Talk:Kluge's law

A rather large catch
The Indo-European word *átta has descendants in many Indo-European languages, including Germanic. But the Germanic descendants also have -tt-, showing that Grimm's law did not affect this geminate. That evidence kind of undermines what is mentioned in this article, in particular the devoicing of voiceless consonants. If Grimm's law did not affect -tt-, did it affect -dd-? And what about the other two, -pp-/-bb- and -kk-/-gg-? CodeCat (talk) 19:45, 21 November 2012 (UTC)


 * Somebody has spread this *átta all over Wikipedia. I think it 1) counts as original research and 2) is highly dubious anyway. I mean, just look at it: it's a mama/papa word, expected to pop up repeatedly and very often in very different languages; it contains not one but two *a, a phoneme so rare in PIE that some have tried to discuss it away entirely; and it contains the only phonemically long consonant in the entire (proto)language!
 * Anyway, did *bb, *dd and *gg even exist in Proto-Germanic, or are they exclusively West Germanic? In the latter case, Grimm's law must have devoiced any voiced long plosives that existed before it.
 * Also also, "Grimm's law" is a cover term for two or three sound shifts; *tt and *dd would have been affected by different ones.
 * David Marjanović (talk) 18:51, 18 August 2013 (UTC)


 * Go to Elamite language: it has atta "father", amma "mother". Ata is all over Turkic, too. Really, it's not a good word to use as evidence for sound shifts.
 * It may be relevant here that onomatopoeia has caused exceptions from the High German sound shift. "Pick" and "peck" are picken and pecken, respectively, and various Upper German dialects have created a second picken which means "glue" (both intransitive, "it sticks (together)", and transitive).
 * To answer my own question, *bb dd gg and the occasional *þþ did exist in Proto-Germanic or at least Proto-Northwest Germanic, but were formed by analogy, says Kroonen (2011).
 * David Marjanović (talk) 15:03, 25 August 2013 (UTC)
 * It has sources, that's why it's spread all over Wikipedia. See Ringe 2006. CodeCat (talk) 15:38, 25 August 2013 (UTC)
 * Ringe (2006) is definitely good enough! I'll leave it in. :-) David Marjanović (talk) 19:24, 6 October 2013 (UTC)


 * David is correct in that this is a rather dubious word because of its sound symbolism. (As was once famously pointed out, Ancient Greek "baaaa" should evolve to /vi/ by regular sound laws, but -- wonder of wonders -- sheep in Modern Greek still say /be/.) Nonetheless, sound-symbolic words sometimes do evolve as normal words: e.g. PIE *wai > English "woe", PIE *mā-ter > German "Mutter", Old Chinese /paʔ/ "father" > Mandarin fù, Old Chinese /məʔ/ "mother" > Mandarin mǔ. (But notice the new English words "ma", "pa", and likewise the new Mandarin words mā "mother", bā "father".) Ringe does cite *atta and claims this may have been the only word with *tt within a morpheme. Benwing (talk) 05:54, 1 September 2013 (UTC)
 * Actually Mandarin bàba "dad", māma "mom"; fùmǔ "parents" (formal), bàba, māma "parents" (informal). – Anyway. When Ringe agrees that it may have been the only word with *tt within a morpheme, it's clear that no safe statements about sound laws can be made! It's statistics with a sample of 1. :-)
 * Better examples might be German – even dialectal Upper German – picken and pecken, unshifted compared to English pick and peck.

David Marjanović (talk) 19:24, 6 October 2013 (UTC)

Kluge's law in non-plosives
The consonant gradation article reconstructs a word with Proto-Germanic *rr and attributes it to Kluge's law, apparently following Kroonen (2011). I expect other long consonants to show up (**ll first of all), but I only just found Kroonen's 400-page book on Google Books and will need a lot of time to read it, so if somebody already knows what it says, please fill in! In the longer run, I want to translate this article into German. Believe it or not, the German Wikipedia completely lacks an article on Kluge's law, and even the article on Kluge himself doesn't mention it! David Marjanović (talk) 18:51, 18 August 2013 (UTC)
 * *ln > *ll is a regular change in Germanic, but *rn > *rr is applied only inconsistently in a few words. CodeCat (talk) 19:04, 18 August 2013 (UTC)
 * I've now read the Google Books preview, which is actually just the first quarter of the book. However, all examples for *ln > *ll it mentions were stressed on the *n- suffix, and Kroonen also seems to wonder about *n + *n > *nn (details hidden in the preview), because clusters of the same consonant never produced long consonants in PIE.
 * Things like learn are explained in the book: the *n was directly in front of *h2 (zero-grade), therefore was syllabic and changed to *un, and then the *u was lost analogically after Kluge's law operated.
 * David Marjanović (talk) 14:44, 25 August 2013 (UTC)
 * Why would the rn of "learn" need explaining? CodeCat (talk) 15:37, 25 August 2013 (UTC)
 * On another note: there's a long tradition in Fennistics of comparing the regular assimilation of *ln but inconsistent assimilation of *rn in Germanic to the exact same situation in Finnic. The one-time explanation of attributing this as G influence in F is not favored much anymore though. The opposite has also been suggested, but I think thus far only as a part of a fringe proposal to explain the entire Grimm's and Verner's Laws as Uralic influence (!) There might be something to be noted based on this, or there might not… -- Trɔpʏliʊm • blah 15:51, 25 August 2013 (UTC)

In any case, "learn" was Proto-Germanic *liznōjan, without *rn at all. If the book gets this wrong, it's a strong sign that the entire book is garbage. But that's not surprising since this whole law is garbage. Benwing (talk) 05:56, 1 September 2013 (UTC)
 * But what makes you think the book gets it wrong? I said it's on Google Books; most of its first 110 pages are here, why didn't you just read them? It's been over a month since you wrote your comment!
 * The confusion is all mine, "learn" is mentioned 3 times in the book (well, the preview anyway): on p. 48 as one of 6 examples that show that *s and *z were not affected by Kluge's law under any circumstances, on p. 101 as a supposed example of an inchoative verb, some of which (not this one) are often cited by opponents of Kluge's law, and on p. 102 to illustrate the suggestion "that the inchoatives continue medial factitives", which is why Kroonen expects "the zero grade of the suffix in the larger part of the paradigm" of the inchoatives, meaning the *n was syllabic and turned into Pre-Proto-Germanic *un before Kluge's law could operate. Later, this *u "was probably removed on the basis of inchoatives with roots in a vocalic element".
 * David Marjanović (talk) 19:24, 6 October 2013 (UTC)

*r-n again
Yes, it's weird. Moulton (1972: 166), a pre-laryngeal work: "(1) Long nasals and liquids. These seem to have arisen through various types of assimilation. Examples: [...] OE steorra "star" < PGmc. /sterr-/ < PIE /ster-no-/, cf. Gk. astēr (but, puzzlingly, unassimilated /rn/ in Go. stairnons 'stars', etc.)." Kroonen (2011) first (p. 23, 56) implicitly attributes the opposition of Old Frisian stera to OE steorra and OHG sterro to PIE *h₂stérō, *h₂strnós, then (p. 174) cites someone speculating if somebody regularized the paradigm to *h₂stérō, *h₂stérnos, on which Kluge's law would not have operated – if indeed Kluge's law operated on *r, which he doesn't explicitly argue anywhere in the whole book. However, with OHG having both sterro and sterno (the latter cited on p. 174 together with ON stjarna), there's definitely some weirdness going on, *handwave* analogy or something. And now I'll go home (it's late enough) and see if I can actually get p. 174 in the preview there, or if his 2009 thesis contains that discussion. David Marjanović (talk) 21:24, 22 September 2014 (UTC)
 * Kroonen (2009: 182, 183) discusses PGmc. "*swambō, *sumpᵖaz ‘sponge, mushroom’". The potentially important part (I'm too lazy to fill in all the italics):
 * "In addition to the roots *swamb- and *swampᵖ-, a root *swamm- is presupposed by Go. swamms (and probably also by OHG swam, MHG swamme and Kil. swamme). In this third variant, the labial stop has disappeared. Consequently, it can neither be explained from *suombʰ-, nor from *suombʰ-n-́, as these root forms in all probability developed into *swamb- and *swampᵖ-. I therefore think that the variant *swamm- continues a root-stressed form *suómbʰ-n-, which, in spite of its nasal suffix, was not affected by Kluge’s law. At a later stage, the labial disappeared between two nasals, so as to give rise to a long m, viz. *swambna- > *swamma-. This development is paralleled by e.g. OHG hunno m. ‘centurion’ < *hunþnan- < *dḱmt-n-, OHG zinna f. ‘merlon’ < *tinþnōn- < *h3d-ent-n- and OHG channa, chanta1168, MHG kanne, kante f. ‘jug’ < *kand-(n)ōn-. Morphologically, the barytone stem *suómbʰ-n- is comparable to *ster-n- as in Go. stairno, ON stjarna f. ‘star’ < *h2stér-n-.1169"
 * Footnote 1168 is just a citation for chanta. 1169, however, reads as follows:
 * "Van Helten (1905: 224) reconstructs *stérnõ (beside *sternṍ > *sterrõ), which he assumes to have arisen as an analogically root-stressed form that arose before Verner’s and Kluge’s law."
 * That implies pretty heavily that he thinks Kluge's law affected *r. Compare this quote from p. 23:
 * "Lühr (1988: 191) further pointed to the fact that n-stems with roots in both stops and resonants were affected in the same way and in the same morphological environments, cf. OHG chnodo : OE cnotta m. ‘knot’ < *ǵnút-ōn, *ǵnut-n-ós, OFri. stera : OE steorra m. ‘star’ < *h2stérōn, *h2st(e)r-n-ós.51 This parallelism confirms Kluge’s view that the gemination of stops is the result of the same process as the doubling of resonants, cf. *fulla- ‘full’ < *plh1-nó-, *wullō- ‘wool’ < *HulH-nó-. As a result of this mechanism, which translated the old PIE suffixal ablaut into a kind of grammatischer Wechsel between roots with and without geminates, the consonant alternations as described in the introduction receive a logical explanation."
 * Trouble is, Lühr didn't think that Kluge's law had anything to do with accent, even though she seems not to have explained why she thought so, except for waffling some nonsense about "consonantal strength". Kroonen explicitly disagrees, but again without mentioning *r.
 * Footnote 51 is the rather uninspiring "“n-Stämme mit *ll < *l-n, *nn < *n-n verhalten sich morphologisch wie die n-Stämme mit Doppeltenuis.”" – "n-stems with *ll < *l-n, *nn < *n-n behave morphologically like the n-stems with double tenuis."
 * Then there's this interesting quote on p. 40: "the creation of the sequence *-n-n- cannot have happened before Kluge’s law, as it would have been simplified before that time limit". That either means he thinks the origin of *nn part of Kluge's law as well, or he thinks *nn was only able to arise once Kluge's law had created other long consonants, which would be strange because *ss must be really old anyway.
 * Anyway: Is there actually an example of *rr, *ll or *nn in a word with PIE root stress? Kroonen (2009, 2011) doesn't explicitly provide any, but for many of his examples of *rr and *ll he doesn't reconstruct the PIE accent at all; the only candidate for *rr from *´rn I can see is *ferrai compared to Lithuanian pérnai, and Balto-Slavic accentology is beyond me.
 * In your edit summary you mentioned "several examples" of *rn in German. Other than Stern, what are they? (Ge)hirn is from *hersō, *hurznaz with leveled ablaut (Kroonen 2009: 144f.), and at this hour (1:52 am) I can't think of any others.
 * David Marjanović (talk) 23:52, 22 September 2014 (UTC)

This law is more-or-less garbage
This is not a generally-accepted law. Something of this sort is in fact mentioned in many reference books on Germanic, but is explicitly disclaimed in such cases. Kortlandt's endorsement is insufficient -- although he's admittedly a hist ling genius, he also endorses a lot of fringy theories and in general is willing to make huge theoretical leaps of faith well beyond the data.

Please also note, Guus Kroonen is at Leiden, just like Kortlandt. I would not be surprised at all to find that he was/is a doctoral student of Kortlandt. The Leiden guys in general have all sorts of interesting but often highly speculative ideas about PIE, which are not generally accepted by others. However, because they make a lot of noise and produce a lot of books, their theories appear to be mainstream to non-specialists. Benwing (talk) 06:08, 1 September 2013 (UTC)
 * "Garbage" and "not generally accepted" sound like different levels of disputable, though… For the former, I'd ask for a "smackdown" critique that could be used to fairly clearly illustrate why the law fails to work. If the latter, and if there are no general critiques, it can be hard to tell if wider support for a proposal is amiss due to it being actually questionable or just due to lack of attention.
 * Also "it's from Leiden so it's fringe" wouldn't really make a tenable source grading principle. Kroonen's dissertation has been out for 4+ years, you'd think there were some critiques to be found if it's poor work. -- Trɔpʏliʊm • blah 18:00, 1 September 2013 (UTC)


 * Wow, Benwing. You've committed a textbook example of an argumentum ad hominem. Please don't take it personally that I have to embarrass you by showing in some detail how I can tell you haven't read the Google Books preview.
 * Yes, the book is from Leiden, it even has "Leiden Studies in Indo-European" on the front cover. Although the preface (p. 12) makes clear Kortlandt wasn't his supervisor, Kroonen thanks him and 4 other people "for their helpful comments on the manuscript". But before the acknowledgments comes the twist: Kroonen was set on a Leiden problem for his PhD, "the Leiden Substrate Theory" as he calls it – and then he found that it was completely unnecessary to invoke substrate influence for "the typical Germanic cross-dialectal interchange of singulate and geminate roots", because "they turned out to be strikingly predictable in nature"! The book is a break with an important part of Leiden tradition. The preface is frankly delightful to read.
 * Kroonen doesn't support the glottalic theory either; check out p. 66.
 * Although acceptance of Kluge's law is low, it isn't zero even outside of Leiden. For his etymologies Kroonen occasionally cites sources that casually say things like "with pp from PIE *bhn". The latest post on this blog, by someone who's working on an etymological dictionary of Old High German (in Jena if you really care), doesn't cite Kroonen, but talks about one of the verbs Kroonen mentions, says Kluge's law is the simplest explanation for why it has reflexes of Proto-Germanic *g in some languages and of *k in others, and explicitly rejects expressive gemination as an alternative.
 * It appears that very, very little has been published on Kluge's law (supporting it or not) in the last 30 years. Keeping in mind that Kroonen disagrees with several of the conclusions in both Lühr (1988) and Kortlandt (1991), there hasn't been much time for the latest version of Kluge's law to either become textbook wisdom or be panned. I'll go check if the Google Books preview of Ringe (2006) says anything about it.
 * David Marjanović (talk) 19:24, 6 October 2013 (UTC)
 * Wow. So, here's Ringe (2006), and he devotes less than a paragraph to it. It's on p. 115, which is (bizarrely) "not part of the preview" but nonetheless accessible this way (I searched for kluge), and it's so short that I'll just quote the whole thing:
 * "Well over a century ago Friedrich Kluge suggested that numerous PGmc forms with unexpected root-final *pp, *tt, *kk had arisen from forms with pre-PGmc *ƀn, *ðn, *ȝn (i.e. with the fricative outcomes of Verner's Law and the first and third part of Grimm's Law, before the voiced stop allophones arose—see 3.2.4 (i) ad fin.) if a stressed syllabic did not immediately precede; he was even able to suggest a relative chronology of the parts of Grimm's Law and Verner's Law that would render such an outcome natural. (See Lühr 1980 with references for fuller discussion.) The problem with Kluge's suggestion is simply that the etymologies are unconvincing in detail: the best examples are assembled at Brugmann 1897: 383–4, and not one must reflect a form with *-n-. On the other hand, perusal of the numerous examples scattered throughout Seebold 1970 strongly suggests that they have been generated by some sort of sound symbolism ('Intensiv-Gemination'), and that is still perhaps the most widely accepted explanation."
 * (Instead of ȝ, Ringe writes ʒ. Clearly an error.)
 * That's all. A brief, vague dismissal of old to ancient papers, without any details on the etymologies, and with the implication that Kluge's exact sequence of sound shifts is needed for his law to work (which is wrong). I'm rather disappointed, and can't take it as arguing against anything Kroonen wrote five years later.
 * David Marjanović (talk) 19:50, 6 October 2013 (UTC)

For the sake of completeness: the thesis itself (Kroonen 2009) says the supervisor was Lubotsky, but Kortlandt was on the thesis committee. David Marjanović (talk) 23:59, 22 September 2014 (UTC)

luttaz
This example is very dubious. Yes, this law as a whole is, but making it rely on examples like this only makes it worse. There is no known sound law for Germanic where *l̥t-nós would give *luttaz. The expected outcome of *l̥t- is *ult-, not *lut-. Thus, the only source for a root form *lut- is as the zero grade of a root such as *lewd-, which does not resemble the purported cognate *laþō at all. CodeCat (talk) 18:36, 24 October 2013 (UTC)

Of course, if you only stick to phonological changes. If you have a word like this, a form with zero-grade in the root would indeed give *ult-, but here of course analogy would have taken place. The same is seen in verbs like "to break", *brek-, past time *brak-, and past participle *bruk-; The past participle should have come from zero-grade *bhŗg-, thus first giving *burk- in PGmc. But analogy made it to *bruk-. By the way, what do you think of the other examples? Just added them because I thought examples were needed. I think reading Kroonen's book would be nice --Tsennoy (talk) 20:40, 24 October 2013 (UTC)

Replaced "luttaz" with "hwitta-". --Tsennoy (talk) 21:39, 24 October 2013 (UTC)

A criticism section at last!
Finally I've written one. I'm aware that a few parts of it don't make sense as long as the rest of the article isn't completely overhauled, which I hope to do tomorrow (or, rather, later today). My main source has been the Google Books preview of Kroonen (2011). It was very late when I discovered that a pdf of Kroonen (2009) is online – I have used it to fill in the gaps in the preview; in the long run, all or almost all citations of Kroonen (2009) should be replaced by Kroonen (2011). David Marjanović (talk) 22:15, 27 August 2014 (UTC)


 * I forgot:
 * 1) Does anybody know any book reviews or other reactions to Kroonen (2011)?
 * 2) In 2013, Kroonen published an Etymological Dictionary of Proto-Germanic, for which I can't get a Google Books preview. Obviously I expect that it upholds Kluge's law, but does anyone actually know? Or is the book just so expensive that nobody has it...?
 * David Marjanović (talk) 10:09, 28 August 2014 (UTC)

Overhauled half of the article
Really have to go to bed now. :-) The law itself, including its regular exceptions, is now presented; I intend to describe the rise and fall of consonant gradation and all the morphological confusion it caused (most notably the origin of voiced and fricative long consonants), and when that is done, I'll remove the tags.

I dropped the *hwitta- example, because I think Seebold (footnote in Kroonen 2011) is right about it: its unshortened long consonant is much younger, coming from the neuter singular *hwīt-t (Old Saxon huuitt, modern German weißes). That explains why only Dutch has a short vowel in that word while a long one is found in Gothic, German and English.

I'd be thankful for references to anything written about Kluge's law after Ringe (2006) came out.

David Marjanović (talk) 23:29, 7 September 2014 (UTC)
 * The problem with your explanation is that Dutch does not have -t as the neuter ending, nor is there evidence of it ever having existed in that language. CodeCat (talk) 23:51, 7 September 2014 (UTC)


 * Ever? When every other Germanic language seems to have (had) it (at some point)? Kroonen himself doesn't use your argument. David Marjanović (talk) 18:34, 14 September 2014 (UTC)
 * I haven't seen it in Old English or Old Frisian either, that I know of. And I believe that even in Old Saxon it's very rare. CodeCat (talk) 19:39, 14 September 2014 (UTC)