Do independent personality traditions describe the same thing?

We embedded 56 archetypes from four traditions in a shared mathematical space. The results were more interesting than we expected.

Western astrology has twelve zodiac signs. Pythagorean numerology has twelve life paths. Chinese BaZi has ten Day Masters. The tarot has twenty-two Major Arcana. These systems developed independently, in different centuries, on different continents, from different philosophical starting points. They share no common ancestor. They do not reference each other.

And yet practitioners who work across multiple systems have noticed something for a long time: certain archetypes seem to describe the same person. Aries and Life Path 1 both describe the pioneer who leads through sheer force of will. Cancer and Life Path 6 both describe the nurturer whose strength comes from emotional depth. Sagittarius and Life Path 5 both describe the restless seeker who needs freedom above all else.

These correspondences get passed around as practitioner wisdom. Nobody had tested them mathematically. So we did.

How we did it

We wrote personality profiles for all 56 archetypes across the four systems. Same structure for each: core nature, strengths, shadow, values. Between 100 and 180 words. Personality portraits, not horoscope predictions.

We fed the profiles into a sentence embedding model, a neural network trained on billions of English sentences that converts text into a point in 384-dimensional space. Similar meanings land near each other. Different meanings land far apart. The model knows nothing about astrology or numerology or BaZi or tarot. It just measures whether two descriptions say similar things.

That gave us 56 points in a shared space, one per archetype. From there we could measure distances, test for patterns, and ask a specific question: do these four systems describe the same personality landscape, or four separate ones?

What we tested

Seven analyses. The short version of each:

We measured similarity within systems versus across systems. If the traditions describe separate things, same-system pairs should be much more similar than cross-system pairs.

We ran permutation tests: shuffle the system labels 1,000 times, recompute everything, see what random chance looks like. This gives us p-values without assuming any particular statistical distribution.

We used the Hungarian algorithm to find the best one-to-one matching between archetypes in different systems. This is an operations research tool for assignment problems. It has zero knowledge of personality traditions. It just finds the pairing that minimizes total distance.

We ran Procrustes analysis to test whether the geometric shape of one system's space matches another's. This goes further than asking whether individual pairs match. It asks whether the whole relational structure is preserved. If Aries is to Libra what Life Path 1 is to Life Path 2, Procrustes detects it.

We extracted 12 structured personality traits from each description (things like assertiveness, empathy, discipline, sociability, intensity) to strip out prose style entirely and measure only personality content. This was the confound control. If our results were just an artifact of writing style, this layer would break them.

We clustered all 56 archetypes and checked whether clusters followed system boundaries or personality boundaries.

And we added 35 profiles from MBTI, Big Five, and Enneagram as controls. If the esoteric systems converge with validated psychology frameworks, that tells us something different than if they only converge with each other.

The results

The algorithm found what practitioners already knew

The Hungarian matching independently discovered correspondences that experienced practitioners would recognize immediately:

We then switched to a completely different embedding model with twice the dimensionality. Eight of twelve alignments held. The algorithm found the same matches from a different mathematical starting point.

That part surprised us. We expected some overlap, not 67% stability across models.

The shape of one system predicts the shape of another

Procrustes analysis asks a harder question than "do individual pairs match?" It asks whether you can rotate and scale one system to overlay the other, whether the internal distances are preserved.

For Western Astrology and Numerology, the answer is yes (p = 0.006). The two systems don't just contain matching archetypes. How far Aries sits from Cancer in one system predicts how far Life Path 1 sits from Life Path 6 in the other. The relational geometry is shared.

Removing prose style made the results stronger

This is the finding we trust most.

We were worried that our results might just reflect writing style. One author wrote all the descriptions. Maybe the convergence was coming from shared vocabulary rather than shared personality content. So we stripped the prose away entirely and extracted 12 personality dimensions per archetype: raw trait scores with no style attached.

The convergence got stronger, not weaker.

All three pairings are statistically significant with style removed. The gap between within-system and cross-system similarity dropped from 0.046 to 0.002. The systems became nearly indistinguishable on personality content.

Some of the trait-space matches feel almost too clean. Bing Fire and The Sun hit 0.996 similarity on personality traits. Life Path 9 and The Star hit 0.994. Taurus and Life Path 4 hit 0.993.

Clusters group by personality, not by system

When we clustered all 56 archetypes, the clusters ignored where each archetype came from. The Adjusted Rand Index between clusters and system labels was 0.057, where 0 means no relationship and 1 means perfect alignment. Basically zero.

One cluster pulled Cancer, Life Path 6, Ji Earth, The Empress, Strength, and The Sun together. Four systems, one personality: warm, nurturing, emotionally grounded.

Another grouped Scorpio, Life Path 7, Gui Water, The Hermit, and The Moon. The depth-seekers. The ones who operate below the surface.

The psychology frameworks mixed right in

When we added MBTI, Big Five, and Enneagram profiles, the clustering got even more cross-system. Seven of eight clusters contained four or more different systems. ENFJ grouped with Cancer and Life Path 6. INTJ landed next to Capricorn and Life Path 22. Enneagram Type 8 grouped with Aries and Life Path 1.

The esoteric traditions and the scientific frameworks are describing the same space.

What we cannot claim

We want to be direct about limitations.

One person wrote all 56 archetype descriptions. Even with style normalization and trait extraction, we can't fully rule out that a single author's understanding of personality is what creates the convergence, rather than the traditions themselves. This is the biggest confound and the hardest to eliminate.

The permutation test on our strongest cross-system pairs was not significant (p = 0.786). The top similarities between systems are no stronger than what you get from shuffling labels. The convergence is structural, living in the shape of the space, not in the peak similarities.

We tested with two embedding models and got different results on some metrics. Procrustes held in the smaller model but not the larger one. The Hungarian alignments were stable in both. Some findings are more robust than others.

The sample size is 56 archetypes. Enough for the analyses we ran. Not enough for sweeping claims about universal personality structure.

What this tells us

Four personality traditions that developed independently, in different parts of the world, in different centuries, describe the same underlying landscape when their descriptions are placed in a shared mathematical space. The algorithms that find their correspondences arrive at the same answers that practitioners arrived at through experience. The shape of one system predicts the shape of another.

This does not prove these systems are "right" in any predictive sense. What it shows is that they are mapping the same territory. Human personality has a shape, and independent cartographers drew maps that agree on where the mountains are.

That is what Sagelon is built on. Not the claim that any single system has all the answers. The observation that when you lay them side by side, they converge on something real.

See the convergence in your own chart

The Sagelon Snapshot runs your birth data through all four systems and shows where they agree. It takes about two minutes. It is free.

Run your free Snapshot →