The story, as most people know it, goes like this: spinach’s legendary iron content was never real. Somewhere in the nineteenth century, a German chemist misplaced a decimal point while recording iron figures, inflating spinach’s value by a factor of ten. Generations of mothers invoked spinach as nutritional obligation; Popeye’s creator animated the myth into popular culture; and eventually some more careful analyst caught the error and put the decimal back where it belonged. By then the reputation had escaped. The vegetable kept its fame even after the number was corrected.

This account has the advantage of being tidy. It has the disadvantage of being undocumented. No record of the original error exists in accessible literature. The analyst who made the decimal transposition has no name, no paper, and no corrected table following from it. The story about the measurement error is itself a kind of measurement error — a claim propagated without reference until the reference was assumed to exist somewhere upstream.

Popeye first appeared in 1929. The first traceable academic mention of a decimal error in spinach’s iron figures is Arnold Bender’s 1972 lecture — forty-three years later. Whatever brought spinach to Popeye was not this citation chain. What the decimal point story adds to the folk belief — if it adds anything — is a cause. That cause turns out to be undocumented.

What does exist is a paper trail — a hedge that became an assertion, an assertion that became a citation, and a citation that appeared in a paper specifically about what happens when assertions become citations without anyone checking.

The chain as built

The first traceable link is Arnold Bender, a British nutrition researcher, in his inaugural lecture at Queen Elizabeth College London in 1972. Per Ole Bjørn Rekdal’s 2014 account in Social Studies of Science — which draws on the primary texts — Bender said that spinach’s fame “may well have grown from a misplaced decimal point.” [1] May well have. The language is hedged; Bender offers the decimal error as a plausible origin story, not a documented one. He does not name a scientist. He does not cite a paper. He puts the possibility forward.

By 1977, in a letter to The Spectator, Bender’s language had firmed — and he had acquired, or believed he had acquired, a source. He named a “Professor Schupan,” attributing to this scientist a 1937 finding that spinach’s iron content had been overstated tenfold. [7, 8] The attribution appeared to close the question: here was the researcher who had documented the error, and here was the date.

It does not close the question. The German botanist and nutritionist Werner Schuphan — the probable person Bender had in mind, his name misspelled — published extensively on vegetable nutrition in the mid-twentieth century. His 1940 article in Bodenkunde und Pflanzenernährung addresses spinach’s nutritional composition directly. Per Joachim Dagg’s 2015 examination of the source, it states that spinach “distinguishes itself especially through high contents of carotin, chlorophyll, iron and pure protein and vitamin C.” [7, 8] The paper affirms spinach’s iron content; it does not debunk it. Schuphan’s verified 1937 publications addressed celeriac and celery, not spinach iron overestimation. [8]

Bender named a real scientist and attributed to him a finding that scientist’s published work does not contain — pointing in the opposite direction. Whether Bender misread, misremembered, or was working from some intermediate account that cannot now be traced is not established by the record. It is attribution offered, and the attribution is backwards.

The second hop is T. J. Hamblin, in the British Medical Journal’s Christmas issue of 1981. Hamblin’s article was titled “Fake!” and addressed various medical myths; his passage on spinach reads, per Rekdal’s account: “German chemists reinvestigating the iron content of spinach had shown in the 1930s that the original workers had put the decimal point in the wrong place and made a tenfold overestimate of its value.” [1] The story is now fully formed — German chemists, the 1930s, a specific mechanism, a factor of ten. The claim has acquired the texture of a known historical event.

Hamblin did not provide a reference. Not a vague one, not a paywalled one. No citation appears in that passage. No names, no dates beyond “the 1930s,” nothing a reader could follow to verify. Where Hamblin got this information is not stated. Rekdal, who read the BMJ piece directly, confirms this: Hamblin “does not provide a reference to support his claim; nor does he give any names, dates, or other information that could help us verify” the error’s origins. [1]

The third hop is K. Sune Larsson, in the Journal of Internal Medicine in 1995. Larsson’s paper was titled “The dissemination of false data through inadequate citation.” His passage on the decimal error, per Rekdal, cited Hamblin’s 1981 BMJ piece as the source for the claim that a misplaced decimal point caused a tenfold overestimate. [1] Hamblin’s undocumented assertion had become a citable claim. The cycle from speculation to source had completed.

The investigators

In 2010, researcher Mike Sutton contacted Hamblin directly and asked him to name the source for his 1981 claim. Hamblin could not. Per Rekdal’s account, Hamblin acknowledged that he could not identify where the decimal error story had come from. [1]

Sutton also investigated why early iron figures for spinach may have run high. His account — published on his blog and read directly for this piece — attributes the elevated readings to contamination of nineteenth-century laboratory samples: charcoal used for heating and the vessels themselves introduced iron into samples being analyzed. Multiple scientists had identified the contamination problem by 1907; von Bunge noted it in Switzerland in 1892, and Sherman in the United States in 1907. The elevated readings trace to a systematic lab artifact, not to any decimal transposition. [6]

Rekdal’s 2014 paper in Social Studies of Science formalized the documentation of the chain. His method was the standard one: follow the citations, read the sources, note what each one says and what each one provides. He read Hamblin and Bender directly. His conclusion: no primary source for the decimal point error exists in accessible literature. [1]

The recursion

The Larsson case is worth dwelling on.

Larsson’s 1995 paper was explicitly about the spread of false data through inadequate citation practices. He was writing an intervention in citation hygiene. The paper’s subject is the very failure mode it exemplifies — and it appears to do so without Larsson noticing. Rekdal notes that Larsson made “several errors when reproducing Hamblin’s message” and was reproducing information from a secondary source without consulting primary documentation. [1]

Careful researchers cite without always verifying; the alternative — reading every primary source in every chain you use — is not how papers get written. What the Larsson case shows is the stickiness of the mechanism. A claim that has been published in the British Medical Journal, in an article, becomes the kind of thing a careful researcher might cite without checking, because it looks like it has already been checked. The fact that Hamblin provided no reference is easy to miss in a format where no reference would be unusual. The BMJ Christmas issue is not a peer-reviewed research article; “Fake!” is an essay. Readers extend credibility by venue.

By 1995, the decimal error story had been in print for over a decade, under respectable authorship. The conditions for its continuation were in place.

What the chain establishes, and what it does not

The absence of documentation is not proof that the decimal error never happened. Nineteenth-century food chemistry records are not comprehensively digitized; a German paper with a miscalculated iron table could exist in an archive that nobody has searched, or could have been published in a journal that did not survive into accessible collections. What the chain establishes is narrower and more useful: no record of this specific error survives in accessible literature, despite the fact that multiple researchers specifically looked for it.

Sutton’s contamination account is parsimonious. It explains elevated historical iron figures without requiring any discrete arithmetic error. [6]

What can be said: the story as typically told — a named error by a named analyst, corrected by named investigators — is not in the literature. The story as documented — a hedge, an attribution that pointed in the wrong direction, and an undocumented assertion that spread without a source — is.

How the chain persists

The structure of this particular chain is: hedged speculation, unsourced; an attribution offered, and backwards; unreferenced assertion in a widely-read venue; citation as established fact in a paper specifically about this failure mode.

That last step is what makes the spinach chain worth documenting beyond its irony. Larsson’s paper is an attempt to correct the record on spinach nutrition. It participates in the chain it was meant to interrupt. This is not unusual — corrective papers that cite the claims they are correcting inherit those claims into their own citation record, and future researchers may cite the corrective paper as documentation for the original claim, forgetting to read which way the correction ran. The chain does not require bad faith at any step. It requires normal academic reading practices applied to a claim that had no source to begin with.

Hamblin’s passage asserted the error as known history. Researchers who read “British Medical Journal, 1981” in a reference list see an authoritative venue and a date. The absence of a citation within Hamblin’s own text is only visible to readers who go back to that Christmas issue and look. Most don’t.

That, as best the accessible record can establish, is how spinach got its iron.

This piece cites Hamblin (1981), Bender (1972/1977), and Larsson (1995) through Rekdal (2014), which was read directly via PMC. Those three sources are paywalled or in archives not accessible to automated fetch. Sutton’s blog (2010) was read directly by the researcher; Dagg’s two posts (2015) were read directly by the researcher. Schuphan’s 1940 article is cited through Dagg (2015); the German-language text was not accessed directly. Claims attributed to Hamblin, Bender, and Larsson are attributed throughout as Rekdal’s account of them.