In these times when the highest ambition of so many is the desire to appear on pop star talent shows or Big Brother to achieve 5 minutes of vacuous celebrity, anyone who might have the least interest in following a scientific career with the aim of doing some good in the world is to be applauded. It seems especially sad therefore when some individuals, no doubt attracted to science for the best of reasons, end up turning scientific gold to base metal.
In 1973 William T. Summerlin announced the solution to a problem that had been dogging the field of transplantation for many years: how to enable tissue grafts from one organism to survive after transplantation to another, even transplanting tissue between different species. The answer was remarkably simple: the tissue to be transplanted should be placed in culture medium for some time prior to grafting. The striking result was substantiated by the use of mice of different colours. An area of skin taken from a black mouse and transplanted successfully to a white mouse gave a highly visible confirmation of this groundbreaking research. The result was of course fabricated; the black grafts were no more than patches of white mice decorated with a black felt tip pen, and Summerlin was exposed as a fraud in 1974.
Thirty years on and the world acclaimed the success of the Korean researcher Woo Suk Hwang who reported the isolation of 11 stem-cell lines derived from human embryos. The result promised much in the way of potential new therapies for human disease, Hwang became a national hero and international celebrity, and attracted millions in government funding for the new research. In the end, like Summerlin's mice, these results too were soon exposed as false, Hwang condemned and now to be prosecuted as a fraud, and the stem-cells that promised so much never existed. Each week we hear of new examples where someone, no doubt drawn to science for the best of reasons, appears to have undertaken scientific misconduct. Sometimes it seems, the temptation to cheat is simply too great: if the facts don't fit the theory, invent new facts.
It seems hard to imagine that the perpetrators of such acts of scientific misconduct could believe they would not be exposed. These are not instances where an oversight led to misinterpretation of results, or problems with the methodology led to the generation of misleading observations. These are cases of deliberate and premeditated falsification. In both cases highlighted above, the 'discovery' reported was of such prominence that it would inevitably attract enormous attention from the scientific community and be subject to the highest level of scrutiny. To most of us, but presumably not to Summerlin, Hwang, or indeed to quite a cohort of others, it is seems almost inconceivable that that fraud at this level would not be discovered and that exposure would effectively lead to the end of a scientific career. This is especially incomprehensible in scientists who are doing very well in any case- there is no doubt that Hwang was a talented researcher and did indeed clone the first dog, Snuppy.
What might be the motivation that behind fraud on such a scale. Frequently it is suggested that fraud occurs as a result of the pressures of a modern scientific career. But all scientists are under similar pressure to publish, to receive grant funding and gain recognition among their peers, yet the vast majority do not perpetrate this kind of fabrication. In the media scientists are portrayed rather differently from the researcher that we know from reality. In general we fit neatly into one of two stereotypes: the social misfit who cannot iron a shirt without burning it; and the Dr Frankensteins who deliberately, or through their misguided efforts, threaten to destroy the world. Indeed in a recent survey among children in the UK by Christopher Frayling, scientists were perceived as ‘mad, bad and dangerous’ working in laboratories marked ‘Secret’ (see nice article in the Daily Telegraph). Such a perception it seems is instilled in the public at a very young age and persists until it becomes part of the cultural fabric of society. Doctors are good: scientists are bad.
The truth is rather more prosaic. While some few may fit either image (especially the misfit category), in general scientists have widely differing personalities and pursue their own personal form of scientific method that reflects their individual characteristics. Some progress cautiously step by step, never going forwards until they are secure of their ground, other may make leaps of imagination and design their expts. according to a hypothesis built upon a few scattered observations in the literature. There may be as many different ways to do science as there are scientists.
Overlaid on each individual personality, a scientist needs a particular combination of skills. Not least is a modicum of intelligence, a good memory, imagination, patience, technical ability, and at a later stage in one's career when success is dependent on lab members, organisational and motivational skills (not that these are unique to the world of science). On top of that, and this is especially important, good science rests on an ability to question the validity of ones own results, and to discriminate between data which reflects a biological reality and that which represents background noise in the system. I stress the word biological here since this particular attribute is less essential in the more precise realms of physics or chemistry. In other words, a set of experiments designed to test a particular hypothesis may generate results that either support or contradict that hypothesis to varying degrees. Some experiments may work while others give a null result. The ability to see the grain of gold in a sea of sand is crucial if science, particularly biological science, is to progress. But the very nature of biology means that there will inevitably be a lot of sand to be discarded. Indeed, it would be impossible to publish a paper in which the true process of discovery is accurately reported; referees and even editors dictate that a submitted manuscript tells a coherent story with a strong punchline, whereas the process of scientific discovery in reality frequently proceeds through a series of meanders where the logic behind the experimentation is sometimes fragmented. One could conclude therefore that as papers do not accurately reflect the process behind the discovery reported, all scientific papers may be regarded as fraudulent1. The added pressure by high profile journals to publish papers rendered as short as possible, means also that much of the truth behind the results reported is lost. Only headlines are publishable; doubts, caveats and limitations to the extent to which the results are applicable, and frequently negative results, are not. The structure of scientific society therefore almost dictates that the process of scientific discovery is almost always misrepresented.
To enhance the likelihood that the headline you aim to publish reflects reality, it is best not to rely on a single experiment. If similar conclusions can be reached from results obtained using an extensive array of different techniques, the data generated is likely to be relatively sound. The problem is that the generation of such data is often a tedious affair. Only a small proportion of experiments might yield meaningful results, yet the hypothesis may nevertheless be sound. To generate sufficient data to convince oneself, let alone the referees and editors of high profile journals that your results are meaningful and important may take years of work.
For most of us, the long process of experimentation is endured in the knowledge that in the end, if the data generated don't support our hypothesis we either need to change the model or start again on a different project. It’s a frustrating process, even more so when the first few results point to a clear path ahead, but still we need to do the experimentation to be able to put a paper together. Occasionally we get a surprise: the initial results clearly support the model but further work reveals that in fact biology does things rather differently, and often more elegantly, than expected. In other words, our ability to predict how biology works is fallible and limited by our experience and breadth of knowledge. Adapting a model that follows the experimentation, rather than trying to shoehorn the data to fit a model is one sign of good science.
Perhaps fraud in science originates with individuals who have such a low tolerance of this tedious process coupled with high self belief and overarching desire to be first. Such individuals might not see the point of wasting time and effort in actually doing the experimentation since their ‘data’ would inevitably be confirmed by others. All that is needed is to produce something that looks like hard won data, and fame, funding, and international recognition of the brilliance of their skills will follow. I do not know whether Summerlin, Hwang or any others fit this profile, but I find it difficult to imagine otherwise how intelligent and successful individuals can publish a high profile paper in which the data are false unless they believe the results will be verified independently by other workers. For most of us, the need to be sure our data are correct drives us to undertake the necessary experimentation no matter how tedious, and the fear of misleading the scientific community by publishing something that is 'wrong', even inadvertently, ensures a high degree of scientific integrity.
Of course, the examples of fraud that are reported are only those of which we are aware. It is possible, even likely, that there are other major papers in the published literature in which the data are largely fabricated, but the 'results' presented correctly reflect biological reality and are subsequently confirmed by others using a more conventional approach. Fabricated data published in more obscure journals would be less likely to be detected given that it would be unlikely to generate as much interest. As a referee, though not yet as an editor, I have come across rare instances where elements of papers have clearly been falsified. In this age, images that might have been almost impossible to fake using traditional photographic techniques may now be much more readily manipulated using digital processing (See the nice commentary on this subject Rossner M and Yamada KM, J. Cell Biol 2004 166:11-15).
In the end it is up to all of us to resist the temptation to adjust images inappropriately or otherwise misrepresent our results. The progress of science depends absolutely on the integrity of the data and the ability of our colleagues to scrutinise it thoroughly. But in the end it rests on the shoulders of all of us to do the best we can to ensure we publish results that are meaningful and sound and interpreted with a careful eye on alternative explanations. Only then can we ever achieve something better than vacuous celebrity.
1. For a more eloquent discussion of the distortion inherent in scientific papers please read: Medawar, P.B. 1963. Is the scientific paper a fraud? In The Threat and the Glory: Reflections on Science and Scientists (1990), ed. P.B. Medawar, 228–233. New York: HarperCollins.
How to get your paper accepted
When I was a PhD student I wrote my first paper. My attempt at entering the scientific literature was so poor that my supervisor took one look at it and decided it would be better that he wrote the whole paper from scratch. For my second manuscript, the effort between us was more evenly divided, and for the third I was able to contribute most. The difficult process continued as a post-doc where again my first attempt was met with a brief ‘The introduction and results will do, but you need to work a lot more on the discussion’. The learning process continued with a submission to PNAS where with another post-doc we had to cut our ‘final’ draft by around 30% to meet the strict page limit. At each stage I learned a little about writing papers. That process has continued and even these days, having been running a lab for more than 20 years, I still feel I am learning. Unless there is someone who will spend a considerable time with you teaching the small things that make a difference between getting a paper accepted and rejected, learning how to write and present your data is a painful path haunted by the ghosts of rejections past. Although the quality of data and the importance of the question being addressed are critical factors in getting a paper published, presentation is also a key component. Poorly presented data and inadequate description may lead to your beautiful results being consigned to the rejection bin. For those of you yet to write up your work, I can tell you that it’s not a nice experience. So, here are some of my personal hints for getting your paper into shape, aimed primarily at junior post-docs and students. It is very incomplete, but it should be taken as a rough guide.
Before you begin
It’s a lot easier to write a paper if you have all the results and they are already made up into figures that follow a logical progression. And to get the results and all the controls in place it is better to start the project with a feeling and a plan of what results you will need to get your project published. Having such a plan means you won’t be wasting time doing experiments that may be interesting, but which are peripheral to the main story, and telling a simple story with one main point often works much better than a paper that tries to tell two or more stories that are only loosely linked. It is also very important not to rely on just one technique to make a point; as far as possible everything you try to conclude should be backed up by multiple experiments that use different techniques. If the results from all these different assays point in the same direction, your conclusion is most likely correct and the referee will tend to believe you.
Read a lot of papers in high profile journals and pay careful attention not just to the content, but spend time examining the structure of the paper and see how the language is used. Especially if you are a non-native English speaker, it may be helpful to make a list of phrases and words from these high profile papers that help emphasise points or make the flow of the manuscript better. The more readable a paper is, the easier job for the referee.
Make the paper exciting for the editor.
Editors are the first port of call for your paper and first impressions count a lot: if they like the manuscript it gets sent out for review; if not, it gets returned to you with comments suggesting that you send the paper elsewhere. Convincing the editor your work is important to them is the first task. Editors are pretty busy and don’t have time to read thoroughly all papers that land on their desk, nor do they have the expertise to make a serious judgement of the importance of each paper that arrives. But they are generally experienced enough to make a rough judgement.
The first impression that is essential comprises the title and abstract. Both need to convey the importance of the work as well as the key results and conclusion. From my own experience as an editor a surprising number of papers sent in have an abstract that just begins ‘We examined….’ Without any background to the topic being explored or rationale for the experiments presented. My first response to such an abstract is ‘So what? Why did you do this and why is it important?’. It is therefore critical to place the work in context and highlight the important question being addressed. If that is done, it creates a favourable impression and sets the tone for what follows. No need for the abstract to describe in detail all the results in the paper. For example, ‘Here we show that factor X is a novel molecular motor implicated in melanosome transport’ seems somehow more concise than as valid as ‘Here we show using a combination of microscopy, biochemistry and molecular and cell biology, that factor X is a novel molecular motor implicated in melanosome transport’. If the reader is intrigued as well as informed they will take a look at the results.
Finally the abstract should also contain a line at the end summarising the significance of the work and in many cases the word ‘mechanism’ works wonders while ‘novel mechanism’ may be even better.
Its also worth remembering that if you want your paper to be read and cited, the abstract is critical; along with the title, it’s the one thing freely available and searchable on the web.
To try to get your data published, you need at least to make it sound as though it is important. For example in my lab, in addition to working on melanocyte, we also work with a specific gene in yeast. If I were simply to focus the introduction on that gene, the work would not appear particularly interesting. That impression can be avoided if your particular gene or system is placed into a broader context. In my case I use ‘understanding how cells maintain or change their identity’ as a broad question of general significance to cover both our mechanistic yeast work as well as our work on pigment cells. Melanocytes and the regulation of the particular yeast genes we use are simply models for understanding this process. In presenting your data, try to find a way to tell the reader how your results fit into the general picture. In the introduction therefore, just progress from the general background down to your particular model system and, within reasonable limits, how your system will explain the meaning of life, the universe and everything.
Make life easy for the referee
Although you might like your data, and your supervisor might think it's fine, the referees and editors of this world may not know you and without their support your paper has more chance of getting trashed. Remember too that the majority of referees find reviewing most papers is incredibly tedious and do it without reward as a service to the rest of the community. So rule number 1 is to make the referee’s life as easy as possible.
i. Try to think who your referees are likely to be and ask whether you’ve cited their papers. Many referees on seeing a paper in their area of expertise will go straight to the reference list to check whether they get a mention - it's sad, I know, but scientists frequently have huge egos that need to be fed regularly.
ii. Try to put yourself in the place of a referee. They want to review your paper with as little effort as possible. That means they don’t want to plough through unnecessary information, and they frequently don’t read parts of the paper e.g. Materials and Methods or even the introduction if they are very familiar with the subject. To assess the science they want to read the results section, see if the figures match, and the data are clean and sound, and conclusions justified. If all those criteria are met, the paper will be well on the way to being accepted.
As part of this, the importance of presenting the data well in the figures cannot be overemphasised. I sometimes see papers where the figures are not straight, arrows point at black spaces instead of the bands they are supposed to indicate, parts or even entire figures are not mentioned in the text, Figures are mislabelled, images are at very low resolution or the whatever the author thinks is there simply isn’t evident. These and more, are the things that annoy referees and suggest that if the author doesn’t value his/her data enough to present it well, they maybe haven’t bothered to do the experiments well either. It's not an impression you want to give to the person who is deciding whether your paper is acceptable for publication.
It's also pretty annoying to have a figure with multiple tracks of a gel labelled 1,2, 3 etc with no indication of what each track represents. It means the referee will have to go to the legend and decipher what is being presented (is it a western or RT PCR; which track is the control?). The golden rule, here is make the figure understandable to the referee (and the reader) without needing to refer to the legend. If you want to see how this is done, take a look at a paper in Cell or Nature.
Also, it is worth mentioning that splitting a highly related result into two figures rather than Figure 1A and 1B doesn’t create a favourable impression. Some authors appear to increase the number of figures in a paper just to make it seem that the paper has more data. It's an obvious, but counterproductive strategy. Better to combine results where they are clearly related into a multi-part figure.
In writing the results, try to present what you have done as a logical but intriguing story in which your aim is to keep the reader’s interest. Make sure you describe well each part of the figure including the controls. There is a simple guide to writing results which is, for each result presented:
What is the specific background and the question you are about to address?
How are you addressing it?
What is the result?
What is the conclusion?
How does this lead on to the next experiment?
It is pretty straightforward. There is also quite a lot of information that is useful to be included in the results that might be considered as ‘introduction’, but by keeping it in the results section it directly links a specific point to the experiment about to be presented. But remember, try not to overstate the implications of your results; in general referees prefer accuracy or description understatement e.g. ‘expression of this protein was severely reduced’ rather than expression of this protein was abolished (unless it was abolished).
Finally, the discussion is difficult as it is much less structured than other sections of the paper. It is worth reminding readers and the referee what was the evidence for the main conclusion drawn from the paper, but the discussion should not be simply a repeat of the results without figures. Just pick the key conclusions of your paper and try to place them into a broad context and again emphasize the importance of the results and how they take our understanding further.