Of dogs and Old Sinitic reconstructions

« previous post | next post »

At the conclusion of "Barking roosters and crowing dogs" (2/18/18), I promised a more philologically oriented post to celebrate the advent of the lunar year of the dog.  This is it.  Concurrently, it is part of this long running series on Old Sinitic and Indo-European comparative reconstructions:

I will launch into this post with the following simple prefatory statement:

Half a century ago, the first time I encountered the Old Sinitic reconstruction of Mandarin quǎn 犬 ("dog"), Karlgren GSR 479 *k'iwən, I suspected that it might be related to an Indo-European word cognate with "canine" [<PIE *kwon-]).

Since that time, as I have continued my Sinological researches, I have come across many other terms in Sinitic that appear to have clear parallels with corresponding words in Indo-European (e.g., mǎ 馬 ["horse"] / mare; gūlu 軲轆 / wheelcycle; Cant. mou4 巫 / magus).  I have never systematically written about all of these words together as a group, and I have never in the slightest asserted that, despite there being so many pairs of words like these that match between Sinitic and Indo-European, the two language families are genetically related.  Rather, I maintain that there has been significant borrowing from at least the Bronze Age and probably beginning already before that time.

Moreover, as in the recent series of Language Log posts on such Sinitic-IE word pairs as those listed above, so far I have focused on phenomena that can be attested by archeological, technological, cultural, and historical evidence (i.e., the correspondences are not merely random, chance similarities, but have convincing empirical data and contextual reasons to support the linguistic comparanda).  However, occasionally I come upon strikingly similar word pairs for which one would be hard pressed to find archeological, technological, cultural, and historical evidence.

One example of the latter sort that has intrigued me for many years is the uncanny parallel between pīmǐ 披靡 (Zhengzhang /*pʰral/ /*mralʔ/) and "pell-mell", both of which mean "in a hasty, disorderly fashion").  Still, I do not put much faith in such comparisons of pairs that resemble each other both in sound and meaning but are lacking hard archeological, technological, cultural, and historical evidence. In the case of pīmǐ 披靡 and "pell-mell", the former was used already more than two thousand years ago, whereas the latter cannot to my knowledge be traced back earlier than Old French (8th-14th c.).

With quǎn 犬 ("dog") and "canine" we are on firmer ground, having a variety of types of evidence in addition to linguistic data.

First the historical reconstructions for quǎn 犬 ("dog"):

Middle Sinitic (MS, ca. 600 AD)    khwenX (Baxter-Sagart [B-S]) ("X" indicates rising tone)

Old Sinitic (OS, ca. 600 BC)    /*kʰʷeːnʔ/ (B-S); /*kʰʷeːnʔ/ (Zhengzhang [Zz])

Karlgren's earlier OS reconstruction was *k'iwən; his MS reconstruction was k'iwεn:.

This is an old word in Sinitic, dating back to the earliest strata of the written language, ca. 1200 BC (see below for paleographical evidence.  This is during the Shang dynasty (ca. 1600-ca. 1046 BC), in the 2nd millennium BC when there were many cultural innovations (almost like a "package"), including the wheel, chariot, utilization of the domesticated horse, cultivation of wheat, bronze technology and specific weapon types, and writing, many of which had evident linkages to the west.

There is another Sinitic word for "dog", viz. gǒu 狗.  Since it is superficially rather different from quǎn 犬 ("dog"), I have often pondered what its origins might be, whether it had a completely different source than quǎn 犬, and so on.

Here are the historical reconstructions for gǒu 狗 ("dog"):

MS kuwX (B-S)

OS /*Cə.kˤroʔ/ (B-S); /*koːʔ/ (Zz) (*C- at the beginning of a B-S OS reconstruction just means some consonant [they believe there is a pre-initial consonant there, but they can't determine what it is])

There is both a Shang period oracle bone inscriptional (OBI) form and a Shang bronze inscriptional (BI) form for quǎn 犬.  They probably are essentially the same pictograph, but the disparate media make them look slightly different.  On the other hand, gǒu 狗, which has now largely displaced quǎn 犬 as the word for "dog" in the modern topolects (except for Min, the most archaic living variety of Sinitic) doesn't seem to have existed before the Warring States period with the clear meaning of "dog".  More details on the paleography of quǎn 犬 and gǒu 狗 are presented below.

Note that this year of the dog is called "gǒu nián 狗年" by nearly all Chinese (12,200,000 ghits), not quǎn nián 犬年 (383,000 ghits).

There is a suggestion here that gǒu 狗 is "from some language ancestral to modern Hmong-Mien languages, from Proto-Hmong-Mien *qluwˣ ("dog"), perhaps from Proto-Austronesian *(u-)(ŋ)kuɣkuɣ("dog")."  Before accepting this claim, we need to step back and take a deeper look at the historical phonological relationship between quǎn 犬 and gǒu 狗.  The following paragraphs are from Chris Button:

Old Chinese 犬 *kʰʷə́ɲʔ is clearly related to (early/pre-) Proto-Indo-European *kʲwon- (or more precisely *kʲwan- since e/o reflects ə/a). The palatal nasal in OC generally comes from an earlier medial "j" palatalising a nasal coda (coronal, as in this case, or velar) but you cannot have medial /j/ and medial /w/ as *kʰwjə́nʔ  or *kʰjwə́nʔ would imply so, short of reconstructing uvular *q as the initial which could have precipitated the rounding, something external was probably going on here just on the basis of the internal OC phonology.

Pulleyblank's (1995) hypothesis that 狗 *káwʔ and 犬 *kʰʷə́ɲʔ are related is probably correct, although I favor your 1998 suggestion about different loans from Indo-European over his phonological permutations. The fact that early PIE *kʲwon- gave standard PIE *kʲwo- provides good support for this since we do not have to account for the disappearance of the "n" in Chinese as it had already disappeared in Indo-European very early on.

The phonotactic constraint in Old Chinese against the co-occurrence of a medial /w/ and a medial /j/ also applied to Old Burmese. So while OC *-əɲ corresponds with OB *-ɐɲ, a combination like *-wɐɲ could not occur in a native OB word in the same way that *-wəɲ smacks of an external source in OC. I'm only aware of one case of <-wɐɲ> in Inscriptional Burmese in the word <klwɐɲ> "serve", that is also written in a more expected form as <klwɨj>. As such the Burmese word *kʰwɨj2 "dog" (tone 2 parallels the glottal in Old Chinese) would be the expected reflex of an external *kʰwɐɲ2 which would then correlate perfectly with the Old Chinese form 犬 *kʰʷə́ɲʔ

Mair 1988 mentioned by Chris above is this:

Victor H. Mair, "Canine Conundrums:  Eurasian Dog Ancestor Myths in Historical and Ethnic Perspective," Sino-Platonic Papers, 87 (October, 1998), 1-74.  (That was published in the days when I had to write in many diacritical marks and non-standard characters by hand.  Consequently, a few schwas got dropped out — they are easily spotted by an empty space.)  Free pdf available online here.

Rather than needlessly inflate this already long blog post by typing in a lot of material that is readily available in that article, I would strongly encourage anyone who is interested in the deep linguistic and cultural linkages between human beings and dogs to read pp. 22-24, where I discuss how the overwhelming majority the daughter languages of the Indo-European family share a common word for "dog", and how the dog played a key role in IE culture.  More surprisingly, I show how the same word for "dog" that was present in PIE from its inception, namely *kwon-, extends far beyond IE to many languages in other branches of the Nostratic macrofamily, but not to Sinitic, which arose later, so that quǎn 犬 OS /*kʰʷeːnʔ/ MS khwenX (remember that final "X" indicates rising tone for B-S) must have been borrowed from an IE derivate of PIE *kwon-.  In the course of my analysis, I show how the hypothetical Nostratic root for "dog", *k[h]wən or *k[h]wan (bracketed "h" should be a small superscript), has not been securely established as genetically pertinent to all of the families normally placed within the Nostratic macrofamily; it is germane only in IE and Afro-Asiatic, but not in Kartvelian, Uralic, Dravidian, and "Altaic".  In Afro-Asiatic, furthermore, the common hypothetical root for "dog" is firmly established only in the Hamitic branch, but not in the Semitic branch.  This affords an approximate idea of the time-depth at which the common word for domesticated dog (*k[h]wən or *k[h]wan) arose and was shared by speakers of the languages enumerated just above:  closer to 6000 BC than to 10,000 BC, but before 4000 BC, by which time IE had established itself as an independent family.

κύων is securely documented in ancient Greek (from Homer on [see Liddell and Scott]) and I begin my "Canine Conundrums" paper with a discussion of how the word "cynic" comes from the Greek adjective kunikos ("doglike") < kuōn ("dog").  In the Mediterranean region, there were undoubtedly many cognates to Greek kuōn among IE sister languages.  Indeed, the Hittite word for "dog" is kun-.  Since Hittite dates to the 16th-13th cc. BC, this puts us in the proper time frame for correlation with OS /*kʰʷeːnʔ/

Similarly, I show how gǒu 狗 ("dog"), MS kuwX OS /*koːʔ/ (Zz), was most likely a later borrowing from Tocharian ku.  This would make good sense, since the Tocharians, though centum speakers, were right at China's back door and are the acknowledged source of the borrowed words for "honey" and "lion" in Sinitic.

Dogs were intimately associated with humans in IE society from the very beginnings of the family,  In J.P. Mallory and D.Q. Adams, ed. Encyclopedia of Indo-European Culture (London and Chicago:  Fitzroy Dearborn, 1997), p. 168a, Mallory states that the remains of dogs are abundant in Mesolithic sites over a wide range of Eurasia inhabited by Indo-Europeans.  Anne-Sofie Gräslund, in "Dogs in graves — a question of symbolism?", in Barbro Santillo Frizell, ed., Man and animal in antiquity (The Swedish Institute in Rome, 2004), p. 167b of 167-176 describes how greatly human beings honored their treasured dogs:

From Mesolithic cemeteries in Denmark and Southern Sweden (c. 5000 BC), we know of separate dog graves, where the dogs have been buried with the same carefulness as human beings.  They lie crouched as if they sleep, covered with red ochre, in single cases they have even got grave goods.  In Swedish Bronze Age cremation graves dog bones from most parts of the body occur, in contrast to, for example, pig bones and sheep bones, which occur only from fleshy parts of the animal, used for food.

Dogs in earliest IE society were used for hunting, herding, and companionship, and as guards.

In Zoroastrian religion, the dog played a key role in rituals associated with death and was treated with utmost veneration for its power in warding off malign forces.

In sharp contrast, I shall never forget an experience I had about twenty years ago in a museum in Xianyang outside Xi'an.  This was at the Yangling Museum just after it had opened at the mausoleum of the Han Dynasty Emperor Jing (188-141 BC; r. 157-141BC).  I was one of the first people to go down into the ingeniously designed subterranean chambers of the museum.  The scale of the mausoleum was enormous, and it included all of the palatial furnishings, utensils, equipment, personnel (figurines), and other accoutrements that the emperor would need in the afterworld.  But the part of the subterranean museum that transfixed me the most were row after row after row of terracotta statues of dogs.  The guide told me that, like the sheep and pigs nearby, these dogs were raised and butchered for meat.  I was especially struck by how bulky the dogs were, as though they were bred with the intent of maximizing the flesh on their bones.  This was still true in 1889, for which see Appletons' Annual Cyclopaedia and Register of Important Events:

Dogs in traditional Chinese society generally were treated with contempt and cruelly abused (of course, there were exceptions, especially among the extremely wealthy and the nobility):

It goes without saying that, in the West, a dog is a "man's best friend".  That is obvious any day of the week when I walk around the little town of Swarthmore where I live, but it's true in any American town.

Turning now from dogs in China and words for "dog" in Chinese to words for dogs in English, I will focus on four main terms:  dog, hound, canine, cur.  For a short, easy introduction to this lexical quartet, here are the entries on them in the Online Etymology Dictionary:

dog (OEtymD)

Old English docga, a late, rare word, used in at least one Middle English source in reference to a powerful breed of canine. The word forced out Old English hund (the general Germanic and Indo-European word, from root from PIE root *kwon-) by 16c. and subsequently was picked up in many continental languages (French dogue (16c.), Danish dogge, German Dogge (16c.)), but the origin remains one of the great mysteries of English etymology.

hound (OEtymD)

Old English hund "dog," from Proto-Germanic *hundas (source also of Old Saxon and Old Frisian hund, Old High German hunt, German Hund, Old Norse hundr, Gothic hunds), from PIE *kwnto-, dental enlargement of root *kwon- "dog." Meaning narrowed 12c. to "dog used for hunting" (compare dog (n.)). Contemptuously, of a man, from late Old English.

canine (OEtymD)

late 14c., "a pointed tooth," from Latin caninus "of the dog," genitive of canis "dog" (source of Italian cane, French chien), from PIE root *kwon- "dog." The meaning "a dog" is first recorded 1869.

cur (OEtymD)

early 13c., curre, earlier kurdogge used of both vicious dogs and cowardly dogs, probably from Old Norse kurra or Middle Low German korren both echoic, both meaning "to growl." Compare Swedish dialectal kurre, Middle Dutch corre "house dog."

Most authorities, such as my beloved American Heritage Dictionary of the English Language, with its esteemed appendix of "Indo-European Roots", trace "hound" and "canine" back to PIE *kwon-.  Similarly, scholars such as Carl Darling Buck, A Dictionary of Selected Synonyms in the Principal Indo-European Languages (Chicago:  University of Chicago Press, 1949), 3.61, pp. 178-179, and Robert K. Barnhart, The Barnhart Dictionary of Etymology (Bronx, New York:  H. W. Wilson, 1988), pp. 139b, 493ab hold that most of the common, general words for "dog" in IE languages derive from PIE *kwon-.

Don Ringe takes a more nuanced, cautious view on "canine":

PGmc. *hundaz seems to reflect a derivative of PIE *ḱwón- ~ *ḱun-, but Lat. canis (the source of canine) cannot, because there's no way to account for (1) the vowel and (2) the loss of *w.  It does seem to be related to Welsh ceneu 'puppy', pl. canawon (< pre-Welsh, maybe Proto-Celtic *kanawū, pl. *kanawones), so there might have been an Italo-Celtic word *kan- of unknown origin.  English 'cur' also has nothing to do with the PIE word (no nasal; no known source for the r).  It seems to be connected to similar Middle Dutch and (not very early attested) Scandinavian words of similar meaning that might be related to a Norse verb meaning 'growl'; if so, then the word was probably onomatopoeic and could have been invented at any time.  'Dog' is even more obscure:  it appears in late OE, apparently out of nowhere, and similar words in other languages appear even later and actually seem to have been borrowed from English.

So it's only 'hound' that goes back to PGmc. and (with modification) PIE; the others are all of unknown origin at one date or another.

As pointed out in my December 23, 2017 post on fá 瞂 ("pelta"), I was inspired by the relevant entry in Paul Kroll's A Student's Dictionary of Classical and Medieval Chinese (Leiden: Brill, 2015).  Likewise, I would say that in the present case his entries on quǎn 犬 ("dog") and gǒu 狗 ("dog"), although brief, reveal an uncanny, innate, subconscious philological awareness of the issues I have explicitly raised in this post:

犬  quǎn    MC khwenX

1. dog; canine; cur.

a. pejorative term for a contemptible person, wretch, knave

狗  gǒu    MC kuwX

1. dog, esp. smaller dog (cf. 犬 quǎn, larger dog); cur; e.g. (med.) ~突 gǒutū, dog-door.

a. hunting-dog, hound.

To conclude with a bit of levity, first, because strict Muslims consider them to be unclean animals, the dog and the pig are whitewashed from the Chinese zodiac in Malaysia:

"Zodiac dog, pig go missing for Chinese New Year", by Trinna Leong, Straits Times (1/10/18)

Second, in this year of the dog, "ten canines" is a pun for "perfect in every respect".

shí quǎn shí měi
shíquánshíměi

十犬十美
十全十美

ten canines, ten beauties
ten complete, ten beauties

Note that, for the pun to work, quǎn 犬 ("canine") is used instead of gǒu 狗 ("dog").

Appendices

1. Paleographical evidence for the earliest known forms of quǎn 犬 ("dog") and gǒu 狗 ("dog") (by Matt Anderson)

The Shang oracle bone inscription / bronze inscription (OBI / BI) forms are the same pictograph, though the BI forms are typically far more elaborated:

Shang BI (from Yan Zhibin's 嚴志斌 Shang jinwen bian):

And here are OBI forms (from Liu Zhao's 劉釗 Xin jiaguwen bian):

You can see that the OBI forms appear in more or less simplified variants, and that a few of the BI forms are as simplified as the simplest OBI variants. And two of the last OBI forms given by Li Zongkun (one of which is also the last given by Liu Zhao) are almost as elaborate as the most elaborated BI forms.

I'm not aware of 狗 being used for gǒu "dog" before the Warring States, but there is an early Western Zhou BI (the Chang Zi Gou ding 長子狗鼎) that appears to use the graph 狗, or at least something structurally identical. It's being used for a given name, though, not as the name of a kind of animal (it's the topmost form below, from Gao Ming's 高明 Guwenzi leibian—zengding ben):

The full inscription reads (with my quick translation):

長子狗乍文父乙尊彝
Prince Gou of Chang made this ritual vessel (in honor of) his accomplished father Yi.

There's no way to know what meaning (if any) the word gǒu (or at least the graph 狗) has here (aside from its use as a personal name of course), but it's perhaps interesting that this vessel is from the south (from the Wuhan area).

2.

For the record, here are the OED etymologies for "dog", "hound", "canine", and "cur".  It's curious that the most opaque one, "dog", is also the most used word for dog nowadays and receives the most extensive treatment from the editors of the OED.  The one that seems the most transparent, "canine", is not much used now and is given the shortest treatment by the OED editors.  Both in terms of use and in editorial treatment, "hound" and "cur" fall between the other two.

dog

Etymology: Origin unknown.

The word belongs to a set of words of uncertain or phonologically problematic etymology with a stem-final geminated g in Old English which is not due to West Germanic consonant gemination and therefore does not undergo assibilation. These words form both a morphological and a semantic group, as they are usually Old English weak masculine nouns and denote animals; compare frog n.1hog n.1pig n.1stag n.1, Old English sugga (see haysugge n.), Old English wicga (see earwig n.), and perhaps teg n.1 It has been suggested that these words show expressive gemination, perhaps due to their being originally hypocoristic forms. (For discussion see R. M. Hogg 'Two Geminate Consonants in Old English' in J. Anderson Lang. Form & Ling. Variation (1982) 187–202.) For some of the words, substratal influence has also been considered (compare pig n.1). Because attestation of these words in Old English is generally rare and confined to glossaries and onomastic evidence (as in the case of dog n.1), if they are attested at all, and also because there is often a better-attested synonym (in this case, hound n.1), it seems likely that the words were stylistically marked in Old English, i.e. considered non-literary or informal.

The word is attested twice as a place-name element (in the genitive plural) in a 14th-cent. copy of an Anglo-Saxon charter of 941 granting land at Buckland Newton, Dorset (doggene berwe is probably to be identified with Dogbury Hill, an ancient hill fort):

a1400  (▸OE)    Bounds (Sawyer 474) in W. de G. Birch Cartularium Saxonicum (1887) II. 500   Endelang stremes on doggene ford þanen up on doggene berwe.

It is also perhaps attested (in the compound doggiþorn ) in a late 12th-cent. copy of another charter purportedly recording a grant of land in Gloucestershire made a959, although it is unclear whether the form here represents this word or its derivative doggy adj.:

c1175  (▸?OE) Bounds (Sawyer 664) in W. de G. Birch Cartularium Saxonicum (1893) III. 113   Of pislege on doggiþorn, of þam þorne to eadingham.

Compare also the following place names: Dogeflod , Surrey (1257; formerly Dogflood, now lost), Doggeworth , Devon (1281; now Dogsworthy), etc.

Also early as an element in bynames and surnames; compare: Syward Dogheafd (a1195), Richard Doggetall' (1201), Robertus Doggefel (1201), Robertus Doggisheued (1204), etc. Compare also Roger le Doge (1296).

The word occurs in a number of other European languages, considerably later than in English, and in many cases with the identifying attribute 'English'. All of these instances probably show borrowing either directly or indirectly < English. Compare Dutch dog (16th cent.; in early modern Dutch also dogge ), German Dogge (16th cent. as dock , docke ; 17th cent. as dogg , dogge ), Swedish dogg (17th cent.), Danish dogge , dog (a1700); French dogue (15th cent. in Middle French denoting a type of hunting dog; 14th cent. as an insult used to a Frenchman by an Englishman), Spanish dogo (1644), Portuguese dogue (1789; 1727 as †dogo ), Italian dogo (19th cent.; a1712 in the diminutive doghetto ). In all of these languages the word is applied more narrowly to particular varieties of dogs, usually mastiffs. This probably reflects the types of dogs which were imported from or associated with Britain, and probably has no bearing on the early meaning of the word in English.

The etymology of the English word is unknown. No likely cognates have been identified with a meaning at all close to that of the English word, and all attempted etymological explanations are extremely speculative. A word of this phonological shape is hard to explain as a regular development from a Germanic base, but nonetheless a number of attempts have been made. One attempt sees a connection with the Germanic base of dow v.1, assuming an original meaning such as 'useful or faithful animal', but this has not met with general acceptance. In this connection an Old English personal name Dycga is sometimes compared as a possible formal parallel from the same base, but it is quite possible that the personal name has no connection with dog n.1 Another attempted etymology takes the word ultimately from the Indo-European base probably meaning 'run' which is probably reflected by Sanskrit dhav- (see prothetely n.), but this poses a number of formal difficulties. Another suggestion is that the word shows a development from an Indo-European base meaning 'to be or become unconscious', but this involves a very large number of unattested stages in the semantic development (assuming a development 'bundle' > 'cuddly bundle' > 'pet' > 'dog'), and also involves a very uncertain original base form.

hound

Etymology: Common Germanic: Old English hund = Old Frisian hundhond, Old Saxon hund (Low German hund, Middle Dutch hont (d-), Dutch hond), Old High German hunt (d-), (Middle High German hunt, German hund), Old Norse hundr (Swedish, Danish hund), Gothic hunds < Old Germanic *hundo-z, generally held to be a derivative of base *hun-, pre-Germanic *kun-, in Greek κύωνκυν-, Sanskrit çwan-çun-, Lithuanian szůszun-, Old Irish cu dog; compare also Latin canis.

For the d (dh) of Germanic hund, the suggestion has been made of association with the verb hinþan to seize, as if the word were understood to mean 'the seizer'.

canine

Etymology: < Latin canīnus, < canis dog; compare French canin, 16th cent.

cur

Etymology: Middle English curre corresponds to Middle Dutch corre 'canis villaticus, domesticus' (Kilian), Swedish and Norwegian (widely-spread) dialect kurrekorre 'dog', etc. The latter is generally associated with the onomatopoeic verb Old Norse kurra to murmur, grumble, Swedish kurra to grumble, rumble, snarl, Danish kurre to coo, German obsolete and dialect kurren to growl, grumble, murmur, coo, compare gurren to coo, Middle High German gürren to bray as an ass. The primary sense appears thus to have been 'growling or snarling beast'. But no corresponding verb appears in English, so that Middle English kurre was probably introduced from some continental source. The combination kur-dogge is met with considerably earlier than the simple kurrecur. Senses 23 are possibly independent echoic formations.

3.

Miscellaneous notes on "dog", "canine", and "cur" from Jichang Lulu:

Dog

"Not a dog's chance, or one more impenetrable etymology", by Anatoly Liberman, OUPblog (5/4/16).

This ("The origin of dog" by Olivier van Renswoude) is also interesting, with two newer proposed etymologies: one related to 'dusk', another to 'duck'.

Canine (and quǎn 犬)

Reconstruction:Proto-Indo-European/ḱwṓ.

Reconstruction:Proto-Sino-Tibetan/d-kʷəj-n.

Cur

The (rare) Dutch form corre/korre, presumably cognate to cur, apparently isn't attested in Middle Dutch (~1200-1500), as per the Middelnederlands(sch) Woordenboek.

The source for it is Kiel=Kiliaan=Kilianus' 1599 dictionary (Etymologicum teutonicae linguae, sive dictionarium teutonico-latinum…); here's the entry from the 1777 edition:

Checking orig just to make sure…

Yup. It's in the 1599 original, p. 256 (Google Books).

On the Swedish side of things, kurre might be derived from kurra 'grumble'.

Hellquist, Svensk etymologisk ordbok, 1922 (via Project Runeberg)

except for this old dialect dictionary that compares it to Finno-Ugric and Turkic words:

Svenskt dialektlexikon, 1862-1867 (also via Project Runeberg).

The verb kurra 'grumble' is well attested in Old Norse (refs in Cleasby & Vigfusson (i.e. Guðbrandur Vigfússon)). No idea where it comes from, maybe imitative of 'grumbling'.

Maybe the mysterious Skt. kukkura (with modern reflexes) is also imitative? Or non-IE? So many questions!

[Thanks to Jim Mallory, Chris Button, Matt Anderson, Don Ringe, Paul Midler, Ben Zimmer, Jonathan Smith, Axel Schuessler, and Chau Wu]



95 Comments

  1. S Frankel said,

    March 7, 2018 @ 10:40 pm

    Oh, "dog" is easy if you're willing to disregard common sense and vowels: Welsh "da gi" = 'good dog'. (There are several possible, if desperate, explanations gemination in the ancestral English form.)

  2. GMan003 said,

    March 8, 2018 @ 12:11 am

    Fascinating read! I had actually noticed this exact similarity not too long ago, since I've been looking for just this sort of similarities in proto-languages, as part of an artlang project of mine. I didn't have the base of knowledge to be able to even guess whether it was just a coincidence, evidence of borrowing, or something else entirely. It was incredibly interesting seeing the actual research behind it all.

  3. Christian Weisgerber said,

    March 8, 2018 @ 12:24 am

    Piotr Gąsiorowski on "The Etymology of Old English *docga":

    ABSTRACT: This article explores the origin of English dog (OE. *docga), generally regarded as a word of unknown origin. It is argued, on the basis of its morphology, that the word is a hypocoristic derivative of dox, an Old English colour adjective. The article suggests that the relation between OE. frox and frocga 'frog' is not an isolated irregularity but an example of a derivational process represented also by dox:*docga and possibly by other such pairs in Old English (e.g. fox:*fogga).

    https://repozytorium.amu.edu.pl/bitstream/10593/2479/1/Dogga.pdf

  4. Jason M said,

    March 8, 2018 @ 2:48 am

    Thanks for this, which I lapped up.

    BTW I noticed the 1922 ordbok also gave a reference at the end of the entry to "turk. chaur, hund". In my 5 minute Wiktionary and Googling, I can't find anything on "chaur". "kurt" is wolf in Turkish but comes according to Wiktionary from a proto-Turkic root.

    Also, seems likely that "kurr" onomatopoeically would be like English "grrr" which would itself seem to be broadly recognized as a "word" for a dog-like growl.

  5. SO said,

    March 8, 2018 @ 3:52 am

    Regarding chaur cf. e.g. the explanation as "infidelles chiens & idolatres" here in L' histoire universelle du monde (1572) and elsewhere in late 16c sources:

    https://books.google.co.kr/books?id=lJg8AAAAcAAJ&pg=RA1-PA63

  6. David Marjanović said,

    March 8, 2018 @ 6:51 am

    I have lots more to say, and several things to read up on first, so just so much now:

    1) Before there were tones, both 狗 and 犬 must have ended in a glottal stop. (Perhaps they still do in Southern Wú or something, I have no idea.) Where did that come from? It's not there in PIE or Tocharian or Iranian or anywhere else in IE, and it's not a known OC suffix either, is it?

    2)

    From Mesolithic cemeteries in Denmark and Southern Sweden (c. 5000 BC), we know of separate dog graves

    …but that's about 2500 years before any IE language was spoken there or anywhere close.

    3) There is another Latin word canis, a color adjective meaning "yellow ~ gray", the color of straw and dust. If dog is related to dusk(y), as Gąsiorowski suggests, perhaps canis is related to canis after all…

    4) Chris Button, is your OC reconstruction published somewhere? It looks very different from Baxter & Sagart, in some respects very different from all others; I'm interested in why that is.

  7. David Marjanović said,

    March 8, 2018 @ 6:54 am

    Also, seems likely that "kurr" onomatopoeically would be like English "grrr" which would itself seem to be broadly recognized as a "word" for a dog-like growl.

    …Also, if you make up a PIE */gr̩/ and send it through the usual Germanic sound changes, you actually get */kur/… :-)

  8. Olivier van Renswoude said,

    March 8, 2018 @ 7:05 am

    Very interesting read. I was notified of it by your link to my article on the word dog. Please note that I have also reviewed the word hound/*ḱuōn recently, with a conclusion about its original meaning that fits well with the role that dogs played in early Indo-European society. You can read the English translation here.

    With regard to English cur and Middle Dutch corre, the apparently related Middle Dutch querie, querre 'bitch' points to an ablaut pattern that excludes a relationship with Old Norse kurra and Middle Low German korren 'to growl', since they themselves rather ablaut with lcelandic karra 'to creak, shriek' and Middle Dutch carren 'to creak'. Therefore I would connect the group of cur with that of Old Norse kvirr, kyrr 'calm', Middle Low German querre 'tame' and High German kirre 'tame, compliant'.

  9. Victor Mair said,

    March 8, 2018 @ 7:32 am

    There's an article by Ilya Gershevitch (in Iran and Islam, in memory of the late Vladimir Minorsky, ed. C. E. Bosworth, Edinburgh 1971, 267-291; reprinted in Ilya Gershevitch, Philologia Iranica, ed. NSW, Wiesbaden 1985, 237-261) which contains an interesting idea about Latin canis.

    I have a pdf of the relevant pages if anyone is interested in seeing them.

  10. Coby Lubliner said,

    March 8, 2018 @ 9:15 am

    "Canine" (sometimes written K9) is standard US police English for "police dog" ("the canine approached the individual…").

  11. Jason M said,

    March 8, 2018 @ 9:29 am

    @SO, thanks. For those who didn't follow the link or the passage is discussing "Tartare" people and their religious practices and views. The passage with Chaur is discussing the Tartar view of Christians: "Ilz appellent le Pape & tous les Chrestiens, Dzinthis,qui signifie Payens, & Chaur, c'est à dire Infideles, chiens, & idolatres, à cause qu'ilz honorent le boys & les pierres." –> They call the Pope and all Christians Dzinthis, meaning Pagans and Chaur, that is to say, Infidels, Dogs and Idolaters.

    Now whence "Dzinthis". Akin to "djinn"? Or Tibetan 'dzin? Or to the Old Chinese *dzin to bring the discussion full circle?

  12. Jerry Friedman said,

    March 8, 2018 @ 10:04 am

    I was a bit surprised that Don Ringe said that Latin canis "cannot" derive from "PIE *ḱwón- ~ *ḱun-". Is the reconstruction of protolanguages really certain enough to call something like that impossible? Are there no known derivations in other languages that are that anomalous? Or should I read "cannot" as "is very unlikely to"?

  13. Olivier van Renswoude said,

    March 8, 2018 @ 10:21 am

    In his Etymological Dictionary of Latin and the Other Italic Languages, De Vaan confidently states that the Proto-Indo-European paradigm *ḱu-ōn (nom.sg.), *ḱuon-m (acc.sg.), *ḱun-os (gen.sg.) regularly developed into Proto-Italic *kō, *kwanem, *kunos, after which *kwanem was leveled into *kanem, which in turn led to a new nominative. Which seems quite plausible to me.

    Ringe's proposed connexion with Welsh ceneu 'puppy, cub' (and Middle Irish cano, cana 'wolf cub') is not unthinkable, however, since Latin recēns 'of recent origin, new, fresh' shows that this root *ken- was also present in Italic.

  14. Jamie said,

    March 8, 2018 @ 10:26 am

    Great article. I assume Japanese 'ken' is also from quǎn 犬

    > and writing, many of which had evident linkages to the west

    Is there any evidence that writing cam from the West? (Obviously not the writing system, but the concept of writing?) I had always assumed it was an independent development.

  15. Olivier van Renswoude said,

    March 8, 2018 @ 10:33 am

    I submitted a response to Jerry Friedman earlier, but it never materialized. Perhaps it was caught by a spam filter, because I used a link (to an online version of Pokorny's dictionary). Here it is again, but without the link:

    In his Etymological Dictionary of Latin and the Other Italic Languages, De Vaan confidently states that the Proto-Indo-European paradigm *ḱu-ōn (nom.sg.), *ḱuon-m (acc.sg.), *ḱun-os (gen.sg.) regularly developed into Proto-Italic *kō, *kwanem, *kunos, after which *kwanem was leveled into *kanem, which in turn led to a new nominative. Which seems quite plausible to me.

    Ringe's proposed connexion with Welsh ceneu 'puppy, cub' (and Middle Irish cano, cana 'wolf cub') is not unthinkable, however, since Latin recēns 'of recent origin, new, fresh' shows that this root *ken- was also present in Italic.

  16. Victor Mair said,

    March 8, 2018 @ 10:36 am

    Question from Jichang Lulu:

    I wonder what any Sino-Tibetanists might think about ST words possibly
    related to 犬, and whether they predate a loan from IE or are
    incompatible with it.

  17. Victor Mair said,

    March 8, 2018 @ 10:41 am

    @Olivier van Renswoude:

    Thank you for your good comments.

    I found your comment with the Pokorny link in the spam filter. Here's the link:

    https://indo-european.info/pokorny-etymological-dictionary/ken-3.htm

  18. Scott P. said,

    March 8, 2018 @ 10:47 am

    …but that's about 2500 years before any IE language was spoken there or anywhere close.

    Yes, but Victor thinks that the IE root comes from a pre-IE substratum, see the discussion above.

  19. Olivier van Renswoude said,

    March 8, 2018 @ 10:53 am

    @Victor Mair:

    Much obliged!

  20. 번하드 said,

    March 8, 2018 @ 11:40 am

    Fascinating.
    I have been using the similarity between French 'chien' and Sinokorean 개 견(犬)(gyeon), http://hanja.naver.com/hanja?q=%E7%8A%AC for its mnemonic value, but had never thought there could be a real connection:)

  21. Jerry Friedman said,

    March 8, 2018 @ 1:01 pm

    Olivier van Renswoude: Thank you.

    I don't know anything about the spam filter here, but I've posted quite a few comments with links, and they've all come through.

  22. Andy said,

    March 8, 2018 @ 1:07 pm

    @David Marjanović: Where did you find this other Latin word 'canis', the one meaning 'yellow/grey'? I looked online and had a rummage through the usual dictionaries, but with no luck. I suppose you're not thinking of 'canus', meaning 'white/grey' (with a long a and from a different root)?

  23. Michael Watts said,

    March 8, 2018 @ 1:14 pm

    (except for Min, the most archaic living variety of Sinitic)

    How is this measured?

  24. James said,

    March 8, 2018 @ 3:13 pm

    For what it is worth, here are Beekes' remarks s.v. (Etymological Dictionary of Greek, Brill 2010 [available through Brill Dictionaries Online, with a subscription]). N.B. that, as far as Greek is concerned, it is already found attested in a compound in Mycenaean:

    κύων [m. , f.] 'dog, bitch' (Il.).

    Forms
    Gen. κυνός, acc. κύνα.
    Myc. ku-na-ke-ta /kun-(h)āgetās/.
    Compounds
    Several compounds, e.g. κυν-ηγέτης, Dor. -ᾱγέτας, -ᾱγός "leader of dogs", 'hunter' (ι 120); see Chantraine 1956a: 83ff.; ἀπό-κυνον plant name 'Marsdenia erecta' (Dsc., Gal.); see Strömberg 1940: 65 and 143; on ► κυνάμυια s.v.

    Derivatives
    Diminutives κυν-ίσκος (Hdt.), -ίσκη (Ar.), -ίδιον, -άριον (Att.); κυνώ [f.] 'female dog', also as a PN (Hdt.); κυνέη 'dog's skin' (Anaxandr.), 'cap, helmet', originally made of dog's skin, later from other materials (cf. αἰγείη, χαλκέη, etc.; Schwyzer: 37, Trümpy 1950: 40ff.); κυνάς [f.] 'belonging to a dog, dog hair, etc.' (Theoc.); κύνειος, -εος 'belonging to a dog' (Ar.), 'shameless, impudent' (Il.), κυνικός 'dog-like, cynical' (X., Men.), κυνώδης 'dog-like' (Arist.); comp. and superl. κύντερος, -ον, -τατος 'more shameless, impertinent'; κυνηδόν [adv.] 'like a dog' (S., Ar.); κυνίζω "to play the dog", i.e. 'to live as a cynic', κυνισμός (Apollod. Stoic .).

    Origin IE * ḱuon- 'dog'

    Etymology
    The name of the 'dog' is preserved in most IE languages: e.g. nom. κύων, Skt. śvā́, Lith. šuõ, gen. κυνός, Skt. śúnas, Lith. šuñs, etc. (the Gr. accentuation is oldest), from IE * ḱuō(n), gen. * ḱun-ós, etc. For Lat. canis, Schrijver 1991: 461 assumes that a development * wo > * wa in open syllable yielded an acc.sg. PIt. * kwanem. The word is also found in Anatolian: Hitt. LÚ kuu̯an- [c.] 'dog-man', gen.sg. kūnaš , HLuw. swan(i)- [c.] 'dog'. The paradigm is strange because of the lack of an old e- grade in the ablaut pattern."

  25. Victor Mair said,

    March 8, 2018 @ 4:19 pm

    @James

    Thank you very much for your comment. I thought it would be in Mycenaean and am very happy to have that confirmed.

  26. martin schwartz said,

    March 8, 2018 @ 5:31 pm

    Dear Victor and all,
    Lots here to digest, many interesting things.
    For now, a few things off the top of my head: For PIE, one should note the initial of the ablauting *kwen-/*kwon-: kun- with ^ over the k to distinguishnthe phoneme from PIE plain k and labiovelar k (with raised w).
    The n disappaears in the naminative in Indo-Iranian and Baltic.
    This word looks like a derivative from a verbal root (the same was said of *dhwer-/dhwor- 'door', for which I found a PIE verbal root; details upon request.
    One should have Tocharianists speculate about the Toch, forms in the
    ostensible period of borrowing into Chinese.
    For Iranian, alongside
    *swan-: sun-, there was also *kuta-, reflected in Iranian languages both of the West and the East. Such a *kuta- may have been borrowed
    into Hungarian (check the form). Ir. *kuta- quite possibly has onomatopoeic ku parallel to that of Skt. kukkura-.
    Interestingly, Russ. sobaka 'dog' is borrowed from Iranian *spaka-
    Pers. sag), cf. the fem. name Spako:, recorded in Herodotus, and various reflexes of *spaka- in West Iranian languages today. The first -a- here is from a PIE syllabic n.
    All best,
    Martin

  27. martin schwartz said,

    March 8, 2018 @ 5:33 pm

    p.s I mistyped nominative.
    M

  28. David Marjanović said,

    March 8, 2018 @ 6:06 pm

    I was a bit surprised that Don Ringe said that Latin canis "cannot" derive from "PIE *ḱwón- ~ *ḱun-". Is the reconstruction of protolanguages really certain enough to call something like that impossible? Are there no known derivations in other languages that are that anomalous? Or should I read "cannot" as "is very unlikely to"?

    You should read it as "is extremely unlikely to, unless Schrijver is right and PIE *-wo- > Proto-Italic *-wa- in open syllables is a regular sound shift"… :-)

    But yes, this and many other PIE reconstructions really are as certain as Ringe presents them. I should add that the two stems *ḱwón- and *ḱun- were not in free or unexplained variation, but represent the o-grade and the zero-grade in the ablaut system.

    Yes, but Victor thinks that the IE root comes from a pre-IE substratum, see the discussion above.

    That's unlikely, because *ḱwón- (nominative *ḱwō, most likely) has a good IE-internal etymology: it's simplified from **pḱwón-, which is an "individualizing" n-stem noun built to *péḱu "livestock" (itself a stress-shifted nominalization of an u-stem adjective to a root *peḱ-). **pḱwón- would then be "the one who has something to do with the herds", like guarding them. It could also explain the Slavic "dog" word pes, which is the expected outcome of *peḱ-. I hope I can find my source for that again.

    Beekes writes *u instead of *w simply because he leaves it to the reader to apply syllabification rules; [u] and [w] were the same phoneme. By writing *ḱuon- he does not mean to imply the stem had two syllables.

    I suppose you're not thinking of 'canus', meaning 'white/grey' (with a long a and from a different root)?

    Oh. I am in fact! Thank you! It's been 20 years, and I only remembered something was different about the declension of these two words…

    The different vowel length most likely rules out any connection at all.

    (except for Min, the most archaic living variety of Sinitic)

    How is this measured?

    It's not measured, of course, and Min has a bunch of very interesting innovations of its own. However, unlike most other Sinitic varieties, Min is not derived from Middle Sinitic and therefore lacks several innovations that MS already had.

  29. martin schwartz said,

    March 8, 2018 @ 6:40 pm

    Re The learned remarks by David Maryanovic', doesn't Slavic simply
    reflect, for 'dog", PIE *pek'u- 'small domestic herd animal'?
    I perhaps wrongly remember that that was the basis for the etymon
    PIE *pk'wVn-. More details and the semantics and the "bridging" of the forms would be valuable.
    The very late Avestan Widêwdâd ( Vîdêvdâd, "Vendidad") has a chapter (13) on real dogs, their varieties (including vohunazga-, which I identified as 'bloodhound', vohu- 'blood' + -zga- < PIE *-sgh2o-, root *seh2g- 'follow the tracks, perceive the spoor', whence Lat. sagire, Eng. seek, etc.), and charming descriptions of the qualities of dogs. The next chapter (14) is about animals which in ancient Iranian folk-taxonomy were classed as '"dogs": foxes, beavers, otters, hedgehogs, porcupines…. For injury to real dogs and the quasi-dogs different punishments are set forth.
    Martin Schwartz

  30. Victor Mair said,

    March 8, 2018 @ 9:18 pm

    From Tsu-Lin Mei:

    Recently I was able to show that Old Chinese 犬*khwinx and Chepang kuy? 'dog' are cognates. There are altogether 16 cases in which Chinese rising tone corresponds to Chepang -? (glottal stop), which shows that the source of Chinese rising tone is -?. This a claim made by Haudricourt, Pulleyblank, Mei (1970) but until recently, no one was able to demonstrate this claim via Tibeto-Burman evidence. Ostapirat 1998 first suggested this idea for Tiddim and Chepang, and Schuessler 2007 extended it to include Old Chinese and Lolo-Burmese (tone B). I added a few examples including 犬屎乳虎。There are now altogether sixteen examples, including 'tiger' 虎OC khla? which is a loan from Mon-Khmer *kla?. The last item is actually Jerry Norman's idea in Norman and Mei (1976). I rediscovered it by reading Harry Shorto, A Mon-Khmer Comparative Dictionary (2006).

  31. Victor Mair said,

    March 8, 2018 @ 9:39 pm

    From E. Bruce Brooks:

    That is correct. There is an Aryan superstrate; same as in India, and by the same means (conquest). The language of the substrate/subject population is from a different language family (Austro-Asiatic). I was trying to find the Chu Language page on my website, but it seems to have gotten lost.

    People make more of Tocharian than the situation warrants, but the situation itself is pretty clear.

    This gives doublets, two words for the same thing, beginning in the 04c when the social structure changed (for military reasons) and the substrate began to be integrated (militarily and later politically) into the superstrate. The philosophies (pl) of Mencius are a direct outcome of this 04c process, and Mencius himself, the guy not the text, was one of its architects. Early Ch8inese lexicography was concerned to work all the doublets into a common language concept, as slightly differentiated synonyms, with sometimes ludicrous results. There are these two words both meaning dog 9to different people). Artificial distinctions are made (one of them is "dog" vs "puppy," a self-refuting bit of fake linguistics), and are still being swallowed by foreigners, and probably also by the natives, who are not prepared to confront the historical situation. Not prepared.

    shr chywaen shr mei

    The first time through it means "ten dogs, ten beautiful women," an unintelligible conjunction and thus a riddle. The second time through, with kanji (and one tone) changed, both halves have other meanings. Not one, both. They have become a compliment to those who believe in the perfection of character, especially their own. Something like "ten aspects in which one's character is fully developed (fully realized), ten kinds of internal excellence." They now have the same context and social referent. The riddle is solved. As for the second mei, see the term "nei mei" in any halfway decent dictionary. Beautiful women are not involved. Not that time.

  32. SO said,

    March 8, 2018 @ 10:06 pm

    Here's a follow-up on "chaur". I was a bit too lazy last time and didn't follow the reference given in the margin of p. 63v of _L' histoire universelle du monde_ (1572) — which leads to Maciej Miechowita (1457-1532) and his _Tractatus de duabus Sarmatiis_. I've meanwhile checked two different editions of that work, one dating from 1517 [*1], the other from 1518 [*2]. Disregarding abbreviations etc. the relevant passage in both interestingly reads as follows: "christianos vero [1517:] dzincis [1518 has "dzintis" instead] id est paganos et gaur id est infideles et sine religione appellant" — no dogs at all here, the "gaur" are merely called "infidels and without religion". So it appears that the reference to dogs here is in itself a later addition (even if going back to 1572 if not to yet earlier times) and in the end simply a reference to the pejorative use of the word (just google for "infidel dogs" to see more of this). Also note the slight differences in spelling: the Tractatus gives dzincis / dzintis (vs. dzinthis in most later adaptations of the passage) as well as gaur (vs. chaur etc.).

    There is quite some variation seen in the spelling of "chaur" from early on, also yielding e.g. Giaur(lar), Gyaur(lar), Kaur(lar) (with plural forms in -lar). [*3]

    More on the word in question: https://en.wikipedia.org/wiki/Giaour

    [*1] 1517 ed.: https://polona.pl/item/tractatus-de-duabus-sarmatijs-asiana-et-europiana-et-de-contentis-in-eis,Mzc5NjY5OQ/9/#item or also https://jbc.bj.uj.edu.pl/dlibra/publication/242954

    [*2] 1518 ed.: https://books.google.co.kr/books?id=Kjs8AAAAcAAJ&pg=PT5

    Readers of German might prefer this 1518 translation (_Tractat von baiden Sarmatien_), reading: "vnd haissen die christen Dzintis/ das ist hayden/ vnnd Gaur/ das ist ungelaubig". See: https://books.google.co.kr/books?id=zZk-1QZ-dvQC&pg=PA19

    A somewhat later Italian translation (_Historia delle due Sarmatie_, 1562, p. 26) misprints "gaur" as "baur" … (while retaining "Dzintis"): https://books.google.co.kr/books?id=cHLYSlgo87kC&pg=PA26

    [*3] See e.g.:
    _De Turcarum moribus epitome_ (here 1558): https://books.google.co.kr/books?id=1kZBAAAAcAAJ&pg=PA113
    _Chronica Turcica (1578)_: https://books.google.co.kr/books?id=JExXAAAAcAAJ&pg=RA1-PT124

  33. Chris Button said,

    March 8, 2018 @ 11:54 pm

    @ Victor Mair

    I have come across many other terms in Sinitic that appear to have clear parallels with corresponding words in Indo-European (e.g., mǎ 馬 ["horse"] / mare; gūlu 軲轆 / wheel, cycle; Cant. mou4 巫 / magus).

    Regarding "wheel, cycle", I'm surprised you did not specifically mention 車 *kɬàɣ "chariot" and its solid Indo-European connection that you talk about in your 1990 article. Incidentally, further support for the *kɬ- initial (in addition to Schuessler's remarks on his *k-hl-) in terms of its evolution into Middle Chinese comes from the relationship between 齒 *kɬə̀ɣʔ "tooth" and 杵 *kɬàɣʔ "pestle" via the ə/a alternation (compare the association of "molar" and "mortar" in English).

    @ David Marjanović

    Chris Button, is your OC reconstruction published somewhere? It looks very different from Baxter & Sagart, in some respects very different from all others; I'm interested in why that is

    Its foundations may be found in Edwin Pulleyblank's numerous works as a modified amalgam of his strongest suggestions along with additional considerations.

    I'm currently compiling "A Derivational Dictionary of Chinese and Japanese Characters" which will contain OC reconstructions for its 6,500 or so characters. Unfortunately progress is exceedingly slow since without funding, nor any academic position to support it, I can only work on it in my very limited spare time.

  34. Jason M said,

    March 9, 2018 @ 12:35 am

    Thanks, SO. Suppose I myself could have also followed the sidenote references also.

    Thus, "chaur" ultimately derives via Persian, likely from the Semitic Aramaic according to Wiktionary: gaḇrā [gbrʾ], "man; person", though the traditional, now contested, origin was from Arabic كَافِر‎ (kāfir, "unbeliever"). If the Persian derivation is right, the pre-Islam Persians referred to the Aramaeans they knew who practiced Zoroastrianism in Mesopatamia using that non-Persian people's own (Aramaic, of course) word for "man", kind of like how Trump refers to Mexicans as "Bad Hombres". And, in the end, as it were, "chaur" has a man bites dog etymology.

    As for Dzinthis, thanks also, as that was enough to find that ultimately this is equivalent to Turkish: "Dinsiz" which is from "din" (religion) + "siz" (lacking): ie religionless people. "siz" is a Turkish suffix for "lacking" or "without". "Din" also comes from Persian by way of Arabic, and the Persian ultimately, according to Wiktionary, from Elamite, which is currently a language dead end.

  35. martin schwartz said,

    March 9, 2018 @ 1:36 am

    Re E. Bruce Broooks' comment: I suspect you're using "Aryan"
    for "Indo-European". The terms should NOT be equated. Arya-
    was used by the Indo-Europeans speakers who settled Central Asia and then \modern Iran, and by the Indo-European speakers who settled in Northern India, as a self-designation. Both groups go back to Central Asia, and has shared a common language and culture in pre-historic times, hgence the common use of Arya -.Because the term Aryan has such a racist history, it is better to speak of Indo-Iranian and put into
    this category the kindred languages of Nuristan, formerly "Kafiristan"
    because the area was Islamized only in the early 20th cent.
    As to whether Northern India was conquered or gradually settled
    (or both) by Indo-European speakers, that's a matter of contention.
    There's also a problem of the reculturation of the Bactro-Margiana
    Archeological Complex (BMAC).
    Kafiristan leads me OS and Jason M on "Chaur, Gaur', etc. Aram. gaBrâ
    means not "bad hombre' but simply 'man', It does not occur as a WORD in Iranian, but (and here arose confusion) as a LOGOGRAM (Aramæogram) in Zor. Middle Persian (Pahlavi) and pronounced /mard/
    'man'. Now, consider the fact that Pers. gabr non-Muslim' (Zoroastrian
    but early also Hindu) occurs only in the Islamic period. Whiile Arabic kâfir 'infidel' was merely transcribed into Classical Persian and pronounced ka:fir, later (in Modern Persian of Iran) ka:fer and ka:fær,
    all with a non-phonemically aspirated k (as in English), and a like æ,
    the Arabic word was HEARD as having a tense, non-aspirated k (like that of French) interpretable to Persian ears as g; the long a in Arabic is actually æ: , interptertable as Pers. a, as against the Pers. back vowel "long" a, and in early Perisan fr, vr, and wr were in flux, as in the word for 'silk', afrêSHum, alternating with abr- and awr-. Since the Arabic word was heard as a salient term markig non-Muslims at a time when that really mattered, it was taken over orally, and not, as per the usual process, merely literally.
    As for Arabic dîn 'religion', this is from Sasanian Middle Persian dên
    (cf. Arabic mahrajân 'festival' from Sas. MPers. Mihragân 'feast-time of Mithra). The MPers. is from Avestan daênâ
    (thoughout %^ = macron), in Old Avestan metrically scanned as trisyllabic) < *dayanâ 'envisionment'. The origin is not Elamite;
    one should believe anything beginnig with Wik- (icncludibng Stig Wikander, q.v.) when it comes to matters Iranian.
    Martin
    (Martin Schwartz, here rankly pulling rank as Prof. Emeritus of Iranian Studies, Univ. Calif. Berkeley)

    This brings me to chaur, gaur, etc. and theremarks of SO and Jason M.

  36. martin schwartz said,

    March 9, 2018 @ 1:45 am

    oops, "This brings me…" at end of my last remark is a dittography
    of sorts. BUT, I add that Pers. gabr, perhaps via Kurdish, gave Turkish
    gâvur (gâur), whence our English giaour, w hich is a sort of word-symbol, like orientalist seraglio paintings, of the Ottoiman East.
    M

  37. Andy said,

    March 9, 2018 @ 3:21 am

    @David Marjanović: **pḱwón- is an ingenious suggestion, but it would be nice to have some trace, somewhere, of that initial p to make it more palatable. On that note, I really think any connection of *peḱ- with the general Slavic word for 'dog', pes, is unlikely. It seems like a pretty long shot that the inital p somehow survived into Slavic while utterly vanishing everywhere else (even in Baltic!! Latvian: suns, Lithuanian: šuõ (genitive šuñs), Old Prussian: sunnis); also, the OCS word for 'dog' is пьсъ (pĭsŭ), and that first jer can't come from the e of *peḱ-, it has to derive from original short *i.

    Let's not forget another pan-Slavic word -suka (bitch)!! I don't think everyone agrees on this etymology, but it can be very plausibly derived it from *ḱwṓ (+ka, the feminine suffix). If that's the case, then *ḱwṓ is in fact attested in Slavic in a tolerably regular form, and incidentally shows there was never an initial p.

  38. Olivier van Renswoude said,

    March 9, 2018 @ 4:44 am

    @Martin Schwarz:

    This word looks like a derivative from a verbal root (the same was said of *dhwer-/dhwor- 'door', for which I found a PIE verbal root; details upon request.

    Yes, Elisabeth Rieken and others have argued that it's a root noun rather than an n-stem. And so I've argued it's derived from *ḱuen-, which Pokorny interpreted as 'feieren, heiligen', but which might as well have been 'to devote, dedicate' vel sim. Hence hound/*ḱuōn means 'devoted, dedicated (one)'.

    Which verbal root do you believe underlies *dʰuer-/*dʰuor-?

  39. David Marjanović said,

    March 9, 2018 @ 7:20 am

    I perhaps wrongly remember that that was the basis for the etymon PIE *pk'wVn-. More details and the semantics and the "bridging" of the forms would be valuable.

    I'll try to look for my source on the weekend.

    also, the OCS word for 'dog' is пьсъ (pĭsŭ), and that first jer can't come from the e of *peḱ-, it has to derive from original short *i.

    True, I should have remembered this – but that's actually an improvement! *i is the epenthetic vowel of choice in Balto-Slavic, so if there ever was variation in how the illegal consonant cluster was repaired (deletion or epenthesis), *i is what we should expect.

    Let's not forget another pan-Slavic word -suka (bitch)!! I don't think everyone agrees on this etymology, but it can be very plausibly derived it from *ḱwṓ (+ka, the feminine suffix).

    Wouldn't that give ***saka instead of suka?

    And so I've argued it's derived from *ḱuen-, which Pokorny interpreted as 'feier[…]n, heiligen', but which might as well have been 'to devote, dedicate' vel sim. Hence hound/*ḱuōn means 'devoted, dedicated (one)'.

    Perfect.

  40. David Marjanović said,

    March 9, 2018 @ 7:25 am

    'devoted, dedicated (one)'

    Rather "the one who devotes (himself)". The other such deverbal root nouns, off the top of my head, seem to designate agents: *bʰōr is the one who carries ( = thief), *spoks > *spōks is the one who gazes ( = owl)…

  41. David Marjanović said,

    March 9, 2018 @ 8:10 am

    ***saka

    Not even – ***svaka.

    *spoks > *spōks

    With *ḱ of course, at least on the morphophonemic level.

  42. Olivier van Renswoude said,

    March 9, 2018 @ 8:44 am

    Rather "the one who devotes (himself)". The other such deverbal root nouns, off the top of my head, seem to designate agents: *bʰōr is the one who carries ( = thief), *spoks > *spōks is the one who gazes ( = owl)…

    Good point. I was fixed on the (pre-)Germanic form, which looks like a past participle developed from an old verbal adjective.

  43. Chris Button said,

    March 9, 2018 @ 9:22 am

    @ Tsu-Lin Mei

    … which shows that the source of Chinese rising tone is -?. This a claim made by Haudricourt, Pulleyblank, Mei (1970) but until recently, no one was able to demonstrate this claim via Tibeto-Burman evidence…

    Ostapirat 1998 first suggested this idea for Tiddim and Chepang…

    Some supporting evidence from Lamkang (a Kuki-Chin language):

    In Tedim (Tiddim) Chin, fortis *s > h > ʔ after a lax vowel but lenis *s > *h > tone3 (falling) after a tense vowel (following previous analyses, Ostapirat treats tense and lax as a quantity distinction but the underlying distinction is one of quality). In Lamkang, tone3 from fortis *s has merged with tone2 (corresponding to OC *ʔ) due to the fortis *s > h > ʔ shift (i.e. the reflexes of the original glottal and the derived glottal merged) whereas lenis *s has remained separate as tone3.

  44. David Marjanović said,

    March 9, 2018 @ 9:47 am

    the (pre-)Germanic form, which looks like a past participle developed from an old verbal adjective

    That makes sense as a reinterpretation once the root noun had become obscure!

  45. Chris Button said,

    March 9, 2018 @ 10:18 am

    @ David Marjanović

    By the way, the etymological relationship between words like 齒 *kɬə̀ɣʔ "tooth" and 杵 *kɬàɣʔ is not acknowledged by Baxter & Sagart where they have 杵 *t.qʰaʔ and 齒 *t-[kʰ]ə(ŋ)ʔ / t-ŋ̊əʔ. They also reconstruct 車 *kɬàɣ "chariot" as *[t.qʰ](r)A whereas I believe it originally would have differed from 杵 *kɬàɣʔ only via the presence or absence of the final ʔ. The notational A in B&S's system for 車 is where Pulleyblank has which would be *kɬàːɣ in my reconstruction. (Something funny happened to the *-àɣ rhyme such that it gave slightly divergent reflexes in Middle Chinese. Pulleyblank was thinking along the lines of phonemic vowel length but my take is that it was probably something more like an Old Chinese version of a TRAP-BATH split such that the a nucleus sometimes lengthened before ɣ. In any case, it has no bearing on the earlier stages of OC so can be safely ignored there).

    B&S's reconstruction of 犬 *kʰʷə́ɲʔ as *[k]ʷʰˤ[e][n]ʔ is not suggestive of a possible earlier phonotactic constraint. Incidentally, the phonotactic constraint on Old Burmese -waɲ to give *kʰwɨj2 "dog" (rather than kʰwɐɲ2) is paralleled in the similar loss of the nasal component in the regular (i.e. non-phonotactically constrained) rhyme *-aɲ in its evolution to modern Burmese.

  46. Robert Ramsey said,

    March 9, 2018 @ 11:37 am

    @번하드: fyi, the Korean word for 'dog' 개 is well attested as a disyllabic form 가히 kahi (low-high tones) in Middle Korean…

  47. Eidolon said,

    March 9, 2018 @ 7:03 pm

    Considering archaeological evidence of possibly domesticated dogs are found in Neolithic China from up to 9,000 years ago – certainly before the rise of any of the language families under discussion – the absence of an indigenous substrate word for, at the minimum, wild dogs would be surprising. But it is possible that dogs, both wild and tame, were considered identical to wolves in Neolithic China, and thus the same word was used for both until they were differentiated through influence from an Indo-European or otherwise "Nostratic" source. What are the oldest words for members of the canis family of animals in East Asia?

    For archaeology, see:

    http://www.kaogu.cn/en/Chinese%20Archaeology/3/The%20Problem%20of%20the%20Origin%20of%20Domestic%20Animals%20in%20Neolithic%20China.pdf

    As well as the following excerpt from "The Archaeology of China: From the Late Paleolithic to the Early Bronze Age":

    "The earliest domestic dog has been reported as present at Nanzhuangtou in Hebei (ca. 10,000 ca. BP) … More evidence for the early domestic dog comes from Jiahu in Henan (ca. 7000 BC), Cishan in Hebei (ca. 6000 BC), and Dadiwan in Gansu (ca. 7000 BC). Complete dog skeletons were intentionally buried in cemeteries or near houses at Jiahu, and on the bottom of some pits at Cishan."

  48. Eidolon said,

    March 9, 2018 @ 7:32 pm

    "…but that's about 2500 years before any IE language was spoken there or anywhere close."

    This is not a problem, as there are many words for dogs in European languages that cannot be traced to the proto-Indo-European root, indicating a diverse substrate of dog related terminology, which would match the archaeological evidence of earlier dogs in Europe such as in Oberkassel and Mechernich. Of course, I agree that these early archaeological evidences from central Europe cannot be used as support for the early association of dogs with an Indo-European culture that has yet to expand from its homeland on the steppe.

    But this just makes the case of kwon- more curious, since domesticated dogs predate the Indo-European language family by many thousands of years across much of Eurasia, and is thus very different from a Bronze Age technology like the wheeled vehicle, which is definitively associated with early Indo-Europeans and were spread all over Eurasia by their expansion.

    Placing the dog into the same category as the wheel, bronze technology, or even the sheep, hence seems problematic to me. Dogs are among the earliest of domesticated animals in most of Eurasia, not a Bronze Age import. We'd then need an explanation as to why both words for "dog" in Old Chinese or even all of Sino-Tibetan are loans.

  49. Victor Mair said,

    March 9, 2018 @ 8:18 pm

    It depends what you do with the dogs (e.g., hunting and herding [that takes a lot of training and close association over centuries] vs. eating them or using them at best as guards) and in what archeological contexts their remains are found. I touched upon some of this in the o.p. It never ceases to amaze me how many different breeds of dogs there are in Europe and the very specific purposes to which they are put. That doesn't just happen automatically. They have to be bred and selected for and then trained to do certain tasks. That's quite different from burying one in a hole in front of your house to scare away evil spirits.

  50. Victor Mair said,

    March 9, 2018 @ 8:43 pm

    Sagacious remarks from Xu Wenkan, who has been studying these matters for half a century and more:

    =====

    Bàidú guānyú dog de dà wén, juédé hěn hǎo, tècǐ zhìxiè. Jǐ nián qián, Zhōngguó hé Ruìdiǎn de yīxiē xuézhě fābiǎo lùnwén, tōngguò DNA fēnxī, rènwéi gǒu de xúnyǎng kāishǐ zài Zhōngguó dōngnán yīdài, dànshì zhège shuōfǎ yǔ kǎogǔ fājué de zhèngjù bùhé, yǐjīng bèi fǒudìng. Hòulái yòu yǒurén shuō zài Central Asia, dàn tāmen bǎ Ménggǔ hé Nepal dōu bāokuò zài nèi, yěyǒu zhēngyì. Kàn lái hái yào jìnyībù yánjiū. Dàn dog de xúnyǎng zài 10,000 nián qián shènzhì gèng zǎo, yīnggāi shì kěnéng de.

    拜读关于dog的大文,觉得很好,特此致谢。几年前,中国和瑞典的一些学者发表论文,通过DNA分析,认为狗的驯养开始在中国东南一带,但是这个说法与考古发掘的证据不合,已经被否定。后来又有人说在Central Asia,但他们把蒙古和Nepal都包括在内,也有争议。看来还要进一步研究。但dog的驯养在10000年前甚至更早,应该是可能的。

    I have read your long post on "dog" and think that it is quite good, so I especially want to thank you for it. A few years ago, scholars in China and Sweden published a paper in which, relying on DNA analysis, they believed that the domestication of dogs began in the southeastern part of China. However, this argument is not compatible with the evidence of archaeological excavations and has been denied. Later, other investigators claimed that [the domestication of the dog began] in Central Asia, but since they included both Mongolia and Nepal, [their findings were] also controversial. It seems that further research is required. But it seems likely that domestication of the dog began 10,000 years ago or even earlier.

    =====

  51. Chris Button said,

    March 9, 2018 @ 9:30 pm

    @ Olivier van Renswoude

    Yes, Elisabeth Rieken and others have argued that it's a root noun rather than an n-stem. And so I've argued it's derived from *ḱuen-, which Pokorny interpreted as 'feieren, heiligen', but which might as well have been 'to devote, dedicate' vel sim. Hence hound/*ḱuōn means 'devoted, dedicated (one)'.

    That is an ingenious etymology! A root noun with with original "n" is indirectly supported by 犬 *kʰʷə́ɲʔ (given the restriction against *kʰjwə́nʔ) as the much older word than 狗 *káwʔ

    @ David Marjanović

    Before there were tones, both 狗 and 犬 must have ended in a glottal stop. (Perhaps they still do in Southern Wú or something, I have no idea.) Where did that come from? It's not there in PIE or Tocharian or Iranian or anywhere else in IE, and it's not a known OC suffix either, is it?

    If there was an *s on the end of the PIE form then we're home and dry (see my post above regarding the shift of *s > h > ʔ versus *s > *h > falling tone)

  52. Chris Button said,

    March 9, 2018 @ 9:50 pm

    Going by this entry regarding "Szemerényi's Law" on wikipedia…

    PIE */ḱwóns/ "dog" > */ḱwṓn/ > *ḱwṓ

    … I'd associate 犬 with *ḱwóns (OC /ʔ/ from PIE /s/) and 狗 with *ḱwṓ (perhaps the OC /ʔ/ here results from analogy)

    I'm getting more than a little out of my depth here in terms of PIE…

  53. Andy said,

    March 10, 2018 @ 12:30 am

    @David Marjanović: You're surely right about *ḱwṓ +ka > ***svaka, but I was rather imagining (I realize I might have actually said so!) a formation from the zero grade *sun + ka (actually it's properly -ъka). (And, in that vein, it's not stretching the imagination to posit a remodelled nominative for the base word; cf. suns and sunnis in Latvian and Old Prussian respectively.) Then *sunъka > OCS *sǫka (Wiktionary suggests that the OCS form is attested, but I can't find any evidence of that), of which suka is the expected outcome in the languages it is found -except Polish! A borrowing? Also, I might be wrong, but I get the impression that suka is not really a 'native' South Slavic word, but is borrowed from the others.

    Having said all that, I could on reflection almost be content to write off suka and pes as examples of animal names which have an onomatopoeic origin (from the sound used to attract their attention). Such etymologies often feel unsatisfactory though, for some reason, despite the fact that there clearly is an awful lot of sound symbolism in animal names universally.

  54. Andy said,

    March 10, 2018 @ 1:45 am

    @David Marjanović: And regarding **pḱwón-, I'm glad to have helped :) But -and I'm really not that well up on Balto-Slavic -I'm only aware of epenthesis as something which occurs early on before syllabic liquids and nasals, not as a process that legitimises inherited consonant clusters. I found a couple of words that initially seemed to have epenthesis, but the vowel turned out each time to be the zero grade of a root in a diphthong. So I guess that either you already have epenthesis by the time there would otherwise be an illicit cluster -in which case there's no reason to delete the p -or you have an illegal cluster, in which case the p has to go.

    I remember reading in Lunt's OCS grammar that in PBS the maximum word-initial consonant cluster was s+stop+sonorant. I assume that reflects general opinion?

  55. Chris Button said,

    March 10, 2018 @ 6:55 am

    @ Tsu-Lin Mei

    Further to my post above, I forgot to mention that it is this same phenomenon that lies behind Weidert's (1987 – TB Tonology) & Matisoff's (1982 – LTBA 6.2) observations regarding an apparent flip-flop of Lolo-Burmese TC-II and Old Chinese TC-III on occasion. This leads Weidert to set up a fourth tone category (TC-IV) to account for cases of such a merger (which is of course unnecessary). I discuss it on pages 47, 62, 66-67 in my STEDT monograph. Note also how Burmese has creaky tone for TC-III which in Inscriptional Burmese is noted with a glottal (simplified to a dot in Written Burmese)

  56. David Marjanović said,

    March 10, 2018 @ 8:37 am

    Just so much for now: I bet Latin cānus (*-adjective) is a root cognate of English hare (*n-stem noun "the gray one"). In other words, I've confused hares with dogs! :-D

    I remember reading in Lunt's OCS grammar that in PBS the maximum word-initial consonant cluster was s+stop+sonorant. I assume that reflects general opinion?

    I'm sure it is.

  57. Andy said,

    March 10, 2018 @ 4:04 pm

    David Marjanović: Lol! It occurred to me to look up any descendants of cānus in Romance, and it has in fact merged with canis in Portuguese cão (though the first sense is apparently quite old-fashioned)!

  58. David Marjanović said,

    March 11, 2018 @ 5:57 am

    … I'd associate 犬 with *ḱwóns (OC /ʔ/ from PIE /s/) and 狗 with *ḱwṓ (perhaps the OC /ʔ/ here results from analogy)

    But… **ḱwóns is an internally reconstructed pre-PIE form, and we just learned that 犬 apparently has a cognate, complete with -ʔ, in Nepal. We'd have to postulate a pre-PIE loan in Proto-Sino-Tibetan if the two forms are to be related. That means we'd have to assume Neolithic contacts, quite possibly predating the invention of the wheel, spanning the whole length of the steppe in the absence of genetic or archeological evidence.

    狗, on the other hand, sure does resemble the Tocharian ku mentioned in the OP.

  59. Olivier van Renswoude said,

    March 11, 2018 @ 7:39 am

    Since *peḱu- 'cattle' derives from *peḱ- 'to pluck wool', we can assume the Proto-Indo-Europeans or their immediate ancestors were originally predominantly shepherds. Possibly, they had been exporting their breed of herding dogs, along with the word *ḱuons (> *ḱuōn > *ḱuō), even before they started using horses and wagons.

  60. Chris Button said,

    March 11, 2018 @ 8:11 am

    It's also worth remembering that OC had a final *s so logically it would have retained the one in *ḱuons had it reached OC with it completely intact. However, the instability of word final *s (compare modern Spanish s~h~∅) suggests that whatever the linking form/language between PIE and OC was, it may well not have arrived as a pure sibilant and hence the OC glottalic interpretation/reflex. Since the OC word 犬 *kʰʷə́ɲʔ is internally suggestive of an external source (albeit not absolutely definitively so due to the labialising effects of uvular initials which I wouldn't reconstruct here anyway) via a phonotactically constrained form like *kʰjwə́nʔ or *kʰwjə́nʔ (the palatal nasal originally has to come from j plus a coronal or velar nasal) it seems highly unlikely that the very close phonological correspondence with PIE is by chance (even more so than in the case of 狗).

  61. martin schwartz said,

    March 12, 2018 @ 2:43 am

    My remarks are addressed to Victor Mair's original posting,
    and to various comments of Andy, David Marjanovic',
    and Olivier van Renswoude. Forgive me if I just write a single
    response, without sorting out the individual comments.
    Nostratic: Apart from Indo-European, our 'dog' in attested , as something like kun, only in one group of Hamitic; there seem to be at least 2 other basic phonic shapres.for 'dog' in Hamitic. one which looks cognate to Proto-Semitic */kalb-/.If indeed the kun words are independant from the latter (I have in mind the fact that l~n within Egyptian, and there seems to be some evidence for */kulb-/ with labial(ized?) vowel in Hamitic, but I'm not an expert there), then we have only one subgroup in Hamitic, which hardly a loaword from Indo-European, woukld be hung on as Nostratic comparand with the latter, whose ablaut/apophony argues for it having arisen within PIE.
    Because of some problems I feel about the Nostratic(s0 enterprise
    a priori, and how some of that scholarship in conducted without
    precise inner-protolanguage phonic and semantic reconstruction,
    I don't think one can use Nostratic arguments significantly, tho there may be something valid to the Nostratic concept(s).
    Since the OSlav. jer of pVs- cannot go back to *e and points to *i,
    maybe (I'm not the first to note this, but I can't cite the lit.) this dog-word may be *'spotted' from PIE root *p(e)ik, which give words for 'spotted' Indo-Iranian (OIr. */paisa-/ 'leprotic', Parth. *pêsak > Arm pisak spotted, mottled', and Skt. animal names in pis'-), and 'variegated in Greek and Slavic. Anyway, nothing to do with *pek'u-.
    Against the latter being related to PIE root *pek' 'to shear/card wool, comb', google Benveniste *peku- *pokos; the google-page should have the gist of Benv.'s argument, which I accept. If we define *pek'u-
    as 'smallish herded animals representing (mobile) wealth', then indeed we can reconstruct ablauting PIE **pk'w-on-:/**pk'un- 'that which (m.)
    attends flocks', i.e. 'dog'. The early loss of p in the initial CCC cluster is no problem; cf. Gr. kteís, kten- ~ Lat. pecten 'comb' (for the record, for all it's worth, I recall Thieme's sugg. that Gr. kuklo:ps is < *pk'u-klop-
    'catle thief' rathee than the patent 'wheel -eye' (but there are even Mesopotamian etymologies as well; I have no opinion here).
    Contra 'dog' < root *k'wen- attested in Iranian and Balto-Slavic words for the numinous, sacred, holy' (verb only in Lat. svinêt; I can't judge if that's not denom.) [and more ambiguously as to protoform OEng. hûsl 'sacrament'], there is no warrant for semantic manipulation "to be devoted to (doggishly').
    As for the dog in very late Avestan Zoroastrianism as connected to rituals of death and warding off evil, we are dealing
    with a ritual involving the glance of a dog (MPers./Pers. sagdId), optimally a 'four-eyed dog' (i.e dog with eye-spots), which repels the corpse-demon. The multiple-eye dog gets us to Cerberus et al,
    guarding the gates of the underworld, a PIE theme.
    Speaking oif Iranian, where I said *kuta- 'dog' is variously attested, noteHungarian kutya 'dog' , cf. Ossetic kudz/kuj, like Hung. híd
    'bridge', Oss. xed/xid (*haitu-) shd. be added to Enc. Iranica Ossetic Ii,
    on Hung. lw.s. from an antecedent of Ossetic.
    Finally, in connection with J. Schindler on PIE *k'won- 'dog'
    and *dhwer- 'door' as representing ablauting nouns for which a verbal
    etymon is expected but not known, my article on the origin of the latter word is "Relative chronology in and across semantic hierarchies: The history of *dhwer(E) in Indo-European", in Rekonstruktion und relative Chronologie, Akten der VIIi. Fachtagung der indogermanischen Gesellschaft … eds. R. Beekes, A. Lubotsky, and Jos Weitenberg,
    Innsbruck 1992, pp. 391-410. Much more than doors here!
    I'll ask a colleague if he has a PDF of it; Pls. tell me if you're interested.
    martin schwarTz (with apologies to D. Marjanovic' for mistyping his name).

  62. martin schwartz said,

    March 12, 2018 @ 3:51 am

    (oo)ps
    That Mesopotamian etymology is for Polyphemos,
    < Bilgamesh = Gilgamesh.
    I'm just reporting, not advocating.
    Martin Schwartz

  63. Olivier van Renswoude said,

    March 12, 2018 @ 6:29 am

    @Chris Button:

    it seems highly unlikely that the very close phonological correspondence with PIE is by chance (even more so than in the case of 狗).

    I know nothing of Sinitic etymology, but the likeness is striking indeed, especially in light of its a priori potential for being a Kulturwort. Cf. German Pferd 'horse' < Old High German pherit, pherfrit, parafred << Late Latin paraverēdus 'extra post-horse'.

    @Martin Schwartz:

    Contra 'dog' < root *k'wen- attested in Iranian and Balto-Slavic words for the numinous, sacred, holy' (verb only in Lat. svinêt; I can't judge if that's not denom.) [and more ambiguously as to protoform OEng. hûsl 'sacrament'], there is no warrant for semantic manipulation "to be devoted to (doggishly').

    From my Dutch perspective the semantic distance is smaller, given the parallel of the common verb wijden, which means both 'to consecrate, make holy' and 'to devote' (e.g. hij wijdt zich aan zijn werk 'he devotes himself to his work'). Similarly, the past participle turned adjective gewijd means 'consecrated, sacred, holy', whilst toegewijd has the worldly sense of 'devoted, dedicated'.

    That said, I do see the need for (other) non-religious derivations from *ḱuen-.

    I'll ask a colleague if he has a PDF of it; Pls. tell me if you're interested.

    I'll swing by the library; I just checked to confirm they have a copy. Thank you!

  64. Chris Button said,

    March 12, 2018 @ 12:43 pm

    @ Olivier van Renswoude

    … the likeness is striking indeed, especially in light of its a priori potential for being a Kulturwort.

    The distinctive internal phonology of 犬 *kʰʷə́ɲʔ also probably accounts for the lack of an associated word family in Chinese. This may be contrasted with 車 *kɬàɣ "chariot" (mentioned above as also having a likely IE source) which seems to have been appropriated by Chinese as a native word and allowed to develop a word-family of its own.

    @ Martin Schwartz & Victor Mair

    There's an article by Ilya Gershevitch (in Iran and Islam, in memory of the late Vladimir Minorsky, ed. C. E. Bosworth, Edinburgh 1971, 267-291; reprinted in Ilya Gershevitch, Philologia Iranica, ed. NSW, Wiesbaden 1985, 237-261) which contains an interesting idea about Latin canis…

    "Relative chronology in and across semantic hierarchies: The history of *dhwer(E) in Indo-European"…

    PDFs would be great – thanks! (dot between my first and last names at hotmail dot com)

  65. Chris Button said,

    March 12, 2018 @ 12:44 pm

    @ Olivier van Renswoude

    … the likeness is striking indeed, especially in light of its a priori potential for being a Kulturwort.

    The distinctive internal phonology of 犬 *kʰʷə́ɲʔ also probably accounts for the lack of an associated word family in Chinese. This may be contrasted with 車 *kɬàɣ "chariot" (mentioned above as also having a likely IE source) which seems to have been appropriated by Chinese as a native word and allowed to develop a word-family of its own.

    @ Martin Schwartz & Victor Mair

    There's an article by Ilya Gershevitch (in Iran and Islam, in memory of the late Vladimir Minorsky, ed. C. E. Bosworth, Edinburgh 1971, 267-291; reprinted in Ilya Gershevitch, Philologia Iranica, ed. NSW, Wiesbaden 1985, 237-261) which contains an interesting idea about Latin canis…

    "Relative chronology in and across semantic hierarchies: The history of *dhwer(E) in Indo-European"…

    PDFs would be great – thanks (dot between my first and last names at hotmail dot com)

  66. Chris Button said,

    March 12, 2018 @ 12:46 pm

    @ Olivier van Renswoude

    … the likeness is striking indeed, especially in light of its a priori potential for being a Kulturwort.

    The distinctive internal phonology of 犬 *kʰʷə́ɲʔ also probably accounts for the lack of an associated word-family in Chinese. This may be contrasted with 車 *kɬàɣ "chariot" (mentioned above as also having a likely IE source) which seems to have been appropriated by Chinese as a native word and allowed to develop a word-family of its own.

  67. David Marjanović said,

    March 13, 2018 @ 7:58 am

    the Proto-Indo-Europeans or their immediate ancestors were originally predominantly shepherds. Possibly, they had been exporting their breed of herding dogs

    Good point.

    the palatal nasal originally has to come from j plus a coronal or velar nasal

    Yes, but why do you reconstruct a palatal nasal? No other OC reconstruction has one in its inventory at all, IIRC.

  68. Andy said,

    March 13, 2018 @ 9:23 am

    @martin schwartz: Thank you for the suggestion of pьsъ deriving from *peik; that's quite convincing, especially given Sanskrit piśá. And for now I'm throwing my lot in with Benveniste in regarding *peku and *pokos as unrelated.

    I'm still not buying **pk'w-on-/**pk'un-. The longer initial clusters in PIE tend to involve resonants or laryngeals, and a sequence of two, let alone three, voiceless obstruents is pretty rare (I'm obviously disregarding thorn clusters). Zero-grade sequences containing stop+stop+resonant (TTR) should develop an epenthetic vowel (schwa secundum) between the first two stops (TeTR) -/i/ in Greek, /a/ in Latin (I've not found much information for other languages). So, in Greek, for example, on the analogy of (zero grade) *kʷₑ.twr̥ > πίσυρες (Epic/Aeolic), one might expect **πικύων from **pk'w-on-. Epenthesis rather than deletion seems to be the way that difficult word-initial onsets are dealt with in PIE itself.

    I was very happy to be reminded of Κύκλωψ <*pk'u-klop-. It's clearly an exceptional development in Greek, but I would still be tempted to go for it because of the corroborating evidence in Indo-Iranian, e.g. Avestan fšūmant 'having cattle'. I don't suppose there's any similar evidence from Indo-Iranian of that initial cluster for our dog word? If not, then the lack of any actual evidence combined with the unusual phonotactics (for PIE) make it quite hard to see the appeal of **pk'w-on-/**pk'un-, apart from the plausible semantics.

    κτείς < *pḱtén-s itself is a very untypical case, and far from straightforward; it can't be used to justify ***pk'w-on-. Various features in the different languages clearly postdate PIE; the main problem to be explained is the extraneous t in κτείς, πεκτέω, pecten and pecto, given that the root is *peḱ (cf. Greek πόκος; also Indo-Iranian, e.g. Pashto žmanj, Persian šāna <*pk-en-). I found one very good explanation in Michiel de Vaan's Etymological Dictionary of Latin and the Other Italic Languages, where he takes up a suggestion by Pinault that there was a reduplicated verb from *pe-pk'- which dissimilated to *petk'- (and thence *pek't, by metathesis). In any case, the loss of the putative initial p in κτείς must have been fairly late. (Old Armenian asr shows the expected result, no clusters here.)

    Many apologies for the long post, but I found all this very interesting :)

  69. Andy said,

    March 13, 2018 @ 9:32 am

    If anyone is interested in schwa secundum and sequences of word-initial obstruents in PIE, I had these papers close to hand while writing my comment above:

    'Schwa secundum' in Greek -Anthony Jakob

    These latter two by A. A. Trofimov (in Russian):
    Initial clusters consisting of two voiceless stops: PIE heritage or dialectal innovation?
    On the etymology of PIE *ku̯sep- 'night'

  70. Olivier van Renswoude said,

    March 13, 2018 @ 11:18 am

    @Andy:

    Yes, If we can go by Pinault's suggestion, we might as well consider κτείς < *kptén- < *ptḱ-én-. That would be a much more natural way for the p to be lost than it being dropped from the anlaut just like that.

  71. Chris Button said,

    March 13, 2018 @ 12:02 pm

    @ David Marjanović

    If you line up Pulleyblank's (1977-8) rhymes against something like Baxter & Sagart's, you will see that rhymes I to IV have a proliferation of rhymes (i.e where the front and rounded vowel hypotheses really disrupt the traditional rhyme categories). However, the rhymes V to XI are much less disrupted. For example, in the "yang" (nasal) set, B&S have split the single traditional category IV into *-an, *-on, *-en but then place *-aŋ, *-oŋ, *-eŋ in the three traditional categories of VIII, X and VI respectively (i.e. without any disruption). The reason for this is something which to my knowledge has not been discussed in the literature.

    Without going into too much detail here, this is because rhymes V to XI do not have bilabial or coronal articulations that lack the ability of velar codas to maintain palatal and velar articulations (just look at examples from languages all over the world as evidence for the ability of velars in this regard). I'll leave the question of uvulars aside for the time being which confused things a little.

    So at least where the source was in PST *-jək or *-jəŋ, as an alternative to OC palatal *-əc and *-əɲ you could reconstruct *-əkʲ and *-əŋʲ which would have been the intermediary stage before palatalisation anyway. These are the palatal counterparts to Li F-K's *-əkw and *-əŋw in which the labial component could equally well be written as superscript (by the way, B&S do admit a *-kʷ in their system where Pulleyblank has *-q but then cannot account for the lack of the nasal counterpart *-ŋʷ whereas the lack of a uvular nasal makes perfect sense from an articulatory perspective). The reason for *-əc and *-əɲ rather than *-əkʲ and *-əŋʲ is because the reflexes of PST *-jət and *-jən (and significantly also *-jəm in certain environments) merged with the reflexes of *-jək and *-jəŋ in Old Chinese. This development of palatal codas neatly parallels the same development in Old Burmese via the effect of *-j- on *-t/k and *-n/ŋ where the palatals *-c and *-ɲ are unequivocally attested in the inscriptions.

  72. martin schwartz said,

    March 13, 2018 @ 7:03 pm

    @Andy, Victor, et al.
    I don't insist on *pk'won-, but maybe the *- loss was
    nudged along by a dissimilatory effect via the following /w/.
    Re 'comb' in Iranian, Yaghnobi has n(i)po$- ($ = s hacek) 'to comb'
    < Proto-Iranian *ni-pâsaya- from PIE root pek', Yagn. noun
    nip$a 'a comb', for which I found a likely Sogdian equiv.
    in the word for 'honeycomb' (in Psalm 19) , but only -$ remains,
    the rest of the word having fallen into the lacuna. I've published
    about in in the ed. of the Sogd. Psalter in N. Sims-Williams, Christian Sogdian texts mainly Biblical…. Btw, Victor, there I have a very long discussion of Sogd. ptsâ∂- 'shield' etc. which I should have mentioned
    when we last discusssed the word; all details on request.
    Meanwhile, thanks to the great kindness of my dear colleague, the marvelous Indo-Europeanist MIchael Weiss, I can beginning tonight to send .pdf-s of my *dhwer(E)- article to all who e-mail me ;
    Chris Button is already on the list.
    Martin Schwartz

  73. martin schwartz said,

    March 14, 2018 @ 2:14 am

    @Andy, David Marjanovic', Olivier van Renswoude, and all:
    More for a possible **pk'- for the PIE 'dog' word: In addition to
    labial dissimilation of **p- by *-w- as a factor, we should consider the basic fact (cf. Pokorny on the orig. nominative) that our word is NOT
    really a root-stem *k'won- (: kun-) if we have an underlying disyylabic
    base *k'uwon- reflected by the Greek nominative (with -ú-), I think the Lith. nominative,the Vedic scansion s'u(v)an- (s' = palatal s) alongside
    s'van, and the Khotanese Saka forms, which seem to show a contamination of *s's'an- (regularly CCuwV-, MAYBE we would thereby get a partial block on **pk'uwon- > *pk'won-
    in the nom.
    As I said, I don't insist on an etym. of 'dog' as having first part 'cattle,
    livestock'– but it would be nice.
    Martin Schwartz
    block on *pK'uwon- > *pk"won-
    retention of the -u-

  74. martin schwartz said,

    March 14, 2018 @ 4:10 am

    Lucubrations: Maybe someone will counter my very putative
    "Sievers" with an "Edgerton", whose status these days I don't know.
    BUT I'm now thinking about the second part of an etymon **pk"u-won-.
    Could this be, instead of an associative suffix, a root-stem?
    I'm thinking of Avestan vãØßa- (v, nasalized a, theta, beta, a-) 'a herd, flock' < PIndo-Ir. *wan-twa-. This should etymologize as 'that which is to be VAN-d, from root van, PIE wen.
    E. Mucciarelli, Changes in the Semantics…,
    PhD dissert. PDF, takes the Vedic root, and its PIE predecessor,
    as 'to appropriate', for which she has a meticulous analysis of the Vedic etc. passages. But MAYBE the root has a more basic mg., 'to gather in',
    which wd work well for Lat. venari 'to hunt' (Meillet and Saul Migron (for Vedic) ,taking different semantic routes, thought that hunting could be ultimately involved with the root). It could also acccount for Irish
    fine 'Verwandtschaft, Stamm, Familie', rather than a word for love and affection from desiring, (wishing to be) acquiring', etc. So, dog as herder, that which *gathers in'??
    Martin Schwartz
    with

  75. Olivier van Renswoude said,

    March 14, 2018 @ 5:53 am

    @Martin Schwartz:

    Not Edgerton, but Lindeman at any rate. Parallel cases such as Vedic diyáuḥ 'sky' beside dyáuḥ would rather suggest a secondary albeit old disyllabicity, or at least the possibility thereof. So I do not see enough reason to reject the notion of *ḱuōn (< *ḱuons) being a root noun originally. As for the labial dissimilation that would have to be assumed, do you have other examples of this for initial clusters? Would the u not be dropped instead?

    Avestan vąθβā- 'herd, flock' looks very interesting, but for other reasons. In fact, formally it looks like an exact cognate of PGmc *windwō 'winnowing' (> OHG winta 'winnowing shovel', OE windwian > MoE to winnow). Kümmel recently added the latter to the LIV under the root *h₁uen- 'hinschütten, ausstreuen' (previously *uenh₁-), which seems best attested in Indo-Iranian. The Avestan noun could belong there under the assumption of an older meaning 'scattering (of livestock)' or something.

    Incidentally, I believe the same root underlies a Germanic word that has long been an etymological nuisance: winter. That is to say PGmc *wintruz < Pre-Gmc *uendrus by epenthesis < quasi-PIE *uenrus, as a word originally referring to snowfall. Cf. the semantics of e.g. Serbo-Croatian vȉjati 'to winnow; to snow heavily' and the phonological development of PGmc *ampraz 'bitter' < Pre-Gmc *ombros by epenthesis < *Homros (Kroonen, 2013).

  76. Olivier van Renswoude said,

    March 14, 2018 @ 6:36 am

    Just came around to reading Benveniste's argument against the derivation of *peḱu- from *peḱ- 'to pluck wool'. It's compelling. He ends with:

    The connection must be abandoned, and *peku-, a vestige of the most ancient Indo-European vocabulary, seems irreducible to any known root.

    I daresay I found a good candidate already: OE ge-féon, ge-feohan 'to rejoice, be glad', OHG gi-fehan 'id.' < PGmc *fehan- < *peḱ-e- (Kroonen, 2013), which would otherwise be isolated, but for which we may posit a more original meaning like 'to enjoy, use', so that it can be tied to *peḱu- 'property, wealth'.

    A good semantic parallel is available in the form of MoDu genieten 'to enjoy, have fun, have the advantage of', ON njóta 'to enjoy, use' etc. < PGmc *neutan-, from which was derived ON naut 'cattle', MDu noot 'cattle, cow' etc. < PGmc *nauta-.

    Moreover, *peḱ- 'to pluck wool' could still be the same root, via the specific, practical sense of exploiting sheep. In other words, Germanic preserved the clue to the primary notion of the root.

  77. Olivier van Renswoude said,

    March 14, 2018 @ 7:53 am

    Also, a more primary meaning for PGmc *fehan- than 'to rejoice' is suggested by its immediate cognates: Go. ga-fehaba 'properly' < PGmc *fēhi- and Go. fulla-fahjan 'to satisfy' < PGmc *fahjan- or rather *fagjan- with undoing of Verner's Law (Kroonen, 2013).

  78. Chris Button said,

    March 14, 2018 @ 10:33 am

    @ David Marjanović

    Just to correct a small typo in my second paragraph above, I meant velars can maintain palatal and labial co-articulations.

    So B&S's *-an, *-on, *-en can overlap in rhyming since the latter two are really just surface assimilated forms of *-wan and *-jan), while their *-aŋ, *-oŋ, *-eŋ cannot since the codas were different as *aŋ, *aŋʷ and *aŋʲ (or *aɲ) respectively.

    @ PIE specialists

    Is there any chance the final stop in "hound" could come from the fortition of an earlier *s ? Or would that be barking up the wrong tree :)

  79. martin schwartz said,

    March 14, 2018 @ 3:41 pm

    @Olivier van Renswoude: OK, Lindeman, but have a look at the
    Wikipedia article on Sievers' law, under new(er)/recent developments,
    where problems with L's law are noted, esp. from Sihler's criticisms,
    which end with the problematic nature of the disyllabic scansions of the Vedic 'sky' word you cite.
    I really can't follow you on a connection of the Av. 'herd' word
    with 'winnowing' via a putative "older meaning 'scattering (of livestock)' or something".
    Martin Schwartz

  80. martin schwartz said,

    March 14, 2018 @ 4:12 pm

    @Oliver van Renswoude (and David Marjanovic' and Andy and
    Indo-European fans (winnowing or not):
    There should be a long apud by me on the PIE word for 'winter' and other words for 'winter' and heavy snowfall' at the end(? non vidi) of Thomas Steer's article on PIE 'earth' and 'winter' in Indogermanische Forschungen, nr. 118, 55-92.
    Unfortunately he omitted my treatment of Russ. metelica 'snowstorm'
    and it cognates (don't google the Russian word unless you like the rock group Metallica). I have a PDF which Steer sent my of my apud, for those who wish .
    I expect I'll soon be ending these kibbles and quibbles and stop running with the dogs and em-bark on other things
    Martin Schwartz

  81. Olivier van Renswoude said,

    March 14, 2018 @ 5:51 pm

    @Chris Button:

    Is there any chance the final stop in "hound" could come from the fortition of an earlier *s ? Or would that be barking up the wrong tree :)

    I'm afraid it would. The accepted reconstruction is PGmc *hundaz, Pre-Gmc *ḱuntos. This is generally understood to be an extention of the original PIE form by the basic and very common suffix *-to- (plus the nom. sg. ending *-s).

    @Martin Schwartz:

    Yes, the mechanism is still not understood. But we seem to be dealing with some kind of disyllabification all the same.

    As for my proposed connexion, it's tricky. The base meaning of the root *h₁uen- as given by LIV is tentative. If the Avestan noun is to fit in, and I'm not adamant that it does, we should consider 'to strew, dispel, drive (away)'. Hence we find a close enough parallel in the derivations from PGmc *drīban- 'to drive', notably ON dreif 'scattering', OE dráf 'driving; flock' < PGmc *draibō as well as ON drift 'snowdrift', ME drift 'driving of cattle to pasture, snowdrift etc.', MHG trift 'herd, pasture' < PGmc *drifti- (Kroonen, 2013).

    And so, my second take: Av. vąθβā- 'herd, flock', PGmc *windwō 'winnowing' < PIE *h₁uentueh₂ 'a driving (away)'.

    I shall look for your apud as soon as pawssible. And can you put me on the list for a PDF of your article on *dʰuer- after all? You can send it to: vanrenswoude@hotmail.com. Thanks!

  82. Chris Button said,

    March 14, 2018 @ 8:52 pm

    @ Olivier van Renswoude

    The accepted reconstruction is PGmc *hundaz, Pre-Gmc *ḱuntos. This is generally understood to be an extention of the original PIE form by the basic and very common suffix *-to- (plus the nom. sg. ending *-s).

    If the suffix was *-to- then what would have been its function here?

  83. martin schwartz said,

    March 15, 2018 @ 12:06 am

    @Olivier van Renswoude:
    I think your reconstruction for the 'winnow' etc. word
    as h1en are a typo for h2en. I haven't gotten around to looking at LIV,
    but I wonder if we don't have an original nasal infix stem
    h2-en-w(h1)- to the root for 'blow wind, air', , h2ew(h1), whence
    Germ. wehen and its Slavic cognates, */h2weh1nto-/ 'wind', etc.
    Or else a Benvenistian élargissement *h2w-en- ??
    The historical semantics of Germ. treiben, Eng. drive, etc. are
    somewhat surprising,, and indeed closely related to meteorological
    phenomena; see apudding to the steer meal, which I'll send you as per your request, as also my *dhwer(E) thing.
    Martin Schwartz

  84. Olivier van Renswoude said,

    March 15, 2018 @ 5:35 am

    @Chris Button:

    When affixed directly to the root, PIE *-to- was used to form primary verbal adjectives going on past participles, which could hence be substantivized, e.g. OHG kind 'child' < PGmc *kinþa- < *ǵénh₁-to- 'begotten (one)', from *ǵenh₁- 'to beget'. It would also suggest PGmc *hunda- 'dog' < Pre-PGmc *ḱun-to- was strictly speaking a different formation from the same verbal root *ḱuen- rather than an extension of the main PIE *ḱuōn.

    However, if the main form was actually an n-stem as has often been argued, i.e. PIE *ḱu-ōn (< *ḱu-ón-s), then the Germanic form would likely continue a present participle of the same verbal root, i.e. PIE *ḱu-nt-, in which case the connexion with *peḱu- 'property, cattle' cannot be maintained. But I would argue against such a connexion either way.

    @Martin Schwartz:

    The renewed entry to *h₁uen- can be read online in its entirety in Martin Kümmel's Addenda und Corrigenda zu LIV² (see Google, because I fear a link would land this message in the spam folder). The initial laryngeal was reconstructed on the basis of Greek εὐνή 'Lager, Bett', of which the earliest context indicates an animal's strewing of leaves and twigs, but I'm not sure whether it merits inclusion, nor the Albanian verb, for which a different etymology is given in Orel's Albanian Etymological Dictionary anyway.

    A nasalization or extension *h₂ueh₁- 'to blow'… Maybe, though I don't know if that final laryngeal would be formally reconcilable with all of the pertinent material. It would probably prevent the (early) epenthesis of the dental in winter that my etymology requires, but of course the appurtenance of that word is still up in the air, so to speak.

    Looking forward to your PDFs!

  85. Victor Mair said,

    March 15, 2018 @ 9:25 am

    Colleagues following this thread may be interested in the comment of Rhona Fenwick on "cow" over at the "Dung Times" post:

    @Ambarish Sridharanarayanan:

    Sanskrit 'go' seems to be cognate with the English 'cow'.

    Indeed. And even beyond the rich IE heritage (virtually every Indo-European branch preserves cognates of PIE *gʷṓws), this is one of those terms that also has similar forms in other unrelated Eurasian languages, perhaps indicating wide cultural borrowing (if the terms aren't simply imitative, which is of course a possibility): Tajik also has гов "cow", Ubykh gʷəmá "cow", Kabardian гуу "bull", Egyptian gw "bull", and Sumerian gu₄ ~ gud "bull", perhaps best known through the mythological Great Bull of Heaven, gu₄.gal.an.na.

    http://languagelog.ldc.upenn.edu/nll/?p=37221#comment-1548195

    http://languagelog.ldc.upenn.edu/nll/?p=37221

  86. Chris Button said,

    March 15, 2018 @ 11:34 am

    @ Olivier van Renswoude

    PIE *-to- was used to form primary verbal adjectives going on past participles, which could hence be substantivized, e.g. OHG kind 'child' < PGmc *kinþa- < *ǵénh₁-to- 'begotten (one)', from *ǵenh₁- 'to beget'. It would also suggest PGmc *hunda- 'dog' < Pre-PGmc *ḱun-to- was strictly speaking a different formation from the same verbal root *ḱuen- rather than an extension of the main PIE *ḱuōn.

    I was figuring something along those lines which was why I was wondering how a noun like "horse" could be suffixed with *-to-. However, if we're treating *ḱuen- as the verbal root (perhaps meaning "to devote/dedicate"?) then I suppose that would make sense. However, there is an interesting comment by Peter Maher over at Anatoly Liberman's blog (cited in the original post here) saying the following:

    I suggest that the D of Gothic hunds originated by anaptyxis of n-s > nds… Cf. the d of English sound, vs French son. Cf. AE Housman's rhyme of pounds and crowns, and the homonymy of dense & dents, tense & tents…

    A hardening of -s to -t is sometimes attested in Tibeto-Burman languages and also sporadically in Old Chinese where it is generally associated with the numerals (e.g. 四 "4" has Middle Chinese readings of both siʰ and sit ) and occasional tangentially related extensions thereof (e.g. 鼻 "nose" as Middle Chinese bjiʰ and bjit ).

    The reason I'm asking all this is that in addition to David Marjanović's comment regarding the source of the OC final glottal, an OC form like *kʰʷə́ɲʔ has ə rather than a as its nucleus which, if we follow the logic that PIE e/o ultimately went back to ə/a, would be expected to have a as its nucleus instead if the source was *ḱuon- (i.e. *kʲwan) rather than *ḱuen- (i.e. *kʲwən)

    It's not a huge issue since Old Chinese ə commonly merged with Tibeto-Burman a (which would have been centralised and often closer to ɐ as the comparison of OC *-əɲ with Old Burmese *-ɐɲ shows in an earlier posting), but OC did nonetheless distinguish ə/a as is clearly attested by its alternation across the whole OC lexicon in a similar way to how PIE distinguished e/o.

    Having said all that, B&S's reconstruction of 犬 as *[k]ʷʰˤ[e][n]ʔ would correspond with *kʰʷjánʔ (based on my comments above regarding their -e- as an assimilated form of -ja-), but this clearly violates the phonotactic constraint against -jw- or -wj- which is why I favor *kʰʷə́ɲʔ as a phonotactically viable version of inviable **kʰʷjə́nʔ that shows the shift of **-jə́n > *-ə́ɲ (the Middle Chinese reflex would have been the same with either ə or a as the nucleus). Then again, languages do of course permit loanwords that do not stray too far from such constraints which allows them to be clearly identified as non-native. In that case *kʰʷjánʔ would then perhaps be the better correlate with PIE *ḱwóns in spite of its externally influenced internal morphology.

  87. 번하드 said,

    March 15, 2018 @ 6:10 pm

    @Robert Ramsey:

    Oh, thank you! I'm learning contemporary Korean only, but you made me look up
    this ancient form at http://stdweb2.korean.go.kr/main.jsp and lo and behold:
    apart from ancient 가히 having had some word stem change to 갛 depending on suffix,
    가히 seems to be still alive as a word for dog in Gyeonggi-do dialect.

  88. Eidolon said,

    March 15, 2018 @ 7:47 pm

    "We'd have to postulate a pre-PIE loan in Proto-Sino-Tibetan if the two forms are to be related. That means we'd have to assume Neolithic contacts, quite possibly predating the invention of the wheel, spanning the whole length of the steppe in the absence of genetic or archeological evidence."

    That's actually not as far stretched as you think. There is archaeological evidence of early Neolithic contacts between different regions of the world, and of course hunter-gatherers could move across vast distances, or else humans would've never made it out of Africa. Many wanderworts may ultimately have originated from the Neolithic period. In such cases, however, the times scale would be so deep that etymological relations would exceed modern criteria for detectable sound change. For that matter, it would exceed the estimated ages of all language families but the most controversial, all-encompassing super families.

  89. Andy said,

    March 16, 2018 @ 2:44 am

    Only just got round to catching up on the developments here…

    @martin schwartz: Could I also please sign up for a copy of the articles on The Doors and Metallica, when you have a moment? My email is hendrix at pidgeme dot com.

    @Chris Button, Olivier van Renswoude: Well, with regard to Peter Maher's suggestion, the d in hunds is general, not just Gothic, so the it would have arisen in Proto-Germanic *hundaz, if not earlier. But yes, I've always considered it epenthetic (cf. OE thunor > thunder); I could never understand what the *-to- suffix would be doing here. (A couple more similar examples of epenthesis in early Germanic would be nice, though.) Having said that, I notice there's an Old Armenian word, skund, for 'puppy, dog', supposedly from the same root, which also has the d (not to mention the surprising onset); potentially an interesting correspondence, but not knowing much about early Armenian, I'll keep myself on a short leash and not speculate.

  90. martin schwartz said,

    March 16, 2018 @ 3:04 am

    Oliver van Renswoude, Andy, David Marjanovic', and all.
    Only because i have deadlines from which I'm being distracted by thinking of alll your interesting posts, (pardon my parting gratuitous
    and corny [sometimes literally involving (British0 corn =Korn, Körner]) puns, which are not meant to mock anybody (except maybe myself).
    I'm dog-tired of this Dang(-uage) Dog of a Language Log, and this will be my (Proto-Iranian) *swan–song to it, and doggone if I respond further–maybe. Thanks, OvR, for pointing me to the LIV2 updates by
    the very admirable Kümmel. Lest we get car(ried)away too far, I don't think we should pile so much on that Greek 'bed'. Formally, we would have a quasi-neo-Benvenistian Theme I **h1ewn: Theme II **h1wen, iin this instance, I think, problematic even beyond Raimo Anttila's objections to going too far with such "paper" etymologies. But my objection really proceeds from semantic considerations: This Greek 'bed' would maybe reflect a Flintstonian concept of the Indo-Europeans heaping up castaway scattered grasses to make a, er. PasturepedicTM bed, as tho the IEs weren't great textile people and didn't heap up blankets. But even if the IEs made beds of grass (yes, I know about Vedic barhis- and other grassy cushions), we would have to posit
    **'scatter' > *strew' "heap up grass and leaves' 'make up a bed' _ 'a bed'.
    But there is what I hink is a simple solution to Gr. euné: 'bed'–
    Pokorny 346 gives s.v. 2eu- various words for getting dress, covering oneself with cloth(e)s, etc. Among these he gives Arm. oganim 'ziehe mir etwas an' < PIE *oú-mim and ot'-oc (recte ot'-oc', I think)
    "Bettdecke'. This would make a nice Greco-Armenian semantic
    pair with Gr. euné: 'bed', unless one wants to add Vedic duróNa-
    (retroflex n), which Pinault, I believe, wished to explain as *bad bed'.
    i.e the bunk on the periphery of the house, on which guests were lodged,
    As for the Iranian words for 'throw up, scatter, throw over' from
    (H)wan(H)-, Cheung Dict. Ir. root 205-206 has a good entry
    (but add the words for 'cover'), and suggests *h2w-. I used to think that
    Gr. aíno: "I winnow, reinige die Körner…aus der Spreu' (Pokorny 82)
    was decent evidence for h2wen- vel sim., but I have no time to
    follow Kümmel's apparent alternative explanation.
    All this gets us far from my suggestion that the Av. word for 'herd, flock'
    attests a root **to gather in', whence E. Mucciarelli's "to appropriate, put into the ritual space', Lat. venari 'to hunt', etc., but
    OvR prefers some kind of scattering of herds/flocks. My suggestion was in the service of a conceivable ablauting root-stem *(-)won- in **p"ku-won-. Speaking of bed, time to go now. I hope in next days
    more on the PIE vb root *k'wen '(?to regard as) holy, numinous',
    god:dog, and more justification of *pk'u-won-, and -t- in Latvian sùnt-ele 'big dog'
    vis-à-vis the dental of "hound'. Then I let the dogs out.
    And I havnt forgotten about PDFs for those who request(ed).
    Martin Schwartz

  91. Olivier van Renswoude said,

    March 16, 2018 @ 5:59 am

    @Chris Button, Andy:

    It seems rather unlikely to me that the dental of the Germanic form arose by epenthesis. As far as I can tell, there were only three places in the original paradigm where this could have occurred, i.e. where there was a cluster *ns, was nom. sg. *ḱuóns before the operation of Szemerényi's Law (> *ḱuōn (> *ḱuō)), acc. pl. *ḱuónns and loc. pl. *ḱunsú. A non-Szemerényified variant of the nom. sg. seems out of the question for lack of parallels, so the dental would presumably have spread to the rest of the paradigm from those two cases, plus the zero-grade would have to have been generalized. A further complication is that it would still be a root noun, whereas the Germanic material shows no sign of that.

    Also, and I should have mentioned this earlier, Germanic had a second form: OHG wint, OFri hwynd, MDu wint < PGmc *hwindaz 'greyhound' < quasi-PIE *ḱuentós. It's possible this was a secondary full-grade or a vṛddhi-formation, but it would be simpler to assume it was a variant from an originally ablauting paradigm.

    All things considered the Germanic material rather points to a different derivation altogether, albeit from the same verbal root. As stated, in my opinion the best candidate so far is *ḱuen- 'to devote, dedicate' vel sim. It would also mean Lithuanian šveñtas 'holy, sacred', Old Church Slavonic svętь 'id.', Avestan spənta- 'id.' < *ḱuentos 'devoted, dedicated' is actually the same word.

    @Martin Schwartz:

    I mentioned Greek εὐνή 'Lager, Bett' only to say I don't know whether I can follow Kümmel in his inclusion of it, so I don't have much disagreement with you there. But for the record: the word originally referred to the lairs of animals and soldiers, thus rudimentary accommodations in the wild as it were and not made beds at home. Also, the point of departure would be 'to strew' (as suggested by the Indo-Iranian verbs), not 'to scatter', so this etymology would really need one semantic step to work, namely 'a strewing (of leaves and twigs)' > 'a lair'.

    I did initially suggest the original meaning of Av. vąθβā- 'herd, flock' was 'a scattering', but changed this to 'a driving (away)' in my second take. With that alternative etymology on the table, the idea that your candidate root *uen- 'überwältigen, gewinnen' had a broader or earlier sense of 'to gather in', which we can then presume also occurs in our dog word, becomes less compelling.

    However, I shall certainly gnaw on this bone some more.

  92. Andy said,

    March 16, 2018 @ 6:50 am

    @Olivier van Renswoude: The connection of *peḱ- 'to enjoy, use' with 'livestock, wealth' is very compelling indeed. I too find it hard to see the connection between flocks and winnowing, though; 'scattering' a flock seems like rather careless shepherding, and all the words I could think of off the top of my head for 'flock' have at their core some notion of multitude/togetherness/unity etc.; there is Greek ἀγέλη, from ἄγω, but even then the idea is still just moving them from a to b rather than throwing them to the winds :)
    I looked at that addendum to the LIV; quite a contrast between the meanings! But the example from Chorasmian meaning 'to cover, shelter' might perhaps help relate vąθβā- to this root; a herd will be kept under the same roof, so we might have here a metonymy comparable to how a collection of racehorses is called a stable.
    I'd always thought that wind and winnow were related, so that was a big surprise -although it sounds like the connection still isn't completely out of the question. Again, the meaning found in the Chorasmian 'to cover' could relate to a good snowfall. I still don't see a likely connection between winnowing and snow though, and the Serbo-Croat word you give doesn't really support the semantics because, of course, the underlying meaning is 'to blow' (and thus 'to winnow', at least in Slavic); the 'blowing with snow' is just a local development (you can also say 'snijeg vije') -though now I think about it, you can say the same in Czech, and there's also a noun, vánice, meaning 'snowstorm' (from the same root), but even so…
    Is it beyond the pale to consider dividing *h₁u̯en- into two distinct roots? Or are there good reasons to keep it as one?

  93. Andy said,

    March 16, 2018 @ 7:34 am

    @Olivier van Renswoude: Hmm, the epenthesis idea might have gone to the dogs then. And that's quite the revelation about *hwindaz. Lots to think about…

  94. Olivier van Renswoude said,

    March 16, 2018 @ 8:28 am

    @Andy:

    You're right, the semantic range of the material under *h₁uen- is a bit much. As I said above, the Albanian verb for 'to put' is better explained through a different root. The remainder can be held together quite nicely though, under the meaning 'to strew, dispel, drive (away)'. Compare for instance how MoE to strew itself involves covering something in the process and how the precursor to MoE to drive has derivations pertaining to herds/droves and things more erratically adrift (in the air as well), like a drift of snow.

    And so please note that whereas I initially suggested 'a scattering' as the original meaning of Av. vąθβā- 'herd, flock', which is indeed not compelling in spite of the familiar sight (to me) of sheep spread out over a field, my second take was 'a driving (away)'. Further parallels can be found in e.g. the PIE root *seuh₁-/*sueh₁- 'to drive, keep moving', with loads of likely derivations in Germanic, such as ON sýsl 'business', OHG swein 'servant, herdsman' (via the stative) and ODu sunista 'herd of horses, cattle etc.' (via the old nasal present).

    I'm pretty confident PGmc *windwō, as attested by OHG winta and especially the denominal OE windwian 'to winnow' cannot be derived from PGmc *windaz 'wind', simply because as far as I can tell the suffix *-wō- was otherwise not used to derive nouns from nouns. Rather we're dealing with the suffix *-dwō- < *-tueh₂-, as found in e.g. PGmc *mēdwō 'meadow, a mowing' < PIE *h₂meh₁-tueh₂ from the root *h₂meh₁- 'to mow' (Kroonen, 2013).

    You make a good point about the Slavic example and the weakness of the connexion between winnowing and snow. I would say English to winnow is also used with regard to snow, but there too it might possibly be through a medial sense of 'to blow'. Either way, it doesn't hurt my etymology for winter, since it is enough that it's the season marked by the strewing and driving of snow across the land.

  95. martin schwartz said,

    March 18, 2018 @ 1:56 am

    @Olivier van Renswoude, Andy, David Marjanovic'…
    As to a root-stem *k'won- 'dog (as devoted/dedicated':
    OK, so you have within Gmc.(Dutch wijden, etc.) etc. a generalization from the cultic sense–AT THE END OF A PUTATIVE SEMANTIC DEVELOPMENT– to being dedicated, so you, in effect (pardon
    this, I can't resist) have a metathetic thematic of man to god and doto man, as it were. But our starting point should be the idea of the holy/sacred, which in langs. I know can be based on separateness,
    vital energy, haleness/wholeness, cleanness/purity (thus with different etyma Persian and Sogdian)and yes, sacrificial procedure, and
    elsewhere awe, the uncanny, etc. We have only reflexes attesting a trace of a verb root *k'wen 'to regard/treat as holy' only in Iranian, Balto-Slavic (or if you like Baltic and Slavic), and maybe Germanic.
    Dutch wijden etc. do not belong to a wide pattern of semantic change
    from 'treating as holy' to 'dedicate oneself'
    and from our evidence a root-stem *k'won- would be defined at best and as most as 'regarding as sacredly numinous'. By the way, "don't get me started" on how PIr. *swanta-, Av. spENta-, whose contexts,
    cognates, and the lack of another word for 'holy', leadus back to
    Bartholomae's 'heilig' (maybe add a bit of creative and benign energy)
    is now variously translated by the Gathologists as e.g. 'virtuous, incremental, prosperous, beneficent' etc.–a story in itself!
    To my root *wen(H), expanding Elena Mucciarelli's van(H) 'to appropriate, bring into the ritual sphere/space' vel sim. to "gather in' (including Lat.
    venari) & the Av. word for 'herd, flock', note e.g. Lat. grex.
    As to connection of Av. 'herd, flock' with the Av. and Khwarezmian root
    'to cover', I think of herds and flocks as free-ranging and numerous;
    being a city boy, I don't know which livestock are lucky enough to shelter in enclosures.
    I take the family of Gr. kheimó:n and other unrelatedwords for 'winter/snowstorm'. I don't remember mentioning any Serbo-Croatian;
    Russ. metelica etc. I take from 'throw, hurl', tho Steer did not include these.
    If Gr. euné: is primarlily 'lair' and then 'bed', I still see it as an -n- nominal deriv. from PIE *h1ew 'to cover oneself'.
    Re the Gmc. hound, Hunt, etc. with -n-t-, MAYBE including Latv. sùnt-ele would push us further back to early PIE nom. *'k(u)won(t)s, bit I don't know.
    I tsake Gr. kheimó;n etc.and many unrelated words for 'winter' from 'snowstorm', and these from
    from impelling, hurling, driving'.
    Finally, as to how and whether an early PIE **pk'u-won-, **pk'won-
    would lose its p-, yes, we don't have many voiceless stop initial cluster because parallels because, I can maintain, such were simplified.
    Greek dial. pisures tells us about Greek, not early PIE. Av. has
    tüiriia- '4th' alongside âxtûirîm 'for the 4th time', tûiriia- 'father's brother' but Khwar. has 'fcwr /eftsur/ id., both < **ptrwiya- < ** pHtrwiya-. anyway, my etym. **pk'(u)won- is just a suggestion;
    maybe *k'(u)won- is the orig.
    I'm outa here.
    Martin Schwartz

RSS feed for comments on this post