Y sent me a link to Cid Swanenvleugel’s The pre‑Roman elements of the Sardinian lexicon (LOT, 2025; free pdf download), saying “It looks ambitious, and if not all true or even verifiable, at least interesting,” and I agree. Here’s the Summary (pp. 535-36):
One of the questions addressed in this study is whether the assumption of a single pre-Roman language, besides Punic, can account for all of the non-inherited lexical material. I have found that there is no geographical patterning in the phonological features found in words of pre-Roman origin. We can, however, discern a near-complementary geographical distribution in the pre-Roman prefixes *k(V)- and *θ(i)-, which have been argued in § 9.1.3 to be variants of one and the same pre-Roman morpheme (cf. also Swanenvleugel 2024). This prefix and other accepted pre-Roman morphemes exhibit an island-wide distribution. Pre-Roman Sardinian words, excluding punicisms with accepted cognates in other languages also occur across Sardinia. All of these findings constitute evidence supporting the hypothesis of a single language, or at least closely related language varieties, having existed all across Sardinia at the time of its romanization. This language coexisted with Punic. The coexistence of other languages with a smaller distribution cannot be ruled out.
The reality of many of the previously proposed phonological and morphological features attributed to a pre-Roman language in Sardinia cannot be confirmed based on the lexical material investigated in this study. This includes the pre-Roman vowel harmony proposed by Serra (1960; cf. § 8.4.2). The same goes for a number of putative pre-Roman suffixes (§ 9.2). What can be maintained is the pre-Roman phoneme *θ, the existence of word-final consonants, and various morphemes, such as *k(V)-/*θ(i)-, *-́Vr, and *-(V)s-.
Based on the evaluation of the previously proposed extra-Sardinian comparanda, it is likely that some Sardinian words derive from a Punic substrate. This mirrors the long-lasting presence of a Punic speech community on Sardinia, which persisted for centuries after the Roman conquest of Sardinia. Previous hypotheses on a linguistic affiliation of pre-Roman Sardinian to either Berber or Basque cannot be maintained on the basis of the lexical evidence. The Sardinian-Berber correspondences are argued to be part of the Punic substrate. The few acceptable Sardinian-Basque correspondences do not point to language relatedness, but rather to independent contacts with a third pre-Roman language. The hypothesis of an affiliation between pre-Roman Sardinian and Etruscan is difficult to evaluate because of the scarce attestation of Etruscan. The positive evidence is restricted to a single convincing Sardinian-Etruscan lexical correspondence, and two possible morphological correspondences.
The analysis of the distribution of accepted comparisons between non-inherited words in Sardinian and in other languages of the Mediterranean indicates an especially close connection between Sardinia, the Iberian Peninsula, and southern France. This is potential evidence that the pre-Roman Sardinian language was related to the ancient Iberian language. There is, moreover, a sizable number of lexical correspondences between Sardinia and the Italian peninsula. The overlap in lexical correspondences between Sardinia and the Iberian Peninsula, southern France, and Italy, is evidence that related pre-Indo-European speech varieties were spoken across the western Mediterranean region.
There is ample evidence for a scenario in which both a pre-Roman “Mediterranean” language and Punic were spoken natively in Sardinia at the time of the Roman conquest. The geographical overlap of pre-Roman Mediterranean words and Punic words shows that these languages were spoken side by side. However, the non-inherited lexicon does not provide much evidence on the sociolinguistic dynamics between these two languages. Likewise, it is impossible to establish based on the extant evidence whether or not either of these pre-Roman languages outlived the other.
I don’t know enough to have useful thoughts, but I expect the Hattery will have things to say.
Downloaded this a few days ago.
SPOILERS
Besides all the to-be-expected Punic stuff, he makes a case for an original single ur-Sardinian language perhaps related to Aquitanian and Basque. Which would make sense.
Hmph. Previous comment eaten by Akismet. But it was wrong: the book actually makes a case for an original ur-Sardinian possibly related to Iberian.
I have enormous respect for those (like Lameen) who do rigorous work on loanwords. Much harder to do really well than first-order comparative work is, it seems to me.
Wonder if posting a third comment will make the second one rematerialise …
How does this compare in quality to, say, Beekes and his Pre-Greek?
Weirdness — comment link on the main page says 4 comments, but I’m only seeing one.
How sure one can be that geographic distribution of lexical or morphological features will survive more than two millennia?
Will download and read later. Just noting this from the summary on the download page:
I think this means that wherever there’s enough evidence of the actual language to tell, no relation is visible, but where there’s only slim onomastic and toponymic evidence, similarities can be found. That might point to a sub- or superstrate.
(I, too, lost a comment. Judging from the above, it will reappear.)
“Closely related to other unattested languages” is quite a phrase. “This thing we have no direct evidence of and can only infer the former existence of from lexical items we otherwise can’t explain must have been closely related to other such things which ditto.”
I didn’t even use any non-standard characters in the title! WTF, Akismet??
Maybe try without the hyphen in title and page address….
https://languagehat.com/pre%e2%80%91roman-elements-in-sardinian
It is the hyphen, I think (see disappeared comment)
OK, done.
“Closely related to other unattested languages” is quite a phrase
The summary is wrong: Iberian is far from “unattested.” What it is, is undeciphered (much more so than e.g. Etruscan.)
Obviously that does indeed lead to a whole new level of speculativeness, but that is not (necessarily) the same as just making random stuff up to fit the hypothesis.
[Hat’s change may have worked. Let’s see …]
It was a non-breaking hyphen rather than a regular one, copied directly from the article title.
For anyone unfamiliar: Non-breaking hyphens are identical to regular ones as glyphs. However, normal hyphens are automatically considered valid locations for line breaks, and a non-breaking hyphen explicitly tells the algorithm flowing the text not to put a line break there. There are also space characters (sometimes known as “sticky” spaces) that cannot be line break locations. However, the first time I encountered a word processor that supported sticky spaces (before Unicode even existed, so they were specially programmed for that piece of software), they were visually different than regular spaces—wider than plain spaces, for most fonts.
How sure one can be that geographic distribution of lexical or morphological features will survive more than two millennia?
Unless earthquakes, fire and brimstone are excluded by definition, uncertainty remains on the table.
It was a non-breaking hyphen rather than a regular one, copied directly from the article title.
Aha, all is explained. I’ve replaced it with a regular one, restoring the natural order.
Now there are two entries for Pre-Roman Sardinian in the threads with most recent comments list.
Which isn’t a problem. I’m just amused by it.
I’ve read the thesis. (Well, mostly. I may have decided to skip the detailed entries somewhere towards the end of the “Flora” section.) The takeaway?
– There’s a common prefix *θ(i)- (that may have been an affricate) and a prefix *kV- that may be in complimentary distribution, phonologically and geographically. Since the prefix(es) also attached to words of Punic and Latin origin, the language that used it/them must have been spoken well into the CE. (I decided for that when I came to *θingòrra “eel”, which Swanenfleugel for some reason doesn’t compare with Lat. anguilla).
– There may also be a couple of suffixes, but those are less convincing.
– There are words with reasonably good comparanda in ancient languages and modern dialects around the Western Mediterranean.
The conclusions from that are cautious and reasonable. I may quote the concluding section of the discussion and the “Outlooks” section of the conclusionn:
Iberian being closer to the language(s) of Sardinia than to Basque makes sense if the “western Mediterranean substrate” results from the coastal or outright marine spread of agriculture westwards by the Cardial Culture, while Basque, i.e. Aquitanian, results from the continental spread by the Linear Ware Culture.
That was almost certainly an affricate; otherwise, some of its further developments become highly improbable at best.
>θ(i)- (that may have been an affricate) and a prefix *kV
More easily explained by substrate IE languages that were centum and thatem. Or possibly the prefixes were passed along differently by k-celtic and θ-celtic groups.
I don’t kinth so.