclic polyatomic rings assembled into a lace doily-like flat structure that could then fold over on itself, leaving amino acid side chains either internal or sticking out. This model was still being defended after the linear structure of polypeptides had been directly determined (in 1951), and the two-dimensional (also in 1951) and three-dimensional structures of globular proteins were solved by X-ray diffraction analysis (about 1957–1959). So the lace doily model was already pathological when Langmuir defended it.

Because this is a brief set of examples of what can go wrong in the process of science, it is useful to set the context and conclusion. Harvard Professor George Santayana famously wrote ‘Those who do not remember the past are condemned to repeat it’. We need to learn from the recent past. Walt Kelly had Pogo say ‘we have met the enemy and he is us’. It is a mistake to think oneself immune to self-inflicted scientific hubris.

A secondary question is, ‘What is the responsibility of the journal process when such a manuscript is submitted?’ Responsibility for beyond the fringe science lies entirely with the authors. Nevertheless, additional responsibility of the journal is not a simple question. Sometimes, the potential of a beyond the fringe problem is not recognized immediately by in-house professional journal editors (who lack recent related laboratory experience and who assign therefore inappropriate reviewers). Then, the outside peer reviewers miss the basic point. This problem will be considered at the end of this article, especially in the context of Nature and Science, two journals that seeking the most innovative new work often give space to beyond the fringe science.

Microbiological examples

Lamarckian-style bacterial training from Hinshelwood to Cairns

Jean-Baptiste Lamarck (1744–1829) introduced a thoughtful innovative theory of inheritance of acquired characteristics (now referred to as phenotype) from one's parents, long before Darwin thought about evolution by selection of the fittest progeny. Understanding of the genetic basis of inheritance and selection came later. Lamarckian thoughts were innovative in the early 19th century and not beyond the fringe. However, such inheritance of acquired characteristics thinking has continued to more recent times, most unfortunately under the influence of Lysenko in the former Soviet Union, but also in Western countries. Microbiology and immunological tolerance became the last bastions of Lamarckian arguments, long after this was understood to be beyond the fringe.

Sir Cyril Hinshelwood was Professor of Physical Chemistry at Oxford, President of the Royal Society, and a Nobel laureate, when he was the last major Lamarckian microbiologist in the UK. John Cairns demonstrated the physical circularity of the Escherichia coli chromosome at a time when some thought that the circular chromosome might be only a mathematical construct from physically linear structures (as is the case for some bacteriophage chromosomes). Cairns also isolated a mutant strain lacking DNA polymerase and found the strain grew well, although it was sensitive to irradiation. Therefore, what was later called DNA polymerase I could not be the DNA replicase. Then why did Cairns later think he had found environmental conditions where bacteria learned from experience which direction to mutate? Hubris, again. Roth et al. (2006) recently summarized this problem, describing the results of Cairns and colleagues in-depth and provided alternative explanations.

Dean & Hinshelwood (1960) and Grant & Hinshelwood (1964) measured the appearance of lactose-fermenting (Lac+) colonies on Petri dishes with non-lactose-fermenting mutants of bacteria including E. coli. Lac cells formed Lac+ colonies over days, which was attributed to the bacteria having ‘learned’ or having been ‘trained’ to utilize lactose from extended exposure (Hinshelwood, 1946; Dean & Hinshelwood, 1964, 1966). Of course, it was mutation, selection, and overgrowth. The wrong overall model was that bacterial cells, being relatively simple, did not require genes, but could have metabolism governed by a series of metastable states, readily described by a series of parallel differential equations (Dean & Hinshelwood, 1966). The results fit the model. While these ideas might have been innovative at the time of Hinshelwood (1946), the explanation was recognizably wrong by the time of Dean & Hinshelwood's (1966) extensive development of the ideas. Dean & Hinshelwood (1966) were familiar with the new microbial molecular genetics, but reluctant to explain their results in that manner.

Hinshelwood also attributed the development of antibiotic resistance to training or learning (Hinshelwood, 1946; Dean & Hinshelwood, 1966). His ideas were generally recognized as wrong by the late 1950s.

It might have been thought that Lamarckian arguments about microbiology would have ended then. However, Gorczynski and Steele published a series of beyond the fringe reports on inheritance of acquired immune tolerance. One appeared in Nature (Gorczynski & Steele, 1981) only 2 weeks before Peter Medawar (whose 1960 Nobel Prize was for demonstrating and explaining the mechanism of acquired immune tolerance) and colleagues (Brent et al., 1981) submitted a debunking report to the same journal. They stated that inheritance of acquired immune tolerance ‘has been faulted by every critical test’ and the ‘experiments executed hitherto to corroborate the Lamarckian interpretation can be faulted’. Why did Nature knowing this was a major problem publish the first report? It should have been stopped.

Several years later, again in Nature, Cairns et al. (1988) measured the mutation from Lac to Lac+ in E. coli cells and found the appearance of mutations continuing over days, only in the presence of lactose. This led to the conclusion that the bacterial ‘cells may have mechanisms for choosing which mutations will occur’. That was a beyond the fringe conclusion. Over the next two decades, Cairns occasionally published additional supporting reports (Cairns & Foster, 1991). Foster took the lead, trying to rescue the phenomenon, including in two Annual Reviews articles (Foster, 1993, 1999) that analyzed the details and responded to negative criticisms and multiple publications from others. Still later, Foster (2004) participated in an argumentative dialog with harshly negative experts, whom she stated had misunderstood or misrepresented directed mutations. Finally, Roth et al. (2006) summarized the overall situation and explained the original data in completely non-Lamarckian terms.

The strains used by Cairns and colleagues contained mutations present on transferrable plasmids and not on the chromosome. Technically precise requirements that were basically irrelevant to the overall claim of an important new mechanism of mutation, selection, and evolution obscured what was happening. Indeed, under the rather special conditions of Cairns et al. (1988), Lac+ mutant clones accumulated during stationary phase and only when lactose was present in the medium. The mutations arose in a normal (or Gaussian) distribution and not in the Nobel Prize-winning ‘jack pot’ distribution found earlier for bacterial mutations. The requirement that the Lac mutant be on a mobilizable plasmid apparently was based on Lac+ mutations arising by a process involving the nicking of plasmid DNA during conjugal transfer of lac DNA and amplification of that DNA (Foster, 2004). In a softening of language, Foster (2004) used and then set aside the original phrase that the ‘bacteria could choose which mutations to make’ and that these mutations are ‘directed’. Later, the mutations were merely called ‘adaptive’.

This series of wasted publications presents an excellent example of how beyond the fringe science moves forward slowly. The original proponents almost never change their minds. The underlying phenomena are not usefully addressed by argument and counterargument. As Kuhn (1962) concluded, the initial claimants just move aside, while newer researchers advance standard explanations. Our purpose here is to enable younger microbiologists to become aware of this recurring historical pattern.

Water with memory of immunoglobulin IgE

Jacques Benveniste opened a major science beyond the fringe episode with a report (Davenas et al., 1988) on the ability of water to alter granule release by IgE-responding white blood cells, which was retained even when diluted 10120 times, so that not a single anti-IgE molecule remained. The water around the original anti-IgE was said to have retained ‘shape’, and the phenomenon called ‘water with memory’. Benveniste referred to this as a form of ‘digital memory’, and a company DigiBio was started to commercialize this phenomenon. Nature published an unsigned caution titled ‘When to believe the unbelievable’ calling the results ‘inexplicable’ on the page just before the Davenas et al.'s (1988) article. Nature also ended the article with a paragraph titled ‘editorial reservation’, stating that the results had raised ‘incredulity’ from multiple readers. Then why did Nature publish this report? There was heavy criticism against the Davenas et al.'s (1988) claim for water with memory immediately on publication. One month later, Nature published seven outside critical letters together with an article by Editor Maddox et al. (1988) referring to the initial report as ‘a delusion’. The claims disappeared quickly from most science, but in a small way reappeared in with the claim for electromagnetic radiation from DNA.

Electromagnetic wave signal as a cause of microbial pathogenicity

Luc Montagnier won the 2008 Nobel Prize for the discovery of the human immunodeficiency virus (HIV). However, since 2009, he has proposed that novel electromagnetic energy signals emanate from the DNA of bacterial pathogens (Montagnier et al., 2009a). The electromagnetic radiation is of low frequency (about 1000 Hz) and survives extraordinary dilution, reminiscent of Benveniste's highly diluted immunoglobulin molecules. Montagnier defended Benveniste's claims (Enserink, 2010) and reported positive effects at dilutions at least 10−18 times, using equipment designed by Benveniste (Montagnier et al., 2009a). The effect passed through filters that would hold back bacterial cells and was attributed to DNA in solution (Montagnier et al., 2011). The electromagnetic radiation passed from the initial radiation-emitting plastic tube to a nearby receiving tube. Montagnier et al. (2009b) also found electromagnetic radiation from DNA of HIV-infected cells from patients with AIDS. Of course, this is beyond the fringe. The negative reaction in France caused Montagnier to relocate to a new institute in Shanghai, China (Enserink, 2010).

GMO plants without a vector

Lucien Ledoux published reports of Arabidopsis thalia plant seeds incorporating naked bacterial DNA, without the need for any specific vector or machinery (Stroun et al., 1967). The newly transferred DNA corrected mutational defects (Ledoux et al. (1971, 1974)). Lurquin (2001) wrote a sympathetic history of this phenomenon titled ‘Green Phoenix’. The title suggested that the dream of genetically modifying plants first arose magically, phoenix-like, in the Ledoux laboratory, and then died from a lack of reproducibility of the data and disbelief about what had actually been done. And finally, the transfer of genes from bacterial cell to plant cell was found again (phoenix-like) by a completely different process, conjugation using the bacterial Ti vector plasmid. Monsanto Company (in St. Louis, MO) in the early 1970s, planning on switching from a bulk agricultural chemical company to one more agribiochemical (now referred to as GMOs) invited Ledoux to fly to St. Louis to explain his results. The technical details and discussions made it clear this was beyond the fringe. And Monsanto waited another decade for the availability of Ti plasmid delivery systems to make gene transfer from bacteria to plant cells feasible.

Ledoux et al. (1971) reported that high molecular weight radioactive bacterial DNA was taken up by Arabidopsis seedlings and that the DNA passed intact into mature tissues, with comparable DNA found in the next F1 generation. The experimental results were very precise, including cesium chloride density gradient analysis showing a difference with added 3H, 2H-containing donor bacterial DNA (Stroun et al., 1967). Ledoux et al. (1974) expanded the experiments and found that the donor DNA from thiamine plus E. coli and not that from a thiamine minus mutant bacterial strain converted the plant to a thiamine plus condition. These results could not be reproduced in St. Louis or elsewhere (Lurquin, 2001), even with seeds and DNA provided by Ledoux. At one stage, Ledoux suggested that maybe cosmic gamma irradiation in the airplanes during trans-Atlantic flights inactivated the activity. Lurquin (2001) provides untestable alternative hypotheses that the data were faked by Ledoux himself or by a staff member in Belgium who wished to please his boss. This is the only one of the five cases we are considering in this report where cheating is thought (but not proven) to have occurred. For the other four, it was probably merely self-delusion, which is the definition of beyond the fringe science.

Arsenic replacing phosphorus in DNA

A recent example indicating that beyond the fringe science is alive and still with us came with the claim that arsenic could ‘substitute for’ or ‘replace’ phosphorus in the DNA of a newly isolated bacterial strain (Wolfe-Simon et al., 2011; published in ScienceExpress online on 2 December 2010). A blog posting criticizing the Wolfe-Simon et al.'s article appeared 2 days later, on 4 December 2010 ( Then, science writer Carl Zimmer ( headlined ‘This paper should not have been published’ and ‘Scientists see fatal flaws in NASA study of arsenic-based life’ just 5 days after online publication in Science. Zimmer used inverted commas to indicate that the first headline was a quote from a source (cited at the bottom of his sixth paragraph). The Slate article covers all of the basic reasoning showing that the article was beyond the fringe. There were other published criticisms. For example, Silver & Phung (2011) wrote a brief summary 3 weeks after the initial online publication stating that the report was ‘science fiction’ rather than science. Much later, Matthew Herper in Forbes magazine headlined ‘New science papers prove NASA failed big time in promoting supposedly Earth-shaking discovery that wasn't’ at The scientific community usually does not use popular magazines as sources of understanding (although beyond the fringe science frequently does), and blogs and tweets are new. However, for arsenic in DNA, it was the authors and their government-funding agency NASA that used rapid Internet vehicles to communicate their ideas, which were then uniformly judged as beyond the fringe. What had happened?

Wolfe-Simon et al. (2011) was published on 2 December 2010, on ScienceExpress (the online ahead-of-paper version of Science magazine). The conclusion that arsenic ‘substituted for’ or ‘replaced’ phosphorus in DNA was not supported by the data. One key example was fig. 2A of Wolfe-Simon et al. (2011), which shows agarose gel electrophoresis analysis with two lanes of crude nucleic acid fractions, one from bacterial cells grown on high phosphate/no added arsenate and the other from cells grown with 40 mM arsenate/no added phosphate. However, there was a measured phosphate contamination level about 1000× less than the added arsenate. This figure has several major disqualifying problems that should have been apparent to the 12 authors and the three outside referees who were sent the manuscript for review. Both lanes show single tight high molecular weight bands characteristic of DNA. Arsenic content of the gel regions containing the DNA bands measured as 1.3 As atoms per 100 000 C atoms under high arsenic conditions and 0.7 As per 100 000 C when grown in the absence of arsenic. That is only a twofold difference. Importantly, the DNA was not eluted from the gel. No explanation was given as to why the DNA was not extracted, as is an easy and needed technique. Of course, the ratio of P to C in DNA is more like 1 : 10 than 1 : 100 000, but agarose gels contain about 1 mg mL−1 agarose, a seaweed polysaccharide. Seaweed products are well and long known to contain high levels of harmless organoarsenic compounds (e.g. arsenic in seaweed‎). My favorite, Nori, contains about 24 mg As kg−1 product, approximately the same ratio of As/C as reported by Wolfe-Simon et al. (2011). A simple negative control measuring arsenic in a region of the agarose gel without DNA would have quickly tested the hypothesis that the arsenic measured by Wolfe-Simon et al. (2011) came from the major carbon source in the gel (agarose) and not the DNA. There are other puzzling and untested questions from fig. 2A of Wolfe-Simon et al. (2011), for example, the failure to measure the arsenic content in the massive ribosomal RNA-containing bands for the high P-/no As-grown cells. These major rRNA bands are not identified as such by Wolfe-Simon et al. (2011), but from staining intensity (not measured), they contain much larger amounts of nucleic acid than the DNA bands. If there were arsenic in nucleic acids, the amount of arsenic also should have been much larger in the RNA bands. To miss such a simple available measurement was an important failure of the authors and the reviewers.

There was a NASA press conference the day Wolfe-Simon et al. was published online, with strong statements indicating that arsenic in DNA was established and that this supported the existence of ‘weird life’, a term earlier used by several of the authors to indicate life forms on the Earth differing from all those more familiar and that these might be useful models for guiding search for alien life on Earth and on other planets.

Announcements at press conferences and a rapid dissemination of claims on YouTube, the blogosphere, and twiiterverse are the new normal for beyond the fringe science. There was an immediate flood of e-mails and blog comments (newer means of science communications), which were without exception highly negative.

Wolfe-Simon et al. (2011) made the new strain available to other researchers including us and cooperated with advice as to how to grow this gammaproteobacterium (the same group as E. coli). Phung et al. (2012) determined the entire 3.5-Mbp bacterial genome and placed the data in GenBank, available to all. Science writer Pennisi commented at Kim & Rensing (2012) analyzed arsenic resistance and arsenic metabolism-related genes, and Elias et al. (2012) used the annotation to isolate a gene for a periplasmic phosphate-binding protein. The protein gene product was purified and shown to discriminate against arsenate and in favor of phosphate by about 10× better than the comparable E. coli phosphate-binding protein. This helps explain the arsenate resistance of the strain, but does not contribute to the question of replacement of phosphate in DNA by arsenate.

Rosemary J. Redfield of the University of British Columbia who had taken a lead in the early blog analysis ( was coauthor of the report by Reaves et al. (2012) showing the absence of detectible arsenic in purified DNA from this bacterial isolate. A second report in the same issue of Science (Erb et al., 2012) also failed to detect arsenic in bacterial DNA.

Peer review and the responsibilities of journal editors

It is accepted that the responsibility for a published report is fully that of all authors and not of the journal publication process or the government agencies supporting the research. However, problems with peer review and publishing could be usefully considered for the examples described above.

For water with memory, I suggest that an additional problem was hubris by Nature Editor John Maddox, who authored a follow-up report (Maddox et al., 1988) entitled ‘“High-dilution” experiments a delusion’ less than a month after the initial beyond the fringe publication. It was known to be a delusion when the initial paper was published, but publishing gets you headlines and television coverage. The results in Davenas et al. (1988) could not be duplicated in Benveniste's laboratory in the presence of the team of Maddox et al. (1988). Benveniste accurately complained on the next page in Nature that it was a ‘trap set by a squad’ (a dog and pony show in American usage) consisting of an editor seeking publicity, a professional magician whose occupation was to catch science fraud and a self-selected USA NIH specialist on fraud and misconduct. The dog and pony show as described by Maddox et al. (1988) seems very vigilante-like, as Benveniste complained.

Sir John Maddox (1925–2009, a physicist by background) used the pages of Nature during his two terms as Editor (1966–1973 and 1980–1995) for similar nonproductive beyond the fringe exchanges. Assignment of fault to Nature comes also from Maddox’ famous fondness for beyond the fringe reports as described in obituaries after his death, for example, Gratzer (2009; During the time between Maddox's terms as Editor, Nature in 1974 published a report supporting the psychedelic parapsychological gifts of an Israeli magician, whose claims for psychokinesis and telepathy were also debunked by the same magician James Randi. As that was not related to microbiology, it is not an example here. However, Maddox was also a broadcaster on BBC radio, and his ‘instinct for publicity pushed science into British newspapers’ ( such as The Times (of London), the Manchester Guardian, and Le Monde (in Paris).

The journals Nature and Science in particular seem to have difficulty in separating their roles as scientific journals for novel technical reports and in journalism seeking the largest popular audience.

For the case of the claim for arsenic replacing phosphorus in DNA, it is documented that peer reviewers were poorly chosen and that the outside referees missed the problem. The anonymous reviews and exchanges between the editor and the authors were released to a reporter in response to a USA Freedom of Information request ( The journal staff actively resisted the negative response (Pennisi, 2010ab) that went viral immediately after online publication. Bruce Alberts (2011a), Editor in Chief of Science and a well-known nucleic acids biochemistry researcher, exacerbated the already-recognized bad situation by obfuscating and arguing for openness and standard processes, which were in this case not used.

There are additional, but somewhat hidden lessons to be learned here. The Wolfe-Simon et al.'s paper published online in December 2010 did not appear in an issue of the journal for 6 months (Wolfe-Simon et al., 2011), rather than the more typical < 6 weeks (that can be calculated from dates given at the end of most Science articles). The authors of Wolfe-Simon et al. (2011) were kept mostly in the dark during this 6 months about the schedule and processes, while the editors of Science considered when the final version would appear and had letters to the editor (in this case called ‘commentaries’) reviewed and revised. The distinction between a commentary and a letter is arbitrary, and Science reporter Pennisi (2011) calls them ‘so-called Technical Comments’. The editors of Science selected eight from more than 20 such negative responses for placement online [not on the journal pages, as was the revised article, plus a cover paragraph from Editor in Chief (Alberts, 2011a) and another commentaries by Pennisi (2011)]. Science delayed in this instance final publication by introducing essentially ad hoc processes and a specialized difference between a ‘letter’ that appears on a numbered journal page or a ‘technical comment’ that appears only online. Pennisi in 2011 told me that Science magazine ‘has a firewall’ that functioned in this case to protect DNA biochemist and Editor in Chief Alberts from more involvement. But that was only when he wished.

There was another microbiology/virology embarrassment in Science at much the same time in 2011, which may provide some basis for understanding. This involved a human disease (chronic fatigue syndrome) with a large number of affected individuals. Science had earlier published that the disease had a retroviral cause, but the retrovirus turned out to be a tissue culture contaminant. Here, Alberts (2011b) published an ‘expression of concern’ doubting the validity of the earlier report and pushing the reluctant authors toward retraction, a process that was completed still 6 months later. None of the 12 authors of Wolfe-Simon et al. (2011) appears interested in retraction even today. Rather the manuscript was altered in unclear ways between the December 2010 acceptance and publication of the online version in ScienceExpress (which legally is considered the actual publication) and the appearance 6 months later in an issue of the journal. Today, with online publication, one can alter a report after publication and remove its initial version from access. The meaning of publication has changed, and the responsibilities of the journal and publisher have become muddied in a new way. Science and Nature (and to a somewhat lesser extent other journals) are developing new and largely ad hoc processes to deal with beyond the fringe reports in the current age of instant communications.


The primary conclusion from the five examples of beyond the fringe published nonsense described here is that the problem exists. Often very capable researchers make foolish mistakes, so we should all be alert to our own susceptibility. Mostly, the problem is not the data, but rather that experienced people claim conclusions from the experimental data that do not in fact follow. Unavoidably, negative opinions have been expressed here that would not be appropriate for a normal science report. Beyond the fringe or pathological science is very old. Unfortunately, it persists and evolves with new means of electronic communication but no change in human nature. It always will be with us, as authors will self-deceive. Journals and reviewers will miss the boat. There is no reason, however, to overly emphasize the bad or make reports of such pathology a major theme. Such mistakes happen rarely. Beginning scientists are generally not aware of the phenomenon, and more experienced scientists wishing the phenomenon to go away think they are helping by ignoring it.