Sorted by lag, shortest first. Bars are scaled to the longest item, 169 days. Eight items at 1, 5, 9, 34, 82, 116, 153 and 169 days, median 58 — the widest spread this page has recorded, and the first edition with no clump at either end. Four verified, three plausible, one contested. Five were read end to end as versions of record; three rest on complete publisher-deposited abstracts and say so. One of the five read in full is graded contested rather than verified, which is the distinction this key is for: reading a paper tells you what it says, not whether what it says is carried by what it did.
Plausible
Published online 2026-10-26
Journal PNAS
Read the deposited abstract, in full
Lag 1 d
Almost everything known about sleep is known about vertebrates, and a spider was asked properly
The abstract’s first move is the one worth keeping: “Sleep is a fundamental behavioral state found across the animal kingdom, yet most of what we know about sleep is heavily vertebrate-biased. Studies are scarce or nonexistent in all but a few invertebrate lineages.” A 2022 report of REM-like twitching in a jumping spider had got a great deal of attention and, as this paper says plainly, “experimental evidence that these behavioral indicators truly represent sleep was lacking in spiders.” So they went and got it, in Portia fimbriata: “diurnal activity, a free-running circadian rhythm, robust sleep architecture, regularly occurring REM-like behaviors, elevated arousal thresholds during those REM-like states, as well as rebound effects following sleep disruption.” Those last two are the load-bearing ones — a raised arousal threshold and a rebound after deprivation are the behavioural criteria that separate sleep from sitting still. The conclusion is stated at exactly the width the evidence supports: “the behavior we observe in spiders is behavioral sleep, providing evidence for an REM-like sleep substate in a terrestrial invertebrate.”
The star stuff frame
The interesting thing is not that a spider sleeps; it is that nobody had checked, and that this is normal. Sleep research built its account of a universal behaviour out of a handful of lineages and then described the result as what sleep is. Every so often somebody tests a creature outside the sample and the definition has to widen — not because the creature is exceptional but because the sample never was representative. A norm assembled from whoever was convenient to measure is a norm about convenience, and this site has an entire shelf arguing that point about human beings. Here it arrives as an ordinary methodological observation in the first two sentences of a PNAS abstract, about spiders, with nobody making anything of it. That is the version we trust most.
What it says, and what the headline asked instead
The wire headline is “Do spiders dream?” and the paper does not contain the question. What it reports is behavioural sleep and an REM-like substate — both hedges are the authors’, in the abstract, and both are doing real work. No neural recording is described in the abstract; the criteria are behavioural throughout, which is why the word is behavioral sleep and not sleep flat. Dreaming is a claim about experience and nothing here reaches it. We did not read the paper. www.pnas.org answered HTTP 403 to the article URL and there is no PubMed Central copy, so the sample sizes, the arousal-threshold protocol and the rebound statistics are all unread, and the species count is one. The paper is CC BY, which bought nothing this week.
Where it goes
A Field Guide to Resting is the obvious door and the fit is sharper than it looks: that guide’s evidence is a penguin huddle, a fur seal, a swift, a bat —
vertebrates almost throughout, with
Cassiopea the one famous invertebrate. The abstract’s own opening is therefore a comment on our guide as much as on the field.
spider returns nothing in that file.
A Field Guide to Nervous Systems is the second door and already carries
Portia. See the
seedbed.
Geiger, N., Capelle, M. Q., Rosentreter, P., Herberstein, M. E., Bahl, A., De Agrò, M., Klein, B. A. & Rößler, D. C. “Behavioral evidence for two-stage sleep in spiders.” Proceedings of the National Academy of Sciences 123(45), online 26 October 2026. DOI 10.1073/pnas.2618030123, CC BY, funded by the Deutsche Forschungsgemeinschaft. Read: the complete abstract as deposited by the publisher at Crossref; every quotation above is verbatim from it. Not read: the paper. www.pnas.org answered 403 and no PMCID exists. Wire: ScienceDaily, 27 October 2026. Lead image: Daniela C. Rößler — the paper’s own last author.
Plausible
Record created 2026-10-19
Journal Current Biology
Read the deposited abstract, in full
Lag 5 d
No wheels, no axle, and a head that stays still anyway
Rolling is nearly absent from animals and the abstract says why in one clause: “Although rolling can confer high speed, it poses a fundamental mechanical challenge, as body rotation destabilizes the head and disrupts sensory-guided adjustment, particularly at high speeds.” A wheel works because an axle separates the spinning part from the rider; an animal has no axle. Dwarf reed snakes of the genus Pseudorabdion roll anyway, to escape, and filming them from several angles shows how: they “stabilize their head during dynamic rotational motion of the posterior body by using precisely coordinated body movements.” The authors’ own summary of what that means is the sentence to keep: it “illustrat[es] how behavioral innovation can circumvent deep morphological constraints and expand the capacities of animals.”
The star stuff frame
The body was never redesigned. The behaviour went around it. There is no rolling organ here, no new joint, no structure that evolved for the purpose — there is a sequence of movements, learned or innate, that makes an impossible thing possible using the parts already present. That is a different kind of answer from the one the question invites, which is always “what structure lets it do that?” This one is: nothing does, and it does it anyway, by timing. The honest reading stops there and does not go on to a human. A capacity can live in coordination rather than in anatomy, and if you only measure the anatomy you will report that the capacity is absent.
The numbers are the release’s, and the dates do not agree with each other
“Four times faster” is not in the abstract. The figures — peak forward velocities of 1.72 to 3.69 metres per second, roughly four times the maximum under lateral undulation, down a racetrack 20 centimetres wide — come from the ScienceDaily release and are marked here as the release’s, not the paper’s. The abstract gives no velocity at all. This is a Current Biology Correspondence, two pages in the magazine section, not a research article, and the abstract is correspondingly short. And the record carries no online-first date. Crossref dates it to the October issue and created the record on 19 October; Semantic Scholar says 1 October. We have used 19 October, which gives a lag of 5 days; from the other date it would be 23. We cannot reconcile them and are not going to pick the flattering one silently. www.cell.com answered HTTP 403 and there is no PubMed Central copy, so the kinematics, the number of individuals filmed and the head-stabilisation analysis are all unread.
Where it goes
Pseudorabdion and
rolling locomotion both return nothing across this site.
More Than Human is the register, and the unspent shape is the one above:
a capacity that is behavioural rather than structural, which is invisible to any survey that asks what an animal is built from.
Nothing here is buildable on a two-page correspondence read at abstract length, and somebody would need the paper first.
Kojima, Y., Fukuyama, I., Hossman, M. Y., Kaliang, C. H. & Nishikawa, K. “Locomotor kinematics and head stabilization in rolling snakes.” Current Biology 36(20), R1061–R1063. DOI 10.1016/j.cub.2026.08.057. Read: the complete abstract; every quotation above is verbatim from it. Not read: the correspondence. www.cell.com answered 403 and no PMCID exists. Dates: no published-online date in the record; Crossref created it 19 October 2026 and assigns the October issue, Semantic Scholar gives 1 October. Wire: ScienceDaily, 24 October 2026. Lead image: Yosuke Kojima — the paper’s own first author.
Verified
Published online 2026-10-21
Journal Science Advances
Read the paper, in full
Lag 9 d
They came through it by never having left the climate they already fit
Crocodile-line archosaurs have been through 250 million years, two mass extinctions and several hyperthermals, and the question is whether their famous attachment to warm wet places is an ancient constraint or a recent habit. This paper builds a time-scaled phylogeny of 402 tips, calibrates niche models on living crocodylians and projects them back about 245 million years, to the Anisian. The answer is conservatism: temperature and precipitation show strong phylogenetic structure, best fitted by a bounded Ornstein–Uhlenbeck process, and the range shifts through time are the animals following a stable optimum rather than the optimum moving. The end-Cretaceous line is the one the coverage wants, and the paper gives it carefully: “Unlike non-avian dinosaurs, low-latitude suitability persisted across all but the most extreme end-Cretaceous scenarios.” And the closing sentence refuses the tidy version of its own result: “Climatic niche conservatism is therefore ancient in crocodile-line archosaurs, but not completely static.” Two regime shifts are identified and they are decoupled — thermal near the Cenomanian/Turonian hyperthermal (~93.9 Ma), precipitation near the Eocene/Oligocene aridification (~33.9 Ma), the latter confined to Australasian mekosuchines.
The star stuff frame
Survival here is not flexibility, and it is not toughness either. The lineage did not adapt its way through the asteroid; it stayed inside a narrow band of conditions it had always needed, and that band did not disappear everywhere at once.
A specialist survived because a refuge persisted — which makes the survival a fact about the world as much as about the animal. The generalist story is the one that gets told, because it reads as a virtue; this is the other one, and it is less flattering and more useful.
Nothing about a crocodylian got better at tolerating anything. What happened is that somewhere stayed liveable, and the lineage was already there. The argument sits one step away from an obvious human application and we are deliberately not taking that step — see the
refusals.
What the paper itself declines to claim
Climate is explicitly not the whole answer. The Discussion names the rest and we are quoting it rather than letting the headline stand: “Non-abiotic factors—including smaller body size, semi-aquatic lifestyle providing access to detrital food webs” also mattered, which means “why did crocodiles survive?” has at least four answers and this paper supplies one of them. The limitations are stated in the paper’s own words: “The main caveats involve coarse temporal bins, non-analog climates, and preservational biases.” Niche models calibrated on 25 or 26 living species and projected across 245 million years are an extrapolation far beyond their training range, which the authors say. One small internal disagreement, recorded because we found it rather than because it matters: the text refers to 26 currently recognised living crocodylian species while the table and methods list 25, and the abstract’s “>400 fossil and extant species” is 402 tips pruned from an original 534. We did not obtain the supplementary materials or the occurrence data.
Where it goes
niche conservatism returns nothing across this site. Crocodilians appear in
The Shell Was Your Calcium (No. 57), for the caruncle they share with turtles and echidnas, and nowhere near this argument. The unspent claim is the uncomfortable one:
persistence can be a property of the environment rather than of the organism, which is the half of any survival story that a survival story never tells. It pairs with the ant paper two items down — same catastrophe, opposite strategy. See the
seedbed.
Chiarenza, A. A., Howard, A. F. C., Varnham, G. L., Farnsworth, A., Valdes, P. J. & Mannion, P. D. “Climatic niche conservatism shapes the macroevolution, extinction, and survivorship of crocodile-line archosaurs.” Science Advances 12(43), eaeg2217. DOI 10.1126/sciadv.aeg2217. Read: the version of record in full at PubMed Central (PMC13685984), through a page-fetching tool — a plain request returned a reCAPTCHA challenge, as described above. Every quotation is verbatim from it. www.science.org answered 403. Not read: the supplementary materials and the deposited occurrence data. Dates: Crossref carries a print date of 23 October 2026 and no online date; Semantic Scholar gives 21 October, which is the earlier and is what the lag uses. Wire: ScienceDaily, 30 October 2026. Lead image: Davide Bonadonna — a palaeoartist’s reconstruction, credited by name and not presented as a photograph.
Contested
Published online 2026-09-23
Journal Science Advances
Read the paper, in full
Lag 34 d
The same asteroid, the opposite strategy — and a verb the design cannot carry
Transposable elements are the mobile, repetitive, long-dismissed fraction of a genome. This paper assembles an unusually broad comparison — “163 genomes spanning 12 of 16 extant ant subfamilies” — and looks at what those elements did over 100 million years. It reports “convergent TE acquisition bursts in ancestors of the most speciose ant clades, preceding rapid diversification” and places that diversification “following the Cretaceous-Paleogene extinction event approximately 66 million years ago.” It also finds that “TEs associate with expansions of odorant receptors and other key gene families,” which in an animal whose social life runs on smell is the specific and interesting part, and that ant genomes are compartmentalised into faster-evolving TE-rich and more conserved TE-poor regions. Set beside the crocodile paper above it is the same catastrophe answered two ways: one lineage came through by staying exactly where it fit, the other radiated into what was left.
The star stuff frame
The junk was the raw material, and that is the oldest correction in genomics still being made. Transposable elements were selfish DNA, parasites, the stuff left over after the genes — and here they are sitting in the ancestry of exactly the clades that went on to become most of the ants. A fraction of a genome gets classified by what it looked like to the people measuring it, and the classification outlives the reason for it by decades. This site already owns the adjacent fact: genome size says nothing whatever about complexity. What it does not own is this one — that the part counted as surplus is where variation to work with came from.
Why this is contested, which is a judgement about a verb
The abstract says the findings “demonstrate the broad-scale impact of TE-dependent genome evolution” and the study is correlational. We read the paper precisely to settle this, and the Discussion settles it against the abstract: “All evolutionary genomic studies are inherently limited in their ability to establish causal relationships,” and the K–Pg event itself “could have independently promoted both lineage diversification and TE proliferation.” That second sentence is a confounder the authors name and cannot rule out — the same extinction could have caused both things this paper connects. Their argument against reverse causation is good: the TE accumulation precedes the diversification in time. Their argument for causation is temporal precedence plus plausibility, which is not the same as demonstration. So the grade is about the word, not the work. The paper is careful in its Discussion and stronger than its design in its abstract and in the phrase “likely provided the raw genetic material facilitating diversification” — and the wire took the abstract. The authors also note, in their own words, that “we cannot exclude the possibility that extinct taxa exhibited different patterns of TE activity,” and 12 of 16 subfamilies is 12 of 16.
Where it goes
The Onion Has More Than You (No. 107) is the only page here that mentions transposable elements, and it mentions them in service of a different claim — that genome size does not rank anything.
The unspent argument is the catalysis one, and it needs the hedge in the previous paragraph attached to it from the first draft, not added in a colophon.
A zine built on this that used the word drive
would be making exactly the move we have graded this paper down for.
Schrader, L., Rinke, J. L., Errbii, M., Teresi, S. J., van den Bos, E., Xiong, Z., Vizueta, J., Boomsma, J. J., Gadau, J. & Zhang, G. “Transposable elements as evolutionary catalysts of ant macrodiversity.” Science Advances 12(39), eaee0306. DOI 10.1126/sciadv.aee0306. Read: the version of record in full at PubMed Central (PMC13600161); every quotation is verbatim from it. www.science.org answered 403. Not read: the supplementary materials and the genome assemblies. Wire: ScienceDaily, 27 October 2026. Lead image: Shutterstock.
Verified
Published online 2026-08-07
Journal Nature Photonics
Read the paper, in full
Lag 82 d
Thirty years of treating one electron as the whole story, and two of them turn up together
High-harmonic generation is how attosecond light pulses are made: drive a gas with an intense laser, an electron is torn loose, driven, and slammed back into its parent ion, emitting a high-energy photon. The whole field has been built on a single-active-electron picture with a hard ceiling — “For over three decades, high-harmonic upconversion has been framed within the confines of a single-active-electron” interaction with a well-defined cutoff. This experiment, in helium driven at 400 nm in a gas-filled hollow-core waveguide, finds light past the ceiling: “We observe a weak secondary plateau that extends the conventional cutoff beyond 120 eV up to the water window at 280 eV.” The new scaling is “up to 5.5 times the ponderomotive energy of the rescattering electrons,” against “the conventional scaling factor of 3.2.” Argon, run for comparison, shows no extended plateau.
The star stuff frame
The limit was a limit of the model, and the model was a simplification everybody knew was a simplification. Treating one electron as if it were alone is a choice made because the alternative is intractable, not because atoms contain one electron; three decades later the ceiling that choice implies has been read as a property of nature. The correction is not that the old physics was wrong — it is that a boundary drawn by what could be calculated got mistaken for a boundary in the world. The water window at 280 eV is the part that matters practically: it is the band where carbon absorbs and water does not, which is what you want for imaging something alive. Correlation between two electrons, which the single-electron picture exists to ignore, is what reaches it.
The observation is solid; the mechanism is the authors’ reading
Double-electron recombination is an interpretation, and the paper says so. It calls that account “the most plausible explanation for the extended high-harmonic cutoff” and states that confirming it requires further quantum simulations — so the measured plateau is the result and the two-electron story is the leading candidate for it. The signal is faint and the paper says that too: “the extended X-ray emission remains intrinsically weak due to low densities of correlated electron pairs.” “Scientists just broke a textbook rule of physics” is the wire’s sentence and no rule broke; a limit derived from an approximation was exceeded by an effect the approximation discards, which is a different and more ordinary thing. We did not reproduce any analysis and the raw spectra were not obtained.
Where it goes
high-harmonic returns nothing across this site. The nearest neighbour is
Beyond the Optical Image, which is about what an instrument can and cannot show; this is the same argument one level down, at the point where
the instrument’s limit turns out to be the calculation’s limit. Unspent and difficult: it would need a reader to care about ponderomotive energy, and the honest version cannot skip it.
Wang, S., Yan, J., de las Heras, A., Song, S., Prodanov, A., Wu, Z., Plaja, L., Popmintchev, D. & Popmintchev, T. “Correlated electrons extend X-ray high-harmonic generation beyond the single-electron limit.” Nature Photonics 20(9), 1092–1098, online 7 August 2026. DOI 10.1038/s41566-026-01976-2, CC BY. Read: the version of record in full at PubMed Central (PMC13533847); every quotation is verbatim from it. www.nature.com answered the article path with 303 to an authentication endpoint. Not read: the supplementary information and the raw spectra. Wire: ScienceDaily, 28 October 2026. Lead image: credited AI/ScienceDaily.com — the one machine-made lead image among the eight filed here.
Verified
Published online 2026-07-03
Journal Nature Immunology
Read the paper, in full
Lag 116 d
A cell can be told what to become by how stiff the place is
A T cell has to move through tissues that differ enormously in how hard they are to push through, and this paper asks whether the cell notices. It does: activated T cells “sensed mechanical cues and responded with changes in cell morphology, nuclear envelope composition” and the initiation of DNA repair, protecting their chromosomes from force-mediated damage. The stiffness range is real tissue, measured: “Liver, lung, kidney and salivary gland were ~10-fold stiffer (2.2 kPa) than LN and spleen (0.2 and 0.4 kPa, respectively).” And the force does not merely stress the cell — it reprograms it, “modulating the expression of the tissue-resident memory T cell-associated transcription factors Klf2, Runx3 and Hic1.” Cells preconditioned on stiff substrates and transferred into mice carried the change with them. The conclusion, in the authors’ words: mechanosensing “impacted T cell fate, promoting a transcriptional program associated with tissue residency.”
The star stuff frame
This site has one founding fact about force and tissue — bone is piezoelectric, and a skeleton turns pressure into charge. Here is a second, in a different cell and by a different route: the mechanical character of a place is an input to what a cell becomes, not just something it has to put up with.
Environment is not the backdrop to the biology; it is one of the instructions. That sentence is as far as the finding goes and it is already enough — a cell in a stiff organ ends up a different kind of cell from its sibling in a soft one, and nobody had to tell it which organ it was in.
The temptation is to carry this straight across to rooms and people and we are not going to; see the
refusals, and note that the paper itself cannot yet say what does the sensing.
The mechanosensor is unidentified, and the paper says so in those words
“While the mechanosensor(s) responsible remains to be determined” — the paper’s own sentence, and the one that should stop anybody reaching for a tidy molecular story. PIEZO1, the obvious candidate and the channel a reader of this site would expect, is not named anywhere in the text we read. What the authors did instead is rule things out: integrins via Talin, YAP/TAZ, cPLA2, myosin II and lamin A/C are each tested and each found not to be the mediator. Mouse first, human second, and only the mouse work is in a living animal. Most experiments use mouse CD8+ T cells; human CD8+ T cells from healthy donors were used in vitro only; the in vivo component is intradermal transfer into recipient mice with analysis in the draining lymph node at day 7. The stiffness numbers for the gels are relative, not absolute — one collagen preparation is reported as “~3-fold stiffer” than another by nanoindentation, and the kPa figures quoted above are for tissues. The authors list reversibility as open: whether these force-instructed states can be undone “will be important to address.” We did not obtain the supplementary data or the sequencing.
Where it goes
Bone Song (No. 1) is the house piece on force becoming signal in a tissue, and it is about bone.
mechanosensing and
stiffness are effectively absent from the rest of the site.
The unspent claim is the general one — that mechanical properties of a place are read by cells as instructions — with bone as the famous instance and this as the second, in an immune cell, in soft tissue.
It is also the claim most likely to be ruined by a metaphor, which is why the seed below carries its refusal before its argument.
Postat, J., Merino, M., Mingarelli, A. R., Cerf, A., Bhagrath, A., Patel, D., Shen, C., Brodbeck, J., Tirgar, P., Rogers, D., Blanc, J., Jeyakumar, T., Schneider, C., Coley, S., Giannetti, N., DePauw, T. A., Textor, J., Ehrlicher, A., Jameson, S. C., Sharif-Naeini, R., Sharma, A. & Mandl, J. N. “Mechanosensing by T cells promotes a tissue-resident memory transcriptional program.” Nature Immunology 27(8), 1708, online 3 July 2026. DOI 10.1038/s41590-026-02581-9, CC BY-NC-ND. Read: the version of record in full at PubMed Central (PMC13414542); every quotation is verbatim from it. www.nature.com answered the article path with 303 to an authentication endpoint. Not read: the supplementary information and the deposited sequencing data. Wire: ScienceDaily, 27 October 2026. Lead image: Shutterstock.
Plausible
Published online 2026-05-27
Journal Royal Society Open Science
Read the deposited abstract, in full
Lag 153 d
One of Europe’s most studied amphibians, and nobody had put a UV lamp on it
The fire salamander is about as closely watched as a European amphibian gets. Under ultraviolet it turns out to glow: “the first evidence of biofluorescence in the fire salamander (Salamandra salamandra), one of Europe’s most studied amphibians.” The fluorescence is “mainly on the ventral and lateral surfaces and in glandular secretions,” and histology matches the emission to the epidermis and the granular glands, with the speckled external pattern corresponding to glandular ducts. Fluorescence lifetime imaging then finds the thing the authors clearly did not expect: the glandular fluorophores differ from those in the skin and are “also ubiquitous in blood components, suggesting that fluorescent compounds may be transported via the circulatory system, a feature previously known only in tree frogs.” A second, yellow fluorescence sits in the dermal chromatophore layer and is excited by UV alone. And the secretions keep glowing after the animal has gone: fluorescence “persisted on substrates for over 24 h.”
The star stuff frame
It was always doing this. The light we look with could not see it. A trait on one of the most-examined amphibians in Europe stayed unrecorded until somebody changed the illumination — not because it was rare, hidden or subtle, but because the instrument defined what counted as the animal’s appearance. This is the umwelt argument without any need to argue it: our visual range is a window, the window is narrow, and an entire channel of an animal’s surface sat outside it. The trail of glowing secretion lasting a day on a substrate is the part that makes the point concrete — there may have been a visible signal in these woods the whole time, in a band nobody was looking in.
Communication is a suggestion, and the authors mark it as one
The abstract’s verb is “suggesting a communicative function,” and the evidence for it is that the trait “fulfilled several criteria for ecological relevance” — not that any salamander was observed responding to it. No behavioural experiment is described in the abstract. A trait being conspicuous under UV does not establish that anything uses it, and biofluorescence in amphibians has a literature full of exactly that gap. We did not read the paper. royalsocietypublishing.org is refused at the network layer and no PubMed Central copy exists, so the histology, the lifetime imaging and the criteria the authors applied are all unread; the paper is CC BY. And the item is 153 days old, carried this week in the present tense, which is the fourth consecutive edition to file or hold something of that shape.
Where it goes
Salamanders are already here:
Nobody in This Pond Is Late (No. 91) uses the tiger salamander complex for paedomorphosis and development, a wholly different claim about a different species.
biofluorescence returns nothing. The unspent argument is
the instrument deciding what the animal is, which this site has argued about images and never about an animal’s own surface.
Not buildable on an abstract: the ecological-relevance criteria are the whole question and we have not seen them.
Burriel-Carranza, B., Brunetti, A. E., Skamnelou, M., Escudero, J., Estarellas, M., Tulloch, S., Riaño, G., Rivera, X., Piulachs, M.-D., Engl, T., Weiss, B., Kaltenpoth, M. & Carranza, S. “Glandular biofluorescence in fire salamanders (Salamandra salamandra): first evidence and ecological implications.” Royal Society Open Science 13(5), online 27 May 2026. DOI 10.1098/rsos.251991, CC BY. Read: the complete abstract as deposited by the publisher at Crossref; every quotation is verbatim from it. Not read: the paper. royalsocietypublishing.org is refused at the network layer and no PMCID exists. Wire: ScienceDaily, 27 October 2026. Lead image: Bernat Burriel-Carranza, Museu de Ciències Naturals de Barcelona — the paper’s own first author.
Verified
Published online 2026-05-08
Journal Science Advances
Read the paper, in full
Lag 169 d
Seventy years of measuring meaning along three axes, and one of them was the wrong axis
Since work in the middle of the twentieth century, the essence of a word’s meaning has been taken to lie on three dimensions: valence, arousal and dominance. That scheme is not a historical curiosity — it is, as the paper says, the basis of “sentiment analysis across many fields,” which is to say it is what software means when it says a sentence is positive. This paper re-examines the English lexicon and then English corpora and arrives somewhere else. The best description is a five-part circumplex, goodness–power–aggression–danger–structure; and the sharper finding is the floor underneath it: “the power-danger-structure framework is the minimal framework that represents essential meaning.” Arousal and dominance, the two that carry most of the weight in a sentiment pipeline, are not in the minimal set. Along the way the corpora say something about their speakers: “large-scale English language corpora reveal a systematic bias toward safe, low-danger words.” Seven corpora are used, from Jane Austen’s six novels and the Sherlock Holmes canon to twenty years of The New York Times, a Wikipedia snapshot, radio transcripts and about a tenth of English Twitter in 2020; the lexicon is NRC VAD, “more than 20,000 English words and phrases,” 20,006 entries in the matrix.
The star stuff frame
An instrument got adopted because it was first, and then everything measured with it inherited its shape. Three axes were proposed, the proposal was good, the axes became the cornerstone — and for decades the question “what are the dimensions of meaning?” was answered by what the standard lexicon happened to score. That is not a scandal and nobody faked anything; it is what happens when a measurement framework is useful enough to stop being examined. The second finding is the one to keep for its own sake: a huge sample of what people actually write leans toward the safe end of danger. What we usually say is quieter than what we can say, which is a fact about the corpus rather than about any person in it, and is exactly the sort of claim that would be ruined by being applied to one.
The lag is wrong by a factor of eleven, and the paper itself names the limit
This was public on arXiv on 13 October 2021, with an abstract identical sentence for sentence to the published one. The journal version appeared 8 May 2026 and the wire reached it 24 October 2026 — so the bar above reads 169 days and the finding had been readable for 1,837. Our lag axis measures journal publication because that is the date records carry, and here it is measuring the wrong event. The paper’s own caveat is the right one to quote: “As any lexicon reflects only the possible but not the used language” — a scored word list is a map of what is sayable, not of what is said, which is precisely why the corpus half of the work exists. The safety bias is reported as a direction, not a magnitude: the text we read gives lexicon medians of −0.019 and −0.038 for the NRC VAD lexicon, and for the corpora says only that “the median danger score is well below 0.” English only, and “70 years” is the wire’s round number; the paper says “the middle of the 20th century” and cites Osgood and colleagues as “published in the 1950s.” Three versions of the preprint exist and we read the first alongside the published text.
Where it goes
valence and
arousal appear across this site only in their ordinary senses; nothing here discusses sentiment analysis or the Osgood framework.
The unspent argument belongs in How We Got Here, and the joints are already visible — a mid-century proposal, an adoption, a lexicon, an industry, and a re-examination seventy years later finding two of three axes dispensable. See the
seedbed.
Dodds, P. S., Alshaabi, T., Fudolig, M. I., Witte Zimmerman, J., Lovato, J., Beaulieu, S., Minot, J. R., Arnold, M. V., Reagan, A. J. & Danforth, C. M. “Ousiometrics: The essence of meaning aligns with a power-danger-structure framework instead of valence-arousal-dominance.” Science Advances 12(19), eadr4039, 8 May 2026. DOI 10.1126/sciadv.adr4039. Read: the version of record in full at PubMed Central (PMC13148342), and the preprint at arXiv:2110.06847, submitted 13 October 2021. Every quotation is verbatim from the version of record. www.science.org answered 403. Not read: the supplementary materials and the ousiometer prototype. Wire: ScienceDaily, 24 October 2026. Lead image: Shutterstock.