Fourteen Kinds of Doing

Take the byline seriously as a piece of engineering for a moment, because that is what it is.

It is a data structure, and it has exactly one shape: an ordered list of names. That is the whole of it. Two degrees of freedom — who is in the list, and what order they come in — carrying everything anyone is ever going to know about who made the thing.

There is no field for what anybody did. Position is made to stand in for it, which is why the conventions differ by discipline and why nobody outside a field can read them: first author means one thing in biology and another in mathematics, and last author means the opposite of itself depending on where you are standing.

A format with one field will record one kind of fact. The question is what happens to every other kind.

This zine is not about a wrong being done to anybody. Field Guide No. 25 catalogues that, and its finding is that credit comes off work in five different ways. This one is about the format underneath all five, and about what to do — which is a different and more cheerful subject.

In May 2015 the two big detector collaborations at the Large Hadron Collider published a joint measurement of the Higgs boson mass in Physical Review Letters.

The paper has 5,154 authors. It runs to thirty-three pages. The research and the references occupy the first nine. The remaining twenty-four are the list of names and the institutions they belong to.

The author list is nearly three times the length of the paper.

It is easy to file that as an absurdity, and the reporting at the time largely did. It is worth resisting. Five thousand people really did make that measurement — the detectors, the trigger systems, the calibration, the software, the years of running. The number is the field being honest about what it took.

What is absurd is the format, which could only record that honesty by printing twenty-four pages of names and still not saying what a single one of them did.

Not the physicists, in the end, but the biomedical journals, and then broadly: a fixed vocabulary for kinds of contribution, to sit alongside the byline rather than replace it.

It is called CRediT, the Contributor Roles Taxonomy. It has fourteen roles, and since 2022 it has been a standard with a number on it — ANSI/NISO Z39.104-2022.

Here they are. This is the entire vocabulary.

The split down that table is ours, not the standard's. CRediT publishes fourteen roles in one flat alphabetical list and ranks nothing, deliberately. The two columns are this zine's reading of which ones a reader could already have guessed at from a name in a byline, and the next spread is about what falls in the right-hand column.

Data curation. Somebody cleaned it, annotated it, and kept it usable.

Software. Somebody wrote the code the result came out of.

Validation. Somebody checked whether it was true.

Visualization. Somebody made it legible — which is often the difference between a finding and a finding anybody understood.

Resources. Somebody kept the instrument running.

Every one of those is work. Not one of them has ever had anywhere to go.

This is the part worth sitting with, because it is not a story about prizes or about anybody being slighted. It is plainer than that. For three hundred years the record of who made a piece of knowledge had one field in it, and the work that field could not describe was not hidden or suppressed. There was simply no box.

The fourteen roles are interesting not because a taxonomy is interesting but because reading them tells you what the old format could not see.

Here is the whole argument, and it comes out of the mechanics rather than being placed on top of them.

A recognition system with one field will reliably recognise the people whose contribution happens to fit that field, and reliably miss everybody else — without anyone intending it, and without any of the missed work being any less real.

Which people? Read the right-hand column once more. The one who kept the instrument running. The one who cleaned the data nobody else could face. The one who wrote the software. The one whose contribution was validation — who checked, and found the number was wrong.

Those are recognisable people. Many of them are us — and not by coincidence. A profile that is spiky rather than flat produces exactly this: enormous contribution in a narrow band, and a recognition format with one box that band does not fit in.

And notice what the fix is not. Nobody needed to become a different kind of contributor. The format needed another field.

A rhyme, not a proof. Nothing about the structure of a byline demonstrates anything about a nervous system, and this zine is not claiming that it does. What the mechanics supply is one honest shape, held steady: a system that can record one kind of contribution will find one kind of contributor, and will go on believing that is all there was. Whether that shape fits your life is yours to say, not ours.

This is where a piece like this usually ends, with the standard as the happy conclusion. It should not.

Once contributions became recordable, they became countable — and people counted them. Vincent Larivière, David Pontille and Cassidy R. Sugimoto read the contributorship statements on 30,054 papers published by the Public Library of Science over 2017 and 2018, and asked how the work actually divides.

It divides along lines anybody could have guessed. 57 per cent of women contributed to investigation, against 49 per cent of men, with a similar gap for data curation. Men were likelier on the roles that travel with seniority — 30 per cent likelier on funding acquisition and supervision, and ahead on resources, software, conceptualization and project administration.

The vocabulary made the division visible. It did not make it go away.

That is what a vocabulary does, and it is worth being exact about the limit rather than disappointed by it. Naming a kind of work is a precondition for valuing it and nothing more than a precondition. Fourteen boxes will faithfully record that the same people keep ending up in the same boxes.

Which is an argument for the taxonomy, not against it. You cannot address a distribution you cannot see, and before 2014 nobody could see this one. But a piece that told you a standard had solved the problem would be doing the thing this house exists to refuse.

And one result in that paper argues the zine’s case better than the zine did. The older, coarser scheme PLOS used had a single category for writing, and under it men appeared to dominate. CRediT splits writing in two — original draft, and review & editing — and the picture inverts: women are 6 per cent likelier to have written the original draft, and men 8 per cent likelier to review and edit it. The authors call the inversion striking, and note the earlier finding had been skewed by how ubiquitous the reviewing half is. So the coarse vocabulary did not merely fail to record something. It produced a false picture, and a finer one corrected it — which is this zine’s whole argument arriving as a measured result rather than a claim. The authors are careful, and so is this: they note women academics are younger on average, while other work has found these differences hold across age and team size.

None of these needs a standard, a committee or a journal. They are the same five whether the thing being made is a paper, a codebase, a school project or a meal.

1 · Name the role, not only the person. “Sam helped” records nothing. “Sam found the error in the third column” records a contribution and a kind of contribution, and the second half is the part that survives being retold.

2 · Put it where the thing is recorded, not only where it is said. Credit given out loud in a meeting evaporates with the meeting. The same sentence in the commit message, the colophon, the footer or the minutes is still there in four years, which is when it matters.

3 · Name who caught it. This is the cheapest and the most neglected. Finding the mistake is work, it is the role CRediT calls validation, and the person who does it is structurally invisible — their contribution leaves no trace in the finished thing except its correctness.

4 · Date the correction and leave it up. A fix made silently teaches everyone watching that being wrong is unspeakable. A fix with a date on it teaches that it is ordinary.

5 · Mark the route. Say how you know — read at the primary, taken from a summary, not read at all. It is an attribution to the source and an honest statement of your own contribution, which is often smaller than a confident sentence implies.

Every one of the five is free, and four of the five take under ten seconds.

A piece recommending a practice should show its own, at the point where it is checkable rather than at the point where it is flattering.

Every source carries its route. Read the colophon of any piece here and each citation is graded — read at full text, from published secondary accounts, or not read. Field Guide No. 25 says on its own face that it never opened the Tesla and Purington patents.

Corrections are public and dated. The changelog carries them by date, in the reader’s view, naming what was wrong and what is right. A per-piece fact-check ledger in the repository is the working half, kept beside the source rather than published as a page.

Errors are attributed to whoever caught them. When a reader or a collaborator finds something, the entry says so by name. That is move three, and it is the one this house most wants to be judged on.

And it is inconsistent, in a way worth printing rather than tidying.

Credit here is much better recorded for sources than for labour. Every quotation has a traced origin; the person who built a diagram or rewrote a paragraph mostly does not appear anywhere. This site has the same one-field problem it is describing, and has only partly fixed it — which is the ordinary condition, and is why the practice is a habit rather than an achievement.

One more thing, and it is the reason this zine does not end on the previous spread.

Attribution is nearly free. That is what makes it available to anyone, today, at any scale — and it is also what makes it a convenient substitute for the things that are not free.

A named contributor who has not been paid is a contributor who has not been paid.

The same holds for a credited person who was not consulted, not employed, not given the next piece of work, or not believed when they said the number was wrong. Recognition is a real good and it is a small one, and it is easiest to offer precisely when the larger things are being withheld.

So: give credit generously, because it costs nothing and it is owed. And notice when it is being offered instead of something. The guide beside this zine has an entry on a woman who was given a magnificent public correction sixty-six years after her laboratory did not have a camera.