Where the idea came from
For most of the twentieth century the settled view was that the adult mammalian brain makes no new neurons. Joseph Altman reported labelled dividing cells that became neurons in adult rats in the 1960s and was largely ignored. Fernando Nottebohm's work on canaries in the 1980s, showing new neurons arriving in song-control regions as birds learned new songs each season, made the phenomenon hard to dismiss in at least one vertebrate. By the 1990s, rodent work had established two clear sites: the subgranular zone of the hippocampal dentate gyrus and the wall of the lateral ventricles, which supplies the olfactory bulb. In rodents these produce neurons in large numbers throughout life, and the rate responds to exercise, enrichment, stress and some drugs.
The question that matters for people is whether any of this transfers to humans, and the difficulty is that the methods that settle it in rodents — injecting a labelling compound and killing the animal weeks later — are not available.
The first human evidence
In 1998 Peter Eriksson, Fred Gage and colleagues, working between Sahlgrenska University Hospital in Gothenburg and the Salk Institute, published in Nature Medicine a study of post-mortem brain tissue from patients with cancer who had received bromodeoxyuridine (BrdU) to track tumour growth. BrdU is taken up by dividing cells and stays in their DNA. In the dentate gyrus, the group found BrdU-labelled cells that also carried neuronal markers. Five patients, but a direct observation.
The second landmark used a stranger clock. Above-ground nuclear testing between 1955 and 1963 raised atmospheric carbon-14, which then declined. Cells born in a given year carry that year's carbon-14 level in their DNA. In 2013 Kirsty Spalding, Jonas Frisén and colleagues at the Karolinska Institute measured carbon-14 in the genomic DNA of hippocampal neurons and, in Cell, estimated that roughly 700 new neurons are added per day in each adult human hippocampus, with the renewing population turning over at a modest annual rate that declines only slightly with age. The estimate depends on a model of which cells renew and how, so the exact figure is softer than it sounds, but the birth-dating itself is hard to argue with.
2018: two papers, opposite answers
Then the field split. In March 2018 Shawn Sorrells, Arturo Alvarez-Buylla and colleagues at the University of California San Francisco published in Nature an analysis of hippocampal samples from foetal stages to old age. Using antibodies against doublecortin (DCX) and PSA-NCAM — markers of young, migrating neurons — they found abundant young neurons in infants, few by mid-childhood, and none they could confidently identify in adults. Their conclusion was that human hippocampal neurogenesis drops to undetectable levels early in life.
A month later Maura Boldrini, J. John Mann and colleagues at Columbia University published in Cell Stem Cell a study of whole hippocampi from people aged 14 to 79 who had died suddenly, with short post-mortem intervals and prompt fixation. They found thousands of intermediate progenitors and immature neurons at every age, with the pool of quiescent stem cells and the extent of new blood vessel growth declining with age but neuron production itself holding up.
The two groups were looking for the same cells with similar antibodies. The obvious differences lay in the tissue: source (surgical resections and post-mortem donors of varying kinds versus rapidly processed sudden-death donors), post-mortem interval, fixation time and the counting method.
Why fixation turned out to matter
In 2019 Elena Moreno-Jiménez, María Llorens-Martín and colleagues at the Centro de Biología Molecular Severo Ochoa in Madrid published in Nature Medicine a systematic test of exactly those variables. They showed that the DCX signal degrades with prolonged fixation in paraformaldehyde and with long delays before fixation; under tightly controlled conditions they found thousands of DCX-positive cells in the dentate gyrus of neurologically healthy people up to their late eighties, with the number falling sharply in Alzheimer's disease. The implication is that the 2018 negative result may be a story about tissue handling rather than about biology — though the UCSF group has continued to argue that DCX alone can mislabel other cell types.
This is the part of the story worth carrying to other fields. An antibody stain is a proxy. When two careful groups get opposite results with the same proxy, the first suspect is the proxy's sensitivity to conditions neither group had fully controlled.
Sequencing instead of staining
The next wave sidestepped antibodies by reading RNA from individual nuclei. Results were again mixed. A 2022 study in Neuron from Nenad Sestan's group at Yale (Franjic and colleagues) found very few cells with a neuroblast profile in adult human hippocampus, in contrast to macaque and pig. A 2022 paper in Nature from Hongjun Song and Guo-li Ming's group at the University of Pennsylvania (Zhou and colleagues) reported immature neuron signatures in adult samples using a differently tuned analysis. In 2025 the Frisén laboratory at Karolinska published in Science a study that trained a classifier on progenitor cells from other species and children, applied it to adult human hippocampal nuclei, and reported dividing neural progenitors located by spatial transcriptomics in the dentate gyrus. It is a strong result that nonetheless depends on the classifier being right about what a progenitor looks like.
What the rate would mean if real
Suppose the Karolinska estimate is about right. A few hundred neurons a day is tiny against the many millions of granule cells already in the dentate gyrus, and far below rodent rates when adjusted for size. Computational models suggest even small numbers of highly excitable new cells could help the hippocampus keep similar memories distinct. That is a hypothesis about function, not a finding. The rodent literature linking exercise, antidepressants and enriched environments to more new neurons has not been shown to hold in people, and no product marketed as boosting neurogenesis has human evidence behind that specific claim. Questions about memory or mood belong with a GP, not with a mechanism that is still being argued over.
Where the balance sits
On present evidence, adult human hippocampal neurogenesis probably exists at a low level and declines with age and with Alzheimer's disease. The disagreement about detection is largely a disagreement about tissue handling and marker choice, which is slowly being resolved as methods converge. What those neurons do, if anything, remains open.