In a letter to Science last week, Dr. Cagan Sekercioglu proposed a plan to quantify the amount of work conducted by each author on a paper, so as to have a better gauge of the role played by each contributor. This would, I assume, be in addition to the “author contributions” paragraph that many publications now list with manuscripts, something Nature started on a voluntary basis almost 10 years ago.
Although I have argued for better quantifiable measures of impact in the past, I can’t support a venture to put a number on author contributions. Trying to understand or better-quantify the scientific impact of an individual study has merit, since grant review boards and faculty search/tenure committees often place too much emphasis on journal impact factor as a measure of individual candidate quality. This is clearly misguided and is nothing short of a short-cut for really exploring the scientific depth of a candidate’s work. But an attempt to actually place a value on the number of times you happened to complete some Western blots for a colleague seems to provide false authority where none should lie. At least in biology, it is well-known that if an author was 5th in a list of 8, s/he was not the driving force behind the project.
In the letter, Dr. Sekercioglu provides the following proposal for calculating author contribution:
I propose that the kth ranked coauthor be considered to contribute 1/k as much as the first author. This way, coauthors’ contributions can be standardized to sum to one, regardless of the author number or how authors are ranked. Author rank can be different from author order, provided that this is declared in the paper. Multiple authors can have the same rank, as long as this is stated and is reflected in the calculations.
This sounds more than reasonable as a calculation. But the fractional authorial assignment would be so arbitrary and variable from lab to lab (and perhaps even within a lab!), that this metric would have little value to anyone without a means to normalize the numbers. Therefore, since number-dropping can often provide “false authority” (see Advertising 101 or Political Campaign Strategy 101), we would be stuck with a value that seems like it should be taken seriously, but for all intents and purposes, is really just muddled by bias and random error. Even if the number could be normalized somehow (I doubt it), I still think that this would be a better measure of how nice open to collaboration one is as opposed to whether one is a good scientist.
The topic of “too many authors” has been lamented for some time now, starting with a letter to Science in 1958 (referenced in the current correspondence). The link from Science is actually screwed up (the downloadable PDF lacks the first third of the letter…), so I did some digging to find the rest and since it’s short, I’ll just post it here:
Too Many Authors
A letter from Z. I. Kertesz [Science 128, 610 (1958)] deplores references which use “et al.” after the first author’s name, particularly when more than three authors are involved. There is cogent argument that, for anything short of a monographic treatment, the indication of more than three authors is not justifiable, in general. In fact, minor contributors should be listed-and their specific contributions shown-in the acknowledgments. A particular report comes to mind that appeared under merely one author’s name. It describes the properties of a rare mineral which had not been adequately characterized or previously reported from localities outside of Russia. This article was written by a mineralogist who used data obtained by a chemist (analytical determinations), a physicist (electron micrographs), and two spectroscopists (minor components). This six-page article might have had five authors, but the fact remains that the over-all responsibility for evaluating the data depended upon a single individual, the mineralogist. In many instances the only justification for the use of more than three authors’ names seems to be the accumulation of bibliographical credit for minor contributions. This situation, if abused (and it has been) can readily become ridiculous. It is discouraged, to some extent, by the use of “et al.” in citing papers that are overloaded with authors.
DUNCAN MCCONNELL
College of Dentistry,
Ohio State University, Columbus
Dr. McConnell would be livid over the situation today, I am assuming. My, my, how things change. Other good commentaries on this issue are here and here.
So on this contribution quantification, I am fully in the camp of a particular forum discussion commenter when he said:
Bad measures of productivity are actively harmful to science, and that is something that their advocates should bear in mind. They are encouraging dishonesty.
I would say that this metric has the potential to be one of these “bad measures of productivity”.
Last updated:
Friday, 24 Oct
2008 - 17:35 UTC
I think finding a fair metric on author contributions is as difficult as finding one to quantify a person’s scientific output based on publications (h-factor etc). In a way, this problem might even be linked.
Author contributions as implemented by “us” are a good step, albeit a subjective one. Something more objective might be if someone independent yet close enough to the authors provides percentages, head of department, dean of school… oh no, I can already see hours of meetings wasted that way.
Overall, I am not sure why everything needs to be quantified. Letters of recommendation, references, networking, all these things although they are more time-intensive could certainly provide a more nuanced picture on someone’s contributions to science. But then we are back to the issues discussed in relation to h-factor, impact factor and so on…
As you can imagine we have had many contributions and discussions about this topic in Nature over the years, some of which are available at Nautilus, the blog for authors.
As you point out, the ‘author contributions’ statement has uses because it tells readers what different authors did to create the paper, which is basically all you need. Many authors now use this option at Nature (see Nautilus link for details). However, I suspect this option is unlikely to be welcomed by bean-counters because it is not a neat, quantifiable measure that tells them what they got for their money. That isn’t possible to do using a simple metric such as a fractional contribution, which has many, many flaws (some of which you point out).
Apologies, a much better Nautilus archive is this one.
Well, the “authors’ contributions” statement might be useful precisely because it is of no use to bean-counters. As well as because it tells you, if you really want to know, what each person on the project did.
It also might serve to decrease “guest authorship” (“Professor X appointed Dr Y to the Department and expects his name to appear on all papers from Dr Y’s lab. Dr Y complies because he is up for tenure review in a few years and needs Prof X in his corner”)
In the course of my editorializing for Physiology News I have written a bit about author statements, which I generally think are a good thing given that
(i) there are now many more papers with >2 authors
and
(ii) it is obviously helpful to be able to figure out what different people did.
To illustrate (i), even in a traditionally small AU numbers field like Physiology, if you look in the May J. Physiol.s of 1987, 97 and 2007 the average no of authors has gone from 2.4 in 1987, to 3.4 in 1997, and to 4.6 in 2007.
Re (ii), I don’t believe a metric will do it, as positions in the AU order has come to mean different things (front end = “experimenter”; back end = “boss”). Hence the joke about the worst place being to be the middle author of five.
I can’t see there being any metric which sensibly summarises what people really contributed. Though sadly, that will not stop people inventing ones which don’t work. To illustrate this, let me tell you a story.
A few years ago a Professor I know (not David Colquhoun!) wrote a column for Physiology News where he was quite uncomplimentary about assessing researchers by citation counting and metrics. To emphasize just how ridiculous it was, he satirically suggested creating an index called the “Centrifugal Factor” – a number that quantified an author’s degree of credit for a paper based on his or her position in the author order. The nearer to the outer ends of the paper the author came – the nearer to either first or last authorship – the higher the “CF”, and the more points the author scored.
Imagine my friend’s consternation a few years later when Imperial College was reported (notably by David Colquhoun, e.g. here ) to be using this precise method in computing “research output metrics” for their staff.
Interesting question, how much should journals be helping readers and how much bean counters? Another case in point is “grant contribution numbers”, which bean counters want stated in articles so that they can assign “outcomes” to work they fund. But these numbers and acronyms are meaningless to readers.