Talk:Centum and satem languages

Tocharian has no centum sound change, explain.
Dbachmann, how does Tocharian not have centum sound change? The palatals become velars. Also, it is likely that none of the centum languages form a node and the same with the satem languages. So, how is it that Albanian is not a satem language, when the term satem now just describes languages which have fricatives or affricates from the PIE palatals. There are no IE subgroups of Centum or Satem. Finally, I think you can keep the ad hominems to a minimum just because I didn't like your maps. Imperial78

Too technical
This article is a mine (¿-field?) of information, but is way too technical for the non-linguist reader to get much out of. Although I've spent many hours studying phonetics, the presentation of information here wearies me. Maybe that's because it dives right into depths of detail after the lead? Perhaps somebody with topic expertise could restructure the article to, first, deliver the main points, and only then, follow up with an exposition at a more detailed level. I'll be happy to help if anyone wants someone to review their suggestions for improvement. yoyo (talk) 15:48, 4 April 2019 (UTC)

Map
While Florian's revert seems to not yet be in effect (the image still displays "Illyrian" as centum and "Palaeo-Balkanic" (?!) as "satem"), I'd like to note a few other issues with it as applies to the "Southern" zone. While the issues with calling Illyrian satem or centum in the lack of evidence have been noted, there are similar disputes for the other "Palaeo-Balkanic languages".


 * The map has Phrygian as satem. Ligorio and Lubotsky (2018) however...


 * Thracian too has now been questioned by Brixhe (2018): . I will note that Brixhe is likely a minority view however, given that he also thinks (same publication, in Klein-Joseph-Fritz 2018 Handbook).


 * Messapic is shown as centum on the map. However, this is is not something we can say with confidence. De Simone (2018, again in Fritz-Klein-Joseph) notes that we really just don't know much about the development of occlusives from PIE to Messapic.

Ideally all of these areas should be neither red nor blue. --Calthinus (talk) 18:12, 21 April 2020 (UTC)
 * The fact is that languages like Illyrian, Messapic, Thracian or Phrygian are too scarcely attested to be confidently classified as either centum or satem. The so-called 'Paleo-Balkanic' languages–which is not a sub-family of languages but merely a geographical grouping (perhaps a Sprachbund) that sometimes even includes Ancient Greek–should be a grey zone on this map, and I don't understand why the editors of the map found the urge to add them. I'm even thinking about creating a new one, at least with a better visual quality. Azerty82 (talk) 18:58, 21 April 2020 (UTC)
 * Another issue is that the map is synchronic and features languages or cultures separated by thousands of years (the Sintashta culture flourished ca. 2000 BCE, while Albanian is not attested before the 15th century CE). But it would take too much time to create an animated one. And Lusitanian is still on the map. We have like 5 shorts texts in Lusitanian. Azerty82 (talk) 19:12, 21 April 2020 (UTC)
 * PS: I'm not seeing Messapic/Thracian/Phrygian on the map any longer (Lusitanian is still there though). Try to refresh the cache of your browser. Azerty82 (talk) 19:17, 21 April 2020 (UTC)
 * Albanian shouldn't be an issue. It isn't on the map I see (and shouldn't be, it is not about Albanian and Albanian is a complicated case). Though I don't think attestation should matter, i.e. we are not removing Baltic just because it wasn't attested until it was almost the Renaissance. --Calthinus (talk) 19:22, 21 April 2020 (UTC)
 * Albanian is on the new version of the map, although it seems to be in a grey area. You're right to point out that the intent of the map is not to feature languages, but sub-families of languages (which existed before the first attestations). It got confusing with the recent additions. Azerty82 (talk) 19:37, 21 April 2020 (UTC)
 * Imo the map should mention Celtic, Germanic, Italic (not colored in Apulia though!), Greek, and Tocharian for centum, and Hittite but not Luwian for centum. We should not have "Anatolic" when centumization is a factor that affected some Anatolic languages but not others! For satem, it should have Slavic, Baltic, Indic, and Iranian, and mayyyybe Armenian but I would lean toward omitting Armenian which is a particularly tenuous case. I don't see the point in OR mentions of anthropological cultures like Srubna etc. Mentions of Illyrian, "Palaeo-Balkanic", Dacian, Paeonian, Thracian, Phrygian, Messapic are not helpful as there is a lot of confusion about these. There is less confusion about Albanian, but Albanian is to be omitted for the same as Luwian because the map is about centum and satem languages not centum and satem languages and also the other Indo-European languages that are not really on either side.--Calthinus (talk) 19:52, 21 April 2020 (UTC)
 * Yes, I totally agree with you. Maybe we could depict Armerian with hatches (perhaps also Albanian and Luwian) as a satemized language. PS: I've noted that you're taking great care of the handbooks, that's good to see ;-) Azerty82 (talk) 19:57, 21 April 2020 (UTC)
 * The distinction between "satem" and "satemized" is sophistry, IMHO. All that's "satem" ultimately really about is whether the dorsal rows for which there is no evidence of them ever having been labiovelar appear as two rather than one rows (secondary palatalisations excluded, which is what seems to have happened in Phrygian as well as in Italic). By that account, Armenian is simply satem, even if it may also show traces of a distinction between the traditionally reconstructed "pure velar" and labiovelar rows (the same is suspected for Albanian, and certain traces of labialisation have even been suggested for Balto-Slavic and perhaps Indo-Iranian, so Armenian is probably not even unique in this regard). Anatolian appears to be essentially centum, per Melchert, but I agree that the case is tricky enough that it's better to not mark it as either on the map, as is already done. I agree it's probably best to leave off poorly attested branches, so maybe we should revert to an even older version of the map.
 * As for the time period shown, it's true that some synthesis is present but also inevitable, and the best-fitting timeframe is somewhere in the late first millennium BC, around 500–200 BC, or perhaps a little later, when Latin loanwords started to enter Primitive Albanian, which still appears to have been before the turning of the eras; the Luwic branch of Anatolian was still not entirely extinct at the time, and Pisidian is directly attested in inscriptions from as late as the second century AD. --Florian Blaschke (talk) 21:01, 21 April 2020 (UTC)
 * Florian -- for me, the time frame is not an issue. Regarding "satem" versus "satemized" (Armenian) or (per Demiraj 2018) "satem-like" (Albanian), imo it is the opinions of RS we should be adhering to.
 * In the Armenian case I am less personally knowledgeable and so I'm unsure (also afaik it's "satem-ness" is based on a mere three etyma which is imvho pretty bad given that it is surrounded by languages with well-known historical shifts between dorsal positions...). Side note with Armenian, the page says, without citation, that Armenian "could be" a centum language with secondary assibilation. This looks suspicious since the page goes on to demonstrate that Armenian has clearly different reflexes for the back and palatal velar serieses, meaning centumization did not occur. It would seem that instead the argument is that Armenian is in fact more like Luwian and Albanian, which are certainly not centum.
 * In the case of Albanian, at one point we had a consensus that referred to it as satem. But now Demiraj in Klein-Joseph-Fritz 2018 states it is definitively "not satem", and also not centum, although since palatovelars were also assibilated it is "satem-like". Instead, it's to be classed with Luwian, since the three dorsal positions have distinct reflexes. That is, if, as this page does, we define satem as the merger of *kw and *k (which Albanian does have specifically before back vowels as well as merging all three dorsal rows in some other conditions). If satem is instead to mean "assibilating/fronting the former palatal velars", then per my original research opinion it still makes sense to call it satem here. I will have to do a quick fix to the page as I have found Byrd (2018) referring to Albanian still as satem, albeit without discussing the matter (this is also the position of Orel, and the former position of Demiraj). However, this page is pretty clearly presenting the definition of satem as merging *kw and *k and fronting *ḱ, stating this in the lede: -- this then, would be the same criterion Demiraj is using, which would per his analysis not apply to Luwian and Albanian (and perhaps Armenian but again maybe the reality for Armenian is the data is just too feeble to tell). --Calthinus (talk) 23:56, 21 April 2020 (UTC)
 * I do think the map should not refer to "satemized" languages, especially if it is unclear what that means. The use of "satemized" for Armenian just meant that it occurred during the development of the proto-language, but I am not sure what the usage in the map for Baltic and Slavic means (all satem languages are satemized per the mainstream view of three dorsal rows in PIE, no?). --Calthinus (talk) 23:58, 21 April 2020 (UTC)
 * Personally, I think that "full merger of pure velar and labiovelar series" is a poor way to define satem as traces of this distinction have been claimed for all branches traditionally defined as satem. "No merger of the palatal and pure velar series, with assibilation of the palatal series" is a better definition. (There do not seem to be any IE branches/languages where the palatal and pure velar series are distinguished but the palatal series has not been assibilated, or for which this has even been proposed, so the condition is strictly speaking redundant, but it also doesn't hurt to include it.) Centum is more straightforward to define: IE languages are centum if the traditionally reconstructed "palatal" and "pure velar" series have identical reflexes and cannot be distinguished anywhere. (Interestingly, Kümmel has suggested – p. 27f. – that PIE may really have been centum, with the distinction in the satem branches having only emerged post-PIE, hence the qualifications.) For Anatolian, see Melchert 2015 (p. 15f.). --Florian Blaschke (talk) 12:12, 22 April 2020 (UTC)
 * Hmm well the Kummel and Melchert (believe I read the Melchert some time back) are certainly interesting, perhaps could be of use to the page. In the 2018 source Azerty sent me (can send to you too if interested) there is much more discussion that could be of use to this page. My understanding is that a number of people including Hans Henrich Hock have argued that an innovative satemization spread from Indo-Iranian to other branches, though idk about the timeline for that. I'm not sure I agree with  as the best definition for satem, but in any case this would also exclude Albanian (there is a merger near /l n r/ of the pure and palatal velars) and some argue that Baltic and Slavic also feature this, though there are counterexamples (Pre Slavic k̑luHsa:te:i > *slu:xɛ:tεi > Russian slyshatj) and other interpretations -- Matasovic (2013) argues instead that these represent a case of "desatemization" in Slavic (see [https%3A%2F%2Fhrcak.srce.hr%2Ffile%2F172848&usg=AOvVaw3Xf9I4IWkWHfHipw4IWoQ-]) and appears to in another publication argue the same for Albanian, while Andersen instead thinks that these "desatemized" (per Matasovic) examples are actually some sort of centum substrate influence. But in the end I think we should be seeking out expert opinion on "what is the most salient aspect of the centum/satem division" and if (probably) there is disagreement among experts, we should say as much.
 * Side note for anyone who might appreciate my nerdy rumination -- this work on the nature of the satem shift in Armenian could be of use |lang_fr&id=f-dFAgAAQBAJ&oi=fnd&pg=PA299&dq=Albanian+satem&ots=ZU1v8LoZW8&sig=xG_Sxlg0Uhj5uzcgwNlwx3Jbnlw#v=onepage&q=Armenian%20satem&f=false. --Calthinus (talk) 18:21, 22 April 2020 (UTC)
 * A depalatalisation of this kind, limited to certain contexts, is also present in Indo-Iranian (namely before liquids), known as Weise's law. However, there is by no means a full merger, as in centum languages. Please read my suggested definitions closely. To claim that "no merger" has occurred in Albanian, Balto-Slavic and Indo-Iranian would be ridiculous. --Florian Blaschke (talk) 18:25, 22 April 2020 (UTC)
 * Okay that's fair. The goal is not to "break" the satem-centum dichotomy with edge cases: if "no complete merger of the palatal and pure velar series, with assibilation of the palatal series" is the best definition and supported by sources, the page should be fixed to at least mention this. Regarding the map, what is your proposal? Azerty and I appear to be in agreement that the (broadly, incl Messapic) Balkan languages (modern and ancient) should not be colored or even marked. Do you disagree with this?--Calthinus (talk) 18:36, 22 April 2020 (UTC)
 * The fact is that satem languages are more easily defined because they seem to constitute a distinct sub-group within the I-E languages, including at best the Balto-Slavic and Indo-Iranian branches, while centum languages exhibit features that are likely to have emerged independently (but even in this situation, Baltic and Slavic languages had already begun to diverge when the distinct satem features appeared, so we are are not talking about an inherited feature that would make a Balto-Slavo-Indo-Iranian branch viable). On the other side, the centum mergers seem to have happened independently in the languages, since they appear in idioms that weren't in contact between each other (Hititte–not proto-Anatolian–and Tocharian, for instance) and is clearly not an inherited sound change. That is why the chronology of the map is important to me, and that is why I think it is decisive to distinguish between 'purely satem languages' (i.e. Indo-Iranian & Balto-Slavic), and 'satemized languages' (i.e. Luwian & Armenian) . Azerty82 (talk) 19:07, 22 April 2020 (UTC)
 * I have more questions at this point. What is the goal of this map? Is it simply to indicate to a reader where satem languages and centum languages are located? Or is it to present Von Bradke's view on how it spread? Or is it to present the modern view of how the dichotomy emerged? I think we need to clarify these things.--Calthinus (talk) 19:16, 22 April 2020 (UTC)
 * As I said above, I agree that it's best to leave poorly documented languages/branches off the map, i. e. to neither name nor colour them. The point of the map, as I understand it, is to give a rough overview of the geographic distribution of satem and centum languages. A secondary point, von Bradke's hypothesis, Dbachmann decided to represent on the map as well, which I personally don't mind, but it could be omitted.
 * Although Anatolian is documented fairly well overall, I'm also fine with leaving it outside of the centum/satem dichotomy for the purposes of the map, given that its case appears to be more complicated and controversial, and Anatolian was already almost on the brink of extinction by the turning of the eras, anyway.
 * Personally, I think that Dbachmann's original version was the best one, and most of the edits to the map have been detrimental, with only two changes being clear improvements: 1) colouring Tocharian blue; 2) changing the distribution of IE in the Iberian Peninsula.
 * If PIE was centum, the problems mentioned by Azerty82 are solved: the satem group is innovative. Even under the conventional three-dorsal-series reconstruction of PIE, the independent centum mergers would be trivial enough if the palatal series consisted of palatal or even palatalised velar stops, or if it was velar while the "pure velar" series was uvular. Also, contra Azerty82, Balto-Slavic is now generally considered to be a coherent branch that did not diverge before the second or first millennium BC, with Proto-Balto-Slavic already reconstructed as satem, despite depalatalisation (or lack of palatalisation) in certain contexts (with loanwords from centum branches, predominantly Germanic, generally only later having intruded especially into Slavic). Azerty82's proposed distinction between "satem" and "satemized" IE is too subtle and too far from consensus for the general reader, I feel (we should be cautious and careful not to propagate our pet hypotheses to laypeople here), although it could be mentioned in the body of the article if it is supported by sources. --Florian Blaschke (talk) 19:28, 22 April 2020 (UTC)
 * We could build off the current version Florian reverted to by removing Lusitanian and  Messapic from centum -- and if we're at it let's remove the territory of Livonians and Volga Finns who were not Balticized/Slavicized until quite recently. The Dacian territory marked "Iranian" can also be fixed. Albanian and Anatolic, I"m equally amenable the current version that has them as gray/white or just omitting them. We could also remove Srubna/Sintashta, and use a separate map that just marks where those cultures were located for the section about Von Bradke's views. This way we are not propagating any hypothesis, I think -- including not propagating Bradke 1890. --Calthinus (talk) 19:39, 22 April 2020 (UTC)
 * Let me clarify my original wording: "Baltic and Slavic languages had already begun to diverge internally when the satem features emerged". I'm not saying that they already formed distinct and unified proto-languages, but that they were clearly dialectalized (or near-dialectalized) by the time satem features spread from linguistic contacts with Indo-Iranians. This is evidenced by the fact that the mergers of labiovelar and velar stops happened distinctively within the 'proto-dialects', and were inherited in distinctive manners in the attested languages. The proposed satem vs satemization distinction rests upon the fact that such features happened independently (and in a limited manner)–in the case of Albanian–or perhaps from later contacts–in the case of Armenian. This is not my "pet (did you mean personal?) hypothesis" but a serious proposition supported (although not universally) by scholars. 19:48, 22 April 2020 (UTC)
 * (Sorry for the length of the quote) . Ringe (2018): The old division between centum languages (those that merged the palatal and velar stops, but not the labiovelars) and satem languages (those that merged the labiovelar and velar stops, but not the palatals) has had to be drastically revised. Melchert (1987) showed that the Luvian subgroup of Anatolian preserved PIE *k, *k, and *kw as separate phonemes. Since mergers are irreversible, it follows that Proto-Anatolian likewise pre- served this three-way contrast. Hittite, however, exhibits the centum merger − and it is impossible that that innovation is historically shared with any of the other centum lan- guages, because Proto-Anatolian had already undergone so many innovations (cf. Melch- ert 1994: 60−91; Garrett 1990: 265−280) that it must have been mutually unintelligible with the other IE languages even if they were still in contact (which is itself doubtful). In other words, the centum merger is a repeatable innovation; and it follows that the centum languages cannot be shown to be a subgroup and need not have been geographi- cally contiguous in the immediate post-PIE period. The satem merger should also have been repeatable; moreover, it appears that not all the languages that have sometimes been said to share the merger actually do. Though the Albanian evidence is difficult to evaluate, it appears that Albanian, like the Luvian group, did not at first merge any of the PIE dorsals (cf. Demiraj 1997: 63−65); that Armenian underwent the merger is unlikely at best, since it appears that some of the PIE labiovelars are palatalized in environments in which the velars are not (cf. e.g. Schmitt 1981: 62−65; Olsen 1999: 805−808 with references) and it is even possible that *kw became *p before *o prior to the Armenian stop shift (Olsen 1999: 805−808). That leaves Indo-Iranian and Balto-Slavic as probable members of a satem group. It appears that those two subgroups share not only the satem merger but also the “RUKI-rule”, a sound change which retracted *s after high vocalics, rhotics, and dorsals (Andersen 1968); that tends to strengthen the case that they constitute a clade. On the other hand, the recent discovery of Gaulish future stems in *-syé/ó- (McCone 1991: 145 with refer- ences) has removed one item from the potential list of satem-language innovations. But the fact that Baltic and Slavic (often separately) exhibit some forms with the centum development of dorsals which are unlikely to be loans from (pre-)Proto-Germanic (cf. already Porzig 1954: 74−75) strongly suggests that the satem merger, at least, spread from Indo-Iranian to Balto-Slavic after the latter had already begun to diversify internally (so e.g. Hock 1991: 442−444). If that is true, then the satem group can be a subgroup in the sense that it remained a dialect continuum after losing contact with the other IE languages, but not in the strictest sense of the term. The fact that the centum merger does not define a subgroup is important for assessing the position of Tocharian, which exhibits that merger (Ringe 1991b: 138−144, 1996: 39− 42). Partly for that reason, some scholars had suggested that in spite of its historical position at the far eastern end of the IE Sprachraum Tocharian was at first a western dialect of IE. Ringe (1991a) examined the available evidence and concluded that Tocharian shares no significant innovations with any other subgroup of the family. (On more recent assessments see further below.) Azerty82 (talk) 19:53, 22 April 2020 (UTC)
 * The second İranian label with the strange dot over the capital I should definitely be removed (along with the second Celtic label and the Lusitanian one), and I consider it to be one of the non-improvements to the map (the intention was apparently to mark Scythian, but that's an unnecessary detail). Can we colour Albanian light red at least? That said, it sort of already is.
 * Azerty82, Proto-Balto-Slavic is generally reconstructed as a completely unified proto-dialect, without any internal dialectalisation. Otherwise, it would not be Proto-Balto-Slavic, and Balto-Slavic would not be a coherent branch. The hypothesis you mention is definitely not consensus, it's only Ringe's proposal, and absolutely a pet hypothesis that is completely irrelevant for the purposes of the introduction and the map. I think Ringe is referring to doublets such as Lithuanian akmuõ vs. ašmuõ, as well as klausýti, for which other explanations have been proposed (with "Proto-Balto-Slavic was centum" definitely not being the best explanation, IMHO). --Florian Blaschke (talk) 19:59, 22 April 2020 (UTC)
 * Well, you and I know well that proto-languages are reconstructed dialect-free because they are reconstructed proto-languages. But let's focus on the map.
 * Why do we need to draw precise borders on the map. Let's just place the names of I-E branches coloured with blue (centum) or red (satem). Azerty82 (talk) 20:08, 22 April 2020 (UTC)
 * I agree with this^. It would be good to put an end once and for all to borders that are anachronistic or just wrong (i.e. the Volga). Regarding Albania and Luwian (or per Melchert "Luvo-Lycian" ), and Armenian, what I gather from the current sources presented is this: both are "special cases", in which distinct reflexes for three rows existed, and assibilation occurred, with reflexes of *ḱ in Proto-Albanian *ts > Albanian th and ts, s in Luvo-Lycian. There are different interpretations among experts for this. Ringe 2018 and Demiraj 2018 (the leading Alb specialist) argue this means Albanian and Luwian exist side by side in their own special group that lies outside the centum-satem dichotomy despite a resemblance to satem in having dental fricative reflexes for the palatal velar series. Meanwhile Melchert 2015 in the above link Florian originally provided has his own view that, and in fact did have a merger of front and back velars but this happened only after the front velars split. What is the NPOV way to draw a map? We should go with the largest common denominator between these two views expressed by leading experts. What is this? Albanian and Luwian are both distinguished by having distinct reflexes for the three rows, and having at least some degree of regular assibiliation that targeted specifically the palatal velars. How does Armenian fit in here, I don't know, we should again consult the experts and not get into lengthy debates about our own views -- but if we agree on this point on Luwian/Albanian, it is progress on the matter. I also think this is a good compromise, we can color the names of the included languages (Luwian, Albanian, maybe Armenian) a shade of lavender  -- this is a concession to Azerty's wish to distinguish "satemized" languages, while at the same time a concession to Florian's desire to note the shared behavior between Albanian and the undisputed satem languages. Do we agree? --Calthinus (talk) 21:56, 22 April 2020 (UTC)

P.I.E. *ḱ ---> Baltic *s
s? Shouldn't it be ś? Muonium777 (talk) 08:48, 5 July 2021 (UTC)

k vs k' minimal pairs?
In the reconstructed vocabulary, do we have anything by way of minimal pairs contrasting k & k' ? --B.Bryant (talk) 09:39, 19 November 2023 (UTC)