User talk:Mellsworthy

Please feel free to contact me in this space about cognitive linguistics, dead languages, and word meaning. Okay, feel free to contact me about anything else too. Mellsworthy (talk) 07:05, 31 October 2011 (UTC)

Reference Errors on 4 June
Hello, I'm ReferenceBot. I have automatically detected that an edit performed by you may have introduced errors in referencing. as follows: Please check this page and fix the errors highlighted. If you think this is a false positive, you can [//en.wikipedia.org/w/index.php?action=edit&preload=User:A930913/RBpreload&editintro=User:A930913/RBeditintro&minor=&title=User_talk:A930913&preloadtitle=ReferenceBot%20–%20&section=new report it to my operator]. Thanks, ReferenceBot (talk) 00:35, 5 June 2014 (UTC)
 * On the Proto-Indo-European pronouns page, [//en.wikipedia.org/w/index.php?diff=611602224 your edit] caused a broken reference name (help) . ([ Fix] | [//en.wikipedia.org/w/index.php?title=Wikipedia:Help_desk&action=edit&section=new&preload=User:ReferenceBot/helpform&preloadtitle=Referencing%20errors%20on%20%5B%5BSpecial%3ADiff%2F611602224%7CProto-Indo-European pronouns%5D%5D Ask for help])


 * As indicated in the report to operator, the reference problem was spurious. Mellsworthy (talk) 20:56, 20 August 2014 (UTC)

ArbCom elections are now open!
Hi, You appear to be eligible to vote in the current Arbitration Committee election. The Arbitration Committee is the panel of editors responsible for conducting the Wikipedia arbitration process. It has the authority to enact binding solutions for disputes between editors, primarily related to serious behavioural issues that the community has been unable to resolve. This includes the ability to impose site bans, topic bans, editing restrictions, and other measures needed to maintain our editing environment. The arbitration policy describes the Committee's roles and responsibilities in greater detail. If you wish to participate, you are welcome to review the candidates' statements and submit your choices on the voting page. For the Election committee, MediaWiki message delivery (talk) 13:54, 24 November 2015 (UTC)

July 2016
Welcome to Wikipedia. We welcome and appreciate your contributions, including your edits to Sintashta culture, but we cannot accept original research. Original research refers to material—such as facts, allegations, ideas, and personal experiences—for which no reliable, published sources exist; it also encompasses combining published sources in a way to imply something that none of them explicitly say. Please be prepared to cite a reliable source for all of your contributions. Thank you. Iryna Harpy (talk) 06:23, 3 July 2016 (UTC)


 * I hope I addressed your concerns on the talk page for Sintashta culture. I did not do original research, I was clarifying what the already cited paper actually claimed. Mellsworthy (talk) 08:58, 3 July 2016 (UTC)
 * Ah, good, resolved in talk page. --Mellsworthy (talk) 05:49, 4 July 2016 (UTC)

Corded Ware


Hi Mellsworthy. So, if "this presents the possibility that the Corded Ware culture might have originated the core-Indo-European language, i.e., the main branch of Indo-European, which excludes only Tocharian and Anatolian", how about the Tocharians having the south-Asian variant of R1a (Underhill 2010)? An early split-off from a Caucasian/Iranian place of origin? Joshua Jonathan  -  Let's talk!   08:50, 3 July 2016 (UTC)


 * This is funny, because I'm probably going to have to cut the assertion since I can't find a source for it. I fear I may be treading on Original Research.


 * The Tocharians, or rather the inferred pre-Tocharians, share the Z93 variant of R1a with one of the Poltavka culture's population and a ton of modern Indic, Iranian, and even (secondarily) Turkic people. But I don't think that all of the evidence actually hangs together yet.  That's why I used the very mealy-mouthed phrasing.  Clearly both Sintashta (with descendant Andronovo) and Afanasevo (with descendant Tocharians) had this haplotype. Given how early the haplotype is attested in the Afansevo population, it is very unlikely that R1a Z93 is derived from the Corded Ware strand of the Sintashta/Andronovo ancestry.  I admit, the possibility is undermined by the very western autosomal DNA of the one Poltavka individual with the haplotype, but hey, something's got to be correct, and it's not Corded Ware dudes teleporting to the Altai, I figure.


 * Mellsworthy (talk) 09:14, 3 July 2016 (UTC)


 * Florian Blaschke was having a dispute on more or less the same topic; he may be interested. Do read this blog, on Iranian origins; it's very interesting.  Joshua Jonathan   -  Let's talk!   11:54, 3 July 2016 (UTC)
 * NB: I placed the Tocarians at the wrong place, of course; it should be little bit more to the south-west.  Joshua Jonathan   -  Let's talk!   12:01, 3 July 2016 (UTC)
 * NB2:read also Dieneke, On Tocharian origins:
 * "Gamkrelidze and Ivanov cited W. N. Henning to the effect that the ancestors of the Tocharians could be identified with the Gutians from the Zagros"
 * Joshua Jonathan  -  Let's talk!   12:06, 3 July 2016 (UTC)
 * Although Henning is a respected researcher, his Tukri speculation is extremely fringe and not taken seriously by any modern scholar as far as I am aware.
 * I can't access the full article (Allentoft et al. 2015), but from what is mentioned in Blond, it seems to assert that the Anatolian branch is derived from the Corded Ware as well. Not sure about Tocharian; but according to Parpola 2012, the Afanasevo culture begins only from 3100 BC and human remains only from 2750 BC, so possibly late enough to be derived from the Corded Ware too in principle, and even if it turns out not to be, keep in mind that the notion that the people of the Afanasevo culture spoke an ancestral form of Tocharian is only conjecture; Parpola mentions that Mallory now seriously doubts it. (One reason why I'm sceptical that the Tocharian branch is derived directly from the language of the Yamna people is the Tocharian speakers' association with light hair and eyes; if the Yamna people weren't light-haired and light-eyed, per Haak et al. 2015, it seems difficult to explain where that trait came from without the Corded Ware as a link. Also, if the Anatolian branch is derived from the Corded Ware, how can the Tocharian branch – in view of its specifically western-IE affinities, if these are not simply all archaisms – not be?) It might be an earlier, 4th-millennium BC Yamna offshoot, whose language might not derive from PIE at all, at least not the stage that all IE languages (including Anatolian and Tocharian branches) are descended from, but an earlier stage of PIE. Be careful not to assume the common notion that Anatolian and Tocharian are not part of IE proper as cut and dried. There's a serious danger that the notion – taken over again and again uncritically – is becoming so entrenched that people engage in circular reasoning without actually stopping to examine if the notion is actually justified. Phonologically at least, Anatolian and Tocharian do not derive from a different stage of PIE than the other branches. --Florian Blaschke (talk) 12:26, 3 July 2016 (UTC)

Allentoft et al. (2015) doesn't even mention the Anatolians! They do write, though:
 * "It seems plausible that Afanasievo, with their genetic western (Yamnaya) origin, spoke an Indo-European language and could have introduced this southward to Xinjang and Tarim"

Which means that the Tocharians were part of the Asian R1a, but took over an IE language? That's relevant in regard to the Dravidians: also Asian R1a, and not an IE language, as far as we know. But then, how come such a high frequency of Z93* at the location of the ancient Afanasevo-culture, which was closely related to the Yamna-culture, that is, western R1a? (see ). Or is that due to later migrations? Joshua Jonathan  -  Let's talk!   13:51, 3 July 2016 (UTC)
 * OK, can you remove the misleading mention of Proto-Anatolians from there then?
 * The Dravidian argument is interesting, haven't heard it before, but the largest concentration of R1a (see figure) isn't anywhere in the Dravidian-speaking area (nor even in Sindh, which may have been Dravidian-speaking formerly) but in the Indo-Aryan-speaking one, and it's specifically R1a1a, so I don't get why you connect R1a with Dravidian-speakers. By the way, isn't Yamna identified with R1b and Corded Ware with R1a? That would support the notion that the Yamna people were ethnically distinct from the Corded Ware people, and that the Corded Ware culture arose from Northern European hunter-gatherers taking over Yamna/Late Tripolye culture and language (though with strong Yamna genetic admixture apparently) and spreading it farther.
 * What do you mean by Tocharians? The people of the Afanasevo culture? In view of the doubts above, I strongly advise against equating them. Just call them Afanasevo people. We just don't know in what connection (if any) they stood with the historical speakers of Tocharian. Maybe they had some genetic input, maybe not. As I said, it's possible that the Afanasevo people spoke a language derived from an earlier stage of PIE (or simply an independent branch of IE, maybe one that was closely related to but not a direct ancestor of Proto-Tocharian), which could mean that they eventually shifted to IE proper (Indo-Iranian?), which would have been easy because of the close linguistic relationship. However, considering Parpola's remarks on Proto-Samoyedic and Kallio 2010, there seem to have been two breaks (and hence potential language shifts): From Afanasevo (IE?) to Okunevo (Paleo-Siberian?) and from Okunevo to Federovo (considered Indo-Iranian). There seems to have been a further language shift, at least in the Minusinsk Basin, to Uralic eventually, specifically East Uralic/Samoyedic (connected with the Cherkaskul' culture), and later to (Yeniseian and) Turkic (and recently Russian). --Florian Blaschke (talk) 14:29, 3 July 2016 (UTC)
 * I am familiar with the issues that both of you are bringing up here. I'm actually an expert on the linguistics and Indo-European side, but I follow the paleogenomics and I know how to do a Monte Carlo Markov Chain simulation myself :)
 * While I'm definitely against oversimplifying, and I agree that we should not necessarily equate pre-Tocharian speakers with the Afanasevo culture, I think that the doubts about Tocharian's affinity with Afanasevo are misplaced; you're right that they could be cousins, rather than parent-child, but the likelihood that the Tocharians were more closely related to some other culture of that time period that has already been excavated is low. Malory's doubts are reasonable, of course, but they are only doubts. Neither he nor anyone else has a replacement theory. The doubts concerning agricultural terminology are particularly ill-founded, since the same sort of agricultural terminology was retained in Proto-Indo-Iranian and its daughters despite very thin evidence for agriculture in the southern steppe and desert regions. Even Berber languages retain agricultural terminology, since a subset of the mostly highly mobile population is always engaged in oasis agriculture. It is not as if Central Asia has been so thoroughly excavated that we should expect to have find traces of such small-scale agriculture yet. --Mellsworthy (talk) 21:56, 3 July 2016 (UTC)

Phonological and morphological stages of PIE and their paleogeography
(One of Florian's statements repeated from above, so that I could move this subdiscussion.) Mellsworthy (talk) 05:59, 4 July 2016 (UTC)

... Also, if the Anatolian branch is derived from the Corded Ware, how can the Tocharian branch – in view of its specifically western-IE affinities, if these are not simply all archaisms – not be?) It might be an earlier, 4th-millennium BC Yamna offshoot, whose language might not derive from PIE at all, at least not the stage that all IE languages (including Anatolian and Tocharian branches) are descended from, but an earlier stage of PIE. Be careful not to assume the common notion that Anatolian and Tocharian are not part of IE proper as cut and dried. There's a serious danger that the notion – taken over again and again uncritically – is becoming so entrenched that people engage in circular reasoning without actually stopping to examine if the notion is actually justified. Phonologically at least, Anatolian and Tocharian do not derive from a different stage of PIE than the other branches. --Florian Blaschke (talk) 12:26, 3 July 2016 (UTC)
 * @Florian I'm not certain why you seem certain that the the Anatolian languages, at least, don't represent a different stage phonologically from the rest of PIE. One could argue that the representation of medial voiceless stops as geminates is actually the original state of things... Not that I'm about to publish that, but it's definitely attractive to imagine PIE *tt > core-PIE *t.  Tocharian is so odd and merges so many things by the time it is actually written that it's impossible to use it in arguments about the details of its ancestor phonological system.  The one piece of data that could be of interest is the loss specifically of PIE *d (not *dh or *t) in many words, but I don't have a strong enough theory for that to say anything, just speculation. --Mellsworthy (talk) 22:05, 3 July 2016 (UTC)
 * It's not clear what the background of the geminates occurring in the Anatolian languages written in cuneiform is and what they might mean, not least because they occur only there and not in the alphabetically written languages, nor in Hieroglyphic Luwian (but that could be due to the defectiveness of the script). There are many ideas, but nothing definite. Melchert's reconstruction of the Proto-Anatolian phonology is almost identical with bog-standard PIE except for the merger of the rows of stops traditionally reconstructed as voiced aspirated and unaspirated (hence, seemingly a deaspiration as in several other branches), and the monophthongisation of *ei. (A loss of consonants is clearly not an archaism either.) Surely, if Anatolian had split off so early, there should be phonological archaisms as well, but I can't see any clear ones, and on this level at least it behaves like any other branch of PIE. Nor does Tocharian show any particular archaisms; instead it can be derived from standard modern PIE (with three laryngeals) just fine.
 * One interesting peculiarity I've seen Kümmel point out is evidence pointing towards short *o being "stronger" or "longer" than short *e (but still shorter than a long vowel), which seems to mean that short *e was [ɛ] but short *o a half-long [ɔˑ], and from typological parallels Kümmel argues that it is likely that, in a stage preceding the rise of vowel quantity in PIE (which is clearly a recent development within PIE), *e goes back to a pre-PIE short **a but *o to a pre-PIE long **ā. These clues are not found in the European branches, but they are found in Anatolian and Tocharian. However, in certain open syllables, as is well known, *o behaves different from *e in Indo-Iranian as well and shows up as a long vowel. But this is the kind of thing I'd expect to find if Anatolian and Tocharian were really outside of "traditional PIE".
 * Surely there are some clear archaisms in Anatolian found nowhere else, but the most convincing ones Kloekhorst mentions are the meanings of individual words, which prove nothing for this purpose, because they are consistent with the possibility that Anatolian simply did not take part in certain innovations shared by the rest (except possibly Tocharian) after the split-up of PIE, sweeping through an already slightly differentiated dialect continuum (say, in the mid-late 3rd millennium BC). But that's not what people usually mean when they talk of Anatolian (and Tocharian) being earlier splits. I can't see any clear evidence, though, that even evidently late sound changes reconstructed for PIE (such as various rules generating long vowels) didn't take place in Anatolian. Let alone that Anatolian vocalism must be directly derived from Kümmel's suggested pre-PIE stage!
 * Also, it is worth noting that from the verbal roots listed in the LIV, there is no evidence that Anatolian is somehow not a regular branch. Its pattern is consistent with the "core IE" branches.
 * This makes Anatolian rather look like Icelandic–Faroese among the North Germanic languages: Mainland Scandinavian has converged secondarily, it doesn't form a node within North Germanic. Genetically, Icelandic–Faroese is closely related to West Norwegian, which has secondarily developped into a direction (with simpler morphology and in some respects phonology) comparable to Swedish and Danish. (This seems to be also due to the shared Middle Low German superstrate, and also the long-standing influence of Swedish and Danish on Norwegian.)
 * So that's why I've never been convinced of the "Indo-Hittite" notion that so many people seem to take for granted now. After all, Indo-Hittite is usually taken as implying not that Anatolian is like Icelandic–Faroese within North Germanic, but like Gothic within Germanic. But the null hypothesis to disprove should not be Indo-Hittite, but Anatolian being a regular branch deriveable from standard PIE as usually reconstructed on the basis of the other branches. For, so far it looks like it can in practice be treated as such, and the success of that practice speaks for itself. At the very least, until "proof by assertion" stops being the strongest argument in favour of Indo-Hittite and compelling evidence turns up that can't be explained differently, one should be careful not to presuppose either scenario as established fact. (Nor should other pet hypotheses be taken as granted.) And the way that is done annoys me. --Florian Blaschke (talk) 00:40, 4 July 2016 (UTC)
 * Well, I agree that your viewpoint is reasonable. Certainly nothing in the published literature proves the phonological archaism of Anatolian. There are huge problems with the defectiveness of the writing systems in the earlier period, and the massive phonological changes in the later period that could mask a lot. In addition, there are a number of features that have traditionally been analyzed as innovations in Anatolian, but might be archaisms:  the final -s in the nominative of n-stems in Hitt., the retention of "thorn" as clusters like *tk.  There's also an interesting alternation between laryngeal and s-final roots in a few cases, which has been mysterious, but might just be another archaism (e.g. Hitt. (a)u(s) 'see', 1s ohhi, 2s autti, 3s auszi, etc.).  It's hard to be certain of phonological archaisms, since in many cases they can be produced by leveling or analogy, and in other cases the lack of a tertio comparationis simply leaves researchers too uncertain to publish.  That's certainly been my situation on these :)  --Mellsworthy (talk) 02:42, 4 July 2016 (UTC)


 * Yeah. I mean there is no question that Anatolian and all that jazz is a huge head-scratcher. I'm nowhere near qualified enough to even understand all the issues en detail involved deeply. But I'm aware that nobody has it all figured out which is why I so loathe premature declarations of certainty.
 * Though root alternations like the one you describe look rather less mysterious than other problems. In the case you mention, LIV does not reconst a final laryngeal in the root at all, so it just looks like an originally separate formation with an -s-suffix became integrated into the paradigm. After all, the ending -zi belongs to the mi and not the hi-conjugation, so this looks like 1sg. *(h1e-)h1óu-h2e, 2sg. *(h1e-)h1óu-th2e, but 3sg. *h1éu-s-ti (vel sim.; a desiderative?). I mean variant roots ending in *-s are not that rare in general.
 * As for Tocharian, if the evidence really points towards western affinities, or even ancient contact with the European branches (I'm myself not convinced that it does; I just wonder if the features in question can't all have other plausible explanations such as being archaisms – but I don't have a good mastery of the evidence right now), then the Afanasevo–Tocharian connection just looks unlikely. Because after all, the Afanasevo culture is so old that it sounds rather implausible that these supposedly specifically "western" features had already developped at the time, let alone that the differentiation had advanced so much that contact with recognisable other branches could be shown. But I'm not very confident on that point in general because I just don't know the evidence and arguments well enough, and the possible explanations that could be adduced.
 * Maybe the (in-?)famous "western affinities" are really just a chimera and Tocharian split off (nearly) as early as Anatolian, in which case of course the Afanasevo people could have been speakers of an ancestral form after all. But the source could either be Yamna or the Corded Ware.
 * I would assume that by c. 3000 BC, "standard PIE" was spoken from the Gulf of Finland to the Aegean and from the Oder (or even Rhine) to the Volga or even (with a gap in between?) to the Altai, both natively by highly mobile steppe nomads with wagons (who probably rode, too, even if direct evidence is lacking) and by others as a second language, which makes it plausible that for some time it could be spoken over such a vast area without any appreciable dialect differences. In fact, it is entirely possible, even plausible, that "non-standard" dialects collateral to "standard PIE", much like other Italic languages existed beside Latin, but were later absorbed. For example, if all IE branches can really be derived from the Corded Ware, "standard PIE" may actually have arosen adjacent to the Pontic steppe, to the north or northwest, perhaps in the Vistula basin. The original steppe people may have spoken more diverse dialects then and used "standard PIE" to communicate with people whose languages were unrelated (including, quite possibly, speakers of a pre-form of Uralic).
 * To speculate, this "standard PIE" could have been a western dialect of "Pan-IE", maybe spoken in the Dnieper/Dniester area, perhaps with a certain substratum influence from the local languages of the Vistula basin or wherever it may have arisen. I use a convenient cladistic analogy here: The "standard PIE" we reconstruct is only the MRCA of the "crown group" that includes all attested IE languages, but must itself have been part of a larger family, analogous to a pan-group. "Stem-IE", then, comprises all the quite possibly numerous (steppe) dialects that have not survived as they were absorbed when "Crown-IE" expanded. This would neatly make sense of the pigmentation puzzle. Even if the known branches do not all go back to the Corded Ware and the PIE we reconstruct was spoken in the steppe (or by the people of the Late/Post-Tripolye culture?) after all, it is useful to keep in mind that there is a vast wealth of languages and dialects, both unrelated and related to PIE, that must have existed prior to the expansion of (the final stage of) PIE and disappeared virtually without trace.
 * As for the genetic evidence, I cannot speak to it directly. I'm not a math/stats wiz and I cannot do a Monte Carlo Markov Chain simulation (whatever exactly that may be, I have only the vaguest idea about it) myself. I'm just a lowly linguist (well, not in the sense of formal employment or post-graduate degree) with archaeological interest on a very amateur level, whose mind has a stubborn emotional need to fit the linguistic evidence into a realistic scenario. --Florian Blaschke (talk) 03:23, 4 July 2016 (UTC)
 * Oh, I realised that since it is generally accepted that Tocharian is not nearly as divergent from the IE mainstream as Anatolian, it should also have participated to some extent in the non-Anatolian "core IE" Sprachbund or convergence. This would supposedly have happened in 3rd millennium BC Eastern Europe. In any case, it seems to make the case of an early split of Tocharian, as early and thorough as that of Anatolian (which supposedly had no or little contact with the rest of IE at least in the period in question, though after the split-up of Proto-Anatolian it may well have had contact with ancestors of Greek, Phrygian and Armenian – Anatolian, whose "centre of gravity" is decidedly west of the Halys, likely came from the northwest, and may have been spoken at the northern coasts of the Aegean originally and contributed to the "Pelasgian" Greek substratum, which may well have consisted of several distinct languages), much more difficult and less plausible. If the people of the Afanasevo culture spoke a language descended from standard PIE, and already in the late 4th millennium BC, and it was an ancestor of Tocharian, it is difficult to see how (if our model is generally correct) the contact with the mainstream or "core" of IE should have worked, as they seem to have been relatively isolated in Southern Siberia (despite the availability of horses and carts), and not in particularly close contact with early Indo-Iranian, either. That's why I think the Afanasevo connection is still uncertain and the precise origin and early development of Tocharian and the people who spoke it remains an unsolved mystery. Contact with Uralic is likely, but not before 2000 BC. --Florian Blaschke (talk) 16:43, 10 July 2016 (UTC)
 * (In what follows I'm again ignoring phonological questions, since these are vexed in both Tocharian and Anatolian, whereas morphology gives us more obvious data, even if people disagree on what it means.) I would sound a note of caution in two directions:
 * Firstly, we actually don't know how much pre-Anatolian diverged from the "crown group"; it may be that the ancestor of the Anatolian languages actually is more similar to "core" PIE, grammatically speaking, than most current researchers assume. I think, for example, that it is perfectly plausible that Anatolian simply lost the pre-subjunctive, pre-optative, and pre-desiderative, having acquired periphrastic means for expressing all of them, as has occurred frequently in other individual groups of PIE languages. Here, I'm only half serious since I think, on par, we do have good reasons to imagine that the common ancestor of Anatolian and the crown group are quite different, especially as concerns the perfect/hi-conjugation, the gender system, enclitic possessives, and sentence clitic subject pronouns. Some of this could have been substratal assimilation, but I actually don't think any of it actually is, except perhaps the above-mentioned periphrastic strategies for future, desiderative, etc.
 * Secondly, we don't actually know how much pre-Tocharian diverged from the "crown group"; while many seem to believe that Tocharian diverges little from "standard" PIE, and even adduce supposed signs of "Western" subgrouping or influence, I think that's probably a mirage. I think that Tocharian's grammatical archaisms are greatly masked by its phonological and grammatical innovations, which means that, so far, we still mostly just have question marks instead of indications that we need to revise PIE.  It remains to be seen whether the past tense with apparent o-grade plural and e- or zero-grade singular (see Malzahn's preterite class I.1-4 in The Tocharian Verbal System, pp. 188-136, 140-148, and 162), reversed from the PIE norm, can somehow be derived from PIE as normally reconstructed, for example.  It's also possible that Tocharian's strange patterns for forming feminine adjectives are actually archaic.
 * On the whole, I believe that Tocharian and Anatolian actually are quite divergent from "Core" PIE, but, as I mentioned above, I don't consider any of this as proven. --Mellsworthy (talk) 09:05, 11 July 2016 (UTC)
 * Yeah, I think we essentially agree here. These are still central open questions, and it's unsound to (conclusively, firmly) commit one way or the other and base far-reaching hypotheses on it. I'm only siding against Indo-Hittite and in favour of a more traditional view as a working hypothesis, per Occam's razor and also for practical convenience. Should it turn out that the Anatolian or Tocharian evidence cannot completely be explained on the basis of the standard reconstruction of PIE as based on the evidence of the other branches, and the other branches show no trace of the necessary revisions to the standard reconstruction, that would indeed seem a serious indication that there is something wrong with the traditional view of Anatolian and Tocharian as "yet more branches like all the rest" ... although, of course, archaisms are by themselves not probative for subgrouping, only innovations are, so you still need common "core IE" innovations to settle the issue, and further you should ideally explain how to get from the revised reconstruction to the situation in the core branches.
 * By the way, I thought I'd once heard that there were (possible?) fossilised traces of at least the optative in Anatolian, but I might be mistaken or misinformed. --Florian Blaschke (talk) 13:21, 24 July 2016 (UTC)
 * I hadn't heard of these traces of the optative in Anatolian, but I wouldn't dismiss it outright. The hypothesis of a split between Anatolian and the core-PIE languages does have some innovations to ascribe to the core-PIE side, the most important potential one of which is the feminine gender.  There is a lot of debate about whether the feminine was lost in Anatolian or innovated in non-Anatolian PIE, and while that would be gang-busters, I don't think the evidence is settled.  The more settled cases are mostly verbal morphology: desideratives (often becoming futures) in *-h1s(y)e-, subjunctive in *-e-, sigmatic aorists, and possibly the aorist vs. imperfect vs. perfect aspectual distinction itself.  The development of the sigmatic aorist is a really nice case, with only the 3s in Anatolian, generalization to some other forms in Tocharian, and a constant *-s- in Italic, Celtic, Slavic, Indo-Iranian, and Greek.  (Clearly the loss in Germanic is an innovation connected to the complete loss of the aorist.  The s-perfect was similarly lost in the Sabellic branch of Italic, again, crucially after a simple preterite was generated from a merger of old perfect and aorist forms.)  I have my own, slightly idiosyncratic theory about what's going on with the aorist/perfect/imperfect, but the subjunctives, desideratives, and the optatives you spoke of are, I believe, cases of univerbations of chain verb constructions that were possible in pre-Indo-European, with a last gasp of productivity in core-PIE.  In other words, as I'm not the first to say, I think these had simply been verbs meaning things like 'want', 'go', and 'be obligated' (the usual sources for such modal forms cross-linguistically), which were put together with a rule that phonologically subordinated a verb to a following function verb (V1(0) + V2).  Other examples are *-(i)ske- (something like 'actively continue, attempt' and 'cause to continue') and *-e- (developing into a simple present class from a 'go and X' type of construction), the first of which is found in Anatolian, alongside the possibly more archaic suffix *-(i)ss(a)- of the same meaning, which took hi-endings in Hittite.  The collapse of the bare chain verb construction is what led to the plethora of infinitive constructions in later languages, as various things rushed in to fill the void.  For non-verb examples, we could point to vocabulary innovations such as the word *kwekwlo- in core PIE, or meaning innovations like core PIE *seH1 'sow' (cf. Anat. *see 'put in'), but any individual one of these is dismissable; in aggregate, I think they are not.  I don't want my position misconstrued as not supporting the Hittite-first position. I do.  I also, however, agree with you that it is not proven.  Still, it has reached a level that is worthy of extended reconstructions and scenarios, all the while keeping in mind the further consequences of other theories (e.g. absolutely coordinate branching of all PIE branches) that cannot yet be dismissed. --Mellsworthy (talk) 22:56, 24 July 2016 (UTC)
 * Yeah, I'm rather resistant to Indo-Hittite, but I try to keep an open mind. I still think it's weird that there seems to be no clear evidence for divergence on the phonological level. So it's still quite conceivable, I think, that these morphological and vocabulary innovations happened in an already dialectally slightly diversified area (that could, then, be identified with the Corded Ware) – but of course it would be necessary to show this diversification somehow to settle the issue. In practice, anyway, working with rather standard reconstructions of PIE still works so well that there seems little point in reconstructing a distinct "Proto-Indo-Hittite" stage – I mean, it would be lovely to learn more about the development of PIE, so I'm certainly not principally disinclined to a "Pre-Proto-Indo-European" reconstruct based on external evidence; that would be mad!
 * As for the feminine gender: It's certainly interesting that non-Anatolian branches preserve evidently archaic traces of the "common gender" situation as well, such as the Homeric epicenes.
 * This illustrates a general methodical problem: Experience shows that attested proto-stages are, as a rule, more archaic than expected from the reconstructions based on descendants, so our reconstructed Proto-Greek, Proto-Indo-Iranian etc. may easily mislead us. "Core IE" innovations could still have gone on – and evidently did to some extent – even after the Proto-Greek, Proto-Indo-Iranian etc. stages, hence in the 2nd and 1st millennia BC. That's why the Vedic, Avestan, Homeric, Mycenaean etc. evidence is so immensely valuable, after all. Without it, we might well have no inkling of certain innovations that eventually spread to the whole dialect continuum.
 * One difficult problem is the existence of so many proposed synchronic phonetic rules of PIE – I find it hard to keep track of them. Many of them are quite special and have rather limited evidence, so it's often difficult to find Anatolian evidence to show us whether the rule is reflected there as well. However, so far, there do not seem to be any examples where Anatolian fails to show the effects of such a rule when there is relevant evidence after all. --Florian Blaschke (talk) 02:22, 25 July 2016 (UTC)


 * I'm not sure if you're familiar with Melchert's essay on the Indo-Hittite problem. Considering how widely you're evidently read on this subject, you've almost certainly read it but probably forgot the details. It sounds pretty authoritative, sound and hard to argue with to me.
 * I see that it's the subjunctive, not the optative, for which the supposed traces are.
 * Personally, I'm fairly contented with its conclusion. I'm of course biased as a sceptic but it does support in my eyes my attitude "much ado about nothing", or at least: not much. And it definitely does leave the possibility open that "non-Anatolian IE" or "inner IE" was never strictly a node, and that innovations in an already dialectally differentiated IE failed to diffuse to Anatolian because of its marginal location.
 * I'd place the immediate ancestor of Proto-Anatolian sometime between c. 3000 and certainly by 2700 BC in Thrace or so, with the bulk of IE, which may or may not have included Tocharian, in Central/Eastern Europe (forming the essentially "inner IE" Corded Ware community), so Anatolian could still have derived from the Globular Amphora community.
 * By the way, Parpola places PIE in the Post-Trypillian culture, 3500–3000 BC, although his argument that the wheel was really invented by the Trypillian culture – while a fascinating idea as it implies that the relatively advanced Old European Trypillians (or Funnelbeaker people?) furnished the Yamna with the crucial technology to "conquer the world" –, has been questioned (the famous "cow on wheels" is rather dubious evidence). --Florian Blaschke (talk) 05:19, 25 July 2016 (UTC)


 * I'm gonna fail to address some of your points for lack of time, but feel free to bring them back up!
 * * While I also agree with the paper, I think you shouldn't be misled by Melchert's cautioning tone. When he delivered an early version of this paper at Berkeley (a long time ago, don't ask!), he responded to push back from the audience by saying that he doesn't think there are no innovations shared by all non-Anatolian, just that it's unclear what to make of them. I'll endorse Melchert--he's generally a careful dude--but you shouldn't think he is specifically a doubter of the basic Core-PIE vs. Anatolian. (He wisely doesn't commit to anything about Tocharian.)  He certainly admits that the feminine is an innovation of non-Anatolian now, so really we're all just haggling over what the differences are between Hittite and Core-PIE and how we can prove them.  I don't trust Melchert on everything, mind; he shouldn't have swallowed Jasanoff's business about the relics of the subjunctive, since it just ain't so. You also should not imagine that Melchert would be all that friendly towards a version of things where the feminine gender was innovated and then spread, via the wave model, through already differentiated PIE languages.  The biggest problem with that is geography, which is not your friend after late Yamnaya when the Tocharians, with their three-gender system, have already hared off to the wild blue yonder, and it just gets worse after Corded Ware.
 * * This is complicated and seemingly tangential, but I believe very relevant to GAC and its relation to Yamnaya and PIE. I think that advances in archeogenetics are going to call a lot of Gimbutas' detailed conclusions into question.  We have already seen that there are no large genetic differences in Corded Ware populations sampled so far, despite her belief that some areas had a small percentage of Yamnaya descendants and others none.  She way underestimated Yamnaya ancestry in Corded Ware.  In totally the other direction, her "Indo-Europeanized" Baden culture looks genetically like Oetzi and the Hungarian late Neolithic genomes, retaining full continuity with Neolithic genomes, other than a resurgence of affinity with the preceding Hunter Gatherer cultures, and, crucially, with no detectable affinity to Yamnaya.  (See Figure 3 of the Haak et al paper, where HungaryGamba_CA represents the Baden culture, with estimated about 80% neolithic admixture, 20% Loschbour, and Yamnaya nuthin'.)  Even the later Bronze Age samples from Hungary from the same paper have less than 20% Yamnaya admixture, in stark contrast to Corded Ware's 75% to 100% well before this.  This to me says that Alpine Europe and Southern Europe and not going to represent the kind of population replacement seen in Corded Ware.  Firstly, this calls into question the whole idea of whether Gimbutas' methodology was consistently picking up on the real signals.  She could have been mostly right for all the wrong reasons.  Secondly, while these errors do not demand particular conclusions, they do invite them.  What if the seeming suddenness and quite low admixture of Corded Ware is exactly what it seems to be: a very sudden population intrusion, probably after some form of population crash or other terrible disaster befall the earlier inhabitants?  I'll say more about Alpine genetics and potential troubles for the Celts in my response to your statements about Celtic from the West below.  --Mellsworthy (talk) 09:25, 25 July 2016 (UTC)

R1a1
About "Asian" R1a (i.e. Z93/M746, R1a1a1b2) vs. "European" R1a (i.e. Z282/280, R1a1a1b1), it's complicated. I know there's a lot of animus about the various Y subclades (do they really call them tribes? yes they do.), and I hesitate to add to the vast overuse of electrons and keystrokes devoted to this topic, but I'll do it anyway, since you're interested. It seems you've read Underhill et al 2014, who definitively identifies the above two branches as belonging to Asia and Europe, but this paper did not know about L664, which differentiates R1a1a1a from R1a1a1b. This is why his name for the two branches differs from the one I give, and which is (correctly) reflected in the current version of Haplogroup R1a1--he didn't know about the upstream marker. Interestingly, archeogenetics now shows that all three major branches of R1a1a1 are attested in earlier populations of Europe: The earliest example of R1a1a1b2 is now found in a sample from the Poltavka culture (a Yamnaya derivative), R1a1a1b1 is found in Corded Ware, and the branch unknown to Underhill, R1a1a1a, is also found in Corded Ware. The simplest interpretation is that R1a1a1, the ancestor of all three, originated in Europe and one specific branch escaped with the Indo-Iranian/Andronovo and Afanasevo expansions from the general North Pontic area, even though Underhill et al probably rightly locate the origin of R1a M420 in Iran. What this means is that R1a and it's subclades are European or W. Asian at various timescales, and we shouldn't essentialize them, or call them "basically" European or "basically" Asian, except as a shorthand. --Mellsworthy (talk) 21:56, 3 July 2016 (UTC)
 * Thanks. The Haplogroup R1a1 article needs further refinement; the tree in there is unclear, and doesn't use the R1a etc naming.  Joshua Jonathan   -  Let's talk!   04:59, 4 July 2016 (UTC)
 * Yeah, the article could be better done. In particular, the latest paleogenomics results are not reflected, even though they're quite interesting. Guess that's on my to-do list. --Mellsworthy (talk) 06:43, 4 July 2016 (UTC)

More musings about Allentoft et al. & languages
I was just about to react to your notification in Talk:Corded Ware culture and started to think and type, but then I realised that my musings weren't really connected to improving the article, so here it goes:


 * I see. But the genetic evidence that appears to show a highly significant difference (only phenotypical, in the form of light pigmentation, or also in terms of haplogroups?) between the people of the Yamna culture and those of the Sintashta and Andronovo cultures, and instead a far greater similarity to those of the Corded Ware, while of course not directly proving anything about linguistic descent, is highly indicative of a significant level of genetic influx ...
 * And given that this is ancient and not modern DNA, important sources of disturbance are eliminated (all the migration going on since the Bronze Age), so there is a much greater chance that there was some linguistic implication as well, at least in the form of an adstrate or so, if not direct descent.
 * The point is, my understanding is that light pigmentation is recessive, and if it's so frequent or even pervasive in Sintashta/Andronovo people (which fits with other indications that Indo-Iranian speakers were originally largely lightly pigmented), the contribution from the direction of the Corded Ware must have been really large to swamp out the darker steppe phenotype (not just a minority – and via élite dominance, even an immigrant minority can change the linguistic makeup of a culture, after all –, so the situation would seem to be analogous to the relationship between CW and Yamna before where there was a strong genetic influx, hence mass migration, as well). And in that case I have trouble with the conventional notion that the language of the Sintashta/Andronovo is directly descended from the steppe people instead of the Corded Ware component, even though it's of course not impossible that the immigrants from the west simply blended in and assimilated linguistically, so the influx would be meaningless for the linguistic origins of Indo-Iranian (in fact, however, there does not seem to be any remotely compelling evidence for a "northwestern Indo-European" component in Indo-Iranian, either).
 * I'm not really sure what to make of all this in linguistic terms, either. I'm reminded of the reconstruction of two layers of Indo-European loanwords in Proto-Uralic by Jaakko Häkkinen, one associated with early stages of Indo-Iranian, thus dating to the 3rd millennium BC, and one which he considers contemporary and links to a "Northwest Indo-European" type which is still close to Proto-Indo-European. Asko Parpola identifies the second type with the Corded Ware and especially the Fat'yanovo–Balanovo culture, although the Sintashta and Andronovo is identified by him not only as Indo-Iranian (as is mainstream), but specifically as Indo-Aryan. (The locus of Proto-Uralic Häkkinen places near the Volga–Kama confluence c. 2000 BC because of the Indo-Iranian layer already in the Proto-Uralic stage, though Parpola locates Proto-Uralic further west and earlier.) So my understanding so far was that Indo-Iranian is traced (presumably via the Poltavka culture) directly to the Yamna culture of the steppe, and certainly not to anywhere farther north or west. But now I wonder what the "northwestern"/CW contribution in Sintashta (which appears to have been assumed already earlier on archaeological grounds) really means. (Is there a non-R1-component in CW- and Sintashta-associated ADNA, such as I1?) If the people of the Fat'yanovo–Balanovo (and Abashevo?) culture were really speaking some ancestral stage of Indo-Iranian instead of "Northwest Indo-European", and Proto-Uralic-speakers may not have had direct contact with this branch, there remains the possibility of the alternative interpretation that the "archaic Indo-European" loanwords could equally have entered Uralic much earlier than 2000 BC – possibly directly from (Pre-)Proto-Indo-European, wherever exactly it was spoken, or (from Pre-/Para-IE?) from the steppe (such as through the Poltavka culture).
 * You know, the point is that Proto-Indo-Iranian and Proto-Balto-Slavic do seem to be close, so it's conceivable, IMHO, that there was a common node even if there wasn't a particularly long phase of common development and Balto-Slavic was later drawn into the "Northwest European" fold (but the allegedly close relationship between Balto-Slavic and Germanic seems to have been largely contact-based and is IMHO wildly overstated – Eugen Hill, who has shown that the *-m/bʰ-isogloss, commonly adduced as evidence against the traditional satem affinities between Balto-Slavic and Indo-Iranian and for a closer affinity of Balto-Slavic to Germanic, is probably not probative after all, has even remarked, IIRC, that the branches are about as different as two branches of Indo-European can be, and we can reconstruct for Proto-Balto-Slavic a fair amount of the verb-morphological complexity typifying Indo-Iranian, including even the *-sye/o-future, while Germanic has at best traces of that complexity, if it ever had such an elaborate system at all), and in archaeological terms this could mean that Indo-Iranian originated together with Balto-Slavic, say, at the eastern end of the Globular Amphora area near the middle Dnieper, and spread to the Upper Volga basin from there instead of directly from the steppe.
 * By the way, and I'm going off on a little tangent now, the existence of traces of endings like *-mos, *-bʰi and presumably even *-mi in Anatolian, especially considering the overall evidence, which points to endings dat. pl. *-m(y)os (~ *-bʰ(y)os), instr. pl. *-mis (~ *-bʰis) and instr. sg. *-mi (~ *-bʰi) in the ancestor of all IE branches including Anatolian (and Tocharian!), apart from spelling serious trouble for any argument based on the *-m/bʰ-isogloss, refutes the view that the Anatolian situation, where nominal morphology shows only an ending *-os for all oblique, non-central cases in the plural, is original and the situation in "core PIE" is an innovation, with the labial endings somehow being turned from numerus-insensitive clitics (along the lines of the extra local case endings in Homeric Greek) to full case endings.
 * In any case, I'm simply unconvinced that Germanic and Balto-Slavic were particularly close (and that Celtic and Italic were particularly close to either branch), so why the Corded Ware (or its Globular Amphora predecessor) should be identified specifically with the origins of Germanic and Balto-Slavic (and perhaps Celtic or even Italic as well) and no other remains not particularly obvious to me, and the identification as the real locus of (Late) PIE makes more sense to me. Or, alternatively, as the locus of only Germanic origins, though that does not preclude the possibility (which is actually quite plausible or even inevitable) of the former existence of other, now disappeared (except for possible borrowings in Uralic), "Northwest Indo-European" dialects, possibly more closely related to Proto-Germanic than to other Indo-European branches, whose origins also lay in the context of the Corded Ware. --Florian Blaschke (talk) 14:18, 24 July 2016 (UTC)
 * I agree that the inference that the language of Sintashta is likely to be more of Corded Ware than of Yamnaya is pretty reasonable, but I think that people are having to chew on it. I also agree that Balto-Slavic is very close to Indo-Iranian, and that there was no "NW-Indo-European".  (I don't agree with everything in the paper you cited, though it may be correct, overall; I have my own crazy reasons for believing the *mi vs. *bhi distinction was poppycock.)  The main problem with a Corded-Ware-spoke-Core-PIE theory, as I see it, is Southern Europe.  Nobody knows anything useful for this question about Albanian (it does not have an archaic retention of *kw vs. *k vs. *k', for example, despite claims), so I'll leave that out for the moment.  I have my own private theory about Italo-Celtic, related to the Celtic from the West thesis, and ultimately deriving it from a Bell-Beakerification of Corded Ware--but the archeological theories for how Greek got where it is are totally murky to me.  In addition, the linguistic affinities are totally bizarre:


 * Half of Greek dialects show the isogloss 1p verb ending *-men with Anatolian, and the other half of dialects just have Standard PIE *-mes.
 * The middle verb is also more like Anatolian, with *-dhH2 endings, but even more than Anatolian, showing no blending or addition of the *-a(r)i, *-o(r)i, *-o(r) set of endings to the original *-dhH2 set in 1st/2nd du/pl and the 3rd du.
 * Together with Armenian (with which it does form a group, pace Clackson), it shows its own special archaism and innovation with respect to vocalization of initial laryngeals.
 * Caveat: Don't believe people like Kortlandt and Beekes who set up a special status for Greek 3s -ei, though. That is the consequence of a special, early palatalization that I've written about, which affected other words like kai 'and' < *kati (cf. Arcado-Cypriot kas) and poi 'towards' < *poti (again, cf. Arcado-Cypriot pos).)
 * There are also resonances in poetic language with Indo-Iranian, such as *klewos *n-dhgwhitom 'imperishable fame'.

This last might well just be due to the fact that both language groups have poetry of the same type attested very early. As far as I can tell, everything that is ascribed to Indo-Iranian + Greek connection, I would ascribe to archaic core-PIE, which was greatly simplified in the West (although we can show at least some of this simplification happened in the individual histories of Germanic, Celtic, and Italic), and somewhat simplified in Balto-Slavic. People hesitate to include features in PIE on the basis of Greek and Indo-Iranian alone, and so assert the commonalities as influences, but I think that will turn out to be illusory. What's left over are the features mentioned above, which say that Greek went its own way very early, although I wouldn't dismiss a partially shared journey with Anatolian, or even a Core-PIE overlay on an Anatolian-like substrate. (By which I do not mean to subscribe to the Anatolian substrate theory as classically encoded, which had Anatolians in Greece itself. I would rather think that the overlay happened much further north, though probably still in the Balkans, and Greece remained at least partially non-PIE until long after Mycenean times.  I don't actually understand how the timing should work out in detail, but I saw a recent mention from a Mycenean scholar ascribing the first Greek entry into Greece to Early Helladic III (abt 2000BCE), which is probably right..

Including speculatively the early splits of Anatolian and Tocharian, and tons of other speculation, the affiliations as I see them are:


 * 1) PIE spoken in Yamnaya territory
 * 2) Yamnaya splits
 * 3) Anatolian from early Southern Yamnaya, via the early Balkan intrusion (no genetic data as yet!)
 * 4) Afanasevo (ultimately leading to Tocharian) from the other end of Yamnaya (near 100% archeogenetic match )
 * 5) Core-PIE is spoken by Corded Ware people (about 75% archeogenetic match to Yamnaya)
 * 6) Corded Ware splits, none of which are genetically confirmed, per se:
 * 7) Germanic group ultimately mostly derived from Battle Axe facies of Corded Ware
 * 8) Italo-Celtic mostly derived from Northern Bell Beaker overlay on Corded Ware and ultimately spreading due to bronze and tin trade starting about 2000BCE.  Potentially and partially confirmed by archeogenetics and by studies of non-metrical dental traits and retention of Corded Ware common ware in the North, showing continuity of the Northern Bell Beaker population with Corded Ware.
 * 9) Somehow, Italic gets all the way to Italy, despite a wall of non-PIE people in Southern Europe (ancestors of Basques, Tyrrhenians, and Iberoi).  I interpret Italic languages to have arrived by sea with the Proto-Apennine culture, heading out from Spain.  Early Proto-Apennine, despite its mountainous name, is found first in scattered fortified sea-side sites in Southern Italy.  This is my own theory--as far as I know, there's no archeologist claiming this, and unfortunately I am not actually qualified to make the claim.  I think people have avoided the association, even if they are migrationist, because Gimbutas fetishized the single grave association with Indo-Europeans so much, and Proto-Apennine people definitely went in for collective burials.  We obviously can't push such associations too far, or Western European medieval ossuaries would be evidence for relic non-PIE languages in Paris and London!  Anyway, I have an elaborated hypothesis on this that would be a whole other thing.
 * 10) Celtic and Para-Celtic Lusitanian remain in the west, along the Atlantic, until they start spreading eastward in the Late Bronze Age (essentially Celtic from the West thesis)
 * 11) Eastern Corded Ware, leading to core-Satem group
 * 12) Clearly, Corded Ware leads to Sintashta leads to Andronovo, speaking some form of Proto-Indo-Iranian language, as per standard theories (archeogenetically confirmed in detail)
 * 13) Although I don't think it's entirely clear, I'm willing to believe a sequence Corded Ware to Trzciniec culture to Lusatian Culture for Balto-Slavic.  I don't believe that the split in Balto-Slavic is older than that.  There is indirect genetic evidence in the high modern affinities amongst Slavic groups for Corded Ware aDNA.
 * 14) Greco-Armenian may have resulted from the scrum that occurred in the Balkans about the same time as Corded Ware, and during and post-Yamnaya, but there's no archeological synthesis, much less genetic evidence.  Seriously, ?????  And Albanian????????  Other minor groups??????????????


 * At any rate, that's my speculations based on my professional, but particular reading of the linguistics, and my non-professional read of archeology and genetics. --Mellsworthy (talk) 20:44, 24 July 2016 (UTC)


 * A couple of comments:
 * I don't understand why Southern Europe is a problem. Personally, while I can't judge the position of Tocharian or even Anatolian linguistically and archaeologically, I'm tempted to declare the Globular Amphora culture the locus of PIE overall, and the origin of all IE branches, pure and simple. (Considering that I'm neither Polish nor from anywhere else in the GA area, which per EIEC does stretch as far as the Elbe and even into Bohemia, and neither are my ancestors as far as I know, I don't even really have a patriotic horse in that race. ;-) ) And I think the final breakup of PIE could be as recent as the first quarter of the 3rd millennium BC, with a short period where PIE was spoken over a very large area but still essentially unitary, and largely a lingua franca-type second language with scattered native speakers like Slavic in the Balkans in the 7th century, though in its GA core area it may have been the main language for most people. (4000 BC is definitely too early even with Anatolian and Tocharian, especially because of Hittite hissa- "pole, shaft, thill" and iukan "yoke".)
 * By the way, I have my own crazy idea that posits an Albano-Armenian branch, which would be pretty handy. I suppose it would simply be the same as Daco-Thracian, as Dacian appears to be very similar to Albanian per Duridanov at least, and Thracian might be closely related to Armenian and the origin of Armenian could plausibly be in Thrace. I like the Graeco-Phrygian idea too, although there are a few hangups. It would result in a neat "Kentum Balkan branch" opposed to a "Satem Balkan branch". Obviously Balkan Indo-European is a big problem because we know so little about the ancient Balkan languages except Greek.
 * Frankly, I just don't buy the Celtic from the West hypothesis. It just makes no sense to me, especially in light of Italo-Celtic, which Schrijver already supported in 1991 with a list of old innovations that I find convincing (and there could be some he only mentioned in class and that aren't listed there; I'd have to check). IIRC, Schrijver has also suggested, more recently, that Proto-Celtic may not have been spoken in Austria or Switzerland or thereabouts, but in Southern France, where a Basque/Aquitanian-related substratum was responsible for the development from /p/ to /ɸ/. Not sure if I buy that one, but it's more western and still close enough to the Alps, which all the homelands I find plausible are situated around. I think it's, among other things, the lack of /p/ typical of Paleo-Hispanic languages that makes people think of the west. I'm doubtful that Lusitanian is relevant because it actually looks more Italic than Celtic-like (even specifically Sabellic because of the *kʷ > p development). The reasons to place Proto-Celtic in Central Europe look pretty good to me, and any arguments to place it much farther west, let alone in the Iberian Peninsula or the Atlantic seaboard of France (or even the British Isles!) have failed to impress me. I suspect that Italo-Celtic was originally a dialect spoken at the western rim of the IE dialect continuum, so probably somewhere in Southern (?) Germany or along the Rhine. Maybe in the Netherlands! It could very well have been involved in the Beaker phenomenon somehow, although there appears to have been a folk-migration from the Tagus estuary to the Rhine, where the migrants encountered IE-speakers. I remain doubtful if the whole area of the Beaker culture, which after all stretched all over the west Europe, was associated with IE; the original migrants in Portugal were hardly IE-speaking. But the eastern half in Central Europe was probably IE. Maybe Proto-Italo-Celtic did spread from the Rhine over a considerable part of Western Europe after all, leaving ancestral stages of Celtic (and perhaps "Sorothaptic"?) in Southern France, of Venetic/Italic in the Padanian Plain, and of Lusitanian in the Iberian Peninsula? I'm really unsure here, even in "crazy-ass speculation" mode.
 * I agree that Anatolian came across the western route and with the Greek connection. I would simply place Proto-Anatolian in Thrace (c. 2500 BC), from where it could spread to western Anatolia via the Bosphorus (I may have mentioned already that the diversity of Anatolian is all west of the Halys) and at least as far as the north of Greece, if not central Greece as well. Parnassos just looks utterly Anatolian (apparently a town Parnassa is indeed attested in Anatolia), and is unlikely to be the result of some late influence from Anatolia – although it certainly looks specifically Luwian or something, not like some separate branch. While I'm not claiming that the Pre-Greek substratum – which could easily have included IE and non-IE languages – was all Anatolian, even Beekes admits that Anatolian did play a role, and the presence of an Anatolian substratum or adstratum in Greece could help explain certain similarities to Anatolian, especially similarities that are limited to certain dialects, such as the -men ending in particular, and perhaps the assibilation of dentals before /i/. My idea is that Greek was originally close to Italo-Celtic (Rix's law is shared by Italic, but awkwardly Celtic does not seem to have it, which may or may not be a problem for Italo-Celtic): Phrygian could simply have preserved the *-tor ending from a common Graeco-Phrygian stage, and *-toi etc. might be an innovation after Greek had split off. Otherwise, however, I agree with Beekes that there was a substratum in Greece that was non-IE and similar (and probably related) to Minoan. Probably part of the "Old European" substratum, which does remind typologically of West Caucasian (and Hattic, apparently, so if a language family spread from Anatolia to Europe with the earliest agriculturists, it was this Hattic-Minoan one). Maybe contact with Anatolian affected Phrygian and Armenian as well, helping to explain some similarities.
 * Something I've got trouble with is making sense of the Graeco-(Armenian-)Aryan affinities. There are a lot of cross-cutting isoglosses and affinities, but if PIE expanded so quickly, contact between distant branches (despite of course possible thanks to horses, carts and wagons) should have been relatively small once the branches were established in their new homes and the first dialectal developments had taken place. And if Greek and Armenian were spoken south of the Danube and Indo-Iranian in Eastern Europe, how could common development between them have taken place? I find it very difficult to come up with a plausible scenario that explains all these affinities between often widely separated branches of IE. Maybe they really had some "Circum-Pontic" sprachbund thingy going. I agree it's also quite possible that Graeco-Aryan is simply bunk and an artifact. I never really liked it.
 * My favourite explanation for the Greek -ei and -eis endings is a series of two proportional analogies. First, in Proto-Greek, you have 2sg. primary *-esi, 3sg. *-eti, 2sg. secondary *-es, 3sg. *-e after *-d has been dropped. Through the proportion 2sg. secondary *-es : 2sg. primary *-es-i = 3sg. secondary *-e : 3sg. primary x => x = *-e-i you get the Mycenaean Greek stage: 2sg. primary */-ehi/, secondary */-es/, 3sg. primary /-ei/, secondary */-e/. In Ancient Greek, this results in the system *-ei, -es, -ei, -e with that awkward homonymy, so it is eliminated by way of a further proportion 3sg. secondary -e : primary -ei = 2sg. secondary -e-s : primary x => x = -ei-s. Neat, huh? No extra sound laws needed.
 * I agree that the Balto-Slavic split is late. 1500 or even 1700 BC sounds implausibly early to me. However, I wouldn't place it later than 500 BC, either. Balto-Slavic was certainly spoken at the eastern bank of the Vistula by the first centuries AD. So the Lusatian culture is an excellent fit for Proto-Balto-Slavic. They even had contacts with and influences from Hallstatt and La Tène, allowing for possible Celtic loans which have been suggested! This would also neatly fill a geographical gap between Germanic/Celtic and Iranian. And of course that's what loan contacts and similarities also suggest where Balto-Slavic was spoken. So it's again Poland, which may have been the (Core-?)PIE homeland as well – Balto-Slavic always turns out to be close to the homeland. And we're certainly nice to the Poles here by lavishing them with so many homelands! Well, Poland plus parts of Belarus and Ukraine, and other adjacent areas, but always centred on Poland; I guess because it's agriculturally very productive, at a crossroads in Europe, and quite suitable as a centre of expansion. (It's easy to forget how big Poland is, too.)
 * Armenian does look like the closest branch to Graeco-Phrygian (which would include Macedonian), so my Armenian-Albanian idea could simply be wrong, but I have an inkling that there might be a rule that deletes prothetic vowels (as in Medieval Greek and Armenian, only more thoroughly), because I've seen the argument that the augment was there but later lost though not always without trace – I recall something about a lenited dh- instead of d- interpreted as a last trace of the augment, although that is admittedly fairly flimsy evidence. I would love for it to be real however, because it could mean that Graeco-Phrygian and Albano-Armenian are both nodes and may form a node together (although that one would cross-cut kentum-satem, which I've still got a soft spot for), which would then be the core of Balkan Indo-European. However, Greek still seems to be unique in having the three-way distinction not only in the word-initial but also the word-internal position when a laryngeal is vocalised between obstruents. I think Armenian and Albanian have both only /a/ there like the other European branches? No idea about Phrygian, though. --Florian Blaschke (talk) 00:51, 25 July 2016 (UTC)
 * Oh wait, I just remembered that Armenian does not show the three-way distinction in word-initial position either, only a-, as far as I am aware. Only Phrygian seems to show it (word-internally too). So Armenian is clearly not as close to Greek as Phrygian is. Too bad Armenian and Albanian are so precariously attested. --Florian Blaschke (talk) 03:52, 25 July 2016 (UTC)


 * I'll answer these approximately backwards:


 * Armenian has a complex retention of initial laryngeals, but it does not merge them all to a-. The prothetic laryngeal before initial *r, reconstructed as *H1, shows up consistently as Arm. e-, as in erek 'evening' < PIE *H1regwos.  There's also a trace of *e in inn 'nine' from virtual PIE *H1nun, an inexplicably Armenian form of normal PIE *H1newn.


 * While I have an opinion on Greco-Armenian, I just don't think we know enough about Phrygian, old or new, to say much, even though I mostly agree with Lubotsky et al.'s reinterpretations of the corpus. Albanian I don't have a strong opinion on.  Albanian linguistics, despite the work of Hamp, is not going to be the way to prove things about PIE for a long time yet.  Just to cover my bases, I have no opinion on Thracian either.  Ancient Macedonian, I have some weirdo hypotheses about, but not too weird...


 * Yeah, the Poles luck out when it comes to homelands, except for the fact that the same thing that makes them a likely homeland makes them get invaded all the time. (See Johanna Nichols' work on spread zones sometime.)  But beware of the wrong form of uniformitarianism:  Before the invention of the heavy iron plow, and especially coulters to turn the soil, it was mighty hard to get a good harvest out of Slavic lands, since the soil is heavy and thick, not thin and sandy.  Agricultural communities set up where there was appropriate soil, and the number of such places in Northern Europe is small.  The early Balto-Slavs were probably heavier on animal husbandry, like many Yamnaya descendants.


 * I have published a different explanation for Greek *-ei, etc. here. It's a shame that I had to cut it down so short, since the longer version discusses, as others have before, just how unlikely the analogy that you (and others) have suggested for *-ei is.  Firstly, we have zero examples, in historical contexts or dialects where we know what's going on, of 3s reformed on 2s alone.  That's not "close to zero", but actually zero.  Analogy is mostly driven by frequency (to regularize rarer forms), and secondarily to avoid homophony or other complexities; neither purpose is served by the first stage of this analogy, and no Indo-Europeanists would ever have considered it, if there were some plausible alternative.  I show in the paper that we need the relevant sound law anyway, so there is no need for the completely unparalleled pattern of analogy.  There are other troubles, including a dialect 2s thematic present tense in -es, homophonous with 2s thematic imperfect -es, that simply does not fit into any reasonable analogy.  It also cries out that there's some form of chopping happening at the end of finite verb forms, presumably because of utterance-final position, lack of accent, or both.


 * Well, I hope our mutual dislike of Indo-Irano-Greek is born out by more positive and definitive facts in the future. Right now, I just have a pile of things that seem to be otherwise explainable, vz. archaism, and Ockham is telling me to cut.


 * So, Celtic from the West. Adding a new section.  --Mellsworthy (talk) 10:46, 25 July 2016 (UTC)


 * OK, just a few remarks:
 * Ah, yes, I forgot. True, Armenian seems to distinguish the first from the other laryngeals at least word-initially before consonant. (Word-internally too? I have no idea, and maybe nobody else has, either.)
 * Well, to put it most cautiously, from what we can tell, the ancient ancestor of Albanian had a lot of affinities and contacts, though especially resembled Core Satem, i. e., Balto-Slavic and Indo-Iranian, and on the other hand the Graeco-Armenian complex (which appears to include Phrygian and what not). So maybe it was originally part of that complex, possibly secondarily affected by Balto-Slavic, although it might also have been the other way round. In any case, it looks pretty central from the point of view of IE dialectology, and it makes sense to seek its origins in between Balto-Slavic (or Core Satem in general) and Graeco-Armenian (which may be traced to somewhere south of the Lower Danube, tentatively, but that's all I'm ready to conjecture so far). Balto-Slavic we have provisionally fixed around Poland and the western Ukraine, and what do we find immediately to the south in antiquity? Dacian. Just saying. It just fits together quite well. Maybe Dacian was really essentially the ancient ancestor of Albanian.
 * Good point on the spread zones and the invasions, although I don't know if that was already true in ancient history too. My suspicion is that the best agricultural land in Poland is found either in the south, or rather around the middle course of the Vistula, south of Warsaw. The north, especially the northeast, appears to have been dominated by foragers for a long time, I think because it was swampy – but that also allowed the Balts to remain largely undisturbed in that corner. Anyway, animal husbandry was the obvious solution for the dearth of good soils, of course.
 * OK. Sounds good. I don't mind that analogical explanation being refuted, it was just the best solution I knew so far. I'm not going to dispute your analysis. Anyway, we already agreed that the ending isn't old. :-) (A reeeeally minor nitpick: I understand that phonetically, a palatalised [tʲ] is most likely to change into an alveolopalatal [tɕ] rather than a palatoalveolar affricate, compare Polish, Irish or Japanese, so I'd rather write ć than č, and the change to [ts] is typologically well-attested too, IIRC. Not that it really matters.) --Florian Blaschke (talk) 03:24, 28 July 2016 (UTC)
 * You're 100% correct about ć rather than č, but I used č because of the Indo-Europeanist tradition of avoiding ć except for the outcome of the palatal velars! That system obviously doesn't work in Slavic where you have to distinguish a zillion different kinds of palatal this that and the other, and in fact the most comparable case for the change all the way to post-vocalic i/j comes from Polish forms like ojca < *oćca < *otьca 'father's'.  There were actually several layers of palatalization in Greek--at least three--so we should use more nuanced symbols there too, not least because reconstruction should pay a lot more attention to phonetic detail than it usually does. --Mellsworthy (talk) 06:12, 28 July 2016 (UTC)

Italo-Celtic from the What?
While I agree with you about Italo-Celtic, and I agree that more isoglosses could be adduced, my skepticism about the traditional theory for Italo-Celtic origins boils down to the following: It is really implausible that the Etruscans and Iberians emerged from anywhere other than near their homes in Southern Europe. Etruscan DNA is very consistent with European Neolithic, and not particularly with more recent Anatolian--in other words, they were not recent immigrants. They were quite like Oetzi, and modern Sardinians, and not that like modern Italians in general, although particular isolated communities retain continuity with Etruscan genotypes. This means several things:


 * 1) Etruscans, and their fellow speakers of Tyrrhenian languages of the Alps and Lemnos, originated more-or-less continuously from the descendants of the Cardium Ware culture, the first farmers of the region.
 * 2) The presence of Rhaetic in the Alps, and the connection of the Etruscans to the Villanova culture, furthermore traceable to the Terramare culture and the Urnfield culture and back to the Polada and the Alps, all seem to indicate that Northern Italy was a pre-Indo-European holdout until probably the end of the Bronze Age, with pockets continuing into recorded history.
 * 3) The Iberians also seem to be first farmer descendants.  I don't think anyone has a different theory.
 * 4) Both Iberians and the Etruscans are the most direct cultural descendants of the Urnfield culture that we have in history.  There's a complicated discussion we could have about why we probably should ascribe the divergent burial traditions vs. cremation traditions in final Bronze Age Italy at least partially to a cultural divide between Southern Italic speakers(!) and Northern Tyrrhenian speakers.  While the Celtiberians immediately neighboring the Iberians cremated their dead, none of the other Celts of Iberia did.  The extent of the Urnfield culture in Iberia matches very nicely with the non-PIE-speaking Iberians, and it is perverse to imagine this culture brought Celtic and Lusitanian, and then quickly was overwhelmed by a different group, the Iberians, with no evidence of cultural break, and with no plausible source for the population. Solid philology points out that the diversity of the increasingly comprehensible Iberian language is probably more in the north, including southern France, than the south.  If you read Spanish, see Ballester's contribution in Palaeohispanica 14 here.  Particularly fascinating is his analysis of the names in the Ora Maritima (p. 71), in which we see that the port-of-call names south of the Ebro around the year 500 look Indo-European, perhaps Celtic, and certainly not Iberian, while in the later Roman period, there are many Iberian city names there, as we can see from lots of coin data and Ptolemy's maps, etc.  Oddly, this indicates a proto-historic expansion of the Iberians southward from Catalunya into earlier Indo-European territory out of the earlier Urnfield territory.
 * 5) This may mean that the main component of the Urnfield culture, at least in Southern Europe, was actually non-PIE-speaking cultures.  Indo-European speakers show up in the historic and proto-historic period exactly where the Urnfield culture was not: most of Iberia except the northeast, most of southern Italy and pockets in Alpine Italy.  The correlation is plenty imperfect, as the Aquitani/Basque region is also Urnfield-free, despite the fact that we should be reasonably certain that non-PIE speakers were in at least parts of the area from pre-antiquity to now.  (There is a Basque expansion hypothesis, whereby especially the westernmost part of their modern territory, and perhaps most of their modern territory, was originally Celtic speaking.  It's something to consider, but the pre-Aquitani/pre-Basques were living somewhere in the vicinity.)  For a variety of reasons, nobody seriously proposes that the Celts and Lusitani of Iberia came overland through the western Pyrenees; the main contenders are 1. via the Urnfield expansion itself, wrongly construed as an Indo-European Celtic macro-family intrusion and 2. via the Atlantic.
 * 6) The positioning and rather solid extension backward in time for these cultures presents a problem:  The PIE speakers, Iberian Celts, para-Celts, Italics, and Veneti, are all on the wrong side of some mountains and/or coasts covered in non-PIE-speaking peoples.
 * 7) Regardless of whether you buy the rest of this, the idea that the Iberian Celts got there by sea in the Bronze Age is essentially inescapable, due to a mountain of recent analyses.  Bronze Age metal finds are quite consistent with this idea, and the complete lack of the expected Hallstatt type of artwork where we know Celts show up in Iberia is completely unanswerable for the standard theories.  Even Ireland presents very, very little evidence for any contact with Hallstatt/La Tene, or anything like it until the Roman period.  We also now know that the major change-over in Ireland (from basically Neolithic to about 33% steppe-derived and not far from modern) occurred in the Bronze Age.  (Nice quote from the paper "At present, the Beaker culture is the most probable archaeological vector of this Steppe ancestry into Ireland from the continent, although further sampling from Beaker burials across western Europe will be necessary to confirm this."  Amen to that.)  While the likelihood that Celtic was spoken in Ireland from 2000BCE is nuts, it is far easier to see Ireland, having first begun to speak an Indo-European language in the Bronze Age, continuing to develop in tandem with neighboring Indo-European languages of the Bronze Age Atlantic trade network to eventually arrive at Proto-Celtic.
 * 8) Tartessian? Lepontic? Where is the earliest Celtic?, a poor man may ask.
 * 9) Dating and attestation:  Even if you don't accept Tartessian as Celtic (I myself am quite skeptical), disagreement about this masks the fact that just about everyone agrees that there are Celtic names in these texts, and there is probably a Celtic name behind the Arganthonios reported by Herodotos as king of Tartessos in c625BCE. If the newer dating of the early texts around 800BCE (based on immediately post-Bronze Age context and paleoepigraphic comparison to Phoenician) holds up, this means that Celtic definitely is attested first in S. Spain.  We should also note that the earliest dates for the competitor for earliest Celtic, Lepontic, supposedly first attested in c. 625BCE, are not at all secure.  The earliest inscription is probably not actually Lepontic.  It remains uninterpretable, and is not included in the listing of David Stifter.  The next oldest, from Ticino c. 575BCE, the isolated Xosioiso, which Stifter does include, is probably a name in the genitive, but it shows no definitively Celtic features, and the form -oiso, generally interpreted as a gen.sg., is found only on two isolated words, this Xosioiso and Plioiso (twice), and then several times in the midst of completely uninterpretable fragments.  Both isolated words are interpreted as names in the genitive, and this is contextually very reasonable, though requiring special pleading for how this unique genitive got that way, but further interpretation is uncertain.  The earliest definitely, definitely Celtic inscription in N. Italy is the Prestino inscription of c. 500, not so long before the Gauls crossed the Alps in c.400BCE.  Even the most secure dating of c. 625 for the Tartessian inscriptions, and the earliest possible dating of Lepontic in c. 575, means that Celtic appears first essentially simultaneously in N. Italy (Lepontic language) and in S. Spain.  If Celtic originated in the Hallstatt culture about 800BCE, they must have teleported to S. Spain, and taken their sweet time presenting evidence of themselves in N. Italy, supposedly next door to their origins.
 * 10) Lepontic shows several features that are best interpreted as resulting from synthesis with a preceding non-PIE culture, in particular the Lepontic patronymic adjectives in -alo-, -ala-, -al, which pretty clearly derive from the Tyrrhenian genitive in -al, used in Etruscan for patronymics.  (Etruscan allowed case-stacking, so it was reasonable for the poor Indo-Europeans to interpret -al and its forebears as an adjective suffix.)  The Italic adjectives in -alis have the same origin, but with a different stem class since they independently adopted the suffix and independently assimilated it to IE-style morphology.
 * 11) Divergence analysis points to Iberia as the center for Celtic differentiation, since the immediate outgroup Lusitanian is found there.  Nowhere else do we find a language so close to, but not quite, Celtic.  In addition, Celtiberian shows quite a number of incontrovertibly divergent features, both innovations and archaisms, compared to the Celtic that we know well from the north of the Pyrenees and in the British Isles.  The ancestors of Old Irish and Brythonic may have been divergent in some interesting ways, but if so, the signals of it are largely lost in the many layers of lenitions and apocopes and other stuff that characterize the immediate pre-literary period.  Gaulish seems surprisingly homogeneous, and even Lepontic has been analyzed as an only lightly divergent Gaulish (see here).  Borrowings from Celtic into Germanic and Slavic are nothing weird, and we should expect these to represent the weirdest Celtic, having supposedly gone north and east from the Alps when most of what we know of Celtic derives from simultaneous migrations to the south and the west of the Alps.  The time depth of divergence is basically 100% wrong for an origin in the Alps.
 * 12) If the early pre-Celts of the Bronze Age were seafarers, what of the Italo-Celtic precursors?  Why did they spread?  If Bronze was re-invented in the vicinity of Cornwall shortly before 2000BCE, after the region had been Indo-Europeanized by a reflux of Bell-Beakers-with-Corded-Ware-substrate, then we know these Bronze Age traders got around.  Someone got Bronze to Southern Iberia.  Somehow Bronze got to Italy, and probably not first from the East.  While there's no necessity, or even plausibility, of imagining direct voyaging to Italy, somehow the trade got there, and the traders spoke something.  Down-the-line trading can have down-the-line effects on language, if there are power assymetries (like sitting on the source of tin) giving advantages to one language group over another.
 * 13) A fun piece of evidence pointing to the seafaring nature of Italo-Celtic is the uniquely Italo-Celtic word *tersa 'land', derived from the PIE root *ter 'dry, thirst, etc.'.  This implies a wet sea vs. dry land dichotomy that is most at home for seafarers.  People hanging out in the Alps, per the traditional Celtic and Italo-Celtic theories, would probably not be thinking this way.
 * 14) After the early Bronze Age, Italy aligned to the east and the Minoans and/or Myceneans.  There were also considerable upheavals with the fall of the Argaric culture in Southern Iberia.  And, as I mentioned above, the probable Indo-European-speaking cultures of eastern Iberia disappeared in the proto-historical period.  These disruptions could have been reason enough to cut the link between Celtic and Lusitanian in Iberia and the Atlantic and Italic in Italy, assuming it wasn't cut much earlier.

Connecting Italic to Iberia is not just my idea. [ifc.dpz.es/recursos/publicaciones/20/39/13villarpedrero.pdf Villar and Pedrero 2001] actually claim that Lusitanian should basically be considered an Italic language. I could say more, obviously, but I've spent hours on this already. Hope all of this makes some sense! --Mellsworthy (talk) 10:46, 25 July 2016 (UTC)


 * We already agree that Lusitanian is basically Italic (or at least very closely related, similar to Venetic, which may or may not be Italic in the proper sense). However, I don't understand how it can be both Italic and the closest relative of Celtic. How does that work? Am I missing something?
 * Heh, the idea that Latin -alis comes from Etruscan is one that I had about immediately on learning that Etruscan has a genitive ending -al. Might even have been before I started studying, don't remember, but definitely a long time ago.
 * My latest conclusion was that there's evidence pointing to a western origin of the Tyrsenian languages, and evidence pointing to the east, with no real way to decide, although I've tended to favour the east because reasons. I suspected that part of the reason for Rhaetian being a bit weird might be a Celtic substratum or something. There seem to be indications that in the early 1st millennium BC, that Etruria was Umbrian-speaking, but I don't know what they were (perhaps really old loanwords in Etruscan? Nah, that doesn't work), maybe only literary tradition. Caere and Falerii in Etruria once had local Italic dialects too, although they were not Sabellic but close to Latin. Anyway, to me it rather looked like Etruscan isn't native to Etruria, but of course that doesn't exclude the idea of an origin in the Padanian Plain. However, archaeological evidence does indicate a limited immigration from Anatolia to the coast of Etruria in the early 1st millennium BC, IIRC, which is quite suspicious. It's easily possible that Tyrsenian (along with the orientalising influences) was introduced to Italy through a thin élite stratum and then spread from the coast inland without much genetic change in the (Italic-speaking but genetically possibly indigenous) native population, all the way to the Padanian Plain and ultimately Tyrol, with Rhaetian imaginable as the result of Etruscan immigrants (separated from the rest after the Celtic invasion of Padania) mingling with (also Celtic?) natives of Tyrol. Maybe that's all wrong, but otherwise I have trouble explaining Lemnian and the Tyrrhenians apparently attested throughout the northern Aegean. I dunno.
 * Rix traces Italic back to Slovenia and ultimately Pannonia. Made total sense to me because of Venetic, and Proto-Celtic supposedly in Austria and Bavaria (see also Lepontic). Can't really judge it, though.
 * I'm not totally convinced yet. Have to process your stuff first, and sleep over it. Hispano-Celtic is a difficult problem, I grant that. Also early Lepontic before 500 BC. (I see Parpola supports Celtic from the West too. Heh.) Finding archaeological correlates for the spread of Italic is notoriously difficult, too, and I have no idea what exactly might have happened, so you might as well be right.
 * Though anchoring Italo-Celtic in Western Europe would be a further blow to Asian homeland hypotheses. I think they're pretty much toast by now. Although that alleged IE substratum in Sumerian (and East Semitic?) is fun to think about. How does that work?
 * Anyway, let's not forget that we agree that Proto-Celtic isn't particularly old, wherever it was spoken. My main beef was with the idea of Celts in the Bronze Age. Just nuts. Italo-Celts, OK. Or, by c. 1500 BC, some Italo-Celtic dialect that was essentially an ancestor of Proto-Celtic but lacked some of its characteristic innovations. Fine. Even if Italo-Celtic isn't a node, some early ancestor at the time should have been in contact with Proto-Italic. Maybe in northern Italy, or maybe in Iberia/Gaul. Whatever.
 * But yeah, Tartessian doesn't look Celtic at all, save for the odd anthroponym perhaps. It looks like a standard Paleo-Hispanic language, judging from Iberian and Aquitanian/Basque. Interesting that you argue that Iberian comes from the north, from Gaul. My impression is that Aquitanian/Basque originates in Roussillon (NESCAS ~ NISCAS) and gradually moved farther west. The "contraction hypothesis", that the area of Basque has simply shrunk and shrunk over the centuries and millennia, is just naive. We know that IE languages haven't always prevailed over non-IE languages, the inverse happened too, even in the Iron and Bronze Ages; especially in the Uralic area, there are really good arguments to this effect. IE-speakers were no master race.
 * (That's also why the idea that IE might once have been spoken in Mesopotamia doesn't sound that weird to me, and it could explain the Semitic–PIE lexical similarities. The IE expansion could have carried tribes as far as Egypt and other centres of civilisation, but they would not likely have been able to establish themselves there as a ruling élite, let alone overwhelm the natives numerically. But why should IE-speakers have spread in every direction except into the Middle East? Nothing kept them from getting there. Horses, carts and wagons could carry them there within a few weeks, I guess.)
 * Not apropos of anything in particular, but lately I learnt that Schrijver has proposed (already in 2009) that Goidelic was actually spoken in Britain originally and exported to Ireland only in the 1st century AD, after (?) the Roman invasion, but back home remained to act as the substratum that turned West Germanic to Old English. Fun stuff. --Florian Blaschke (talk) 04:55, 28 July 2016 (UTC)
 * Schrijver's hypothesis of pre-Indo-European substrate in Ireland or Scotland, a hypothesis of his that I am willing to entertain, does not require that the Irish lived outside of Ireland, so I'm surprised that he has argued that. Found the ref, and I'll have to read that later. Schrijver is, upon occasion, crazy, but also smart, so I'll try to reserve judgement until I read it.
 * If Indo-Europeans originated as a land power in Eastern Europe, there's good reasons why they would not be seen in the Middle East until they are. The Anatolians are already in Eastern Anatolia by about 2000 BCE, since their names are attested in Assyrian documents.  From the other side, the Mitanni, with rulers from the Indo-Iranian branch, are found doin' stuff in the North Middle East by 1500 BCE, which is pretty fast work getting there from the Sintashta centers in 2000BCE.
 * There's something complicated to say about the Kassites. I had thought that a number of people had believed that the rulers of the Kassites may have been Indo-European as well, but the paper that purports to show that (Ancillotti 1981, La lengua dei Cassiti) is not well received, and I can see why. There is a recent paper [diachronica.pagesperso-orange.fr/TMCJ_vol_3.2_Fournet_Kassite_kings.pdf here] that tries to connect some fragments of the Kassite language to Hurrian, Urartian, and NE Caucasian (also known as the Nakh-Daghestanian languages).  The paper presents the sources of information very poorly, and I had to swim up the source chain to Clay's book on names in Kassite documents here (pp. 36-41) to see what the heck was going on. It does seem that the names are built in un-Indo-Iranian and un-Indo-European way, with a verb followed by an agent; this is obvious for several names that have elements glossed in Akkadian, e.g. Šagarakti-šuriyaš 'the Sun saves him', with šagarakti = Akk. napšaru 'release, salvation' and šuriyaš = Akk. šamaš 'sun/sun god'.  However, Fournet's paper is ridiculous for not recognizing or not acknowledging the Indo-Iranian theonyms that are embedded in these non-Indo-European names, namely marattaš (better maruttaš ) = Akk. ninurta (mistranscribed the old way as ninib in the Clay book) the god of hunt and war, which isn't half bad to compare to Vedic Maruts, and the aforementioned šuriyaš equated with the god of the sun, identical with Vedic Suriya. Before we get too excited, just about nothing else is very plausibly IIr.  (The stem -bugaš does occur in the right position to be a theonym if it's built like the other names, but it has no gloss and it's awfully uncertain to connect it to Ved. bhaga- .) You might think that would make it impossible that these names contained IIr elements.  But wait!  The Kassites made lots of names with verbs in Kassite and divine names borrowed from the Mesopotamian religions, so it seems they were quite open to that kind of thing.  Interestingly, there's also clearly another word for sun or sun god, since the name element sah- is also glossed in Akkadian as Šamaš 'sun/sun god'.  This second root, which is found in Kassite names that are not Kassite rulers, is probably from the very-not-PIE native Kassite language.  While the main conclusion is that Kassite was not Indo-European, this risks burying the fact that they did, at least in my view, borrow some Indo-Iranian gods to name their rulers!  To me, this does seem more likely if they had Indo-Iranian rulers to start with, who were quickly and thoroughly linguistically assimilated, insisting only on their own special name for the Sun God and such like.  This is not surprising in the Horse, Wheel, and Language perspective, since the Kassites were the culture that brought horses and chariots to Babylonia, and they had to have gotten them recently from Sintashta descendants.  --Mellsworthy (talk) 09:16, 28 July 2016 (UTC)
 * Oh, and about the position of Lusitanian, Celtic or Italic: Lotsa people say lots of things, but most people think it looks very similar to Celtic, except for the retention of p and the extremely fun word indi for 'and'. It certainly went the way of Celtic, not Italian, with the voiced aspirates, at least.  Still, to my mind, it doesn't look specifically Celtic or Italic, but just Italo-Celtic.  --Mellsworthy (talk) 09:29, 28 July 2016 (UTC)