Jews Behind The Mountains

For the last year or so, I’ve been collecting Iberian, Hispanic, and Sephardic DNA matches. My goal: to map out the segments of my genome inherited from Jews who fled Spain and Portugal, trace them to specific branches of my family (most of whose paper trails run out in the early 19th century), and maybe, just maybe, establish connections to specific places on the Iberian Peninsula.

In my last post, I conjectured that deep analysis of my Y-chromosome subclade could help elucidate my direct male line’s path across the Mediterranean, into the proto-Ashkenazi community, and out to the Polish-Ukrainian border shtetl of Luboml. In all likelihood, this is the path of most of my ancestors. But Central and Eastern Europe have never held much appeal for me; by contrast, I speak Spanish and Portuguese, and have visited both Iberian countries more than once. Focusing on my minor Sephardic ancestry keeps things interesting, and allows me to work at a manageable scale.

With the help of Kevin Brook of, I’ve managed to validate 30 significant segments of Sephardic origin, with several others of lower confidence. Each of them matches at least a few Iberians or Latin American Hispanics, plus the usual boatload of Ashkenazim, and perhaps the odd Balkan or North African Sephardi. So far, I’ve managed to find matches in all but a few Latin American countries, plus several regions of Spain and Portugal.

Some segments of likely Sephardic origin.
The red-and-green one on Chromosome 5 is the subject of today’s post.

Not all of these matches have a story to tell—for what it’s worth, we now know that Jewish ancestry is common across Latin America, and Sephardic segments are anything but rare among Ashkenazi Jews (though my 30 breaks Brook’s previous record).

Here I’d like to focus on one genetic connection, on the paternal strand of my 5th chromosome, which proves to be an exception. A simple e-mail to I.C. revealed the story of a family with deep roots in the secret Jewish community of Trás-os-Montes in far northeastern Portugal, their tenacious preservation of a forbidden heritage, and their far-flung genetic ties to the Americas, Eastern Europe, and the East Mediterranean.

I.C., it turns out, is something of a lay expert about the Jewish community of Trás-os-Montes (“Behind the Mountains”, centered around the county town of Bragança), from which her mother’s mother’s mother hailed. When A.A., a Jewess from Duas Igrejas (a village on the border with Old León), married outside the secretive community, it caused a stir—and on the eve of World War II, she and her husband raised their daughter, N.C., non-Jewish.

Relevant locations, Trás-os-Montes.

A.A.’s family, whose pedigree extends back to the early 17th century on multiple sides, hailed from a collection of villages along the Douro with a documented history of Jewish settlement and inquisitorial persecution, chief among them Lagoaça and Vilarinho dos Galegos.

The earliest of her documented ancestors: F.G. and A.R. of Quintela de Lampaças, likely born around 1600. On October 2, 1634, the Jews of their village decided to hold Yom Kippur services—after which authorities found out, issuing arrest warrants for 19 of the participants, 9 of whom they managed to capture.

Several centuries and inquisitions later, the Jews of Trás-os-Montes forgot all but a few normative Jewish practices, developing their own code of ritual in its place. In any event, the consciousness of difference persisted down into the 20th century, where in dozens of northeastern Portugal villages, everyone knew which families were Jewish and which were “Old Christian”, as well as the limits of inter-communal comity. For example, like ought to only marry like.

But considering that starting around 1500, all secretly-professing Portuguese Jews were officially baptized alongside their gentile neighbors, can we really assume that inter-communal hanky-panky was effectively socially policed and rare?

Existing studies, limited to uniparental markers, present a mixed bag. Curiously, researchers had concluded that relative to the better-known crypto-Jewish community of Belmonte, farther south along the Spanish border, Jews from Trás-os-Montes seemed to have mixed more with the Portuguese general population. But without autosomal analysis (that is, a closer look at the 23 chromosomal pairs of recombinant DNA), it can be hard to draw conclusions about population history.

With N.C.’s genome, we have a chance to elucidate things. Now, for a reminder: N.C.’s father was born to a gentile family in Duas Igrejas in 1890; her mother was born to a Jewish family in the same village in 1916.

Using a simple population admixture Monte Carlo simulation program, we can get an answer to our basic question: what percentage of N.C.’s ancestry is Jewish, as opposed to Portuguese? (Sephardic Jews are quite genetically distinct from Portuguese and Spanish people, as Ashkenazi Jews are from German and Polish people, and so on.)

If N.C.’s father were entirely typically Portuguese, and her mother were entirely Sephardi Jewish, we would expect to get a model of 50% Portuguese and 50% Sephardi Jewish, give or take a few percentage points.

On the other hand, if her mother had picked up some gentile Portuguese admixture over the course of 13-15 generations since the original edict of forced conversion (1497) and the establishment of the Portuguese Inquisition (1536), we would expect a Sephardi Jewish share substantially lower than 50%. There is also the chance that her father might have had some Jewish ancestry, which suggests a Sephardi Jewish “maximum” above 50%.

Knowing nothing about the details of population dynamics among the secret Jews of Trás-os-Montes during the 16th to 19th centuries, one would probably assume that N.C. would come out to 25 or 30% Sephardic, maybe more, but maybe less.

As it turns out, one would be wrong. Using two-way mixes of potentially representative population averages, we yielded the following best fit in Eurogenes K13:

And its rough equivalent in MDLP K23b:

(A note: Algerian Jews are heavily Sephardic, and Galicians (from northwestern Spain) speak a dialect of Portuguese. Working with Eurogenes K13 population averages, nMonte mildly prefers Algerian Jews to Balkan Sephardim as a proxy for “Sephardic”, and prefers Galicians to the given Portuguese sample as a proxy for “northeastern Portuguese”. MDLP K23b prefers the 2 obvious populations.)

The crux:

Two models using the best amateur “admixture calculators” out there estimate that N.C. is between 47% and 58% Sephardic, and between 42% and 53% Portuguese. This corresponds strikingly to the null hypothesis of a “fully Jewish mother, plus a gentile father”—and might as well be taken as direct confirmation of a 50-50 blend.

N.C.’s mother, A.A., was likely of ~100% Jewish origin. While other Jews from Trás-os-Montes might test differently, this result is of broader relevance than just one individual’s Jewish limpieza de sangre. Because A.A.’s family originates in a wide range of locations across Trás-os-Montes, this result suggests the existence of crypto-Jewish kinship networks across northeastern Portugal that maintained very strict endogamy for ~400 years. This possibility can be further illuminated by testing more individuals of recent crypto-Jewish ancestry (it’s unlikely that there are many of full Jewish stock left, but half- and quarter-Jewish individuals, like N.C., will do).

And when you accept this finding, things begin to fall into place. Secret Jews, known pejoratively and with a wink as Cristãos Novos (New Christians), were viewed by their neighbors as necessary partners in trade, but were otherwise considered a race apart, too disreputable to marry—while they themselves took to shunning members of the clandestine tribe who got too chummy with members of the outgroup.

A caveat:

Both of these two models exceeds the ideal 2-3% distance limit, meaning that while the proportions of our models make an impressive case for Jewish endogamy in Trás-os-Montes, the degree of fit could be a bit better.

For starters, I would guess that none of the Sephardic reference populations are pure representations of the medieval Spanish/Portuguese Jewish population. Turkish, Greek, and Bulgarian Sephardim are likely to have mixed with Ashkenazi and Romaniote Jews, and all North African Jewish populations are a mix of Sephardic migrants and veteran Jewish communities.

Plus, it is possible that Trás-os-Montes Portuguese are genetically distinct from other Portuguese groups (Portuguese regional variety is not well-characterized, and Trás-os-Montes is the most isolated corner of the country), or that modern Portuguese have some Jewish ancestry, on average, making them a poor stand-in for the Portuguese gentiles of a few centuries ago.


The segment I share with I.C. and her grandmother, N.C., is also shared with several Ashkenazim, Mexicans, and New Mexico Hispanos, as well as a Chilean, a Brazilian, and an Alexandrian Sephardi Jew. It is very likely that the most recent common ancestor was a Spanish Jew, and that at least some share of N.C.’s Jewish ancestry derives from the wave of Spanish Jews who fled to Portugal around the turn of the 15th century. However, the segment of DNA we share is quite long to be 500 years old, so the possibility of a later flight from Portugal to Eastern Europe can’t be ruled out.

Excavating E-Y6923


Having just reformatted the website, I’d better get my money’s worth—what better excuse to rekindle my writing practice? But for the time being, it’s going to be easier to write about trivia than picking up the issues.

My latest trivial pursuit? Tracing my direct patriline all the way back to prehistory. Weird, right? Especially considering that the paper trail stops with my great-great-grandfather.

Turns out, there are thousands of people on the same sort of mission, a large share of them Jews with a limited paper trail. We live in the paradox of knowing (at least in theory) where our ancestors were in 1000 BCE and in 1000 CE, even if we can’t find documentation from before 1850.

Remarkably, genetic data have mostly confirmed the received narrative of Ashkenazi Jewish history: origins in Israel, classical-era migration to the West Mediterranean, emergence in Germany in the centuries following Charlemagne, and later diffusion into the Slavic lands.

In the last 7 to 10 years, the focus of Jewish population genetics has shifted from uniparental markers (especially Y-DNA) to autosomal DNA, giving a fuller picture of the mixed origins of European Jews. But with the help of better and cheaper sequencing technologies, hobbyists have honed in on the Y-chromosome, assembling a detailed catalogue of Ashkenazi patrilineages that for the most part date back to the Late Classical and Early Medieval periods.

Why does this matter? Well, as it turns out, rather than resorting to pie charts claiming that x% of Ashkenazim have Haplogroup J1, and y% have Haplogroup R1b, we can now make assertions like: 9 Early Medieval men account for the direct paternal ancestry of a full half of Ashkenazi Jews.

Courtesy of

For each of these 9 men, and for any other Ashkenazi Y lineages from the same time horizon (there are something like 100, but because of power laws, the vast majority of Ashkenazi men trace their direct patrilineages to the first few dozen), we can look at YTree, a tree of all charted human Y-chromosome mutations, check out its sister-lineages, and hypothesize about its classical, ancient, and ultimately prehistoric origins.

For some of the largest Ashkenazi groups (again, I mean specific Early Medieval subclades, like Q-Y2200, not “J1” or “J2”, each of which comprises about 20% of Ashkenazi Y-chromosomes, but across dozens of subclades), extensive work has been done to find non-Ashkenazi connections in the Jewish world, the Middle East, and beyond.

I’d like to do the same for E-Y6923, my own patrilineage, which holds the distinction of 5th largest Ashkenazi subclade, found in 5.6% of Ashkenazi men.

On E-Y6923: the big picture

For starters, consider that E-Y6923 represents a man who lived in the Late Classical or Early Medieval period, from whom 1 in every 18 Ashkenazi Jewish men directly descend on their direct male lines. Only four other men from this era contributed a larger share to the Ashkenazi gene pool.

Most Ashkenazim who belong to the Y6923 subclade don’t know it, even if they’ve tested their DNA before. Why? Because Y6923 wasn’t characterized until a few years ago, and 23andMe will still default to labeling a Y6923 man as “E-M34”, the macro-group under which Y6923 falls.

E-M34 is a geographically scattered, but well-characterized haplogroup with Paleolithic roots in either Northeast Africa or the Levant—its direct ancestor was found in human remains from Epipaleolithic Israel, and its earliest exemplar belongs to a man dug up in Neolithic Jordan.

Curiously, E-M34’s most famous member is Napoleon Bonaparte, whose membership speaks to the haplogroup’s low frequency in Mediterranean Europe. While not a member of the “classically Jewish” subclade Y6923, Bonaparte’s direct male-line relative did test positive for Y4972—which interestingly enough, is the closest sibling of the branch we’re discussing today, with representation among Europeans, Arabs, and Romaniote Jews.

A final note before we dive into the landscape of Y6923 itself: Y6923, as well as its Napoleonic sibling branch, descend from E-L791 (itself the main descendant of E-Z841), which appears to be the less successful son of E-M34, relative to E-M84.

M84 lineages account for 46% of Ashkenazi M34 (the majority belonging to Y6923), but they tend to outnumber L791 branches everywhere else that both are found. This is a question for another post.

Suffice it to say, L791, our intermediate father-clade, appears to have a most recent common ancestor ca. 3600 BCE—a date that falls smack in the middle of the Ghassulian period of the Copper Age, and seems to align with the dating of Proto-Semitic (or late pre-Proto-Semitic). Its descendants range across the Middle East and Europe, a saga that will one day hopefully be unrolled.

Y6923: the branch at hand

For now, our goal is limited to unrolling the saga of Y6923, which for at least the last 1,500 years (and in all likelihood, much longer), has been a Jewish saga.

First, one last technical clarification: we’ve been talking about Y6923, but we actually ought to be saying Y6938. Why? Because all Ashkenazim who’ve tested as members of Y6923 are also positive for Y6938, a downstream marker that’s only been discovered in the last 2 years.

As it happens, while Y6923 has a most recent common ancestor ca. 400 CE, Y6938 has a most recent common ancestor ca. 650 CE, which might be a helpful clue to tracking the origins and movements of this important line.

For what it’s worth, 650 CE dates a few centuries before the first attested, continuous Jewish settlements in old Ashkenaz, suggesting either a very early, pre-documentation origin in Germany, or more likely, an origin somewhere in France or Italy, in the generations prior to the foundation of the Rhenish Jewish community.

Until recently, the “closest relatives” of Y6923 Ashkenazim hadn’t been much help: E-Y6923 and E-Y4972, sister-descendants of Y4971, appear to have split during the Copper Age.

In the next few years, I expect the gap between Copper Age and Late Classical Period to narrow with& the detection of new markers and new branches, but who knows where and among whom.

The latest bombshell gives Y6923 a second child, and Y6938 a sibling: Y102667, exemplified by two non-Ashkenazi individuals—a Puerto Rican and a Libyan Jew. While we can’t make any detailed guesses about where Y6923’s direct ancestor was during the time of the Prophets, we now finally have a historical-era connection outside of the Ashkenazi community.

Courtesy of YFull.

On a fundamental level, these particular connections—Spanish and Libyan Jewish—should come as no surprise. We are talking about classical Mediterranean populations, which lived, along with the proto-Ashkenazi community, under Roman rule.

Nonetheless, a common Late Classical origin with these individuals rules out a few possibilities: namely, that Y6938 entered the Ashkenazi community via Rhenish or Khazar converts (given the heavily Levantine stamp of E-M34, these were never likely scenarios).

In addition, because the Puerto Rican and Libyan Jewish samples share a branch, we can rule out the possibility that Y6923’s presence in Spain is the result of an Ashkenazi back-migration to Spain. Y6938 is, to date, exclusively Ashkenazi, while Y102667 is, to date, exclusively non-Ashkenazi. This bifurcation also leads me to doubt the suggestion that Y6923 might have originated among Jews in Spain prior to arriving in Early Medieval Ashkenaz.

I am assuming that the Puerto Rican individual carrying Y102667 is of ultimately Jewish origin. While the simplest story would be an Early Modern converso introgression, we really can’t be sure of this—indeed, it’s possible that Jews were marrying out into the gentile Iberian community at a low rate for at least 1,500 years. Given the phylogeny and time depth of Y6923, a native Spanish origin for our Puerto Rican sibling seems very unparsimonious.

Wherever Papa Y6923 (or at least the most recent common ancestor of both Y6938 and Y102667) was located, it’s a safe bet that he was Jewish. Jewish, but not Ashkenazi, as the Ashkenazi community as such didn’t yet exist in 400 CE.

The puzzle of Y102667

By YFull’s estimate, the Puerto Rican Y102667 individual and the Libyan Jewish Y102667 individual share a most recent common ancestor dating to around 400 CE, contemporaneous with the split between Y102667 and Y6938. Superficially, this suggests no more recent affinity between the two individuals—but a more sophisticated reader might be able to correct me on this.

A disclaimer: the body of work on Late Classical/Early Medieval Jewish migrations around the West/Central Mediterranean, and on the origins of the Sephardic community of Iberia, is more diffuse (or possibly, more opaque to me) than that on the origins of Ashkenazi Jews in Italy and France. Either way, we’re dealing with a lot of conjecture and spotty sourcing.

My surface-level interpretation of the TMRCA date supplied by YFull suggests that Y102667 isn’t particularly likely to have originated in either Spain or Libya. That is, unless Jews were migrating from Spain to Libya, or from Libya to Spain, around 400 CE.

Rather, the common ancestor most probably came from an area that was supplying Jewish settlers to both Iberia and Libya during the Late Classical period. I will have to do some digging in my free time to find a country that matches this description (and I don’t expect a high-confidence answer), but off the top of my head, I can think of Northwest Africa and Italy as plausible candidates.

An alternative explanation dismisses the TMRCA, based on 2 samples, and narrows our theater of inquiry to Europe: since Late Medieval/Early Modern Sephardi Jews fled Spain for every possible Mediterranean shore, including Libya, it might be the case that Y102667 arrived in Libya in the last 500 years, via Spain. In this case, Y102667 would almost certainly have been born on European, likely Iberian, soil.

Wrapping it up: where Y6923 stands

There are other intermediate scenarios and timelines worth discussing, but the difference between these two theories of Y102667 presents a stark enough choice of possibilities for the origins of Y6923.

Flipping the order of the previous section, if Y102667 in the Libyan Jewish community is indeed of Sephardi origin (I asked Doron Behar for information on the sample’s origin, and no dice), then we can postulate that Y102667 is a marker of Iberian Jewry. Given that Y6938 is a bona fide marker of Ashkenazic Jewry, a common Late Classical origin would most likely trace back to some territory between Spain and Northern France/Germany—Southern France, but even more likely, Italy.

According to the most popular genetic models, both Ashkenazi and Sephardi Jews owe a substantial minority of their ancestry to Italy, a conclusion that is very much in step with the conventional historical narrative. In this case, Y6923 (or some downstream descendant of it, at least) is a Roman Jewish export from the Italian Peninsula, with likely Early Classical origins in Israel—for which no direct proof yet exists.

If on the other hand, Y102667’s Libyan connection has nothing to do with the Sephardi expulsion from Spain, we are left to ponder: where might the a most recent common ancestor of Ashkenazim, Sephardim, and Libyan Jews have lived, ca. 400 CE. Since we are only talking about one individual among many, rather than some averaged ‘central tendency’, the possibilities are various.

A piece of poorly-verified circumstantial evidence for this less-parsimonious path exists in the presence of a purported (via STR testing) Y6923+ Tunisian Jew in FTDNA and Semargl datasets, but until we can get ahold of him and test him for Y6923, Y6938, and Y102667 directly, it is too early to make any assumptions.

However, Y6923 in multiple North African Jewish communities, plus the Sephardi community, might suggest either a wide footprint of Y6923 across the Western Jewish world, or at the very least, some kind of Roman-era South or East Mediterranean origin for the common ancestor of Y6938 and Y102667. And either way, Italy remains an option.

Remember, at the end of the day we’re investigating the origins of 1/18 of all Ashkenazi patrilineages—not the Ashkenazi community at large. But maybe, just maybe, we can extend the reach of our family histories beyond our wildest dreams, and supply a new migration map for the next atlas of Jewish history.

Questions to pursue

What share of the Libyan Jewish gene pool is of Spanish Jewish origin? If 10%, it is not particularly likely that Y102667 arrived from Spain. If 30%, all bets are off.

Does the internal phylogeny of Y6938 (which is surprisingly well-characterized) offer any clues to origins, migration routes, and timing of the original Ashkenazi community?

Can we trust the TMRCA of 400 CE given for Y102667’s 2 individuals?

Are there any other non-Ashkenazi Y6923+ samples waiting to be catalogued?

What might’ve been a common source for Late Classical Jewish migration to both Spain and Libya?

What might’ve been a common source for Classical Jewish migration to Spain, France/Italy, and Libya?

“Evidence never works, stop trying to respond to it”

… as the world becomes even more random and confusing, the brain very slowly adjusts its highest level parameters. It concludes, on a level much deeper than consciousness, that the world does not make sense, that it’s not really useful to act because it’s impossible to predict the consequences of actions, and that it’s not worth drawing on prior knowledge because anything could happen at any time. It gets a sort of learned helplessness about cognition, where since it never works it’s not even worth trying. The onslaught of random evidence slowly twists the highest-level beliefs into whatever form best explains random evidence (usually: that there’s a conspiracy to do random things), and twists the fundamental parameters into a form where they expect evidence to be mostly random and aren’t going to really care about it one way or the other.

Antipsychotics treat the positive symptoms of schizophrenia – the hallucinations and delusions – pretty well. But they don’t treat the negative symptoms much at all (except, of course, clozapine). Plausibly, their antidopaminergic effect prevents the spikes of aberrant prediction error, so that the onslaught of weird coincidences stops and things only seem about as relevant as they really are.

But if your brain has already spent years twisting itself into a shape determined by random coincidences, antipsychotics aren’t going to do anything for that. It’s not even obvious that a few years of evidence working normally will twist it back; if your brain has adopted the hyperprior of “evidence never works, stop trying to respond to it”, it’s hard to see how evidence could convince it otherwise.

— Slate Star Codex, “Treating The Prodrome


  1. “When will I go and see the face of God?”
  2. “come join the youth and beauty brigade”
  3. “Let me take you down”
  4. “the gate is straight, deep and wide”
  5. “inciting and inviting me”
  6. “my heart has become used to itself”

Let us meet ourselves…

Let us meet ourselves exactly where we are.

Let us salute ourselves,
and each other,
and the moment,
and everything that is.

Let us meditate on the words of the teacher:
“All forces have been steadily employ’d to complete and delight me,
Now on this spot I stand with my robust soul.”

Let us give thanks for everything that has brought us here.

Let us give thanks for this vessel.
Let us pray for an auspicious voyage.
And let us make an offering,
of our hearts,
of our souls,
of our bodies,
of our minds.

Let us approach with reverence, joy, and a lightness of heart.
Let us hold nothing back.
Let us make full contact.

Let the stories fall away,
Let the geometry speak for itself,
Let every impulse and thought be accepted and beautified,
Let the wall between God and not God be revealed for what it is.

For indeed we are this creation,
And the Ground of Being is the ground of our being,
And the processes of nature cannot be evil,
And nothing in this universe is alien to us.

And let us meditate on the moment,
And on the journey ahead,
And on where we will find ourselves tomorrow.

And let us dedicate this experience to the betterment of all beings.

Let there be peace.
Let there be light.
Let there be love.

Let it happen!

Because no one argues for them, no one argues against them

Stances are very simple, and don’t require any specific beliefs or practices. No one explicitly promotes them. You pick them up automatically from our cultural “thought soup.” They are the ways people talk about meaning in soap operas and cafes.

Confused stances are insidious, because they are unnoticed. Because no one argues for them, no one argues against them. They are memes, mental viruses that people propagate by talking, without awareness of them.

— “Stances trump systems“, Meaningness

Genetic genealogy culminates in the once-thought-impossible

Just used DNA Painter for the first time. Astonishing. With new genealogically-known cousins popping up every few weeks on my lists at 23andMe and MyHeritage, I can now localize DNA segments to 1 grandparents, 4 great-grandparents, one (non-overlapping) great-great-grandparent, and one (overlapping) great-great-great-great-grandparent — for a grand total of 14% of all possible segments securely identified with an ancestor (if I chose to be looser about it, the % would approach 20).

This awesome free utility makes it possible to visualize them all in one chromosome map, which you can compare against matches of unknown provenance. Something I never thought possible.