Everything Else
Is Commentary

The problem this chain is about is stated, plainly and early, in The Descent of Man. Not as a puzzle Darwin left for later — as a mechanism he named.

He then describes the tribe that wins — its members "always ready to aid one another, and to sacrifice themselves for the common good" — and finishes the thought with five words that the next hundred years spent arguing about:

So the structure was there from the start. A behaviour can cost you something relative to the people beside you and still spread, because the group carrying it does better than the groups that do not. Two levels of accounting, running at the same time, pulling opposite ways.

Everything that follows in this chain is an argument about which level you are allowed to do the arithmetic at — and about what happens to a hedge when a sentence gets shorter.

In 1962 V. C. Wynne-Edwards published Animal Dispersion in Relation to Social Behaviour, arguing that animals restrain their own breeding to keep populations within what the habitat can bear. Good for the group. Costly to the individual.

It was taken apart, and it deserved to be. George C. Williams's Adaptation and Natural Selection (1966) is the standard reference for the demolition, and the argument is short enough to state in a sentence — which is why it won.

Put one animal in the restrained group who does not restrain. It leaves more offspring. They inherit the not-restraining. Repeat, and the group is no longer restrained. A trait that is good for the group and costly to the individual gets eaten from the inside, every time, unless something specific stops it.

Dawkins puts the same objection in the first chapter of the book this chain later spends four links on, and it is worth quoting because it shows how settled it was:

This joint is the one that keeps the chain honest. The rejection of naive group selection was not an ideological purge and it was not a mistake. It was a correct argument against a specific bad model. Any account of this history that skips it — including a sympathetic one, including ours — is selling something.

With the group-level explanation discredited, altruism needed an account that survived the selfish-rebel objection. W. D. Hamilton gave it one, in two papers in the Journal of Theoretical Biology in 1964.

The idea: a gene does not care which body it is sitting in. If helping a relative propagates copies of the same gene, helping can pay even when it costs the helper. The condition is compact enough to write on a napkin — a benefit to the recipient, weighted by how related they are, exceeding the cost to the actor.

This is good science and it remains good science. It made predictions, they were tested, a great deal of them held. Ants and bees and naked mole-rats and helpers at the nest all became legible. Nothing in this chain is an argument against inclusive fitness, and if a reader finishes it thinking otherwise we have written it badly.

Two things to keep, though, for later links. First, Hamilton's answer is an accounting device: it re-books the ledger from the body to the gene, so that a cost to the body can still be a profit on the account. Second — and this is the part the popular version dropped — re-booking the ledger to the gene does not by itself say anything about which level the interesting behaviour is happening at.

It says where to add things up. That is not the same claim, and the difference is about to matter enormously.

Six years later George Price wrote down a piece of algebra that describes how any trait's average changes under selection — and it splits the change into two parts.

One part is the selection happening within groups. The other is the selection happening between groups. Not as a rival theory. As two terms in the same expression, either of which can be positive while the other is negative.

So by 1970 the mathematics for both accounts existed, in one formula, written by one person. The framework that would be called kin selection and the framework that would be called multilevel selection are both derivable from it. The dispute that ran for the next forty years was, at the level of the algebra, already settled before it started.

An honest note about a temptation. George Price's life ended in circumstances that are often told as a parable about altruism, and we are not going to tell it that way. It is not evidence about anything in this chain, it belongs to him rather than to our argument, and reaching for it would be exactly the move Too Good to Check exists to refuse. The equation is the contribution.

And a lead we are not calling evidence. Kropotkin's Mutual Aid (1902) had already gone looking for cooperation in nature and found it everywhere, sixty-eight years earlier. But he was making a different kind of claim — natural history in the service of a politics, without any of the accounting that would later settle the question either way. He is a lead, not a citation. Being early is not the same as being right, and a chain that recruited him as an ally would be doing what this collection was built to catch.

Here the chain leaves the mathematics and picks up the other thread, which is the one that actually reached most people. It starts with a conversation about a word.

Dawkins tells the story himself, in the introduction to the thirtieth anniversary edition. In 1975, through Desmond Morris, he showed the partly-finished book to the London publisher Tom Maschler.

Dawkins kept his title. And then, thirty years later, in the same passage:

He goes further, and names the reason: "I can readily see how the title could be misunderstood, and this is one reason why I should perhaps have gone for The Immortal Gene. The Altruistic Vehicle would have been another possibility."

Hold that third option for a moment. The book we know as The Selfish Gene could, on its author's own later account, have been called The Altruistic Vehicle — and it would have described the same mathematics, the same arguments, the same chapters. Only the level named in the title would have changed.

This is the link the chain exists for, and it is entirely quotation. Everything the popular reading of The Selfish Gene gets wrong is corrected inside The Selfish Gene.

On where the emphasis goes:

And then, explaining why — in a sentence that is multilevel selection in everything but name:

On the naturalistic fallacy, in chapter one:

And on what he wanted the book used for: "if you wish, as I do, to build a society in which individuals cooperate generously and unselfishly towards a common good, you can expect little help from biological nature."

So: the author says the operative word is gene, that the book is mostly about altruism, that altruism can win at other levels, that he is not deriving an ought from an is, and that he personally wants a generous and unselfish society. All of it printed. None of it travelled.

The misreading arrived immediately and Dawkins spent decades on it. The philosopher Mary Midgley's attack in Philosophy opens with a sentence he still quotes back:

Which is true, and beside the point, and he replied at length. But the more interesting event is what he did to his own text.

The most-quoted line in the book is probably "Let us try to teach generosity and altruism, because we are born selfish." In the anniversary introduction, Dawkins takes it back:

He explains that he had not yet drawn the distinction between vehicles — usually organisms — and the replicators riding inside them, and asks readers to "mentally delete that rogue sentence and others like it."

Consider the shape of that. An author identifies his own most-repeated sentence as misleading, prints the retraction in the front of every subsequent copy, and asks readers to delete it. The sentence is still what most people can quote. The retraction is not.

A correction printed in the same book as the error does not automatically win. Nothing about being right makes a sentence travel.

Here is the claim you have probably met: that the selfish gene supplied a biological alibi for a politics of self-interest — that it arrived in 1976 and was extremely convenient for what followed.

We are marking this joint contested, and we want to be exact about why. Several things here are documented and one is not.

Documented: the phrase escaped biology and became common speech. Documented: it was widely read as a claim about human nature rather than about levels of accounting — Dawkins says so repeatedly and irritably, and this chain has spent two links quoting him. Documented: the 1970s sociobiology debates were explicitly political on both sides.

Not documented: that the metaphor caused a political shift. That is a causal claim about ideas moving through a culture, of the sort that is very easy to assert, very satisfying to believe, and extremely hard to establish. We have not seen it established. We would like it to be true, which is precisely the reason to say we have not shown it.

What we will say is narrower and survives. A phrase that compresses the gene is the unit of accounting into selfishness is fundamental is available for that use. It is pre-loaded. Anyone wanting to naturalise self-interest did not have to distort it much — they only had to stress the first word, which is the exact stress its author asked them not to use.

Availability is not causation. But it is not nothing, and it is the most we can support.

Back to the mathematics rail. In 2007 David Sloan Wilson and Edward O. Wilson published Rethinking the Theoretical Foundation of Sociobiology in The Quarterly Review of Biology — and the abstract does not hedge about what it is doing.

Read what that does and does not claim. It does not say the 1960s rejection was wrong — the paper's own account of Darwin, two pages later, is that "selfish individuals might out-compete altruists within groups, but internally altruistic groups outcompete selfish groups." The refutation of Wynne-Edwards stands. What is revived is a more careful formulation: not "animals do things for the good of the species," but a specific claim about when between-group selection can outweigh within-group selection.

Which, if you have been following the rail, is Price's two terms with names on them.

The paper is where multilevel selection gets its modern statement as a foundation rather than a heresy — and it is also, in one specific and instructive way, the place this chain turns back on itself.

Wilson and Wilson state their rule once in the body of the paper, and once at the end as a summary. The two are not the same, and the difference is this chain in miniature.

In the body, stating the rule that all the models obey:

At the end, after invoking Rabbi Hillel — who summarised the Torah standing on one foot and ended "everything else is commentary" — they offer their own one-foot version:

Between the two, single has gone and the exception has gone. Not dishonestly — the second is openly labelled a summary, and a summary is allowed to compress. But it is the compressed one that travels, because it is the one built to.

And it travelled to us. The Stimpunks glossary entry on Altruism quotes the one-foot version — reached second-hand, through Jorn Bettin's The Beauty of Collaboration at Human Scale, rather than from the paper. No single. No exception. We are three links down the chain we are describing, and we did not notice until we went to the paper to write this spread.

The glossary is being corrected. This zine is not the place that fixes it, and publishing the error before the fix is the point of saying so at all.

If this chain ended at 2007 it would be telling you that the question is settled in favour of multilevel selection. It is not, and the evidence that it is not was published in Nature.

In 2010 Martin Nowak, Corina Tarnita and Edward O. Wilson published The evolution of eusociality, arguing that inclusive fitness theory is "a viable alternative to the direct fitness of standard natural selection only for a well-defined subset of situations — it is not generally applicable."

The response was one of the largest collective rebuttals in modern biology: Abbot and more than a hundred co-signatories, in Nature the following year, under the title Inclusive fitness theory and eusociality. Their objection is blunt — that the argument rests "upon a misunderstanding of evolutionary theory and a misrepresentation of the empirical literature." Several separate rejoinders ran alongside it.

A note on a number we are not printing. The author count for that reply is usually given as 137. We counted the author list in the manuscript PDF and got 118, with no duplicates, alphabetical from Abbot to Zink. We have not resolved the discrepancy — it may be the difference between manuscript and published versions, or our extraction dropping names. So the zine says more than a hundred, which both counts support, and does not print a figure it cannot stand behind. This chain is about compressions that lose their hedges. It would be an odd place to invent one.

Where the dispute actually sits, as far as we can tell: not on whether cooperation evolves, and not on the data. On which formalism is more general, and which one shows you the causes.

And here is the finding that makes forty years of public argument look strange. For a wide class of cases, kin selection and multilevel selection are mathematically equivalent. Not similar. Equivalent — two ways of partitioning the same covariance.

The kin-selection route takes the covariance between an individual's trait and its fitness across the whole population and splits it by regression into direct and indirect components. The multilevel route takes the same total and splits it into within-group and between-group components. Both are rearrangements of the same terms, and both fall out of Price. The literature notes, with some weariness, that the equivalence has been pointed out repeatedly and the argument has continued anyway.

Why this joint is contested rather than documented. The equivalence itself is established for the standard models — that part is not in doubt. What remains genuinely disputed is its scope: whether it holds in full generality, and whether two formally equivalent accounts are therefore equally good, or whether one of them shows you the causal structure and the other buries it. That is a live disagreement among people who agree on every equation, and we are not going to adjudicate it in a zine.

But notice what the equivalence does to the fight the popular versions staged. Selfish gene against altruistic group, individual against collective, one side's summary against the other's — and underneath, the same sum, added up in a different order.

Both compressions were true to their arithmetic. Both lost the part that said this is a choice about bookkeeping, not a discovery about what things are really like.

Nothing in twelve links licenses anything about how anyone should be treated. Marking this leap is not modesty; it is the accurate tag. We are stepping from a fact about bookkeeping to a claim about people, and the step is ours.

Here is the step. If you do all the accounting at one level, the contributions of every other level become invisible — not disproved, invisible. Not because anybody lied. Because they were never entered.

Account for a child's difficulty entirely at the level of the child, and the ledger balances: the deficit is in the child, and the available repair is to fix the child. Account at the level of the room — how many transitions it demands, how loud it is, whether anyone in it is safe-enough to co-regulate with — and different entries appear, and different repairs become thinkable. Neither ledger is lying. They are recording different columns.

Which is why where is the problem located is never only a technical question. The choice of level is made before any evidence arrives, it decides what counts as data, and it is almost never stated out loud. A whole apparatus of diagnosis, remediation and grit sits on top of a bookkeeping decision that presents itself as simply looking at what is there.

So: everything else is commentary — and the commentary is where the hedges live. The exception clause. The single. The introduction saying stress the second word. The retraction of born selfish. Every one of them printed, by the author, in the same volume as the claim. Every one of them lost on the way to us.

Read the commentary. It is not the part that was too long. It is the part that was true in more cases.