The problematic of substrates – A case study of Iberia


The following post is written with two main purposes: The first one is to explain some of the problematic when dealing with substrates in a way that is accessible for anyone to understand and get a better perspective about this complicated subject. The second one is to take a look at how ancient DNA (aDNA) can help in solving part of this problematic (in this sense, it’s a first addition to the “introduction” I wrote a while back).

For these purposes I’ll be looking at the subtrates in Iberia. There are two main reasons for choosing Iberia. The first one is that being part of the peripheral area of prehistorical West Eurasia is offers a relatively simple and straightforward population history, but unlike the rest of the periphery is also offers relatively early information about the languages spoken there (going back to the Iron Age). The second reason is that the latest paper on the subject it’s one of the best sampled areas we have today when it comes to aDNA. A third, less important reason, is my own knowledge of its history and language, which makes is easier for me to write about it than it would be to write about any other place.

I will try to keep things as concise as possible, with a few short linguistic notes provided by Kristiina (see acknowledgements) marked with numbers (in red: note¹, note², etc… to avoid confusion with any phonetic symbol) and some other more extensive ones to add clarifications and additional thoughts marked with red asterisks (note*, note**, etc…, idem), the latter of which should not distract from the main text.


Basque and Iberian: and overview of their relationship

I’ll start with a short summary about the subject of the relationship between Basque and Iberian languages, since most of the literature about is available in Spanish and therefor less accessible for many readers.

The history of Vascoiberismo (the hypothesis of these two languages being related) goes very far back in time, but it’s not until quite recently that the hypothesis has become widely accepted even by previously sceptic linguists. The main reason for this broader consensus has been the research on the numerals by reconstructing the proto-Basque ones and finding close similarities (far beyond coincidence) in the Iberian inscriptions, and in places (within those inscriptions) where one could (or would) expect to find a numeral.

Iberian and Basque numerals, followed by the Spanish translation. From Wikipedia (in Spanish)

Together with the same system for constructing the higher numbers, it has left few sceptics when it comes to accept the close relationship between both numeral systems. But the next questions was if this was a case of a wholesale borrowing of the numeral system from one language by the other or if it meant a genetic relationship (same origin) of both languages. Here is where there has been more debate, but in most cases the genetic relationship has been the preferred explanation*.

Moreover, given the aDNA that we’ve been getting in the last few years, the only plausible sources for these languages are two: the Early European Farmers (EEF) from Anatolia and the Bell Beaker Culture (BBC, ultimately from the steppe). With the relative proximity of East (including NE) Iberia and SW France (Aquitaine) – even with the Pyrenees as a barrier to strictly direct contact – and the very close genetic relationship of the involved populations, it becomes very easy to think that they are genetically related. I would go as far as to say it’s even necessary, since if they weren’t, that would force one to come from EEF and the other from BBC, preventing the possibility of the BBC having brought IE languages to Western Europe**.

However, if we accept a genetic relationship between the two languages, the following question might come to any readers mind: if these two languages were indeed closely related, why can’t we understand the Iberian inscriptions still? And that’s an important and very legitimate question, to which I can offer the following answer: The main reason why we can’t understand the Iberian inscriptions by using Basque as a reference is that modern Basque is a language that has survived rather miraculously into the 21st century. It’s not long ago that it was an endangered language mostly spoken in rural areas, some isolated from others, with different dialects not always easily intelligible among them. So it’s a very “drifted” language, only revived and standardized quite recently (as Euskera batúa, or Standard Basque), and with a limited native vocabulary (the larger part being borrowed from its neighbouring IE languages). So it’s not surprising that it’s still really hard to understand Iberian, even in the -more likely than not- case of it being genetically related to Basque. Another reason is probably the nature of the Iberian inscriptions that is mentioned further down.


A hypothesis about non-IE preroman Iberia

I will call this a hypothesis because it is not something completely proven. However, with what we know today it’s a solid hypothesis and not just pure speculation.

The main point of it is that the non-IE speaking areas of Iberia were first Indo-Europeanised by the Romans.

Given the population movements that we know about, it’s difficult to reconcile the idea that those areas were once IE and that later they became non-IE***.

So what follows will be based on this hypothesis, and it will be future research that should confirm or deny its veracity (though even if this hypothesis is falsified, it won’t necessarily invalidate many of the concepts I’ll comment about. It would, however, invalidate the specific examples used here based on it).


What is Indo-European?

The first thing we should clarify is what do we mean by Indo-European. It might seem a trivial question, but defining the meaning in a specific way is important to avoid misunderstandings while reading this text. So basically there could be two different definitions of IE:

  1. The language spoken by a population from a specific place at a specific time, before the dispersal/expansion and subsequent diversification of the language. I will refer to this a Proto-Indo-European (PIE)
  2. The language that expanded after that initial stage, succeeding and, to different degrees, absorbing many other now dead languages encountered along the way, most of which are completely unknown to us (and therefor have only reached us through the IE languages that absorbed them). I will refer to this as Indo-Europeanised (IE-ised).

While the second definition might be useful for some other fields, it’s clearly not too useful when it comes to determine the origin of the IE language (PIE) and its dispersal. So it’s important to keep in mind the first definition as the one that we’re going to need for this kind of research.


A preliminary look at the substrates in Iberia

There has been a traditional divide in the Iberian toponymic areas based on the Celtic –briga and the Iberian Il-/Ili-/Ilti- respectively which matched the historical IE and non-IE speaking areas:

-Briga Toponyms in the Iberian Peninsula. García Alonso 2006.

However, the situation became more complicated as more research was done. A big turning point was Francisco Villar’s work on toponymy (emphasis mine):

Recently, however, Francisco Villar (2000) has offered a new, somewhat revolutionary approach according to which there might have been a very old Indo-European layer that was particularly strong in the south. In my own work (see for instance 2003) I have also interpreted place-names as Indo-European that are found in regions generally considered non-Indo-European.

García Alonso, J.L. 2006.  -Briga Toponyms in the Iberian Peninsula

This research brought back the Paleolithic Continuity Theory (PCT) which argued that IE languages were very old in Iberia (not exclusively in Iberia, needless to say), probably going back to the Mesolithic, and that non-IE ones were only a much later arrival. Now, this may sound very outdated in the aDNA era, but that’s just because of how fast things have moved in the last 5 years. In any case, this particular theory (the PCT) doesn’t really matter for the analysis of the substrates done by several authors.

To briefly show the results that this line of research has given, I’ll refer to a study which is in English and freely available (follow the link), by Leonard A. Curchin: NAMING THE PROVINCIAL LANDSCAPE: SETTLEMENT AND TOPONYMY IN ANCIENT CATALUNYA, 2006. ****

And I’ll start by just showing the statistical summary of the 97 toponyms (including hydronyms) analysed:

  • Iberian names: 10 (10% of total)
  • Indo-European names: 49 (51%)
  • Greek names: 10 (10%)
  • Latin names: 22 (23%)
  • Unclear: 6 (6%)

Important note: Indo-European names refer to those that cannot be associated to any specific IE branch (including Celtic or para-Celtic), so they are just of generic/unknown IE origin.

So now one might wonder why is this. Why would Catalonia (in this specific case, but it’s a similar situation with the rest of eastern and southern Iberia) have only 10% Iberian toponyms? And why 51% IE ones, when it’s so unlikely that it was ever an IE speaking area before the Romans arrived? And which IE language would that be, unrelated to Celtic or any other specific branch, some sort of PIE?

Before moving onto those questions, let’s ask ourselves once again:


What is Indo-European? (Part II)

As already mentioned in the first part above, what we are interested in is the language spoken by the original population before it expanded. That is, PIE. So let me quote here something that Kristiina, a regular commentator well knows by the readers here, has mentioned a few times:

The amount of roots having more or less the same meaning is very high in the presumed IE vocabulary. In Encyclopedia of Indo-European culture by Mallory and Adams there are 18 roots for ’bend’, twelve for ’bind’, seven for ’branch’, twelve for ’burn’, six for ’fear’, seven for ’field’, seven for ’goat’, seven for ’grain’’, eight for ’grow’, five for ’axe’, five for ’water’ just to mention a few examples. Modern developed languages usually have only one or two.

So it seems like a very pertinent question to ask if all those roots belong to what I’m referring to as PIE or to what I’m referring to as IE-ised. And while I’m sure the answer is more or less obvious to any reader, let’s consider a few issues here.

First, what is it required for a word or root to be considered PIE? Is there a standard widely accepted by all the experts in the field to accept or reject a proposal for considering a word as PIE? As far as I know there isn’t such thing (and knowing how linguists tend to go their own ways rather than cooperate in solving problems, it’s hardly surprising if there isn’t any standard for such a basic and important thing). And if there is, what is it? Is it enough if at least two words in two different languages can be reconstructed following the regular sound changes in each to the same hypothetical PIE root? And if two languages are enough, can they be any two languages (say, German and Russian or Latin and Lepontic)?¹

So let me do a little experiment by proposing this simple hypothetical standard:

  • A word/root can be considered PIE if it is attested (with regular reconstructions from the putative PIE root) in at least one language in each of these two groups (Group A and group B), and being attested in a minimum of three of them in total.
  • Group A: Greek (ancient, here and onwards unless specified as ‘modern’), Latin and Celtic.
  • Group B: Anatolian, Indo-Iranian (again, ancient unless specified otherwise) and Tocharian.

The reasoning here is that the languages in these two groups are less likely to share post IE expansion vocabulary, because they are ancient ones (even if Tocharian is attested quite late, but its isolation helps to keep the chances of later borrowing from Group A languages) and there’s not much evidence of later contacts (though the case of Indo-Iranian might be debatable until we don’t figure out its specific origin).

With this hypothetical standard in place, how would some of the presumed IE roots fair when measured against it? Well, this is a labour that I can’t do here in any great extent, so I’ll just look at a few cases based on the following resource which is easy and accessible to everyone:

Holm, Hans J. (2016, in progress): >Indo-European Universal Concepts List (M. Swadesh’s 1971=final meanings). With “unmarked” translations in 17 representative extinct and modern IE languages. From

  • Water: 7 different roots are mentioned. The first one², wód-r̥ {sing}, auwed-(r)– < h₂wédōr {koll}, would comply with the requirements mentioned above, as it is attested (from that list of 17 languages) in Russian, Lithuanian, Old Icelandic, Norwegian (Bokmål), Old Irish, Modern Irish, Latin, Albanian, Greek, Hittite and Sanskrit. The second one, h₂ekweh₂– ‘(running) water’, only appears in Italian and Latin (aqua, also present in other Romance languages) in that list, so it wouldn’t comply. From the other 5, if we accept what are marked as “deviant meaning”, one of them, ap– < eh₂p– ‘water, river’³, appears in Greek, Hittite, Tocharian B, Avestan and Sanskrit, so it would qualify too. Overall, 2 out of 7.
  • Sun: 6 roots are listed. One of them would qualify (1 out of 6)
  • Leaf: 1 out of 12 listed would qualify
  • Woman: 1 out of 7.
  • White: 1 out of 12.
  • Knee: 1 out of 3
  • Moon: 1 out of 8
  • Green: 1 out of 8
  • Man: 2 out of 10
  • Liver: 1 out of 9

I certainly can’t certify the accuracy of either the list provided above (it’s marked as a “work in progress, for lexicostatistical purpose only”), nor the validity of either the example standard I proposed or the precise counting of how many roots conform with it if confronted with additional resources. However, I think that the picture is clear enough to say that those details would only increase the accuracy (probably adding a small number of words to those that qualify) but wouldn’t change the outcome overall.

So to reiterate: PIE is not the same as IE-ised. And while setting some standard can mean that we leave a few real PIE roots out because we can’t prove them reliably enough to be PIE (that’s always going to be the case, not just with this specific subject), it’s still a better situation than throwing everything into the (same) basket.


Back to the substrates

So with this in mind let’s get back to the substrates in Catalonia and hopefully we can now understand better the reasons for those statistics. But first let me point out an additional methodological limitation which is mentioned in the paper (emphasis mine):

Such an inquiry [about the substrates] is not without difficulties. For one thing, we have only a limited knowledge of the vocabulary of Iberian, which is not related to any other known language, and can only identify toponyms as “Iberian” if one or more of the name elements appear in Iberian inscriptions (which consist largely of personal names).

This makes it probably easier to understand why we can only identify 10 toponyms as Iberian. You simply can’t say that something is Iberian unless you have an Iberian inscription with an equivalent name (root, prefix, suffix,…)to prove it. Given the very limited Iberian corpus, you can expect to find very few coincidences.

How about Indo-European? Well, the Indo-European corpus is enormous, from ancient to modern languages. With the problem of considering everything as IE (PIE and IE-ised), the amount of presumed IE vocabulary is so extensive that it’s difficult not to find coincidences, specially if you don’t restrict yourself to any specific language(s) known to have been spoken in the area or surroundings. This results in a very strong bias that only has gotten worse with time:

In fact the situation is much more complicated. In recent years, several Catalunyan place-names previously assumed to be Iberian have been reinterpreted as Indo-European by F. Villar (2000), raising questions about early Indo-European settlement in this supposedly non-Indo-European zone.

The result of this methodology is the one we already know from the aboe statistics: 51% Indo-European toponyms (but of an unknown branch) vs. 10% Iberian and even 23% Latin. When you consider that this is in an area where we know securely that non-IE was spoken and have attested and readable inscription of it, It’s hardly surprising the difficulty in finding non-IE substrates in places where there is not even an attested language before Indo-Europeanisation (like the rest of Western Europe).

Now, for the same of completeness, let’s take a quick look at the toponyms themselves. The paper lists them in alphabetical order, starting with river names. So to avoid any case of cherry picking, I’ll follow this same order:

Alba (Pliny III, 22). This clearly comes from the IE hydronym *albho– (IEW 30). Parallels include the river Albis (Elbe) in Germany (Tac. Germ. 41) and the river Albe, Albas or Albula in Italy, an early name of the Tiber (Pliny III, 53; Steph. Byz. s.v. Albas).

The IE root *albho means ‘white’, which is unrelated to water, flow, etc… We have many red rivers, which have an evident explanation due to the colour of their waters, be it due to oxides or anything else. So I guess there’s something (freezing in winter, snow?) about these rivers that can justify them being called ‘white’? But more importantly, is the root PIE or IE-ised? Checking again the aforementioned list, it appears in Russian, Lithuanian, Latin and maybe Greek (Ἀλουίων, Albion?) and Armenian (aɫauni, ‘pidgeon’, ‘dove’). Apparently there’s also the Hittite (alpas) meaning ‘cloud’, and the Sanskrit ऋभु (ṛbhú) meaning ‘skillful’, ‘expert’, ‘master’.

Anystus (Avienus 547). While Pokorny saw this name as Illyrian, comparing the Bulgarian river Andzista, Schulten more reasonably interprets it as Greek anystos [ᾰ̓νῠστός] “practical”; thus, “the useful (river)”. However, the possibility remains that it is a hellenized transliteration of an indigenous name: cf. the river Anisus (modern Enns) in Noricum, which Anreiter et al. relate, not very convincingly, to a supposed IE *on– with hydronymic suffix *-is-.

Not sure if this counts as Greek or Unclear. Probably the latter, so not much to comment.

Arnus or Arnum (Pliny III, 22). Pliny gives the name in the accusative, which leaves the gender uncertain. Various hypotheses have been advanced: Pokorny made it Illyrian, Garvens Basque, while Jacob derived it from a supposed theonym Airo. Its true root is surely the IE hydronym *ar– with secondary suffix –no-. Cf. the Italian river Arnus (modern Arno).

If its true root is surely the IE hydronym *ar I guess there’s not much to discuss either.

Baetulo (Mela II, 89). See below on the city of the same name.

Baetulo (Mela II, 90; Pliny III, 22; Ptol. II, 6, 18). Like Baecula, this name could be formed from IE *gʷhei-. However, the word baites which appears repeatedly in Iberian inscriptions on lead shows the possibility of an Iberian origin. The suffix –ulo is a latinized form, as shown by the orthography baitolo on the town’s pre-Latin coinage; cf. the classical spelling Castulo for indigenous kastilo in Oretania.

Let me complement it with the referenced Baecula which appears just above it in the paper:

Baecula (Pliny III, 23; Ptol. II, 6, 69). Villar derives the element bai– in various Hispanic toponyms from IE *gʷhei– “to shine, be white” (IEW 488-489), though the Iberian personal name baikaŕ may argue for an Iberian root *bai– or *baik-. In any case, there is no guarantee that all bai– toponyms (e.g. Baetis, Baedunia, Baesucci, Baelo) come from the same root. Polybius (X, 38, 7) mentions another Baecula in Bastetania.

So again I’m not sure if these two count as Indo-European or as Iberian or as ‘Unclear’. The proposed IE etymology doesn’t look very solid, if one has to be honest. Once you allow for such speculations you’re in a very muddy terrain. The Iberian one is not much more solid, but we have to realize the difficulty of finding such coincidences in the limited Iberian inscriptions. And yet they do look more similar, I’d say?

I won’t go on, since the paper is available for anyone interested in it. This was just to show how it works to try to figure out the true etymology of an ancient toponym and how difficult it actually is.



Since I think the above should be self explanatory it seems unnecessary to summarize it here. Instead, I’ll go back to the proposed hypothesis about the non-IE speaking areas of Iberia having been Indo-Europeanised only with the arrival of the Romans. If this is true (and it most likely is), then this is an opportunity to rethink some paradigms. Except for the 10% Greek toponyms and the 23% Latin ones, everything else would be non-IE. What should we do with that ~50% presumed IE substrate in these areas? For a start, we could use it to clean up roots we consider IE and probably shouldn’t. As a real world vocabulary attested before any IE speaking population set foot on the areas, that would be a pretty strong argument for disproving the presumed IE etymology of them (though this does not preclude the necessity of having a standard for what can be considered PIE, as that would make a big difference too). Equally important is the affiliation of such substrate. And here again we’d have to favour quite strongly an Iberian origin of most of it (though there’s the risk of ascribing pre-Iberian substrate to the Iberian one – a risk that is unavoidable but not that big considering the possibilities it opens up). Once we have a much larger non-IE lexicon from Iberia, we could go and compare that to the rest of Western Europe (to start with) and see what happens. It might be the way forward to finally shed some light on the obscure European prehistorical linguistic situation.

aDNA is speaking to us. We might still not fully understand the message, but we have to try so that progress can be made. This post is just an amateur attempt to show its possibilities. It’s now the experts’ job to make good use of all the new information we’re getting on a monthly basis and start building new models, which I hope will happen soon so we can all learn from them.



Special thanks to Kristiina for her help with this post, for sharing resources, providing linguistic notes and valuable feedback.


Linguistic notes:

1 – As a reference, Kristiina points out to me that the criteria for a Proto-Uralic for being considered as such is not very clearly defined either. She recalls the minimum requirement being that the cognate word exists in an Eastern Uralic language (Ugric or Samoyedic) and a Western Uralic language (Permic, Volgaic, Baltic-Finnic and Saami). I’ll leave it to her to elaborate further on this problematic at some point.

2 – However, she points out that the Pan-Uralic word for water is almost identical and shows regular sound changes (link)

3 – Though for this one it could be relevant to mention the Sumerian ‘ab‘ (‘sea’) or the Basque ‘*ɦibai‘ (‘river’).

4 – This root can hardly be considered the common word for ’white’ and ’light’ in IE languages. The basic meaning appears only in Latin, Umbrian and Greek. The Baltic and Slavic cognates mean ’lead’. The Celtic distribution is very scarce, the root ‘elbid‘ is only found in ancient Welsh and in no other Celtic language. It could also be a substrate word

5 – It should be noted, however, that Basque possesses the roots ur ’water’, jario ’flow’ and gernu ‘urine’.


Extended notes:

* For example, Francisco Villar, a prestigious Spanish linguist specialised in IE and strong proponent of a some sort of “everything IE” (more about it in the main text), changed his sceptic view about Vascoiberismo due to the recent advances made in that field:

En los últimos años se han producido ciertos resultados de la investigación en el ámbito de los numerales que han llevado la cuestión a un terreno más firme. Con los numerales incorporados al elenco de coincidencias, el parentesco entre ibero y euskera me parece ya la única hipótesis sostenible.  La amplísima coincidencia en el sistema de los numerales señalada primero por E. Orduña (2005, 2006, 2011) y ampliada y consolidada por J. Ferrer i Jané (2009), especialmente (aunque no sólo) en los 10 primeros numerales, descarta en mi opinión cualquier explicación por préstamo. De hecho encuentro entre los numerales ibéricos y los euskeras no menos ni peores que las que en realidad se dan entre las lenguas indoeuropeas históricas.

Por añadidura, la hipótesis de los préstamos del íbero al vasco, aparte de su inviabilidad para el sistema de numerales, siempre ha tenido el punto débil de la falta de evidencias en favor de un contacto real entre ambos ámbitos incluso desde el punto de vista geográfico.

Villar, F. 2014. Indoeuropeos, iberos, vascos y otros parientes

My translation [and emphasis]:

In recent years there have been some results from the research in the field of numerals that have brought the question to a firmer ground. With the numerals added to the list of coincidences, the relationship between Iberian and Basque seems to me the only sustainable hypothesis. The extremely broad similarity in the numeral system pointed out first by E. Orduña (2005, 2006, 2011) and extended and consolidated by J. Ferrer i Jané (2009), specially (but not only) in the first 10 numerals, discards in my opinion any explanation by loan. In fact, I find between the Iberian and Basque numerals no less nor worse [coincidences] than the ones that actually exist between historical Indoeuropean languages.

Moreover, the hypothesis of a loan from Iberian to Basque, aside from its infeasibility for the numeral system, has always had the weakness of the lack of evidence in favour of a real contact between both cultures even from a geographical point of view.

The infeasibility referred to in the emphasised text refers to the extremely rare nature of a wholesale borrowing of the numeral system.

You can find more (in Spanish) by reading the authors mentioned in the above quote. For example:

E. Orduña, 2013. Los numerales ibéricos y el vascoiberismo:

Abstract: In this work we examine the implications of the existence of a great coincidence between the Iberian lexical numerals and the Basque ones, thus applying this proposal to the Iberian lead of Ensérune, where we can observe that the possible lexical and morphological coincidences between both languages are not limited to the numeral system. Besides, some possible loan words from Greek to Iberian are proposed, and some aspects of the structure of the Iberian numeral system are revised.

Regarding the critics, I can mention two authors. Javier de Hoz has admitted the validity of some of the correspondences in the numerals, but his own hypothesis about Iberian was that it was a lingua franca (due to it being the first one written and acquiring some prestige) and it was only native to the south eastern part of Iberia, so its use throughout a much larger extension (esp. the NE) was just for trading purposes. So he could obviously not accept a genetic relationship between Basque/Aquitanian and Iberian without first dropping his preferred hypothesis. I can’t discuss here the problems of his hypothesis about Iberian as a Lingua Franca, but anyone interested can check out this paper (in Spanish). Suffice to say that no one has really accepted it as valid except Joseba Lakarra, the other author who rejects the relationship between Basque and Iberian. His main argument (apart from supporting the one from Javier de Hoz) is that the reconstructed proto-Basque numbers proposed that match the Iberian ones are not a valid reconstructions according to his own ones. However, Orduña has argued that those former reconstructions match the proto-Basque most widely accepted and with a more secure chronology (Koldo Mitxelena’s), while Lakarra’s own reconstruction has a vague chronology and is much more insecure.

Francisco Villar (whom I’m quoting here because he’s a prestigious linguist, but mostly because of two other reasons already mentioned, namely, his previous scepticism about VascoIberismo and the fact that he’s neither dedicated to the study of Basque or Iberian, but an Indo-Europeanist who has no dog in this fight) has sided with Orduña in this last criticism too:

Por otra parte, la evidencia de los numerales me parece tan consistentemente apoyada en el Método Comparativo que me atrevería a afirmar que si resulta incompatible con el paleo-euskera reconstruido hay que proceder a corregir esa reconstrucción, que es hipotética y perfectible, como toda reconstrucción.

My translation:

Furthermore, the evidence of the numerals seems to me so consistently supported in the Comparative Method that I would dare to say that if it turns out incompatible with the reconstructed paleo-Basque one should proceed to correct that reconstruction, which is always hypothetical and perfectible, as every reconstruction.

Overall, while I can’t have a strong opinion about such technical debate, nor any preference over the relationship (or lack thereof) between Basque and Iberian, I do think it’s by far the most simple and consistent explanation. And as I said before (more about it in the second note) probably a necessary one.


** The possibility of Bell Beakers bringing IE languages to Western Europe, given the data that we have already for a while (and now with more detail thanks to the latest study recently published: Olalde et al. 2019) is a low probability one. However, it’s still possible and I personally prefer to leave that possibility open until we get a more detailed information about the period between 1500-700 BCE in Iberia, which is currently poorly sampled (plus some extra details from the transitional period ca. 2400-2200 BC).

The situation with the current data would be like this: Non-IE languages have 3 possible sources:

  • Bell Beakers
  • EEFs
  • WHGs

While I would caution against the necessity of language shift with large population replacement, especially in the male lineages (as FrankN already wrote a while back), I’d also say that good reasons are needed to prevent language shift from happening in such scenarios. One thing usually mentioned when it comes to language shift is that language is imposed by the “winners” (be them a small elite or a larger part of the population). This is not completely accurate. Languages are rarely imposed. The main driving factor in language shift is convenience. People, whether elite or commoners, majority or minority, change their language to another one when, and if, they see some benefit (for their own interests) in doing so. The exceptions to this rule are mostly due to some sort of “nationalism” (in older times better referred to as “strong ethnic identity” – usually as a reaction to what is felt as an aggression from a different ethnic group) where ideological reasons would be placed above the practical ones.

In the case of this large population replacement throughout Western Europe during teh transition from the Copper Age to the Early Bronze Age, convenience does not seem like it might have been a strong factor in the incoming Bell Beakers to shift from their language to those of the previous populations that they were largely replacing. Therefor, the probabilities of those non-IE languages coming from the Bell Beaker side is very clearly higher than the other two (EEFs and WHG, the latter having close to zero probabilities).

When we get more samples from the mentioned periods, if there is any significant surprise then things could change and push the probabilities of Ibero-Vasconic coming from EEF. Difficult to say how much without knowing which surprises those may be. If, on the contrary, there are no surprises, then those chances would go further down, leaving Bell Beakers as the only realistic option.

One problem not mentioned yet is the relationship between Tartessian and Iberian. The reason is that there’s no real answer to it: whether they are related or not is unknown due to the very poor knowledge of the Tartessian language. If they could be proved to be non-related, then that would leave us with the only possibility of assigning Tartessian to that EEF population and Ibero-Vasconic to the Bell Beakers. But that’s really just speculation. It’s much more likely that Tartessian is actually related to Ibero-Vasconic, and that would still allow for the possibility of Bell Beakers to having brought IE languages to Western Europe (if we assign the non-IE ones to EEFs, that is). Those IE languages, in turn, would have gone extinct by the Iron Age without us having any notice about them, which requires a selective replacement of them by the para-Celtic and, specially, Celtic expansions. That is, these latter expansions would have completely replaced the IE languages brought by Bell Beakers to Western Europe, but left the non-IE ones originally from the EEFs untouched – once again a very low probability scenario.

Lastly, I will comment on another issue: the language diversity of the two main populations involved (EEFs and BBs, leaving WHGs out of the picture).

The case of EEFs is harder to analyse due to its depth in time. The idea that this population was quite homogeneous at the start of the European colonization seems quite solidly based. However, as they moved slowly throughout Europe, each group quickly lost contact with others due to the low mobility. We should expect that the Danubian and Cardial expansions diverged (linguistically) from each other quite significantly during the subsequent two millennia. The diversification in Western Europe was probably not as high, due to the later arrival and apparently being mostly of Cardial origin. In trying to quantify the degree of divergence we could take into account a couple of factors: Low mobility and isolation would increase divergence rate when compared to populations keeping closer contacts*****. But then not as much as interaction with native populations would do. And this lack of strong external influence would clearly slow down divergence when compared to scenarios with stronger interactions with local populations. Overall, I’d say that the languages of EEFs throughout Europe (maybe leaving aside complicated and less sampled areas like SE Europe), might have been in the range of Indo-European languages. That is, mutually unintelligible in most cases (geographically conditioned). Just like a Spanish speaker cannot communicate at all with a German speaker (or Russian, or Greek, or Armenian, or Hindi speaker), populations across Europe would probably be in a similar situation. Still, though, languages would be theoretically related if they could be studied by linguists.

When it comes to the Bell Beaker Culture, again we should assume an homogeneous population at the beginning of their expansion. Higher mobility would have contributed to lower divergence after the expansion. But interactions with local populations would have clearly accelerated it (the exception might be the British Isles, where little interaction seems to have taken place. But in Central Europe, France, Iberia or Italy the interaction and influence from locals would have been greater). The idea of the existence of a pre-Celtic language ca. 2500 BCE in Central Europe and the maintenance of a sort of language continuum throughout the subsequent 2500 years (something like the Roman Empire keeping Latin as a stable language, but for much longer – until the Romans themselves disrupted it) is completely unsustainable, though. The presumed IE substrate found throughout Western Europe (IE, but non-Celtic) would just add to the impossibility of such scenario.


*** Such scenario is clearly problematic. For a start, it lacks any sort of evidence, be it archaeological or genetic, and no one has ever postulated such hypothesis which means it does not have any support from any any point of view. But leaving that aside, we can try to check if it might have been possible in some way given that here we are interested in language, and that’s something that we have to analyse anew, with the latest data we have.

The possible scenarios that would be needed to justify the presumed linguistic substrates (shown in the main text), would be something like:

  • EEFs bringing some early form of IE language with the arrival of the Neolithic to Western Europe. Then we would have the Bell Beakers from Central Europe (ultimately from the steppe) coming in and replacing those early IE languages with an Ibero-Vasconic ones, but nevertheless, and in spite of the language shift being accompanied by a large population replacement, leaving most of the place names intact, with only a minor impact in the subsequent 2000+ years.
  • EEFs having brought non-IE languages and then Bell Beakers having replaced them with an early form of IE throughout Iberia (and Western Europe as a whole), and with it, replacing most of the place names, as one would expect. However, at a later point (800-700 BCE the latest, but excluding significantly earlier dates), a non-IE language expanding from “somewhere” all along Southern and Eastern Iberia, and reaching Aquitaine in South Western France.

I won’t extend in explaining the extremely unlikeliness of the first scenario because I’m sure that no one will agree with it. So let’s look at the second one to see how likely that can be.

I guess that the survival of some non-IE language in Iberia in the case of Bell Beakers bringing IE languages, with them and replacing most of the non-IE ones, is something perfectly possible. Some sub population could have adopted the local language for one reason or another. One problem though, is that after that, we need such population to keep that non-IE language for the subsequent ~1400 years in spite of them being surrounded by IE speakers (who in their surroundings would have spoken the same IE language and would have been able to communicate easily). This scenario requires a geographical (more likely) or some sort of ideological (less likely) isolation of the non-IE speaking population. Genetically, we could have them as either clearly shifted to the EEFs (low impact from the R1b/steppe Bell Beakers), or as just identical to the rest of the population (but then isolated and inbred for those ~1400 years, making them at least partly distinctive). Then it would be required for that population to have expanded -at least linguistically- during the first third of the 1st mill. BCE throughout the areas mentioned above.

While we still don’t have the kind of fine grained sampling required to judge in all its fairness such scenario from a genetic point of view, what we have so far does not suggest it in any way. We should be talking more about a cultural-linguistic expansion, something for which there isn’t any specific archaeological evidence either. So this would be a case of an “invisible” (genetically and archaeologically) linguistic expansion from an isolated population suddenly leaving it’s longstanding situation to somehow cause a linguistic shift throughout a vast, well connected (geographically and culturally – possibly linguistically), well populated area like the Mediterranean coasts of Iberia, and even reaching the quite less connected (geographically and culturally) Atlantic coast of Southern France.

Possible? Yes, but clearly not very likely. Probably below the realistic threshold already. And that’s without even factoring in the last step needed to complete this scenario, which is the selective replacement of old IE languages from the Bell Beakers by para-Celtic and Celtic expansions leaving non-IE ones intact, already mentioned in the previous note. And to add one more problem, with this late expansion of an IE language, it would seem unjustified that the Iberian and Aquitanian languages would have been called by different names, since they should have been very clearly related (just like the rest of Iberian ones). All of which makes an already unrealistic scenario become basically impossible.


**** This study is specifically about Catalonia, so some may wonder if the results can be explained by the Urnfield Culture influence in that area. But no, not really. In short:

  • The Urnfield influence in the area is real, but it was a relatively short phase. Clearly not long enough to explain the results.
  • The nature of the influence is mostly seen as a process where the Urnfield newcomers must have blended with the locals, resulting in an uninterrupted transition to the period of the Iberian Culture. If Urnfield people came speaking an IE language, it seems like it didn’t have any continuity and probably they adopted the local Iberian language early after their arrival.
  • If Urnfield was an IE speaking culture, there’s ample agreement in that it must have been some sort of pre-Celtic, which would explain some para-Celtic linguistic influence in Catalonia. But yet such influence doesn’t show up in the study (nor Celtic itself) instead being an unidentified, generic IE substrate we would be dealing with.
  • Whatever the case about Urnfield and Catalonia, we’re not talking about a phenomenon seen in Catalonia specifically, but also in the rest of non-IE speaking areas. So Urnfield cannot really explain anything in Catalonia, but even less so outside Catalonia (so it’s really irrelevant for the bigger picture).


***** I will note that Kristiina mentions the increase in WHG admixture seen in MN Iberia (a phenomenon, the increased WHG admixture into farmer communities that correlates with distance -from Anatolia- and time throughout Europe). Indeed, the possible linguistic consequences of these interactions will have to be investigated in the future. Unfortunately, right now it’s still difficult to say much without having further details about the nature of the interactions, how did the admixture occurred, how did male lineages from WHG come to replace most of the Anatolian ones, etc… For this, we’ll need even more dense sampling with lucky finds and in those use isotopic values to get more info too (we actually have one of this lucky finds since early on, the Hungarian sample from a farming context who turned genetically to be a WHG. It would be interesting to have isotopic values from this sample, code name: KO1, ID: I4971, to know if he grew up in the same place as the other samples and if his diet was the same, which could tell us if he was part of the community since a child -even though both parents must have been genetically WHG-  or if he joined as an adult). Interesting in this respect is the very recently published study about the Megalithic phenomenon that we didn’t have time to comment (even when one of the authors is a collaborator of this blog), that helps to bring some insight into the societal structures of the Megalithic builders and opens the path for more detailed knowledge that could allow us one day to be able to say something meaningful about the possible linguistic influence of European hunter-gatherers in the farmer’s languages (though being very likely that the latter are long gone without any knowledge about them, it might all be restricted to loanwords that made it into successive languages).

Share it

104 thoughts on “The problematic of substrates – A case study of Iberia

  1. @all

    Busy days, but I hope I’ll be able to put up a post about the Estonian aDNA this week.

  2. Yeah, Alberto – that idea of mutual borrowing, or “overlaying” of IE branches can of course be extended further into the past, beyond the Iron age cases of Slavic-Skythic, Greek-IndoIranian (Persian Wars, Alexander the Great), Celtic-Germanic and Celtic-Daco-Thracian interaction.
    E.g., one of the issues that complicates the IE phylogeny are Anatolian – ItaloCeltic isoglosses (formation of passive voice, etc.). Now, we know since quite some time of the presence of a “Cypriot-like”, CHG-enhanced ANF signal across the Mediterranean. The new Iberian studies, and also the fresh Sardinian/ Sicilian one, suggest that in the W. Mediterranean this signal post-dates the EBA. OTOH, we are now also able to tell it apart from Iron Age Greek colonisers.

    While we still lack a chronologically fine-tuned aDNA record from BA Italy, and also BA Anatolia for confirmation, an Anatolian expansion through the Mediterranean during the MLBA now looks like a possibility worth considering to me (maybe the legendary descent of Romans from Troy turns out to not have been a complete fiction…). One custom that certainly spread out of Anatolia (ultimately out of Caucasia) through the Mediterranean during the MLBA was pithos (amphorae) burials, the earliest attestation of which I am aware of is Areni1 (Armenia_CA).
    A “dual origin” of Italic languages, in a blend of Anatolian/Aegaean influence on the MBA Appenine Culture, and Central European Urnfield influence on the LBA Proto-Villanova Culture would explain the high diversity of Italic languages already during their early attestations sometime around the late 7th century (and Etruscan influence on both groups their apparent similarities, such as the “proto-Italic” bh->f/v and gh->h sound shifts). Such a model has already in the 1970s been proposed by Italian archeologist M. Pallottino.

    Ultimately, this kind of “multi-layering” may be extended further back in time: Urnfield, e.g. combines Tumulus/ post-Unetice and S. Balkans traditions, with the latter also being suspected to have strongly influenced Mykenean and Armenian [There are, btw, a couple of isoglosses between Albanian and S. German dialects, e.g. alban. dósë, Franconian Dausch “pig”].

    And Sintashta of course doesn’t represent the earliest cultural wave that reached the Indus from Central Asia – long before we already had Jeitun influencing heavily on Harappa. Jeitun ultimately goes back to Sang-e-Chakmak in the Alborz foothills, and the S. Caspian (Hotu etc.) Neolithic The S. Caspian origin of Kashmiri goats and walnuts has recently been genetically demonstrated. [How Holstein-Frisian cows obtained there ca. 10% Indian Water Buffalo admix, however, is still somewhat mysterious].

  3. @ Frank
    Good points. I think that ties in well with some of Garrett’s points, that final forms of *proto-Italic, or *proto-GReek came together in their respective geographies.
    BTW: do you know of any references for Boleraz influences on Salzmeunde phase of TRB ?

    @ Egg
    That might make sense. Sherrat’s model seems feasible to me. But still early to tell. I think the Irish teams might be publishing more data from Beli Manastir. With this we might more definitively see how steppe & farmer groups interacted in the Balkans.

  4. @Frank

    I agree with the superstrate terms. Other than warlords, perhaps these early IA were traders of horses, weapons and lapis lazuli?

    Sum. urud(a) is very interesting and is kinda confirming a hunch I’ve had for a while.

    Bagh is pretty common in II names, and appears in many Persian cities from Georgia to Afghanistan. It’s highly likely that the Maryannu warriors had the city built for them.


    More linguists in favor of the steppe hypothesis prefer a clade with Balto-Slavic over Greco-Aryan.

    I don’t argue for a simplistic J2 scenario or any Y-DNA one. It could’ve been R1 from Steppe Maykop, J2, E-V13…etc. Though I am still not sure why you think R1 is more likely than J2 when there is no R1 in either Mycenaeans, Hittites and Swat.

    When I look up a linguistic theory I like to find supportive evidence outside of that linguistics, and that’s why I prefer Greco-Aryan.


Leave a Reply

Your email address will not be published. Required fields are marked *