Inhistorical linguistics,Weise's law describes the loss ofpalatal quality that some consonants undergo in specific contexts in theProto-Indo-European language. In short, when the palatovelar consonants*ḱ*ǵ*ǵʰ are followed by*r, they lose their palatal quality, leading toa loss in distinction between them and theplain velar consonants*k*g*gʰ. Some exceptions exist, such as when the*r is followed by*i or when the palatal form is restoredby analogy with related words. Although thissound change is most prominent in the satem languages, the change probably occurred prior to thecentum–satem division, based on an earlier sound change which affected the distribution of Proto-Indo-European*u and*r. The law is named after the German linguistOskar Weise, who first postulated it in 1881 in order to reconcilecognates inAncient Greek andSanskrit.
TheProto-Indo-European language is thereconstructed ancestor of all theIndo-European languages. It was spoken around the 4th millennium BC or earlier.[1] No record of the language exists, but its forms have beenreconstructed through thecomparative method.[2]
Weise's law describes a sound change that affects the palatovelar consonants of Proto-Indo-European, sometimes called dorso-palatal or simply palatal consonants.[3] These sounds are articulated both with the back part of the tongue and thehard palate of the mouth, represented with*ḱ,*ǵ, and*ǵʰ, where theasterisk signifies a reconstructed or unattested form.[4] They are contrasted with plain velar consonants, also referred to as dorso-velar or simply velar consonants, which are articulated with the back part of the tongue and thesoft palate, represented by*k,*g, and*gʰ.[4] Both of these sets were further contrasted with thelabiovelar consonants, likely pronounced with a simultaneous articulation with the back part of the tongue against the soft palate and therounding of the lips, represented by*kʷ,*gʷ, and*gʷʰ.[5] These three contrastive sets are often known collectively asguttural consonants.[6]
Although only one branch of theIndo-European language family – theAnatolian languages – maintained a distinction between all three sets of consonants,[7]historical linguists divide the Indo-Europeandaughter languages into two categories based on how these sounds developed, namely thecentum and thesatem languages. In the centum languages, the palatovelar sounds lost their palatal quality and merged with the plain velars, creating only a two-way contrast between the plain velar and labiovelar sounds.[8] The terms centum and satem are derived from the Proto-Indo-European word*dḱm̥tóm, later shortened into*ḱm̥tóm, meaning 'one hundred'.[9][10] Centum languages, named after theLatin word for 'one hundred', are those languages in which the palatovelar sounds underwent depalatalization – that is, lost their palatal quality – thereby merging with the plain velars, creating only a two-way contrast between the plain velar and labiovelar sounds.[11] By contrast, satem languages, named after theAvestan word for 'one hundred' (𐬯𐬀𐬙𐬆𐬨satəm), are those in which the labiovelar sounds lost their labialization, causing a lack of differentiation with the plain velar sounds called amerger. The palatovelar sounds, on the other hand, underwentassibilation – also called satemization in this particular context – whereby these palatovelars becamesibilant consonants.[12] Sibilant consonants compriseaffricates, such as[t͡ʃ] (as inchat), andfricatives, such as[s] (as insunk).[13][14]
Oskar Weise first described a problem in correspondences betweenAncient Greek andSanskrit cognates in his 1881 article "Ist anlautendes γ vor λ abgefallen?" ('Is initial γ dropped before λ?').[15] In it, he notes an imbalance in the relationship between Ancient Greek and Sanskrit cognates, writing:
If we examine the Indic words beginning with guttural +r orl and compare them with their Greek reflexes, we will notice that all those which have retained the guttural in Indic intact show guttural +ρ, whereas Greek guttural +λ only occurs regularly when the palatal sibilants [ś ],j,h appear in Indic. The absence of exceptions in this rule automatically prohibits the assumption that coincidence prevailed here. Of course, this excludes cases wherer (orl ) is not immediately after the guttural, but there is a vowel in between, although the rule stated above often applies here too.[16]
According toAlwin Kloekhorst in 2011, Weise's original article has "been largely forgotten by the scholarly world", but its findings have appeared sporadically in linguistic literature with some of it needing revision in light of other research.[17] In 1894,Antoine Meillet described the law and defended it as established fact in a dissertation for theSociété de Linguistique de Paris on the difficulty of determining gutturals in Proto-Indo-European, citing Weise as its progenitor.[18] In 1978,Frederik Kortlandt similarly considered Weise's findings strong but limited in scope, citing both Weise's and Meillet's works on the law in his own research on theBalto-Slavic languages.[19] In 1995,RobertS. P. Beekes also described the process derived from the law, but did not reference its origins with Weise.[17][20] Kloekhorst presented a defense of the law in 2008,[21] followed by a more complete account –Weise's Law: Depalatalization of Palatovelars before*r in Sanskrit – in 2011.[22] The 2011 defense conglomerates several different sources on the topic, some referencing Weise and some not, and summarizes its general characteristics, its relative chronology, and possible violations.[23] Kloekhorst has been credited with reviving interest in the law.[24]
Weise's law describes a depalatalization which affects the palatovelar consonants of Proto-Indo-European:*ḱ*ǵ*ǵʰ. In the satem languages, as well as inAlbanian andArmenian,[a] these sounds became fricatives, such as[s] or[z], rather than remainingstops. Weise's law, however, demonstrates that these sounds depalatalize before*r, thereby merging with the plain velar stops*k*g*gʰ.[17] Because the palatovelar sounds becamesibilants through the process ofassibilation in the satem languages while the plain velars did not, the merging of palatovelars with plain velars explains why these words have plain velarreflexes in words that share a common Indo-European root containing a palatovelar. In other words, while the palatovelar stops were made into alveolar sibilants in most cases, Weise's law explains many exceptions, though not all.[17] The effects of the law are commonly found inzero-grade stems – that is, stems without a vowel[26] – which may receiveinserted vowels in thedaughter languages. One such example may be found inOld Avestan𐬐𐬆𐬵𐬭𐬞𐬇𐬨 (kəhrpə̄m; 'figure, body'), derived from*ḱrp-os- 'body'.[27][b] Although the original palatovelar*ḱ does not immediately precede*r in the Avestan reflex, it is still in accordance with the depalatalization described by Weise's law since*ḱ immediately precedes*r in the zero-grade form*ḱrp-os-.[27]
Sanskrit contains many apparent violations of the rule, particularly where the surface representation of the word containsśr- orhr-, implying a derivation from an unmodified*ḱr- or*ǵʰr- source. However, these are often the result of later sound changes particular to a language orlanguage family. For example, where/l/ became/r/ in many circumstances, such as inश्रवस् (śravas, 'fame'), which is derived from Proto-Indo-European*ḱleu-es-, andह्राद् (hrād; 'to resound, to make a noise'), which is derived from*ǵʰleh₃d-.[30] Thus, these apparent counterexamples do not actually represent exceptions to the rule.[31]
Other apparent violations occur in contexts in which the palatovelar consonant and*r cross a morphemic boundary, such as between anaffix and the root it modifies, or share a clear derivational relationship with another word that would not have been subjected to the sound law, leading to ananalogical change.[32] With respect to the first apparent violation, the Sanskrit wordअज्र (ájra; 'field, plain') is derived from*h₂éǵ-ro- ('field, pasturage'), where the expected outcome is*अग्र (*ágra).[c] However, the typical reflex of the palatovelar consonant has been restored because the palatovelar and the*r are separated by a morphemic boundary, represented here with a dash.[31] In another apparent violation, Sanskritशृङ्ग (śṛṅgá, 'horn') is derived from the zero-grade form*ḱr-n-go-. Although the expected reflex is*क्ङ्ग (*kṛṅgá), the attested form may have been restored based on a relationship with related words likeशरभ (śarabhá, 'a kind of deer'), which is derived from a full-grade form of the root,*ḱer-n-bʰó- where the*r is not immediately preceded by the palatovelar.[35]
All other violations of the rule appear in the particular sequence*Ḱri, whereḰ represents any palatovelar sound. Kloekhorst suggests that thehigh front vowel*i may have palatalized the preceding*r, giving no motivation to depalatalize the initial palatovelar sound.[36]
The chronology of Weise's law is the subject of some debate. The depalatalization described by the law must have occurred by at least the time theIndo-Iranian languages diverged from the rest of Proto-Indo-European (c. 3000 BC).[38] Kloekhorst argues that it probably occurred much earlier, after the divergence of theAnatolian languages (c. 4500 BC),[39] since the distribution of late Proto-Indo-European*u and*r underwent an exchange in placement, ormetathesis, which only occurred after both the Anatolian language family had diverged from Proto-Indo-European and palatovelars had undergone the depalatalization described by Weise's law. This explains exceptions such as Sanskritश्मश्रु (śmáśru, 'beard'), which derives from the form*smóḱ-ru- rather than from the earlier*smóḱ-ur, attested inHittite𒍝𒈠𒀭𒆳 (zama(n)kur, 'beard'). The Sanskrit form does not show depalatalization because the depalatalization occurred only in environments where the palatovelar was followed by*r prior to this metathesis.[d][41]
In 1978,Frederik Kortlandt noted that, while it is tempting to assert that Balto-Slavic and Indo-Iranian depalatalizations derive from a common innovation, the scope of depalatalization is much larger in the Balto-Slavic languages and there is positive evidence that this kind of depalatalization never occurred in Armenian, pointing to examples likeսրունք (srunkʿ, 'leg') from Proto-Indo-European*ḱrūs-ni- andմերձ (merj; 'near, close to') from*méǵʰ(s)ri.[42] Instead, Kortlandt suggested that Weise's law applies only to Indo-Iranian languages and, although Indo-Iranian languages and Balto-Slavic languages restored palatal features in a similar fashion, these restorations occurred independently of one another.[43] Based on Albanian and Balto-Slavic agreement in depalatalization, he considers Albanian to have been atransitional dialect of Balto-Slavic and Armenian during the same period.[43]Robert S. P. Beekes, disputing some of Kortlandt's etymologies,[44] wrote that depalatalization is assumed to have taken place before*r in Armenian as well.[45]
Although the effects of the law are most clearly demonstrated in satem languages, Kloekhorst suggests that this sound change occurred before the centum–satem split, arguing that it almost certainly occurred in late Proto-Indo-European after the departure of theAnatolian languages.[40] Because their reflexes appear to be in accordance with the law, Kloekhorst groupsAlbanian andArmenian in with the satem languages.[46][e] He further notes that it is likely that secondary depalatalizations took place at a later date in each of the satem language families outside theIndo-Iranian languages, as depalatalization is more extensive in those languages.[40]
