Nine men, all young to middle-aged, buried together in a moat dug into the eastern side of medieval Sidon, in a pit dated to the thirteenth century and dug alongside a mass grave holding at least twenty-five bodies with signs of violent death. Frankish belt buckles and a silver denaro minted between 1245 and 1250 lay with the bones. In 2019 Marc Haber and colleagues sequenced the whole genomes of these nine men, and the result was not the homogeneous army of French and German knights the word "Crusader" usually conjures. Four of the nine are genetically indistinguishable from the Roman-period and modern population of Lebanon. Two carry a genuine, measurable mixture of European and Levantine ancestry. The remaining three are European, but not from a single place, and this article's own G25 reconstruction of where each of the nine actually sits puts real, checkable numbers behind that diversity, including a floor test on the two claims that matter most: how much European ancestry the two admixed men actually carry, and whether any of it survives in Lebanon today.
A pit, a denaro, and a name that hides more than it says
Sidon changed hands repeatedly during the Crusades. Captured by Baldwin I in 1110, it fell to Saladin in 1187, was briefly retaken, and was lost for good with the fall of Acre in 1291. Somewhere in that span, in a dry moat on the eastern fortification of the city, a group of men died together and were buried in a single arbitrary deposit. The archaeology, radiocarbon dated to a broad 1025 to 1283 calCE window for the wider burial complex, and the Frankish material culture found with the bodies, identified the smaller deposit as Crusader dead. What the archaeology could not say is who these men actually were before they became a battle casualty on a Levantine coastline. That is the question ancient DNA answers, and answering it precisely requires treating the word "Crusader" as a military and religious category, not a population.
Data and method
All nine sequenced individuals from the Sidon pit are used here: SI-38, SI-39, SI-40, SI-41, SI-42, SI-44, SI-45, SI-47 and SI-53, drawn from the realigned Global25 entries in the dated ancient sample file, with coverage ranging from 34.4 to 97.9 per cent. The reference panel covers Lebanon's Roman period and its three modern confessional populations (Christian, Muslim, Druze) on the local side, and France, Basque France and Spain, Andalusia, Sardinia, northeastern Italy and Greece on the European side, the same regions the original 2019 PCA implicated. Distances are scaled Euclidean distances multiplied by one thousand, and ancestry proportions come from non negative least squares with a sum to one constraint, the conventions used throughout this site.
Two honest notes before any number is used. Six of the nine genomes sit below 50 per cent coverage, which is the usual range for this kind of degraded medieval DNA and adds individual level noise that a population average would not carry. And a Global25 coordinate is an autosomal summary with no chromosome of origin attached, so nothing below can independently confirm or contest the Y-chromosome evidence the original paper also reports; that evidence comes entirely from the published study, and this article does not repeat haplogroup claims it cannot itself verify from G25 coordinates.
Where the nine actually sit
Global25 distance from each of the nine Sidon individuals to the nearest of the three modern Lebanese confessional populations (Christian, Muslim, Druze). The gap between the local cluster, the two admixed men and the three Europeans is not subtle.
The split falls cleanly into three bands, and it reproduces the original study's own PCA classification without borrowing its numbers. SI-45, SI-38, SI-44 and SI-42 sit between 26.8 and 45.0 units from modern Lebanon, the same range occupied by the Roman-period Lebanese population itself: these are the four local men. SI-40, SI-47 and SI-39 sit between 159.3 and 172.5 units away, a gap roughly four times wider than the local cluster's own spread: these are the three Europeans. SI-53 and SI-41 sit in between, at 85.5 and 105.0, closer to Lebanon than any European but far too distant from it to be local: these are the two admixed men, and the gap between them and both other groups is wide enough that this is not a borderline call.
The three Europeans are not one population wearing three labels. SI-40's nearest independent neighbours in the full 30,072-entry panel are Basque Spain (37.2), Basque France (40.5) and French general (43.1): a Franco-Iberian border profile. SI-47 and SI-39 both sit closest to French Alsace and French general (30.4 and 32.1 for SI-47; 36.5 and 42.7 for SI-39), a broadly northern French and western European profile rather than anything specifically Iberian or Sardinian. None of the three shows a clean Sardinian signal on this reconstruction, which is worth stating plainly since the original paper describes one of its three Europeans as Sardinian-clustering on a two-component PCA; a full 25-dimension Global25 distance does not always agree with a two-component projection, and this article reports what its own method finds rather than forcing agreement.
The two admixed men, and what they are actually mixed with
A raw distance places SI-41 and SI-53 between Europe and the Levant. It does not say what their European component actually is, and treating "Europe" as one undifferentiated source would waste the most interesting part of the result. Modelled as a two-source mixture of Lebanon's Roman-period population and a single European source, substituted six ways, the fit is not equally good for every candidate.
| Individual | Best-fitting European source | Levant / Europe split | Residual |
|---|---|---|---|
| SI-41 (AD 1240, 95.0% coverage) | Spanish, Andalucia | 25.7% / 74.3% | 0.0378 (lowest of six candidates tested) |
| SI-53 (AD 1099, 97.9% coverage) | Greek, Central Greece | 6.4% / 93.6% | 0.0247 (lowest of six candidates tested) |
For SI-41, Andalusian Spanish beats French, Basque, Sardinian, northeastern Italian and Greek sources on residual every time it is tried, and it is also his single closest neighbour in the full panel (Spain, Andalusia, Early Medieval, at 25.0 units). His European ancestor most plausibly came from Iberia, not from northern France or the Rhineland the word "Crusader" usually implies.
SI-53 is the more surprising case. Buried around AD 1099, the year Jerusalem fell to the First Crusade, his single closest population in the entire 30,072-entry panel is not French, Iberian or Sardinian but Greek, Central Greece, at 25.7 units, closer to him than anything else including the Roman-period Lebanese baseline. A two-source model gives Greek ancestry 93.6 per cent of his profile with the lowest residual of any source tested. That is not a small Levantine admixture on top of a French or Norman base. It is a man whose ancestry looks predominantly Aegean, with a real but modest Levantine component, at the earliest date in the entire pit. Byzantine Greek soldiers, sailors from the Italian maritime republics operating throughout the eastern Mediterranean, and Greek-speaking Christian communities already resident on the Anatolian and Balkan coasts are all more consistent with this profile than the Frankish knight of popular memory, and the date makes a Byzantine or Aegean origin considerably more plausible than a French one for this specific man.
The placebo test: is 9 per cent European actually European?
Every one of the six candidate sources returns some non-zero European weight even against SI-45, one of the four men already established as local by every other measure. Reported without a check, that would look like a third admixed individual. It is not, and the reason is worth showing rather than asserting.
| Target | Weight range across six sources | Residual range | Reading |
|---|---|---|---|
| SI-41 (admixed) | 60.3% to 74.3% Europe | 0.0378 to 0.0727, varies sharply by source | Real signal: the model cares which source it is given |
| SI-53 (admixed) | 45.8% to 93.6% Europe | 0.0247 to 0.0762, varies sharply by source | Real signal: one source fits far better than the rest |
| SI-45 (local control) | 8.5% to 15.4% Europe | 0.0343 to 0.0357, flat regardless of source | Noise: the model does not care which source it is given |
| Lebanese Christian, modern | 8.2% to 16.2% Europe | 0.0176 to 0.0213, flat regardless of source | Same noise floor as SI-45, seven centuries later |
The distinction is not the weight the model assigns. It is whether the fit actually improves when the source changes. For SI-41 and SI-53, swapping the European source moves the residual by a factor of two or three: the model is being genuinely picky, and it is picky in a historically coherent direction (Andalusia for SI-41, Greece for SI-53). For SI-45, the residual barely moves no matter which source is offered, between 0.0343 and 0.0357 across all six: the model has nothing to be picky about, because there is no real European component to detect, and the eight to fifteen per cent it reports is the floor noise of fitting two sources to twenty-five coordinates on a single degraded ancient genome, not a signal. A spike-in test on the Roman-period Lebanese average confirms the method itself has no such floor when the target is clean: injecting one per cent of French ancestry into that average and refitting recovers essentially exactly one per cent. The floor appears only once individual-level genome noise is introduced, and it sits at roughly nine percentage points for a single ancient genome in this particular two-source system.
No lasting trace: the pulse checked against Lebanon today
The same test run on the three modern Lebanese confessional populations returns the identical pattern as SI-45: eight to sixteen per cent apparent Europe, flat residuals regardless of which European source is offered, no source the model actually prefers. Modern Lebanon sits inside the same noise floor as a genome already established as locally Levantine seven hundred years before any of these three populations existed as such. That is the Global25 face of the conclusion Haber and colleagues reached independently in 2017 using entirely different methods: present-day Lebanese ancestry derives overwhelmingly from the local Bronze Age population and a Eurasian steppe-related admixture around 1750 to 170 BCE, with no significant Crusader-era European signal surviving in any of the three confessional groups tested here, Christian, Muslim or Druze alike. Two centuries of intermittent European rule, and the demographic contribution of the Franks who died at Sidon and elsewhere, left behind an admixed man or two inside a single military mass grave and nothing measurable in the population that came after.
What this method cannot say
Three limits, stated before anyone else states them. Six of the nine genomes sit below 50 per cent coverage, which is the normal range for degraded medieval remains but adds real individual-level noise; this article treats no single percentage as more precise than the placebo test shows it to be. A Global25 coordinate carries no chromosome-of-origin information, so nothing here confirms or revises the original study's Y-chromosome findings, which come entirely from the 2019 paper itself. And the nine men in this pit are nine men, not a random sample of the Crusader states' population; a mass grave from one battle at one city on one date cannot stand in for the demographic history of the Crusades as a whole, only for what happened to this particular unit.
Myth and measurement
| The claim | What the evidence shows |
|---|---|
| The Crusaders buried at Sidon were French and German knights, a single homogeneous force from western Europe. | Four of the nine are genetically local Levantines. Of the three who are European, one is Franco-Iberian, two are broadly French/western European, and the earliest man in the pit, dated to 1099, is predominantly Greek/Aegean rather than Frankish at all. |
| A mixed European-Levantine ancestry in two of the nine means the Crusaders were becoming genetically indistinguishable from the local population within a generation or two. | The two admixed men, SI-41 and SI-53, carry 74.3% and 93.6% ancestry from a specific, checkable European source (Andalusian Spanish and Greek respectively) alongside a real but minority Levantine component; they are not halfway to local, and a placebo test confirms the local men in the same pit show no comparable European signal above the method's own noise floor. |
| If Crusader men fathered children with local women, that ancestry should still be detectable in Lebanon today. | A century-scale event this large would need to leave a population-wide trace to be detectable at all. Modern Lebanese Christian, Muslim and Druze populations all sit inside the identical noise floor as a genome already established as purely local seven centuries earlier: no source the model tries fits any better than any other, the signature of nothing to detect. |
References
- AJHG 2019 Haber M, Doumet-Serhal C, Scheib CL, Xue Y, Mikulski R, Martiniano R, Fischer-Genz B, Schutkowski H, Kivisild T, Tyler-Smith C. A Transient Pulse of Genetic Admixture from the Crusaders in the Near East Identified from Ancient Genome Sequences. American Journal of Human Genetics 104:977-984. https://doi.org/10.1016/j.ajhg.2019.03.015
- AJHG 2017 Haber M, Doumet-Serhal C, Scheib C, Xue Y, Danecek P, Mezzavilla M, et al. Continuity and Admixture in the Last Five Millennia of Levantine History from Ancient Canaanite and Present-Day Lebanese Genome Sequences. American Journal of Human Genetics 101:274-282. https://doi.org/10.1016/j.ajhg.2017.06.013
- Press 2019 Wellcome Sanger Institute. Crusaders made love and war, genetic study finds. Read the release
The coordinates
Global25 scaled coordinates for the nine Sidon individuals and the reference populations used above, for anyone who wants to reproduce the models in Vahaduo.
Lebanon_Sidon_SI-53__AD_1099__Cov_97.89,0.118376,0.14319,-0.003771,-0.02584,0.010771,-0.006972,0.00094,-0.002308,-0.007772,0.012028,-0.001299,-0.00015,-0.013677,-0.007294,-0.008822,-0.002121,0.007953,0.00114,0.001131,-0.001,0.007487,-0.002349,0.003821,0.008073,-0.001916
Lebanon_Sidon_SI-38__AD_1150__Cov_43.36,0.093335,0.146236,-0.050911,-0.089148,-0.022773,-0.030399,-0.001645,-0.012923,-0.000205,0.014397,0.006658,-0.007044,0.016204,-0.001376,-0.010586,0.021082,0.016559,-0.003167,0.005154,-0.009505,-0.001872,0.008285,-0.015036,0.009519,-0.004311
Lebanon_Sidon_SI-47__AD_1150__Cov_35.45,0.12862,0.136081,0.061471,0.026809,0.033852,0.007809,-0.005875,0.007154,0.010022,0.002005,-0.006658,0.007493,-0.00892,-0.01734,0.014929,0.001856,-0.000913,-0.004054,-0.00729,-0.005253,-0.000998,0.007172,0.00037,-0.001928,0.002994
Lebanon_Sidon_SI-44__AD_1150__Cov_39.17,0.103579,0.153345,-0.041106,-0.079458,0.000615,-0.0251,-0.00376,-0.011999,-0.002045,0.023873,0.010718,-0.002098,0.005352,0.0,-0.009772,0.001724,-0.005607,0.000887,0.012696,-0.008254,-0.007362,-0.004451,-0.000739,-0.008435,0.005987
Lebanon_Sidon_SI-40__AD_1150__Cov_34.37,0.126344,0.147252,0.048271,0.002584,0.045547,0.003626,-0.008225,0.010153,0.022907,0.033349,-0.001786,0.013638,-0.015609,-0.023946,0.010043,0.007425,0.008736,0.011402,0.011439,0.007879,0.011105,-0.002597,0.0053,-0.006145,-0.001197
Lebanon_Sidon_SI-42__AD_1225__Cov_49.63,0.068294,0.153345,-0.045254,-0.090763,-0.008617,-0.0251,-0.0047,-0.014307,0.014521,0.010205,0.012179,-0.010341,0.013231,0.004817,-0.0076,-0.009281,-0.014603,-0.001267,0.004148,-0.015507,0.000499,0.011994,-0.009613,0.005904,0.009221
Lebanon_Sidon_SI-41__AD_1240__Cov_95.01,0.075123,0.150298,0.015462,-0.031331,0.038776,-0.013387,-0.013396,0.005769,0.022293,0.033349,0.003897,0.013488,-0.000595,-0.014313,0.002307,0.000928,0.000261,-0.003294,-0.004022,0.005127,-0.001373,-0.002349,0.0053,-0.000602,0.003592
Lebanon_Sidon_SI-39__AD_1248__Cov_47.66,0.122929,0.131003,0.050157,0.04199,0.035699,0.016733,0.00423,0.008538,0.004909,0.005103,-0.007795,0.006894,-0.011001,0.001239,0.011129,-0.011933,-0.014603,-0.003167,-0.001257,0.005878,0.009608,0.01422,-0.002835,0.003133,0.002395
Lebanon_Sidon_SI-45__AD_1252__Cov_96.83,0.083091,0.151314,-0.058454,-0.07752,-0.005539,-0.031794,-0.01081,-0.008077,-0.006545,0.013485,0.000162,-0.007643,0.002527,-0.003991,-0.008415,-0.000133,-0.001434,-0.002154,0.000126,-0.002251,0.002995,-0.001855,-0.001725,0.001205,-0.000239
Lebanon_ERoman,0.0838497,0.146575,-0.05506,-0.092055,-0.016516,-0.0304923,-0.00235,-0.0073843,0.0068173,0.0093547,0.010555,-0.008942,0.0155597,-0.0015137,-0.011446,0.0152037,0.010909,0.0031247,-0.000461,0.0019593,0.0011647,0.0046163,-0.0064497,0.0060247,-0.0050693
Spanish_Andalucia,0.108132,0.146076,0.035707,-0.003434,0.045256,-0.005857,-0.00193,0.00266,0.024715,0.031162,-0.002154,0.007272,-0.013207,-0.011408,0.010243,-0.002498,-0.006176,-0.0011,-0.004261,-0.002133,0.002929,-0.001302,-0.001972,-0.005334,0.001885
Greek_Central_Greece,0.118945,0.13861975,0.0028285,-0.02236775,0.014695,-0.005787,0.00088125,-0.00294225,-0.0037835,0.01599125,0.00158325,0.00037475,-0.00297325,0.00406,-0.01516675,-0.0005305,0.01131075,0.00351575,0.00490225,-0.00600275,-0.00446075,-0.00139125,0.002126,0.00198825,-0.000419
French_general,0.12725066,0.14206298,0.048035763,0.019803939,0.040454594,0.0057259707,0.0028751553,0.0034433537,0.010964412,0.01792176,-0.003531958,0.0069664864,-0.014029829,-0.0092185363,0.011249951,0.0035757718,-0.0015258768,-0.00023754557,0.0011234343,0.00084802532,0.0027821352,0.0018239047,-0.0013037431,0.0059363903,-0.0013490234