Revised Estimates for the Number of Human and Bacteria Cells in the Body

For fifty years, one number defined how we talked about the human body and its microbial passengers: ten bacterial cells for every one human cell. You've heard it in textbooks, in documentaries, and in the opening paragraphs of a thousand microbiome articles. It became a reflex. The problem is that it was never really measured. It came from a single back-of-the-envelope calculation made in the nineteen seventies — essentially one sentence of arithmetic — and then it just traveled, unexamined, for decades. Ron Sender, Shai Fuchs, and Ron Milo decided to actually check it. What they found wasn't that the number was slightly off; it was built on assumptions that were quietly, significantly wrong. The original estimate, traced in their PLoS Biology paper, worked like this: take the alimentary tract, assume it holds about one liter of material, assume a bacterial density of around ten to the eleventh bacteria per gram — roughly what you'd measure in stool — multiply the two, and you get ten to the fourteenth bacteria in the body. That's a hundred trillion. The ten to one ratio followed directly. But Sender and colleagues point out two problems immediately. First, treating the entire alimentary tract as though it has the bacterial density of stool is wrong by orders of magnitude. The stomach and upper small intestine are acidic and fast-moving; their bacterial concentrations are closer to ten thousand bacteria per milliliter, not a hundred billion. Second, the volume assumption of one liter was never defended. The colon — which is where the bacteria actually live — holds far less than that. So they went organ by organ: the stomach, small intestine, and colon. They collected measured bacterial concentrations for each compartment, multiplied by organ volumes, and compared. The result was unambiguous: the colon does almost all the work. Everything else contributes negligibly to the total bacterial count. The colon is the only compartment with both high bacterial density and meaningful volume, so that's where you have to look. Measuring bacterial concentration in the colon is done indirectly — you count bacteria in stool. Early studies used direct microscopy on diluted samples. Later work used fluorescent techniques: DAPI staining, which lights up nucleic acids in all cells, and fluorescent in situ hybridization, or FISH, which targets bacterial 16S ribosomal RNA specifically. Sender, Fuchs, and Milo compiled data from fourteen studies and converted everything to a common unit: bacteria per gram of wet stool. The geometric mean across all those studies came out to 0.92 times ten to the eleventh bacteria per gram, with a standard error of 19 percent. The median was nearly identical at 0.91 times ten to the eleventh. The fact that the geometric mean and median agree closely is reassuring — it means the central estimate isn't being pulled around by outliers. For colon volume, they leaned on MRI-based measurements. One height-normalized dataset gave about 480 milliliters of inner colon volume for a 1.70 meter reference man; another cohort suggested around 430 milliliters after accounting for the fraction occupied by stool. Weighing these together, the team used 0.4 liters as their best estimate of colon content, with a standard error of 17 percent. Multiply concentration by volume, and the total bacterial count for the reference seventy kilogram adult male comes to 3.8 times ten to the thirteenth. Call it 38 trillion bacteria, with a bacterial mass of roughly 0.2 kilograms — about 0.3 percent of body weight. Now for the other side of the ratio. Sender, Fuchs, and Milo built a census of human cells by type. They focused on the six types that together account for 97 percent of all human cells: red blood cells, glial cells, endothelial cells, dermal fibroblasts, platelets, and bone marrow cells. And here is where the accounting gets genuinely surprising. Red blood cells dominate the count completely. The average blood volume of 4.9 liters, multiplied by a mean red blood cell concentration of 5.0 times ten to the twelfth cells per liter, gives 2.5 times ten to the thirteenth red blood cells. That single cell type makes up about 84 percent of all human cells by number. Add platelets and the other hematopoietic — blood-forming — cells, and the entire hematopoietic lineage accounts for roughly 90 percent of the total human cell count. This is strange to sit with because red blood cells don't look like what most people picture when they think "human cell." They have no nucleus, no DNA. They are, as Sender and colleagues note, essentially bags of hemoglobin. But the team counts them as cells, and the numbers are what they are. What the census also revealed is how badly previous estimates had overstated several other cell types. Glial cells — the support cells of the brain — were previously quoted at 3 times ten to the twelfth, based on an assumed ten to one ratio of glia to neurons. The revised figure is 8.5 times ten to the tenth, a reduction of more than thirty-fold. Endothelial cells, which line blood vessels, dropped from 2.5 times ten to the twelfth down to 6 times ten to the eleventh. Dermal fibroblasts fell by a factor of about one hundred, once proper dermal thickness and layer-specific densities were applied. Combined, those three cell types were previously estimated at 7.5 times ten to the twelfth; the revised combined figure is 0.9 times ten to the twelfth. The team also cross-checked their census with a mass-based approach, using total body potassium measurements to partition how much of a person's weight is actually inside cells versus outside them. That sanity check confirmed a counterintuitive result: the cells that dominate by mass — fat cells and muscle cells — are not the same cells that dominate by number. Adipocytes and myocytes together account for about 75 percent of total cell mass but only around 0.2 percent of total cell count. The body, numerically, belongs to tiny red blood cells. Structurally and metabolically, it belongs to large, comparatively rare cells. Put the two sides together. Bacteria: 3.8 times ten to the thirteenth. Human cells: 3.0 times ten to the thirteenth. The ratio — bacteria divided by human cells, what the paper calls B over H — is approximately 1.3. Not ten. Not one hundred. One point three, with an uncertainty of plus or minus 25 percent and a population variation of 53 percent across standard adult males. That ratio isn't fixed. The authors identify colon volume, bacterial density, blood volume, and hematocrit as the four parameters that drive it. Women, at a reference weight of 63 kilograms, have a B to H ratio closer to 2.2. Infants, the elderly, and obese individuals each land at different values, ranging from about 1.4 to 2.3. The number moves with body composition. There's also a definitional question that cuts to the core of the finding. If you exclude red blood cells from the human cell count — which some would argue is reasonable, given that they have no nucleus — then you're comparing 3.8 times ten to the thirteenth bacteria against only 0.3 times ten to the thirteenth nucleated human cells. That ratio is back to ten to one. Sender and colleagues are transparent about this: they regard red blood cells as cells, but they acknowledge the debate. The answer you get depends on what you decide to count. And there's a timing dimension too. Because roughly a third of the colon's content is expelled in a single bowel movement, that one event can drop the bacterial count from 3.8 times ten to the thirteenth to around 2.5 times ten to the thirteenth — below the human cell total. For a brief window after defecation, bacteria are actually outnumbered. None of this diminishes the biological significance of the microbiome. The bacteria are there, they are metabolically active, and their influence on human physiology is real. But the specific framing — outnumbered ten to one by our microbial passengers — was always a rhetorical artifact, not a measured fact. The broader lesson from Sender, Fuchs, and Milo's work is methodological: Fermi estimates are useful, but foundational numbers need to be revisited as better data accumulate. The unglamorous work of checking a half-century-old one-sentence calculation turned out to matter quite a lot. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.