The following discussion can be divided into to two main categories: (1) who to include as a co-author, and once the list of co-authors has been determined, (2) in what order should they be listed?
Before launching into discussing the issues related to Category 1, it is prudent to declare that there are as probably as many conventions as there are publishing scientists, and each discipline’s most general conventions differ across the scientific spectrum. I’m sure if you asked 10 people about what they considered appropriate, you could conceivably receive 10 different answers.
That said, I do still think there are some good-behaviour guidelines on authorship that one should strive to follow, all of which are based on my own experiences (both good and awful).
So who to include? It seems like a simple question superficially because clearly if someone contributed to writing a peer-reviewed article, he/she should be listed as a co-author. The problem really doesn’t concern the main author (the person who did most of the actual composition) because it’s clear here who that will be in almost every case. In most circumstances, this also happens to be the lead author (but more on that below). The question should really apply then to those individuals whose effort was more modest in the production of the final paper.
Strictly speaking, an ‘author’ should write words; but how many words do they need to write before being included? Would 10 suffice, or at least 10%? You can see why this is in itself a sticky subject because there are no established or accepted thresholds. Of course, science generally requires much more than just writing words: there are for most papers experiments to design, grants to obtain to fund them, data to collect, analysis and modelling to be done, figures and tables to prepare and finally, words to write. I’ll admit that I’ve co-authored many papers where I’ve done mainly one of those things (analysis, data collection, etc.), but I can also hold my hand over my heart and state that I’ve contributed more than a good deal to the actual writing of the paper in all circumstances where I’ve been listed as a co-author (the amount of which depends entirely on the lead author’s writing capacity).
It is my personal view that if someone contributed to any of the components required to complete the article, then they should be given the opportunity to claim co-authorship. The only proviso to that policy relates to two of these sub-components: (i) collecting data and (ii) funding the research.
Data collection: In most cases, it’s pretty clear that that this process takes a lot of time and effort, so if someone who collected the data you need passes it along to you, you should probably offer authorship on any articles arising. That said, there is now a huge number of freely available databases out there, so it would be completely unfeasible to offer authorship to all the players involved. In a legal sense, if the data are publicly funded and available online, you have every right to publish papers based on them without necessarily including the custodians as co-authors. Most such databases give very clear indications of their specific rules on this matter, so do read up on them. If it’s a colleague with a less formalised database, then authorship is a little more straight-forward. My rule of thumb? If in doubt, include as co-author.
Funding: Scientists spend a huge amount of time begging (errhm, … requesting … ) funding to do their research, which inter alia requires original thought, collaboration, novelty, good communication skills, impressive track records and certain flair for impressing funding agencies. If someone funded the work that led to your paper, you should probably consider asking them to co-author your work. That said, if it’s the 500th paper arising from a major grant with scores of collaborators, it might not always be sensible or necessary to do so. The main rule of thumb here is that you should always communicate openly and sincerely with anyone who might be affected by this issue.
Another problem that can rear its ugly head is when someone is paid to put together a dataset or collect data. In most cases I know, the person specifically employed to do so is not generally expected to be a co-author, unless she/he demonstrates exceptional effort, novelty, independence and foresight. I currently have a research assistant that meets all these latter criteria and more, and so she is not only included in any papers arising from ‘her’ dataset, she is also leading a few papers of her own.
All these considerations, complexities, social contracts and expectations in mind, the best piece of advice I can give on authorship is this: it rarely hurts you to include someone of marginal contribution, but it can really hurt you to exclude someone who at least thinks he/she deserves authorship. Engrave that sentiment into your frontal lobe for posterity, because following it will make a huge difference to your long-term publication success.
Now for the stickiest issue of all — author order.
The main reason this is so sticky is that there are even fewer rules of thumb to follow. I’ll tell you what I tend to do, but it really depends on the discipline and the culture of the labs to which you belong. Of course, first authors should be the ones that tend to do the most work, and especially the actual writing. There are times when this isn’t always the case, but it’s so standard that there shouldn’t be many situations where it’s difficult to determine who this should be.
In ecology anyway, there’s an increasingly recognised trend that the second-most important position in author order is the last. This tradition probably goes back to medical research laboratories where the lab head generally occupied the final position in the list of authors (whether or not she/he actually wrote any of the paper). The culture was strengthened when the British Research Assessment Exercise ratified the practice by placing emphasis almost exclusively on first and last author positions for papers counting toward one’s track record. In Australia anyway, the practice is becoming more and more mainstream, hence one of the reasons that I tend to take the last position in most papers coming out of my lab and for which I arguably have the second most important contribution to overall authorship.
Talking to colleagues in France, USA and elsewhere, I’m not sure the practice is quite as strong, but you should at least be aware that it exists. After the ‘last position’ rule of thumb, the next most important position is certainly second author. After that, it tends to be a case of diminishing contributions up to the penultimate author. In cases where this isn’t necessarily clear, and certainly in many cases of many (i.e., > 10) authors, the ones between position 2 and n-1 are often listed alphabetically by surname. The latter trick is a good way to avoid infighting amongst co-authors.
Increasingly there are situations where two or more authors have equal contributions. It is perfectly acceptable and even now conventional that such authors are indicated by some sort of superscript symbol as having contributed equally. This often includes the first two authors, or the last two, but I’ve even seen it applied to positions 2 and 3, or to even more than two authors. For anything but the lead and second author, however, the paper will always be referred to as SURNAME1 et al., so take the ‘equal contribution’ convention with a grain of salt.
In conclusion, make sure you discuss with all your authors any issues that could arise BEFORE they do. Heading off any future problems by discussing them well beforehand is a good way to avoid woe later on. If you follow our paper-writing protocol and circulate to all potential authors from the outset of the idea, you will also avoid many of the worst problems from the outset. Happy (co)writing!