Each dot is where a tradition begins, sized by living adherents, and the curves between dots are descent, including syncretism where a tradition has more than one parent. Time runs left to right on a compressed scale, so the last three centuries get as much room as the previous ninety thousand years. Click any dot for its full vector, lineage and adherent history.
Click a dot and the rest of the chart drops away, leaving its family two generations out: direct parents and children in bold lines, then their relatives (grandparents, grandchildren, siblings) in thinner faded lines. Click it again to bring everything back. A line also extends showing how long it lasted: to the present edge if still practised, which is most of them; fading out if a descendant carried it on under another name, like Nicene Christianity into Catholicism and Orthodoxy; ending in a hard stop if it died out, like Manichaeism or the Aztec religion.
All 197 traditions are always drawn, in both modes, so nothing shifts or appears when you click. Fit to window scales the page to fit and labels only the traditions that ever reached about a million adherents; the smaller ones stay on the chart as faint dots and pick up their names when you select, search for, or click something they are related to. Turn it off to label all 197. Either way you can drag the canvas to pan, and the panel is capped to the window height so both scrollbars stay reachable.
Every tradition in the database, 197 of them. Click one for what it is, who founded it, and its core beliefs about the universe, a higher power and the soul. Click a belief for how it fares: what the evidence says, the best case for and against, how many unrelated lineages reached the same idea, and who still holds it. The same scoring as the Belief scores tab, at the weights set there.
What is left of the beliefs about a higher power once two rules are applied. First, drop anything scoring below the threshold. Second, where two ideas cannot both be true, keep only the higher-scoring one. The result is a set that is internally consistent: every idea in it can be held alongside every other, and nothing in it is contradicted by anything else in it.
What survives
Ruled out by something that outscored it
Below the threshold
Answer as much or as little as you like, then press Build it. What comes back is a single statement of the higher power you have described, assembled from the beliefs that hold up best. Every line comes from a real tradition, carries its rating for evidence and for argument, and cannot contradict any other line: where two things you asked for cannot both be true, the better-supported one stays and the other is shown underneath with what displaced it. Leave a question blank if you have no view on it.
Anything else you would want it to be
Matched on the words you use against what each tradition actually claims. No text is sent anywhere; this runs in your browser.
Read this before trusting any number here. Only the 1900 and 2025 columns rest on anything like enumeration. Everything earlier is regional population estimates crossed with the tradition understood to have dominated that region, and the pre-500 CE figures are little better than informed guesses. One significant figure is the real precision. The shape of the curves is the finding; no individual number should be quoted. Allocation is exclusive, so schools counted inside a larger body (Ash'arism within Sunni Islam, Advaita within Hindu sects, most Pentecostals within Catholic and Anglican totals) are marked overlay and excluded from the stack to avoid double counting.
Family rollup
Headcount is not probability. These are present-day figures: how many living people hold each position, and how many living traditions teach it. That is a measurement of belief, not of support. For exactly these propositions there is good reason to think human belief is produced by cognitive defaults rather than by tracking reality, so a 73% column is not a 73% credence. What should move a credence is on the Evidence page. The gap between the two is the thing worth looking at.
The third leg of the framework: for every position on every axis, how much does current science and serious theoretical work support it being true? Three levels only. Predicted: what you would expect if the standard scientific account is roughly right. Argued: serious scientists or philosophers make the case on mathematical or conceptual grounds; live and respectable, and not evidence. Unknown: nothing in current science speaks to it either way.
Deeper costs, like a position needing a mechanism nobody has, are deliberately NOT a fourth level: those costs are already charged on the Evidence page, and charging them twice would double-count. Where they exist, the reason line says so. Each row also shows how many living people hold that position, so the belief-support gap is visible per row.
Only claims that are, or once were, empirically checkable appear here. The central metaphysical claims of nearly every tradition are unfalsifiable and absent by design. That matters for reading the scores: low damage usually means low exposure, not vindication. Deism takes almost no hits because it asserts almost nothing checkable.
Scorecard by tradition
Checkable claims is everything the tradition ever asserted that could be tested. Of those, Now false is established wrong, and splits into dropped (taught then abandoned) and still taught. Unresolved is genuinely open. Supported is what survived testing. Click a column to sort, or click a tradition to read its claims.
How widely each claim is still taught
Solid bars are what is still taught; faint bars behind them are what was once taught, as a share of living humanity in traditions that held the claim or descend from one that did. The gap is retreat.
Every claim, and what happened when it was tested
One row per belief system, and every claim it has ever made, sorted into five piles. Got there first: asserted before anyone could check, and it held. Wrong: evidence establishes the opposite, or the effect was looked for and was not there. Still argued: decidable in principle, and specialists genuinely disagree. Untested: checkable in principle, but nobody has ever checked it. Untestable: no observation could count for or against it, in either direction. Click any row to read all of it.
Scope: core beliefs only, by default. A claim counts only if its truth bears directly on the nature of the universe, a higher power, or the soul. Mundane historical events (a tunnel, a census, a founder existing) test whether a tradition kept accurate chronicles; practice claims with mundane mechanisms (meditation, diet) test practitioners. Both are excluded. The line that matters: prayer stays in — it tests whether the power acts; meditation does not — it tests the human. Toggle "Core beliefs only" off to see the full catalogue.
A high percentage usually means low exposure, not insight. The rate is computed over tested claims only, so a tradition that asserted three checkable things and got them right scores 100% while one that asserted eighty scores far lower. Rates over fewer than eight tested claims are greyed and annotated with their denominator for exactly this reason. Read the counts before the percentage.
Two disciplines carry over from the earlier version. A confirmation only counts as a prediction if it came before the proof and was a real bet: right answers copied after the fact, and claims practitioners could verify by ordinary observation, sit in a footnote group inside each row, counted nowhere. And the higher-power filter is on by default: belief systems that positively deny anything above the human are excluded until you toggle them back in.
A single "at odds with science" score would collapse three completely different situations. A tradition can be at peace with science because it was right, because it revised when the evidence arrived, or because it never said anything checkable. Deism and Advaita score as clean as a physics department, and not because they passed a test. So what is plotted is how much a tradition got wrong against the share of that it still teaches, with the trail showing how far it has revised. Founding year is on every point, because the obvious objection, that this is really just a measure of age, is testable here and it fails.
Horizontal is how much it got wrong. Vertical is the share of that it
still teaches, so 0% is fully revised and 100% is fully unrevised. Every
tradition starts at 100% and moves down as it drops things, and the
faint trail is that journey, so its length is exactly how far it has
revised. Circle area is living adherents; click any point.
The x-axis used to read "exposure", meaning checkable claims ever asserted. That
was misleading: across 72 religious traditions it correlates with failure at
r = 0.978, because essentially everything a tradition said that could be
checked turned out false. It was one variable drawn twice, so it now says what it
measures. Note that x = 0 holds two different things: traditions that said almost
nothing checkable, and the few that made checkable claims and got them right.
By family
The same claims, judged by a different court. The scientific-evidence tab asks what OBSERVATION says; this tab asks what ARGUMENT says: is the claim coherent, does it sit well with the rest of what its tradition holds, what is the strongest case for it, and what is the strongest case against. The two are independent. A solid firmament is refuted but perfectly coherent; the Trinity is beyond evidence but philosophically rich. Science went silent on 864 of the 1,813 claims; this is the court that can still hear them.
What the three grades mean, and what they cannot mean. Philosophy rarely settles truth, so the scale measures how a claim fares under argument. Strong: the weight of argument favours it, and it is rare by design. Moderate: real arguments both ways, held by serious philosophers today, and the honest verdict for most metaphysics, theist and naturalist alike. Weak: coherent, but nothing recommends it beyond the tradition's say-so. Naturalist claims are held to exactly the same standard: "there is no God" and "mind is entirely physical" sit in Moderate with burdens of their own, not in Strong by default.
Internal contradictions are a second, overlapping measure, not a fourth grade. Each row has two bars. The upper bar is the grade profile and divides all of a tradition's claims into strong, moderate and weak. The lower red bar, on the same scale, is how much of that same doctrine is caught in a contradiction. They overlap on purpose and do not sum: a claim can be graded moderate and still sit inside a conflict, and across the corpus 584 claims are tangled, of which 367 are moderate or strong. Read the pair as a picture of the tradition rather than as a partition of it. A grade describes one claim. A contradiction is a conflict between several claims a tradition holds at once, so no single claim can carry it: the problem of evil is not a property of "God is omnipotent", and only appears when omnipotence, perfect goodness and real suffering are put together. A separate pass took each tradition's whole body of doctrine and looked for subsets that cannot all be true, finding 212, of which 14 are strictly inconsistent, meaning something in the set has to be false. Each card names the conflict, lists the claims that collide, and states the tradition's own answer at its strongest, because a contradiction a tradition answers well is still a real finding and hiding either half would misrepresent it.
Positive beliefs about a higher power or the soul, scored on three axes at equal weight and ranked. Denials (no creator, nothing survives, ritual does nothing) and plain descriptions of nature (the earth is a sphere) are filtered out by default: they are judged in the other tabs, but they are not candidate beliefs. Each row names the tradition the belief came from. Scientific evidence and philosophical support bear on whether the claim is true. Reached independently is the number of unrelated lineages that arrived at the idea behind the claim without inheriting it, shown as the actual count. Each claim is filed under one of a few hundred ideas, because traditions word the same idea differently and the wording would otherwise hide the convergence. People who hold it is shown on every claim but not scored. Each idea's Science and Philosophical bars are the mean of its claims, Lineages is its own count, and Adherence counts everyone whose tradition still teaches any form of it. Open an idea to see the claims behind it. Advanced options let you rank the individual claims instead, rank by lineages, restrict to one subject, or include the denials and descriptions of nature.
The four weights share 100%: move one and the others adjust to keep the total. Equal shares give equal say in the ranking. Getting that right takes an adjustment, because what reorders a list is how much an axis varies, not the number on its slider: nearly three quarters of these beliefs are untested, so science varies less than argument and would otherwise be quietly outvoted by it. How the weighting works.
How the weighting works
Science and argument are each scored on the same seven points, 1 to 7, and a point is turned into a 0-100 figure so the four axes can be averaged together. The composite is the mean of whichever axes apply to a belief, at the weights you set. An axis that does not apply is left out rather than counted as zero, so a belief science cannot address is judged on the rest instead of being penalised for the gap.
The adjustment. Two axes given the same share of the 100% would not otherwise carry the same influence. What moves a ranking is spread: an axis on which almost every belief scores the same cannot reorder anything, however heavily it is weighted. On this corpus science is bunched, because 1,231 of the beliefs are untested and 421 are wrong, while argument spreads across most of its scale. Left alone, argument decided about two thirds of the order while appearing to count equally. So each axis is divided by how much it varies across the beliefs being ranked before your share is applied. A percentage point then buys the same amount of reordering on whichever axis you spend it. The divisors are measured from the data each time the page loads, so they stay right as the judgments change.
What that does not fix. The adjustment makes the sliders honest; it does not give science more to say. Evidence bears on under a tenth of these beliefs, so argument is left carrying the ranking whatever the sliders read. Adherence starts at 0% because how many people hold a belief is not a reason to think it true; it is shown on every row so you can see it without it moving anything, and you can give it a share if you want it to count.
The mirror of the Evidence page: where science, mathematics or philosophy points toward a position some tradition holds. The three tiers are not interchangeable. Tested and held up is evidence. Live argument is a serious case made on mathematical or conceptual grounds that has never been tested, and an argument with unmeasured premises cannot function as a result however distinguished its proponents. Claimed but does not hold is where most reasoning in this territory goes wrong, which is why it is here.
Compatibility is the share of active axes where two traditions either agree or at least one is silent. Commitment is how many axes a tradition takes any position on. Plotting them together separates the two reasons a belief can look broad.