The observation readers arrive with
The same surnames appear over and over in their tree. Cousins marrying cousins, four and five generations deep. Families that stayed "Spanish," married other "Spanish" families, and kept doing it for three centuries. And then the DNA test comes back with a Jewish percentage that their neighbours don't have.
They conclude: my family kept the bloodline.
The truth is better, and stranger.
What the published science actually says
The landmark study: Chacón-Duque et al., Nature Communications (2018), "Latin Americans show wide-spread Converso ancestry."
- 6,500+ individuals from Brazil, Chile, Colombia, Mexico and Peru (CANDELA).
- Haplotype-based (SOURCEFIND), not ADMIXTURE — 546,780 SNPs against 2,359 reference individuals from 117 populations. This matters: haplotype methods read shared segments, which is what actually survives from a distant ancestor.
- ~23% of individuals carried >5% East/South Mediterranean ancestry, averaging 12.2% among those; the Sephardic component averaged 7.3%.
- By country: Colombia highest at ~10%, Chile ~4%, Mexico ~4%, Peru ~2%, Brazil ~1%.
The single most important sentence in that paper, for us: the estimates are
"values over and above those present in present-day Spanish/Portuguese
reference individuals." They ran it against a control panel. That is
exactly what score_contrast.py refuses to ship without. When a study does the
contrast properly, the finding survives — and that is why we can cite this one.
The older Antioquia study: Carvajal-Carmona et al., Am J Hum Genet (2000).
- A staggering sex bias: ~94% of Y chromosomes European, ~1% Amerind — against ~90% Amerind mtDNA. Immigrant men, local women, at scale.
- And: "~14% of the Antioquian haplotypes could have a Jewish ancestry", including the Cohen Modal Haplotype.
That 14% needs handling with care. The authors themselves wrote that greater confidence "requires a more extensive evaluation of their frequency in Mediterranean populations," and that their work was limited by how few Iberian and North African populations had been sampled. Beyond that: it is a six-STR haplotype study from 2000. The Cohen Modal Haplotype is a short motif that is common among non-Jewish Arab populations too. By modern standards this is low resolution. "Could have" is carrying the whole sentence.
It is best read as the question that prompted better work, not as a result.
Costa Rica — the gap
The Central Valley is a genuine, well-described founder population: small colonial founding group, high endogamy, repeatedly studied as a possible genetic isolate, and unusually well documented genealogically (Sanabria Martínez, Genealogías de Cartago hasta 1850).
But: Costa Rica is not in CANDELA. The 2018 study did not sample it. We found no peer-reviewed study estimating converso ancestry in Costa Rica with modern haplotype methods.
So the honest statement is: Costa Rica has the strongest genealogical documentation of any of these populations and the weakest published genetic evidence. That is a gap, not a finding — and it is exactly the kind of gap the site should name out loud instead of quietly filling with Colombia's numbers.
The mechanism
Endogamy does not create Jewish ancestry. It does three things to ancestry that is already there.
1. Pedigree collapse. In an endogamous population the same founders occupy many positions in your tree at once. A single converso ancestor fifteen generations back is not one slot in 32,768 — they are dozens of slots, because your parents' lines keep rejoining. The expected contribution is far higher than the naive halving math predicts.
2. The segments stay long enough to see. This is the real answer. Every generation, recombination chops inherited DNA into shorter pieces. Ancestry from 1500 is normally shredded into fragments too short for any method to call — that is why most Latin Americans, and most Iberians, carry converso ancestry that is simply undetectable. Endogamy keeps re-pairing people who descend from the same founders, so those segments get refreshed rather than only ever diluted. The signal survives into the range where software can see it.
3. Variance goes up. Two siblings' cousins in an endogamous population can differ far more in measured ancestry than in an outbred one. A dramatic result in one relative and a flat one in another is expected, not contradictory.
So the right sentence is not "my family kept the bloodline." It is: "my family's structure preserved the evidence." Their converso ancestors are probably no more numerous than their neighbours'. They are simply still readable.
And this is the same mechanism as Quebec. French-Canadian descendants are the other textbook endogamous founder population, and they get Jewish percentages that puzzle them for precisely this reason. One explanation covers both, and the site should tell them so.
What this does NOT license
- A consumer percentage is not the 2018 estimate. 23andMe and Ancestry report against an Ashkenazi reference panel. Sephardic is detected poorly. Below ~2%, any figure is noise. A Costa Rican seeing "3% Ashkenazi" has not measured the thing Chacón-Duque measured, and the two numbers must never be set side by side on this site.
- Mediterranean is not Jewish. The component these methods pick up sits in a region of genetic space shared with Lebanese, Syrians, Greeks, southern Italians and plain Iberians. The 2018 authors flag their own overestimation risks: drift in Native reference groups, post-colonial admixture in references.
- Endogamy inflates match lists. You will match thousands of people and the lists will look dramatic. That is population structure, not close cousinhood, and it will mislead anyone doing segment triangulation without accounting for it.
- The repeating surnames still carry nothing. They are the signature of endogamy — genuinely informative about your pedigree's shape, and therefore about detectability. They say nothing whatever about origin.
The takeaway to end on
If you descend from one of these families, the endogamy your ancestors practised for religious and social reasons of their own has done something none of them intended: it kept a five-hundred-year-old signal legible.
That is worth knowing. It is not a document. What it earns you is a better prior and a smaller haystack — a defensible reason to go looking in Cartago, Antioquia or Monterrey parish registers for a named person in a dated record.
THE CITATION CARRIES THE CLAIM, NEVER THE MARKER — but a well-built contrast against a control panel tells you which archive to sit down in.
Sources
- Chacón-Duque et al., "Latin Americans show wide-spread Converso ancestry and imprint of local Native ancestry on physical appearance", Nature Communications 9:5388 (2018)
- Carvajal-Carmona et al., "Strong Amerind/White Sex Bias and a Possible Sephardic Contribution among the Founders of a Population in Northwest Colombia", Am J Hum Genet 67 (2000)
- Morera & Barrantes, "Is the Central Valley of Costa Rica a genetic isolate?" (2004/2007)
- Sanabria Martínez, Genealogías de Cartago hasta 1850