ms.tex

\documentclass[letterpaper,twocolumn,superscriptaddress,showkeys,longbibliography]{revtex4-1}
\usepackage[utf8]{inputenc}
\usepackage{color,dcolumn,graphicx,hyperref}
\hypersetup
{
    colorlinks = true, linkcolor = blue, citecolor = blue, urlcolor = blue,
}

\begin{document}

\title{Developing a preprint culture in biology}

\author{Philippe Desjardins-Proulx}
\email[E-mail: ]{philippe.d.proulx@gmail.com}
\affiliation{Theoretical Ecosystem Ecology laboratory, Universit\'e du Qu\'ebec \`a Rimouski, Canada.}
\affiliation{Quebec Center for Biodiversity Science, McGill University, Canada.}

\author{Ethan P. White}
\affiliation{Departement of Biology, Utah State University, United-States of America.}

\author{Joel J. Adamson}
\affiliation{Ecology, Evolution and Organismic Biology, University of North 
Carolina at Chapel Hill, United-States of America}

\author{Karthik Ram}
\affiliation{Environmental Science, Policy, and Management. University of California, Berkeley, United-States of America.}

\author{Timoth\'ee Poisot}
\affiliation{Theoretical Ecosystem Ecology laboratory, Universit\'e du Qu\'ebec \`a Rimouski, Canada.}
\affiliation{Quebec Center for Biodiversity Science, McGill University, Canada.}
\affiliation{International Network for Next-Generation Ecology.}

\author{Dominique Gravel}
\affiliation{Theoretical Ecosystem Ecology laboratory, Universit\'e du Qu\'ebec \`a Rimouski, Canada.}
\affiliation{Quebec Center for Biodiversity Science, McGill University, Canada.}

\keywords{Publishing; Preprint servers; Green Open Access.}

\maketitle

\section{Introduction}

Public preprint servers allow authors to make manuscripts publicly available
before, or in parallel to, submitting them to journals for traditional peer-
review. The rationale for preprint servers is fundamentally simple: to make
the results of research available to the scientific community as
soon as possible, instead of waiting until the peer-review process is fully
completed.  Sharing manuscripts using preprint servers has numerous advantages
including: 1) rapid dissemination of work-in-progress to a wider audience; 2)
immediate visibility of the research output for early-career scientists; 3)
improved peer review by encouraging feedback from the entire research
community; and 4) a fair and straightforward way to establish precedence.

Open preprint servers offer a great opportunity for open science, especially if
the community embraces the idea of discussing preprints. Initiatives like
Haldane's Sieve (\href{http://haldanessieve.org/}{http://haldanessieve.org/}) \cite{coo12}, a
new blog discussing arXiv papers in population genetics, will help make arXiv
attractive for scientists looking to promote their work \cite{lom12}. These
initiatives are important to fully exploit the potential of open preprint
servers. Posting preprints online increases the community of available informal
peer reviewers, and uses the internet for its original community-building
purposes.

Preprints began to gain popularity 20 years ago with the advent of arXiv, an
open preprint server widely used in physics and mathematics \cite{gin11}.
Preprints are also integral to the culture of other scientific fields.  Paul
Krugman noted that, in economics, the \emph{traditional model of submit, get
refereed, publish, and then people will read your work broke down a long time
ago. In fact, it had more or less fallen apart by the early 80s} \cite{kru12}.
In addition to a section in arXiv, economists have the RePEc (Research Papers in
Economics) initiative, which aims to create an archive of working papers,
manuscripts, and book chapters.

Despite the success of this approach in other fields, most manuscripts in
biology are not posted to preprint servers and are therefore not seen by more
than a handful of other scientists prior to publication. In this article, we: 1)
highlight the advantages of open preprint servers for both scientists and
publishers; 2) discuss the preprint policies of major publishers in biology; and
3) review the most popular preprint servers currently available.

\section{The case for public preprints}

The first and most often discussed advantage of arXiv and open preprints is
speed (Figure~\ref{fig:map}). The time between submission and the official
publication of a manuscript can be measured in months, sometimes in years. For
all this time, the research is known only to a select few: colleagues, editors,
reviewers. Thus, the science cannot be used, discussed, or reviewed by the wider
scientific community. In a recent blog post, C. Titus Brown noted how posting a
paper on arXiv quickly led to a citation (arXiv papers can be cited) and his
research was used by another researcher \cite{bro12}. The current system of
hiding manuscripts before acceptance pose problems for both scientists and
publishers. Manuscripts that are unknown cannot be used and thus take more time
to be cited. It has been shown that high-energy physics, with its high arXiv
submission rate, had the highest immediacy among physics and mathematics
\cite{pra05}. Immediacy measures how quickly articles are cited. 

Public preprints can be crucial to early-career scientists. The delay before
publication is seldom compatible with the pressure to show an impressive
publication record when applying for a scholarship or a position. Increasing the
perceived value of pre-prints as close, or equal, to journal articles will allow
young researchers to put their research outcome in the open, and build a
reputation for themselves through the diffusion of their work without fear that
this work will not be recognized by grant or job committees. In this
perspective, mixed publication systems like F1000Research [\ref{f1000r}] should
be encouraged as they combine the best of both worlds: contributions are
initially considered preprints, but acquire full article status after positive
reviews are received.

Posting manuscripts as preprints also has the potential to result in higher
quality science by allowing prepublication feedback from a large pool of
reviewers. Prepublication reviews by a small network of colleagues are common in
the biological sciences and form an important part of the scientific process.
These ``friendly'' reviews increases the chances of errors being caught prior to
publication. Furthermore, the formal peer-review process as a whole is
critically over-loaded. As the number of active scientists increases and the
pressure to publish increases, it is becoming difficult for journals
to find reviewers \cite{hoc09}.  At the same time, rejection rates are high in
most journals \cite{aar08,roh09}, and when not invited to submit a revision,
authors must start the process over again at another journal. As a result,
initiatives to reduce time from submission to publication have emerged across
the scientific community. Rohr et al.  \cite{roh09} called for the recycling and
reuse of peer-reviews: by attaching previous reviews and detailed replies to a
new submission, both the editor and the referees can gauge the work done on the
manuscript, and perhaps evaluate it with less prejudice. A widespread use of
preprint servers can achieve the same goal of reducing the time spent in review.
With a rich enough community of scientists depositing preprints, and commenting
on them, the process of an open pre-review can become widespread and will
overall increase the quality of first submissions \cite{hoc12}.

\begin{figure}[ht!] \centering\includegraphics[width=0.50\textwidth]{map.pdf}
\caption { It can take several months, and even a few years, before a submitted
paper is officially published and citable.  The average time to publication
varies greatly between journals and can be as low as 104 days (Evolution for
2011) to 213 (PLOS One in 2010).  Meanwhile, few people are aware of the
research that has been done since, typically, only close colleagues are given
access to the preprints. With public preprint servers, the science is
immediately available and can be openly discussed, analyzed, and integrated into
current research. It benefits both science and publishers. Both want the papers
to be well-known and cited, and public preprints make it possible to integrate
research even before publication, greatly improving immediacy.  }
\label{fig:map} \end{figure}

Finally, public preprint servers offer a fair way to establish intellectual
priority by making the work available as soon as it is complete. Some
manuscripts will spend much more time than others in the review process and/or in
production after acceptance. This means that publication and
acceptance dates do not accurately characterize who came up with an idea
first. For this reason, mathematicians and physicists have embraced arXiv in
part to establish priority in a fair way \cite{gin11,cal12}.

\section{Preprints in biological sciences}

In contrast to other disciplines, the field of biology has effectively no
preprint culture, with the exception of small pockets of primarily highly
quantitative research (\emph{e.g.}, epidemiology, population genetics).
Submitting to preprint servers has become more common in the past few years,
and the quantitative biology section in arXiv is experiencing the fastest
growth in submissions \cite{cal12}. Yet, the number of biology papers
submitted to preprint servers still represents only a small fraction of the
total research produced in biology.

There are a number of reasons why biologists did not develop the reflex of
sharing preprints, many of which are based on common misconceptions. For
example, in contrast to other fields there is a perception in biology that
public preprints make it easier to steal ideas \cite{gin11}. However, there is
no evidence of this happening in the numerous other fields that have adopted
preprint servers, and since preprint servers create a clear record of who had
the idea first, and when, this appears to be a largely unfounded concern. In
other fields preprints serve the opposite role, they allow straightforward
establishment of precedence, letting research lay claim to an idea thus
preventing it from being ``stolen'' \cite{gin11}

Another major concern is based on a certain interpretation of the Ingelfinger
rule: scientists should not publish the same manuscript twice \cite{alt96}. A
preprint is simply a document that allows ideas to spread and be discussed, it
is not yet formally validated by the peer-review system. This is why the
majority of publishers do not see arXiv and similar services as a violation of
the Ingelfinger rule. Almost all of the major publishers in biology are
preprint-friendly, including: Nature Publishing Group, PLOS, BMC, PNAS,
Elsevier, and Springer \ref{table:policies}. This year, both the Ecological
Society of America and the Genetics Society of America changed their policies to
allow public preprints. \emph{Nature} even felt compelled to respond to the
rumour that they refused manuscripts submitted to arXiv by saying that
``\emph{Nature} never wishes to stand in the way of communication between
researchers. We seek rather to add value for authors and the community at large
in our peer review, selection and editing'' \cite{nat05}.  Still, a few journals
adopt a ``by default'' hostile attitude towards preprints, mostly due to the
lack of clear policy of the publishers, or perhaps because a preprint culture
has not developed in biology and the practice is still considered unusual. As an
example, Wiley-Blackwell, which publishes some of the leading journals in
biology, has no official policy on the matter.

\begin{table*}
    \centering
    \begin{tabular}{|ll|}
    \hline
    Publisher                                   & Policy \\
    \hline
    Springer                            	& Accept \\
    BMC                                 	& Accept \\
    Elsevier                            	& Accept \\
    Nature Publishing Group             	& Accept \\
    Public Library of Science           	& Accept \\
    Genetics Society of America                 & Accept \\
    Royal Society                       	& Accept \\
    National Academy of Science (USA)           & Accept \\
    Ecological Society of America       	& Accept \\
    Oxford Journals                             & Accept \\
    Science                             	& Ambiguous \\
    Wiley-Blackwell                       	& No general policy \\
    British Ecological Society                  & No answer to our query \\
    \hline
    \end{tabular}
    \caption{Policies for important publishers in biology. Some publishers
tolerate preprints except for a few of their medical journals, for example the
\emph{ Journal of the National Cancer Institute} from Oxford and \emph{The
Lancet} from Elsevier.}
    \label{table:policies}
\end{table*}

\section{Current offerings}

We briefly discuss the main options to submit preprints to open servers:
arXiv.org, figshare, and the upcoming PeerJ and F1000Research.

\subsection{arXiv}

arXiv (\url{http://arxiv.org/}) is the most widely-used preprint server today,
and its use is almost universal in some branches of mathematics and physics.
arXiv provides a reliable citation system for all eprints and is especially
popular in high-energy physics. Physicist Paul Ginsparg created arXiv in 1991
for theoretical high-energy physicists to communicate preprints via email and
ftp, and soon thereafter adopted the newly created world-wide web
\cite{jackson2002preprints}.  arXiv now receives over 7 000 submissions per
month (\url{http://arxiv.org/show_monthly_submissions}) and divides its
submissions into subcategories of physics, mathematics, computer science,
quantitative biology, finance and statistics.  The quantitative biology category
includes subcategories for Populations and Evolution, Quantitative Methods and
other categories that may be of interest to biologists.

One aspect of arXiv that differs from other options is that it has a moderation
system, which requires that papers must be categorized by an endorser.  At least
one author of a paper must be an endorser that has previously submitted a paper
or has received permission to submit to a particular category. arXiv is now
administered by the Cornell University Libraries, with funding coming from
voluntary pledges by academic institutions along with matching funds from the
Simons Foundation \cite{arxiv_future}. This approach to financing is one of the
numerous measures that arXiv takes to ensure that the repository will remain
permanently available and submissions will be readable.

\subsection{figshare}

figshare (\href{http://figshare.com}{http://figshare.com}) is an open server
that allows scientists to submit any research output: manuscript, figures,
datasets, videos, theses, presentations, and so on. There are no rules to limit
what constitutes a research output and, unlike arXiv, there is no endorser
system. All figshare content has a unique digital object identifier (DOI) like
any journal article, thus offering a permanent and stable link to the content.
A flexible tag system is used to classify each item. Comments can be made on all
content allowing for centralized discussion related to the material.

One of the advantages of figshare over arXiv for biologists is that is it not
limited to quantitative sciences. arXiv.org has sections on quantitative biology
but might not be appropriate for non-quantitative work. With its flexible
approach to preprints, figshare offers an alternative to arXiv for empirical
biologists. Furthermore, by allowing all types of content, figshare also
provides an archive for early results (\emph{e.g.}: figures, lab presentations).

\subsection{PeerJ}

PeerJ (\href{https://peerj.com/}{https://peerj.com/}) is a new open access
publisher that combines both a preprint server and a peer reviewed journal.  It
is focused on the biological and medical sciences, which may help overcome the
perception that preprints do not have a place in biology. Like figshare this is
an advantage relative to arXiv for biologists doing non-quantitative work.  Also
like figshare, PeerJ allows commenting on posted preprints, improving the
potential for pre-publication dialog. In addition, preprints can be made private
if the authors choose, and shared only with selected colleagues. While this
reduces many of the benefits of preprints described above, it may allow some
researchers who would not otherwise post preprints to begin exploring the
possibility in a manner appropriate to their current circumstances.

In contrast to other preprint servers, users cannot post unlimited public
preprints for free. One preprint per year can be posted for free, with a onetime
(\emph{i.e.} lifetime) fee of 99 dollars required to allow an author to post
unlimited public preprints. The preprint server is not tied to the journal, so
preprints can be posted regardless of where they will eventually be submitted
for publication.

\subsection{F1000Research}\label{f1000r}

F1000Research is not a public preprint server like the previous three servers.
Whereas arXiv, figshare, and PeerJ offer an option to submit a manuscript
without having it reviewed, papers submitted to F1000Research will eventually be
reviewed. Thus, F1000Research offers a hybrid model with publicly available
manuscripts at time of submission and standard peer-reviews that occur as part
of the submission process. Manuscripts are considered ``accepted'' and will only
be indexed after two positive referee responses. F1000Research works closely
with data providers to integrate raw data to the paper. For instance, upon
submitting a paper, authors are asked to upload their data, which are then
integrated in \emph{e.g.} figshare widgets, the DOI of which are given in the
paper when the data are first mentioned.  By integrating data to the paper,
F1000Research is working to make science more reproducible and open.

\subsection{GitHub}

This manuscript was developed entirely as an open project on GitHub. GitHub is
one of several hosting services for collaborative development using the Git
version control system (VCS).  Git is a decentralized revision control system
created by Linus Torvalds and is used primarily to develop software, including
the Linux kernel. Git provides powerful features that allow numerous
contributers to work asynchronously on the same project, often in parallel
branches, all of which can be effortlessly merged and version controlled.  While
Git is created primarily for software development, where the use of version
control systems is standard \cite{aru12}, it is ideal for academic research
since it provides a way to collaborate on every step of the manuscript
development process, from data manipulation and analysis to writing and
revision. For example, during the development of this manuscript, each author
would clone the project (\emph{i.e.} make a personal copy), modify it, and then
merge their changes into a master branch. This takes the preprint process to an
new level, where the entire writing process is open from the beginning.

\section{Conclusion}

The ongoing discussions on the publication process, peer-reviewing and
alternative publication models are all symptoms of the current uneasiness of
scientists with the ever growing obsession with bibliographic metrics such as
the impact factor \cite{Fisher2012}. There is pressure on researchers to orient
their publication strategy in order to maximize their number of publications and
total citations. A well-known consequence is to submit manuscripts first to the
most prestigious journals, and then resubmit to ``lower level'' journals as they
are rejected. The numerous negative impacts of such behavior have been discussed
in depth \cite{hoc09} and include a long delay between the time a manuscript is
finished to its publication.  They all contribute in a general slowing down of
scientific progress.  Research activities and the publication process are
drifting away from their fundamental object, namely the diffusion of novel
scientific discoveries. 

Developing a preprint culture in biology will not solve all problems with the
current publication process. However, it might significantly reduce its negative
consequences. The role of peer-reviewing is to judge the scientific quality of a
study. It is the first barrier against the fraudulent and poor quality science
susceptible to impede scientific progress. In practice, the peer-review system
is not only used to evaluate scientific quality but also to judge pertinence. On
the other hand, preprints are not filtered, neither for their quality nor their
pertinence. Widespread adoption of preprint servers has the potential to shift
the diffusion strategy: journals would remain important to validate
publications, but the relevance of a study, its contribution to science, should
only be judged by many more readers than the typical two-four anonymous
reviewers. With a shift in the diffusion strategy, the role of traditional
journals and their editors would be to showcase scientific discoveries for
specialized readership. 

One of drawback of making publication easier is the proliferation of studies of
uneven quality. A trade-off between the intensity of the peer-review filtering
and the benefits to science has been hypothesized \cite{Aarssen2012}.  With
increasingly stringent peer reviewing, the quality of published papers and
easiness of finding discoveries might increase, at the cost of censorship, an
increased load on authors and reviewers, and greater delays for publication.
Preprints are simply bypassing this model for what we believe is the progress of
science: they speed up the dissemination of scientific discoveries, impede
censorship and put on reader's shoulders the responsibility to judge originality
and pertinence.

\section{Funding}

PDP is supported by an Alexander Graham Bell scholarship from the National
Sciences and Engineering Council of Canada. EPW is supported by a CAREER Award
from the National Science Foundation (DEB-0953694). JJA is supported by NSF
DEB-0614166 and NSF DEB-0919018.  TP is supported by a FQRNT-MELS post-doctoral
scholarship and 25 cents found by a coffee machine.  KR is supported by NSF
DEB-1021553. DG is funded by a Discovery Grant from the National Sciences and
Engineering Council of Canada and by the Canada Research Chair program.

\section{Acknowledgements}

We thank Carl Boettiger, Mark Hahnel, and Hedvig Nenz\'en for helpful comments
on an earlier version of this manuscript.

\newpage
\bibliography{refs}

\end{document}