How to use toponyms to argue against nationalistic ideas
The idea that I strongly believe mathematics supports if applied to Croatian
toponyms, that is, that Illyrian (the language spoken in Croatia in ancient
times) was a centum Indo-European language, is a very politically
incorrect idea. Here I present some rhetorics which you can use to
argue for it -- and, more importantly, to fight the claims made by politically motivated people (mostly nationalists).
First, let's address the elephant in the room, namely, the fact that Albanian nationalists as well as left-wingers all over the Balkans (for the reasons having to do with political correctness -- they believe arguing otherwise could be used to justify the genocide of the Albanians on Kosovo in the late 1990s) believe that Illyrian was not only a satem language, but also a direct ancestor (or at least a language closely related to the direct ancestor) of Albanian. Here is how I would argue against it:
Though, to be honest, the claims typically made by Croatian nationalists seem to make at least some degree of sense, unlike the statements made by Serbian nationalists and Albanian nationalists. I can at least understand Croatian nationalists, whereas I have trouble conceiving how somebody might believe what other nationalists here on the Balkans claim to believe. Have they even looked at the ancient toponyms?
UPDATE on 24/08/2026: Look, in case it is not obvious, any honest linguist will admit that the study of the toponyms is a very soft science. In other parts of historical linguistics, you don't even need to calculate the p-values, because the patterns are so strong that the mere suggestion that they are coincidental must appear absurd. Take, for instance, the evidence for the High German Consonant Shift, such as English 'd' regularly corresponding to German 't': "day"-"Tag", "cold"-"kalt", "old"-"alt", "dream"-"Traum", "red"-"rot", "do"-"tun", "death"-"Tod", "dale"-"Tal", "ride"-"reiten", "blood"-"Blut", "dew"-"Tau", "deep"-"tief", "good"-"gut", "god"-"Gott", "bed"-"Bett", "drive"-"treiben", "drink"-"trinken", "drop"-"Tropfen", "dove"-"Taube", "deaf"-"taub", "daughter"-"Tochter", "bread"-"Brot", "door"-"Tür", "word"-"Wort", "world"-"Welt", "shoulder"-"Schulter", "hard"-"hart", "loud"-"laut"... The mere suggestion that it's a coincidence appears absurd. The p-value is so low it's not even worth thinking about. But, in the study of the toponyms, the most striking pattern in the Croatian toponyms, that is the k-r pattern in the Croatian river names, has a p-value of somewhere between 1/300 and 1/17. And most of the time you do not even have a weak p-value, but you have to rely on anecdotes of the worst kind. And people seem to intuitively realize that, without having studied historical linguistics.
Low-hanging fruit in linguistics is gone. Has been since the 19th century. These days, if you want to possibly discover something new, you have to rely on statistics and bad anecdotes. In some sense, the same thing happened in physics. Ernest Rutherford said 100 years ago that if your experiment requires statistics, you ought to have done a better experiment. These days, all the ideas for an experiment in physics without statistics have been tried, so physics these days heavily relies on statistics. Except that, in physics, the p-values of those high-hanging fruit are in the neighbourhood of 1 in a million, whereas, in historical linguistics, the p-values of the high-hanging fruit is such that it's considered to be right on the margin of error even by the social science standards. It's not at all obvious what the right thing to do is. Leave it to nationalists who apparently never bother to do any mathematics to get closer to the truth?
First, let's address the elephant in the room, namely, the fact that Albanian nationalists as well as left-wingers all over the Balkans (for the reasons having to do with political correctness -- they believe arguing otherwise could be used to justify the genocide of the Albanians on Kosovo in the late 1990s) believe that Illyrian was not only a satem language, but also a direct ancestor (or at least a language closely related to the direct ancestor) of Albanian. Here is how I would argue against it:
So, you believe that Illyrian, the language spoken in western Balkans and southern Italy in ancient times (the Messapian dialect of Illyrian) was a direct ancestor of the modern-day Albanian language, right? So, one of the first sound changes (if not the very first one) that occurred from Proto-Indo-European to Albanian is the change from 's' to 'gj'. And that change occurred before the satemization -- the 's'-es in Albanian coming from Indo-European 'kj' are unaffected by it. And, while we do not know when centumization and satemization as areal sound changes occurred (or even if they occurred at roughly the same time -- glottochronology of the Celtic languages suggests centumization probably preceded satemization by a millennium), I think we can all agree that, by the time of Ancient Roman historians who recorded the Illyrian toponyms, they were ancient history. But, if so, how come do recorded Illyrian toponyms show the Indo-European 's' intact? If there are any two toponyms that all Illyrologists agree are Illyrian in origin and what they meant, it is Salapia (the ancient name for the salt marsh Salpi in Apulia in southern Italy) and Serapia (the ancient name for the river Bednja). Namely, as far as I know, all the scientists who have studied the problem agree that Salapia was Illyrian for "salty water" coming from *seh2l (salt) and *h2ep (water), and that Serapia was Illyrian for "flowing water", from *ser (to flow) and *h2ep (water). If Illyrian was really the ancestor of the Albanian, how do you explain those toponyms?Now, when it comes to Croatian nationalists, who believe that Illyrian was some generic satem language but not an ancestor of the Albanian, here is how I would respond:
This is a problem that everybody seems to ignore, even if it is, when you think about it, a huge problem. Everybody seems to focus primarily on whether it's possible that Illyrian was a satem language, which is, don't get me wrong, a problem (and I will focus on it later in this web-page), but not as serious as this one. When we are talking about whether a fragmentarily attested language belongs to the same branch of Indo-European as some language with known sound changes, we should first be looking for the traces of the earliest sound changes (in this case, the one with the 's' sound), rather than the sound changes that came later (in this case, the satemization), right? The claim that Illyrian was an ancestor of Albanian suffers from basically the same problems as the claim that Lusitanian was a Celtic language -- the sound changes do not match at all, not even the earliest ones. Making such claims appears to be contradicting one of the basic principles of historical linguistics, namely, that the sound change has no memory. It is as close to a belief in magic as a historical linguistics statement can get.
One of the counter-arguments I've seen repeat on the Internet (Quora...) is the claim that Serapia was named after an Egyptian goddess. Why that is way less likely than Serapia being a transparent compound meaning "flowing water" should be obvious. Similar toponyms appear all across Croatia and elsewhere in the Balkans: both the ancient name for Sremska Mitrovica and the early medieval name for Srima near Šibenik was Sirmium. Sirmium is also widely agreed to mean "to flow" in Illyrian. Though, Serota (the ancient name for Virovitica) is also agreed to have meant "to flow", but, based on the 'o' in the suffix rather than the expected 'a', it is often suggested to be Celtic in origin. Osseriatis (a town on the mouth of the river Vrbas into Sava) contains the element -ser-, but is usually considered not to come from it. And there is nothing close to a consensus on where Serenae (the early medieval name for Sveti Đurađ near Donji Miholjac) comes from.
Look, I know where you are coming from. I understand how somebody can think the toponyms such as the toponym Osseriatis on the mouth of the river Vrbas into Sava (supposedly meaning "lake" and related to Croatian "jezero", as there are many lakes there) and Birziminium (the early medieval name for Podgorica in Montenegro, appearing to come from the famous Indo-European root *bhergjh meaning "fortress", or perhaps from a homonymous or near-homonymous Indo-European word for the birch tree) prove that Illyrian was a satem language. But, first, you need to understand that this etymology of the name Osseriatis is linguistically rather implausible, as the Slavic word "jezero" is almost certainly not Indo-European in origin. Why ignore the more obvious possibility of Osseriatis meaning "place of flowing into" or "river mouth"? That would allow us to explain the geminated 's' as a result of an assimilation from a prefix meaning "into". As for Birziminium, I understand that it sounds compelling at the first sight, but you need to understand first of all that it is only attested in the 9th century, so it being affected (corrupted) by Slavic sound changes (the Second Slavic Palatalization turning 'g' into 'z' in front of 'i') cannot be ruled out, and, second, that those toponyms are, once you take a broader picture, clearly cherry-picked. Here is some evidence that Illyrian was a centum language, of at least the same quality as the evidence usually cited that Illyrian was satem (and the first evidence is of way greater quality):Now, for the Serbian nationalists, many of whom believe Slavs are native to the Balkans:I hope I've at least made you think with this. Like I've said, I know where you are coming from, but I think you are misled. There seems to be only one scientifically rigorous argument (with a p-value) we can make about whether Illyrian was centum or satem, and it argues that it was a centum language.
- The k-r pattern in the Croatian river names. Seven major Croatian rivers have names starting with consonants 'k' and 'r', in that order, some with a vowel between them: Krka, Korana, Krapina, Krbavica, Kravarščica, and two rivers named Karašica. If you assume the Illyrian language was a centum language, there is a relatively obvious explanation for that: that *kurr was Illyrian for "to flow" coming from *kjers (to run). There is no obvious explanation for that pattern if you assume Illyrian was satem. So much so that linguists have to insist it's a coincidence even when shown a relatively-rigorous calculation suggesting its p-value is somewhere between 1/300 and 1/17 (calculated using basic information theory, by the method with birthday paradox and collision entropy). I've published a paper about it in 2022 in Valpovački Godišnjak called Etimologija Karašica, and I've also made a video about it a year ago, which is in the Latin language with English slides. And let me say that it seems, purely from the phonological perspective, that Karašica coming from Illyrian *Kurr-urr-issia (flow - water - suffix), borrowed into Proto-Slavic as *Kъrъrьsьja > *Karrasja (Havlik's Law) > *Karaša (loss of geminates and yod-coalescence) >> Karašica (adding a Slavic suffix), seems to be a lot more plausible than the mainstream etymology of it coming from Turkic "kara sub" meaning "black water". And that there is independent evidence that *kurr really did mean to flow in Illyrian, namely, the fact that there is a verb "kuriti" in the Silba island dialect of Croatian meaning "to flow" (in the phrase "vrime kuri" meaning "time flows"). The k-r pattern in the Croatian river names is not anecdotal evidence, this is evidence with a p-value (the p-value of that pattern is somewhere between 1/300 and 1/17). If you will claim that we should still for some reason expect the patterns such as the k-r pattern in the Croatian river names to occur by chance, you are making an informatical statement that the p-value of that pattern is weak. And that statement should be supported by informatical means. Go ahead, try to support it! I am almost certain that you cannot. And if you will try to support it using some model which supposedly takes the Sonority Sequencing Principle into account, make sure you've studied my visualization of the Sonority Sequencing Principle in Croatian, showing just how complicated that is (naively taken Sonority Sequencing Principle barely predicts the consonant pairs at the beginnings of the words, and it does not predict them at the end of the words correctly at all -- the correlation is positive rather than negative). And the other supposed reasons why my calculations might be wrong are way less plausible. This alone should be enough, in my opinion, to prove to any reasonable person that Illyrian was probably a centum language, but, as if that wasn't enough, I will show some other, mostly anecdotal, evidence.
- Why is the ancient name for Krk, the northernmost big island in the Adriatic sea, "Curicum", if that did not come from the Indo-European *(s)kjeh1weros for northern wind (whence Latin "caurus")? The most compelling counter-argument I heard is that I am ignoring the possibility that the nesonym Curicum comes from the Venetic language. But, if so, why isn't the 'c' in the beginning aspirated? The first 'e' in *(s)kjeh1weros obviously disappeared (zeroth ablaut?), so the *kj and *h1 were right next to each other. Venetic had aspirated stops.
- What did the ancient name for Požega, "Incerum" mean if not "heart of the valley" (*h1eyn-kjer(d))?
- What did "Cibalae", the ancient name for Vinkovci, mean if not "strong house" (*kjey meaning "house" and *bel meaning "strong")? Connecting it to Greek word for "head", κεφᾰλή, is incompatible with modern Indo-European linguistics, modern Indo-Europeanists almost unanimously agree the 'k' in the beginning of the Greek word is due to the Grassmann's Law, and that it comes from *gh.
- Look into the Messapic inscriptions. If there is one phrase that every scientist who has studied it agrees on what it meant, it's that "klauhi Zis" meant "Listen, God". But, if so, klauhi has to come from *kjlew (to hear, to listen), does it not? OK, I realize that in some rare satem languages *kj depalatalizes before *l and *r, but not in nearly most of them.
Why do you think almost no ancient Balkan toponyms make sense as Slavic? The only ancient Balkan toponym that I am aware of that looks plausibly-Slavic is Tergestica (you can claim it's related to Slavic "trg" meaning street crossing or square in a city). And perhaps the toponym Tininium, from "tanati" meaning "to cut". What about all the others? What did they mean? If they were Slavic, it should be relatively easy to tell what they meant, right?Anyway, I hope I've helped somebody fight against those nationalistic forces. Though, in my experience, your arguments will mostly be not "refuted", even with blatant ad-hoc hypotheses, but rejected.
And why aren't all changes from Balto-Slavic to Slavic visible in the toponyms? Why didn't "Curicum" change into something like *Krac (following the progressive palatalization, the Third Slavic Palatalization), but to Krk? The Vulgar Latin syncopa of the middle unstressed vowels? How could it possibly affect the name Curicum if it were Slavic to begin with? What are you imagining happened? Borrowing the toponyms back-and-forth between Romance and Slavic until they don't even resemble the original? Please, apply the Occam's Razor.
Though, to be honest, the claims typically made by Croatian nationalists seem to make at least some degree of sense, unlike the statements made by Serbian nationalists and Albanian nationalists. I can at least understand Croatian nationalists, whereas I have trouble conceiving how somebody might believe what other nationalists here on the Balkans claim to believe. Have they even looked at the ancient toponyms?
UPDATE on 24/08/2026: Look, in case it is not obvious, any honest linguist will admit that the study of the toponyms is a very soft science. In other parts of historical linguistics, you don't even need to calculate the p-values, because the patterns are so strong that the mere suggestion that they are coincidental must appear absurd. Take, for instance, the evidence for the High German Consonant Shift, such as English 'd' regularly corresponding to German 't': "day"-"Tag", "cold"-"kalt", "old"-"alt", "dream"-"Traum", "red"-"rot", "do"-"tun", "death"-"Tod", "dale"-"Tal", "ride"-"reiten", "blood"-"Blut", "dew"-"Tau", "deep"-"tief", "good"-"gut", "god"-"Gott", "bed"-"Bett", "drive"-"treiben", "drink"-"trinken", "drop"-"Tropfen", "dove"-"Taube", "deaf"-"taub", "daughter"-"Tochter", "bread"-"Brot", "door"-"Tür", "word"-"Wort", "world"-"Welt", "shoulder"-"Schulter", "hard"-"hart", "loud"-"laut"... The mere suggestion that it's a coincidence appears absurd. The p-value is so low it's not even worth thinking about. But, in the study of the toponyms, the most striking pattern in the Croatian toponyms, that is the k-r pattern in the Croatian river names, has a p-value of somewhere between 1/300 and 1/17. And most of the time you do not even have a weak p-value, but you have to rely on anecdotes of the worst kind. And people seem to intuitively realize that, without having studied historical linguistics.
Low-hanging fruit in linguistics is gone. Has been since the 19th century. These days, if you want to possibly discover something new, you have to rely on statistics and bad anecdotes. In some sense, the same thing happened in physics. Ernest Rutherford said 100 years ago that if your experiment requires statistics, you ought to have done a better experiment. These days, all the ideas for an experiment in physics without statistics have been tried, so physics these days heavily relies on statistics. Except that, in physics, the p-values of those high-hanging fruit are in the neighbourhood of 1 in a million, whereas, in historical linguistics, the p-values of the high-hanging fruit is such that it's considered to be right on the margin of error even by the social science standards. It's not at all obvious what the right thing to do is. Leave it to nationalists who apparently never bother to do any mathematics to get closer to the truth?