What the spectra actually constrain
The measurement is narrow and it is solid. Nine massive galaxies that had already stopped forming stars — seen as they were about seven billion years ago, at redshift 0.7 — were observed with the James Webb Space Telescope and its NIRSpec instrument under the IMFERNO programme and joined to deeper, bluer spectra from the LEGA-C survey on the Very Large Telescope, covering roughly 3700 to 10780 angstrom in the galaxies' own frame. The team fitted all of it at once with the full-spectrum code alf, letting the age, 18 elemental abundances and the two low-mass slopes of the initial mass function move freely. What comes out is a single ratio: the mass-to-light ratio with a free mass function, divided by the mass-to-light ratio with the Milky Way's. The light stays the same; the only thing that changes is the accounting of which stars emit it.[1]
Several of the nine sit well above a ratio of one, which is what a bottom-heavy mass function looks like from the outside: more small, faint stars than the Milky Way's recipe would lead you to expect. The excess moves with velocity dispersion and iron abundance, so the same trends already known in nearby massive elliptical galaxies are now pinned seven billion years further back. One galaxy, 1158527, sits at a ratio of 5.36 and follows none of those trends. Its axis ratio of 0.97 suggests a disc seen face-on, in which case its line-of-sight velocity dispersion of 133 km/s misses much of the gravitational well, and what stands there is an undercounted mass rather than overcounted stars.[1]
Where the factor of 3 to 4 comes from
None of the nine is one of JWST's "impossibly early" galaxies. Two of them are the oldest in the sample, with formation redshifts above 6, and on a star-formation timescale of a hundred to two hundred million years they would have stopped forming stars by redshift 5.5 — which makes them plausible descendants of that early population. The mass correction is carried backwards along that descent: if the descendants have bottom-heavy mass functions, the ancestors probably did too, and the masses reported for them rise by a factor of 3 to 4. That is an argument about lineage, and it inherits every uncertainty in the claim that these particular galaxies came from those particular ones. The prosaic alternative is simple: a galaxy's mass function can change between the epoch when it forms and the epoch when it is observed, and the descendant link is a statistical match more than a tracked history.[1]
There is one internal check, and it is worth naming precisely. Correcting the stellar masses by the measured ratio makes them rise towards the virial masses derived from the velocity dispersions and — except for 1158527 — they stop short of exceeding them. That is a feasibility test: a mass function that made the stars outweigh the whole gravitational well would be ruled out, and this one survives. It does not amount to independent confirmation. The authors note that the dispersions go uncorrected for slit losses in the NIRSpec micro-shutter array, and that rotation would make them a poor tracer of the gravitational well; both effects lower the virial masses and leave the agreement looking tighter than it is.[1]
The fixed slope that could reverse the sign
The method is sensitive to the ratio of low-mass to high-mass stars more than to the high-mass end on its own, so the slope above one solar mass is held fixed at the Milky Way's value throughout. The study says plainly what that costs: if the real slope up there is shallower, stellar masses at high redshift come down rather than up. Both can hold at once — an excess of small stars and a shortage of the largest ones — and that combination is exactly the concordance mass function van Dokkum and Conroy proposed, which reconciles high-redshift observations with galaxy formation models instead of straining them further. The paper's headline result and the assumption most likely to undo it sit inside the same fit.[1]
The same shape appeared in the previous column here. In MoM-BH*-1 the dense-gas model that explains an extreme Balmer break also bends the ruler used to weigh the black hole; here too the spectrum is the measurement and the mass is the model. This case is cleaner, because the assumption is written down and testable: extending the spectra to redder features more sensitive to the mass function would tighten the low-mass constraint, yet the stellar population models needed to fit those features cannot yet reproduce even nearby galaxies. Until that changes, a factor-of-4 correction to the early universe's mass budget rests on nine galaxies at redshift 0.7 and a slope nobody measured.[1], [2]