The Tower of Babel

Sergei Starostin's Home Page

Russian
<< Home Page << Articles and Books



    1.10. From PWC to modern West-Caucasian languages.
    First we will describe the development from PWC to PAT (Proto-AbkhazTapant or Proto-Abkhaz-Abaza), PAK (Proto-Adyghe-Kabardian) and Ubykh languages, and then give a short characterization of the development from PAT and PAK to modern languages.

    1.10.1. Consonantism.
    The multitude and extreme complexity of correspondences between the consonants of West-Caucasian languages are due to the PWC features of palatalization and labialization, as well as their combination. The interaction of these features (for the most part transferred from following vowels, see above) with local and laryngeal features of consonants creates a very complicated and, at first sight, chaotic system of reflexes. Below, however, we will try to show that this system can be explained.
    For PWC we reconstruct the following system of consonants (not including the additional features of palatalization and labialization):

Labials                p       p:       b       ©       f            w       m
Pharyngealized labials       pI    p:I    bI    ©I                          mI
Dentals                t       t:       d       ®                      r       n
Hissing                c       c:       ½       ˆ       s          z
Hushing                ‰       ‰:       ¾       Š       «    «: ¼    j
Laterals             œ       œ:       ž              ›    ›: L    l
Velars                k       k:       g       ™       x       
Uvulars                q       q:       G       ª       »       
Pharyngealized uvulars      qI    qI:           ªI    »I        ’I


    Pharyngealized consonants in PWC are reconstructed mostly on the basis of Ubykh evidence, and tense ("preruptive") consonants - on basis of the Adygh evidence, though other languages often have specific reflexes of these classes of consonants as well. The tense fricatives «: and ›: reveal specific reflexes in Adygh (Shapsug), see below, and in some cases their reflexes differ from the reflexes of respective lax fricatives in other languages as well (Ubykh, PAT). However, in combination with certain quality features, the reflexes of lax and tense fricatives have apparently merged in all WC languages: thus, it is impossible to distinguish PWC *›´ from *›´: (although there is a distinction *› - *›:), or *« from *«: (although *«´ is opposed to «´:). It is quite possible that PWC also had distinctions of other tense and lax fricatives (*f-f:, *s-s:, *x-x:, *»-»:, *»I-»:I), but, because of the specific character of Adygh reflexation (the Shapsug dialect preserves only two pairs of opposed fricatives at the present time, «-«: and «´-«´:), their reflexes have merged wth each other in all modern West-Caucasian languages.
    We reconstruct the following correspondences between the subgroups of West-Caucasian languages:

PWC          PAT          Ub       PAK
*p             *p          p          *p
*p:          *b          b          *p:
*b             *b          b          *b
*©             *©          ©         
*f             *f          f          *x(º)
*w             *w          w          *w
*m             *m          m          *m
*p´             *p          tº       *t
*p´:          *b          (dº¬tº)    *t:
*b´             *b          dº       *d
*©´             *©          ®º      
*pº          *p          f          *p
*bº          *f          f          *b
*p´º          *c´         
*b´º          *½´          dº       *d
*©´º          *ˆ´          ®º      
*pI          p          vI       *p
*p:I                       bI       *p:
*bI          *b          bI       *b
*©I                       ©I      
*mI          *m          mI       *m
*t             *t          t          *t
*t:          *d          d          *t:
*d             *d          d          *d
*®             *®          ®         
*r             *r          d-,r¬’    *t:-,r
*n             *n          n          *n
*r´             *r          L         
*tº          *tº          tº       *t
*t:º          *d(º)       tº       *t:
*dº          *dº          dº       *d
*®º          *®º          ®º      
*t´º          *tº                    *c
*®´º          *®º          ®º      
*c             *c          c          *s-,c
*c:          *c          c          *c:
*½             *½          ½          *½¬z
*ˆ             *ˆ          ˆ         
*s             *s          s          *s
*z             *z          ½¬z       *z
*c´             *c´          c´          (*s)
*c´:          *½´(?)       c´          *c:
*½´             *½´          ½´          *½¬z
*ˆ´             *ˆ´          ˆ´         
*s´             *s´          s´          *s
*z´             *z          ½´¬z´       *z
*cº          *cº          cº       *sº-
*c:º          *cº          cº       *c:º
*½º          *½º                    *½º
*ˆº          *ˆº          ˆº       *ˆº
*sº          *sº          sº       *sº
*zº          *zº          ½º¬zº    *zº
*c´º          (*‰)                 *c´
*½´º          *¾¬¼       ¾´          *½´
*ˆ´º          *Š          Š´          *ˆ´
*s´º          *s´          «º       *s´
*z´º          *½´¬z´       ¼º       *z´
*‰             *c´          c          *s
*¾             *½´          ½          *½¬z
*Š             *ˆ´          ˆ          *ˆ´(?)
*«(¬*«:)       *s´          s          *s
*¼             *z´          z          *z
*‰´             *‰´          ‰´          *‰´
*‰´:          *½´(¬‰´)       ‰´          *‰´:
*¾´             *¾´          ¾´          *¾´¬¼´
*Š´             *Š´          Š´          *Š´
*«´             *«(?)       «´         
*«´:          *«          «´          *«:
*¼´             *¼          ¼´         
*j             *j          j          *j
*‰º          *‰          ‰          *‰´
*‰:º          *zº          ‰          *‰´:
*¾º          *½º¬zº       ¾          *¾´¬¼´
*Šº          *Š          Š          *Š´
*«º          *sº          «         
*«:º          *«          «          *«:
*¼º          *½´(?)       ¼         
*‰´º          *‰          ‰´          *‰
*‰´:º          *¾¬¼       ‰´          *‰:
*Š´º          *Š          Š´         
*«´º          *«º          «º       *s´
*«´:º          *«          sº       *«:
*¼´º          *¼º          ¼º       *z´
*œ             *x          «´          *‰´
*œ:          *c          (c´)       *‰´:
*ž             *l          L          *t“
*             *x/ˆ       ˆ´          *Š´
*›                          s´          *«´
*›:          *x          s´          *«´:
*l             *l          d-,0¬j    *t“-,L
*œ´             *x´          «´          *‰´
*ž´             *‘´          ’(¬z´)    *‘
*´             *«´                    *
*›´(¬›´:)       *«´          ›          *›
*L´             *¼´          L          *L
*l´             *r-,l¬‘´    L          *d-
*œº          *cº          cº       *‰
*œ:º          *cº          cº       *‰:
*žº          *l          w          *“
*º          *ˆ(º)       ˆ(º)      
*›º          *«          sº       *x(º)
*›:º          *«          s(º)       *«´:
*œ´º          *‰º          f          *x
*œ´:º          *¼º          ¾´          *‰´:
*ž´º          *¾º          ½         
*´º          *Šº          ˆ          *Š´
*›´º(¬›´:º)    *«          «º       *x(º)
*L´º          *¼          ¼º       *‘´
*k             *k          k´          *k´
*g             *g          g´          *g´
*™             *™          ™´          *™´
*x             *x          «´          *x
*‘             *‘          ‘¬’       *‘
*k´             *k´                    *k´
*g´             *g´          g´          *g´
*™´             *™´          ™´          *™´
*x´             *x´          s´          *«´
*‘´             *‘´          z´          *¼´
*kº          *kº          kº       *kº
*k:º          *gº          gº       *k:º
*gº          *gº          gº       *gº
*™º          *™º          ™º       *™º
*xº          *xº          x          *x(º)
*g´º          *gº          g´          *gº
*™´º          *™º          ™´          *™´º
*x´º          *sº          x´          *xº
*‘´º(?)       *zº          ’´          *’º
*q             (*“)       q          *q-
*q:          *q          q          *q:
*G             *‘          ’          *’
*ª             *ª          ª          *q:
*»             *“          »         
*’             *          ’          *’
*q´:          *q         
*G´             *‘´          ’´
*ª´             *ª´          ª´          *‚
*»´             *“          x´          *»´
*’´             *          ’´          *’
*qº          *“º          qº       *qº-,q:º
*q:º          *qº          qº       *q:º
*Gº          *‘º          ’º       *’º
*ªº          *ªº          ªº       *q:º
*»º          *“º          »º       *»º
*’º          *º          *’º       *’º
*q´º          *“(º)       x´          *qº-
*q´:º                       q´          *q:º
*G´º          *‘(º)       ’´          *’º
*ª´º          *º(*ªº?)    ª´          *‚º
*»´º          *“(º)       »´          *»º
*’´º          *(º)       ’´          *’º
*q:I          *                    *q:
*ªI          *“(?)       ªI       *q:
*»I                       »I      
*’I          *          ’I       *’
*q´I          *q          q(I)       *“
*q´:I          *‚          ª          *‚
*ª´I          *“          ª          *‚
*»´I          *“          »(I)       *“
*’´I          *          ’          *j
*qIº          *“º          »Iº       *qº-
*qI:º          *º          qI(º)    *q:º
*ªIº          *ªº(*qº)    ªIº       *q:º
*»Iº          *“º          »Iº       *»º
*’Iº          (*“º?)       ’Iº       *’º
*q´Iº          *qº          »º       *“
*q´I:º          *º          ªº       *‚º
*ª´Iº          *“º          ªº       *‚º
*’´Iº          *º          w          *w¬’º


    Comments.
    1. Labial consonants. As we see from the table, labial resonants, the rare fricative *f and labialized pharyngealized consonants do not have any palatalized or labialized correlates. Other consonants have palatalized and labialized variants. While evaluating the given reconstructions one must keep in mind that the labialized dentals are articulated as biphocal explosives (i.e. phonetically = /tp´, db´, ®p´/) in modern Abkhaz and Ubykh, and have a pronounced palatalized character; therefore, the development of the type *p´ > tº (with a possible further delabialization tº > t) seems quite possible to us. The reconstruction of the series of "palatalized-labialized" labials is rather hypothetical. One could reconstruct here a series of "palatalized-labialized" dentals as well, but some external correspondences lead us to assume the labial character of original PWC phonemes. In Ubykh and Adygh they give the same reflexes as palatalized labials proper; in PAT we should suppose, at first, the depalatalization of palatalized labials (*p´ > *p), and afterwards - the delabialization of "palatalized-labialized" with a following affrication (*p´º > *p´ >c´).
    2. Dental consonants. For a specific rare correspondence "PAT *r: Ub. L: PAK *½" we suggest a PWC palatalized *r´. The liquids *r and *l (on the latter, see below) are thus the only PWC resonants that have palatalized correlates (correlation in labialization in the system of resonants is altogether absent). An interesting feature of the dental series is the possible presence there of at least two "palatalized-labialized" consonants (*t´º and *®´º), while the respective plain palatalized ones are missing. However, it is not to be excluded, that they were present in PWC, but were subject to depalatalization (or, less likely, merged with the reflexes of affricates) in all descendant languages.
    3. Hissing consonants. Plain hissing consonants are well preserved in descendant languages (except the fricativization *c > s in PAK and the ½-z variation in the place of *½ in PAK and in the place of *z in Ubykh). Palatalized hissing consonants are preserved in PAT (though we should note the depalatalization *z´ > z) and in Ubykh, but depalatalized in Adyghe. Labialized hissing consonants are generally well preserved in all three subgroups (in modern languages labialized hissing consonants are phonetically articulated as dentolabialized, see below).
    The suggested "palatalized-labialized" hissing consonants have specific reflexes in subgroups. In Adygh a consonant shift occurred: after the depalatalization of original palatalized sounds (see above), the delabialization of "palatalized-labialized" hissing sounds apparently happened (*c´º>*c´, *s´º>*s´, etc.). In Ubykh a development of palatalized-labialized hissing sounds into palatalized-labialized hushing ones apparently occurred (*½´º>¾´º, *s´º>«´º, etc.), with a following delabialization of affricates (there are no hushing labialized affricates present in modern Ubykh), but with the preservation of labialization in the subsystem of fricatives (it must be noted that the phonemes, transcribed in Ubykh as «º, ¼º, according to their descriptions, have an evident - though phonologically insignificant - palatalization, i.e. phonetically they are /«´¹, ¼´¹/). A similar development had apparently happened in PAT, i.e. first all palatalized-labialized hissing affricates developed into palatalized-labialized hushing ones; then affricates were simultaneously delabialized and depalatalized - cf. a similar development that occurred much later in some Abaza dialects, where the process (*tº) > ‰´º > ‰, *«´º > « happened (see [Lomtatidze 1976]). As for palatalized-labialized fricatives, they have apparently simply lost their labialization (i.e. the same process as in Adygh languages occurred). As a result, there we have a specific system of correspondences, wherein the original labialization of all regarded consonants is hinted at only by the fricative reflexes «º, ¼º in Ubykh. Below we will see that complicated processes of delabialization, accompanied by a shift in other quality characteristics of consonants, are typical for other local series of West-Caucasian consonants as well: it was these processes that have caused the extreme complexity of phonetic correspondences, observed among modern WC languages.
    4. Hushing consonants. PWC palatalized hushing consonants are comparatively stable in WC languages. They are usually preserved in descendant languages (except the palatalized hushing fricatives *«´, *«´:, *¼´, which have lost palatalization in PAT and in PAK; as we see from the table, the palatalized hushing fricatives of PAT and PAK go back only to PWC lateral or velar consonants).
    The matter is more complicated with the reflexes of PWC non-palatalized and "palatalized-labialized" hushing consonants. Here the following situation is observed:
    a) there are several rows of correspondences where the PAT palatalized hissing ("hissing-hushing") consonants correspond to Ubykh and PAK non-palatalized hissing ones. In this case we cannot reconstruct palatalized hissing consonants for PWC, since their place is already occupied (see above: palatalized hissing consonants are reconstructed on the base of the correspondences "PAT, Ub. palatalized hissing: PAK non-palatalized hissing");
    b) despite the fact that non-palatalized hushing consonants are present in all three subgroups of West-Caucasian languages, the correspondences between them are never uniform; thus, on one side, we have rows of correspondences where the non-palatalized hushing consonants in PAT and Ubykh correspond to palatalized hushing consonants in PAK; on the other side, we have rows of correspondences where non-palatalized hushing consonants in PAT and PAK correspond to palatalized hushing consonants in Ubykh. The parallel fricative rows of correspondences usually reveal traces of original labialization. Thus, the row "PAT *‰ : Ub. ‰ : PAK *‰´" is parallel to the row "PAT *sº : Ub. « : PAK *«" (we must keep in mind that the non-palatalized PAK *« is a reflex of an earlier *«´, see above); on the other hand, the row "PAT *‰ : Ub. ‰´ : PAK *‰" is evidently parallel to the row "PAT *« : Ub. sº : PAK *«:", etc. Here the Abkhaz and Ubykh evidence serves as an argument for reconstructing original labialization in those rows of hushing correspondences, where only non-palatalized (and non-labialized) hushing reflexes are found nowadays.
    In our opinion, this situation should be interpreted as follows. The original non-palatalized hushing *‰, *Š, *«, etc., have lost their hushing feature in all three subgroups, changing into non-palatalized hissing consonants in Ubykh and PAK (only the glottalized *Š, for some unclear reason, has yielded PAK hissing-hushing *ˆ´), and into hissing-hushing consonants in PAT. Here we should note that the so-called "hissing-hushing" consonants in West-Caucasian languages have a dual phonological and phonetical nature; on one hand, they occupy the place of palatalized correlates for nonpalatalized hissing consonants (and therefore merge with plain hissing consonants in case of depalatalization). On the other hand, they form a sort of intermediate series between hissing and hushing consonants. Because of the last circumstance, the direct development of non-palatalized hushing into hissing-hushing consonants seems phonetically justified. The fact that, in the regarded rows of correspondences, one must reconstruct hushing consonants, is confirmed by external comparison with East-Caucasian languages as well.
    After the loss of old non-palatalized hushing consonants their place in PAT, Ubykh and PAK started to be filled by means of delabializing originally labialized consonants. In PAT reflexes of labialized and "palatalized" hushing affricates merged (all of them developed into non-palatalized hushing consonants, except the affricate *¾º, that apparently fricativized early and developed similarly to the fricative *«º; the variation zº¬½º in the place of *¾º is probably a rather late occurrence). The tense fricatives *«:º, *«´:º developed similarly ( > PAT *«); however, lax fricatives preserved labialization, though they have been locally shifted (*«º > sº and *«´º, *¼´º > «º, ¼º). We must note that the latter shift is, in a way, "orthographic": the PAT labialized hushing *«º, *¼º, as well as the respective consonants of the modern Bzyb dialect, undoubtedly had been phonetically palatalized (= [«´º], [¼´º]), though they had no non-palatalized correlates. The reflex ½´ in the place of PWC *¼º is unclear (one should expect a *zº).
    In Ubykh the labialized hushing affricates were subject to a uniform process of delabialization (*‰º > ‰, *‰´º > ‰´, etc.). Nonpalatalized labialized hushing fricatives developed similarly (*«º, *«:º > «, *¼º > ¼), but palatalized-labialized hushing fricatives have preserved labialization. Lax *«´º, *¼´º > «º, ¼º (just as in PAT, this rule does not really mean any phonetic change; on the phonetic character of these Ubykh consonants see above, page ), but the tense *«´:º has lost its hushing character and developed into sº.
    Finally, in PAK the palatalized and non-palatalized labialized hushing consonants seem to have "swapped places" after delabialization. Apparently, there was a whole series of successive processes that led to this result. We can suppose that the palatalized-labialized hushing consonants, as a result of delabialization, first changed into hissing-hushing (traces of this stage are preserved as hissing-hushing fricatives s´, z´ in the place of PWC *«´º, *¼´º). Non-palatalized labialized hushing consonants consequently became the only labialized hushing series and underwent a phonetic palatalization (if there is no phonological distinction in palatalization, the palatalized articulation of hushing consonants is always more typical for Caucasian languages), as a result of which, after delabialization, they merged with the reflexes of palatalized hushing consonants. Only afterwards did the hissing-hushing consonants (except lax fricatives *s´, *z´ < *«´º, *¼´º) were apparantly shifted into the non-palatalized hushing series. Thus, here we also see a phonetic closeness between non-palatalized hushing and hissing-hushing consonants, which was already mentioned above (concerning the development like *‰ > c´ in PAT).
    5. Lateral consonants. The reconstruction of lateral consonants in PWC is extremely complicated. PAT has altogether lost laterals as a local series; PAK and Ubykh have only three lateral consonants (›, L and ), whose correspondences with the PAT fricatives «´ and ¼´ were already established by N. S. Trubetskoy (see [Trubetskoy 1922]). However, we may show that these series of correspondences are not at all the only ones that reflect PWC lateral consonants.
    First, we should note an evident palatalized character of those PWC consonants that were preserved as laterals in PAK and Ubykh (this is already testified to by the palatalized character of the PAT reflexes «´, ¼´). Therefore, we reconstruct PWC palatalized laterals *›´, *L´, *´ here.
    The presence of laterals in PWC can be also postulated in those series of correspondences where PAT has the lateral reflex l. Therefore, it is comparatively easy to reconstruct the PWC resonants *l and *l´ (though in some languages and positions we observe the specific reflexes r, d, t“ here; the reflex L in PAK and Ubykh is not surprising, because L is the only voiced lateral phoneme here). However, we have two more series of correspondences (PAT *l: Ub. L: PAK *t“ and PAT *l: Ub. w: PAK *“) in which we should reconstruct some PWC lateral affricates (the latter row evidently represents a labialized correlate of the former one, and the labialization of resonants is impossible in PWC). The most natural solution would be to reconstruct the voiced affricates *ž and *žº, the reflexes of which appear to be rather close to those of the resonant *l.
    In addition to the examined corespondences, there is still a rather large number of specific rows of correspondences, in which descendant languages either have sibilant reflexes (that do not at all fit into the ranks of the rows of correspondences of original hissing and hushing consonants examined above), or mixed reflexes, with front affricates or fricatives of some languages corresponding to velar fricatives in others. The latter circumstance is already a direct hint at the possibility of reconstructing original lateral consonants here (because velar fricatives are typical reflexes of laterals in many East-Caucasian languages, see above). The reconstruction of PWC laterals in all of these series is well confirmed by East-Caucasian correspondences.
    All types of correspondences that are of interest to us here are divided into four types (which is by itself already an argument for the reconstruction of a standard PWC four-way distinction of qualitative features):
    1) Correspondences of non-palatalized PAT fricatives (*x) or hissing affricates (*c, *ˆ) to the Ubykh palatalized hissing consonants s´, ˆ´ (or to the hushing «´ in the series PAT *x: Ub. «´: PAK *‰´) and PAK palatalized hushing consonants. Here we reconstruct PWC non-palatalized laterals (since the places of palatalized laterals are already occupied - see above, and because of the nonpalatalized character of PAT reflexes). We should note that PAT has a double reflex (x or ˆ) of the glottalized *, the former being present if a following hissing consonant is present in the word (i.e. in a dissimilative position), and the latter - in independent position. The phoneme *ž, reconstructed above, is included in the series of PWC nonpalatalized lateral consonants as well, despite the fact that it has specific reflexes in descendant languages (which is natural for the least stable element of the lateral series - the voiced lateral affricate).
    2) Correspondences of the palatalized PAT fricatives «´ and ¼´ to PAK and Ubykh lateral ›, L, . Here, as we have already pointed out, we reconstruct PWC palatalized laterals. This series should also undoubtedly include the correspondences of PAT palatalized fricatives x´, ‘´ to Ubykh «´, z´(¬’) and PAK *‰´, *‘ (i.e. the rows of correspondences "PAT *x´ : Ub. «´: PAK *‰´" and "PAT *‘´ : Ub. z´(¬R) : PAK *‘"). In the two latter rows it is natural to reconstruct the palatalized PWC affricates *œ´ and *ž´ (which have also lost their palatal character in PAK and Ubykh).
    3) Correspondences, in which PAT and Ubykh reveal labialized hissing consonants, but PAK has hushing reflexes. It is evident that we also should add here the rows of correspondences with nonpalatalized hushing fricative reflexes in PAT, labialized hissing fricative reflexes in Ubykh and either hushing or velar fricative reflexes in PAK. These are most probably reflexes of PWC labialized lateral affricates (*œº, *œ:º, *º; *žº, as well as the respective non-labialized *ž, has specific reflexes, see above) and fricatives (*›º, *›:º).
    4) Correspondences of PAT labialized hushing consonants to different (velar or front) consonants in Ubykh and PAK. There are exactly four such correspondences (by the number of affricates: *œ´º, *œ´:º, *ž´º and *´º), and we should apparently reconstruct the PWC "palatalized-labialized" lateral affricates here (we must note that in PAT labialized hushing affricates, as well as fricatives, on which see above, had probably been phonetically palatalized). There are also two fricative rows, in which Ubykh «º, ¼º correspond to non-palatalized hushing consonants in PAT and to velars in PAK, and in which we reconstruct the lateral fricatives *›´º and *L´º respectively.
    We should once more emphasize that for all the above-examined rows of correspondences quite reliable external PEC data exists, pointing to the fact that all these rows actually reflect original lateral consonants. Without such external data the reconstruction of the PWC system of laterals would certainly be absolutely hypothetical. Let us note that the correspondence of Adygh front affricates to lateral consonants of Daghestan languages was also mentioned in the works of A. I. Abdokov (see [Abdokov 1976, 1983]); these works, however, do not contain a reconstruction of the PWC system of laterals.
    We see that the reconstruction of three local affricate series (hissing, hushing and lateral) and a four-way distinction of qualitative features (plain = nonpalatalized, palatalized, labialized and palatalized-labialized consonants) allows us to find a place and to suggest a phonetic interpretation virtually for all - absolutely chaotic at first sight - correspondences between West-Caucasian affricates and fricatives.
    6. Velar consonants. At the present time the opposition of palatalized and unpalatalized velars exists only in Abkhaz and Abaza, and is reconstructed on their evidence. In Ubykh and PAK all velars underwent a secondary palatalization. The "palatalized-labialized" velars are reconstructed for those rows of correspondences, in which Ubykh has non-labialized (palatalized) velars corresponding to labialized velars in PAT and PAK. The reflexes of the palatalized fricatives *x´, *‘´ (preserved in PAT but subject to sibilantization in Ubykh and PAK), as well as of *x´º, *‘´º (yielding normal reflexes in Ubykh and PAK, but developing into *sº, zº in PAT) are somewhat specific (though easy to explain).
    7. Uvular consonants are best preserved in the Ubykh language (where we should note only the usual processes: weakening of tense uvulars and delabialization of "palatalized-labialized" uvulars, completely similar to the analogical process in the system of velars, see above). We must note that the signs x´ and ’´ in Ubykh are somewhat misleading: there is no distinction of velar and uvular palatalized fricatives, and therefore we could write x´, ‘´ as well as »´, ’´ (thus, the development *»´ > x´ is pure orthography, not a real change). Among other changes in Ubykh we should also note the fricativization G > ’ (in combination with all qualitative features) and *q´º > x´.
    In PAT we observe a characteristic process of laryngealization *», ’ > “, (in combination with all qualitative features). A similar laryngealization has also occurred here with the lax uvular *q (and its qualitative correlates), obviously, as a sequence of the preceding fricativization *q > *». Among other processes that have affected the PAT system, we should mention: 1) the weakening *q:>q; 2) the fricativization *G > ‘ (we must note that, in PAT, uvular and velar fricatives were not opposed, therefore we could write ’ instead of ‘). The original palatalization is preserved only by PAT reflexes of the PWC phonemes *G´ and *ª´; palatalized q´, “´ and ´ are missing in PAT, therefore in the place of PWC *q´:, *»´ and *’´, PAT has non-palatalized reflexes.
    In PAK lax q, qº are well represented only in initial position; judging by the available examples, the PWC lax uvular affricates were strengthened in intervocalic position in PAK. Palatalization of uvulars (except the relic fricative »´; on its reflexes in Adygh languages, see below) has been lost in PAK. We must note, however, that while the non-palatalized glottalized *ª, *ªº lose their glottalization in PAK ( > q:, qº:), the respective palatalized consonants yield the specific laryngeal reflexes ‚, ‚º. We should note that A. Kuipers ([Kuipers 1963]) reconstructs here PAK *ª, *ªº, which does not seem quite legitimate to us (see below).
    Uvular pharyngealized consonants are only preserved in Ubykh. In this language pharyngealization was lost only by palatalized pharyngealized consonants, whose presence in PWC can, however, be postulated on basis of specific rows of correspondences, where Ubykh uvular (sometimes preserving their pharyngealization) consonants correspond to PAT and PAK emphatic laryngeals.
    In PAT most pharyngealized uvulars developed into emphatic laryngeals; the uvular character was preserved only by PWC *q´I, *q´I and *ª´Iº. In PAK the pharyngealized uvulars (except the palatalized ones) develop just as their respective non-pharyngealized counterparts. However, the suggested uvular pharyngealized palatalized (as well as palatalized-labialized) consonants have all developed into emphatic laryngeals (we should note the development *’´I > j, *’´Iº > w, that had apparently had an intermediate stage , º).


    1.10.1.1. Consonant clusters.

    Consonant clusters are extremely widespread in modern West-Caucasian languages. Historical analysis shows, however, that in most cases these combinations are secondary, arising as a result of vowel reduction. Only in comparatively few cases can we reconstruct real consonant clusters for PWC (mostly on the basis of Ubykh and PAK data; in PAT all the old clusters were probably altogether lost) in the following correspondences:

PWC    PAT       Ub          PAK
*P-C    C          PC          PC
*M-C    m(V)C    mC          C
*T-C    C          TC          TC
*n-C    C          nC          n(V)C
*r-C    C          C          r(V)C
*s-C    C          «C/«´C       «C/tC


    In the latter type of clusters, Ubykh and Adyghe have « preceding a uvular consonant, but «´ and t respectively preceding other consonants.
    All the listed types of clusters are mostly attested in the beginning of PWC roots and have probably historically developed from the sequences *CVC-, where vowel reduction had occurred already on the PWC level. In some cases, however, the initial PWC clusters probably reflect grammatical prefixation (see above, page 85), the traces of which are almost lost in nominal PWC roots. Grammatical prefixation is probably reflected in some other cases as well (e.g., there is a set of roots with the initial sequences PC- or TC- in PAK, but plain C- in Ubykh and in PAT; these cases are difficult to interpret in any other way than the reflection of old prefixation).
    In the medial position of PWC non-monosyllabic roots, consonant clusters are extremely rare (on the simplification of old consonant clusters in this position see above, page 63). However, there is one class of cases that should be specially mentioned. In some roots with initial labial consonants there are variations of nasal and non-nasal reflexes in descendant languages. We tentatively reconstruct nasalized vowels here and establish the following rows of correspondences:

PWC                   PAT       Ub    PAK
*wV~-                   *w-       w-    n-
*bV~-                   *m-/P-    P-    P-
*p:V~-,*pV~-,*©V~-       *m-/P-    m-    P-


    In many cases like this nasalization probably reflects old lost nasal or liquid medial resonants. However, the reconstruction of nasalization for PWC is rather tentative - first of all, because there are no traces of nasalized vowels after other initial consonants or in monosyllabic structures. The exact phonetic character of this phenomenon in PWC is yet to be investigated.
    PAT has the reflex m- usually preceding fricatives and resonants, while a non-nasal reflex is present in other cases. Laryngeal features (voice / voicelessness / glottalization) of the initial labial in PAT, Ubykh and PAK reflexes depend on the following consonant (see below).


    1.10.1.2. Variations of laryngeal features of consonants.

    In some cases in West-Caucasian languages we observe a violation of regular (see above) correspondences of features, such as voice/voicelessness, tenseness/laxness and glottalization. Variations of voice/voicelessness and glottalization are not rare in the development from PNC to PWC (see above); but sometimes such variations are observed as well within West-Caucasian languages themselves. Sometimes the reasons for these variations are difficult to establish; in most cases, however, the probable reason is the activity of various assimilative processes.
    One of the most typical cases is the violation of correspondences of laryngeal features in structures of the type PVCV, where the first consonant is usually labial, and the second consonant is usually a front one. Most frequent are the following types of correspondences between languages:

PWC          PAT       Ub       PAK
*pVC:V       *p(V)CV    pCV    *p:C:V
*pVC·V       *p(V)CV    pCV    *©C·V
*pVZV       *p(V)SV    bZV    *bZV
*©VZV(?)    *b(V)ZV    ©C·V    *bZV


    In some other types of roots we may also suppose the activity of various assimilative processes; they all, however, require individual comments.

    1.10.2. Vocalism.

    West-Caucasian languages are known for their poor vocalic systems. For PWC we should apparently also reconstruct a system consisting of two vowels: * and *a, preserved in all descendant languages. In Ubykh and in the Adygh languages there is an additional long vowel ƒ. In the Adygh languages this vowel is evidently secondary: it appears in the first syllable of the word in the place of the short *a in case there is another a in the next syllable (in case of in the next syllable there is no lengthening). In Ubykh, ƒ in the first syllable can also appear only before the vowel a of the second syllable. In the same position, however, we can meet the short a as well, thus the distribution between a and ƒ is not quite clear here (though it would not be reasonable to reconstruct the opposition *a - *ƒ for PWC on the basis of Ubykh data alone).
    In many cases the PAT, Ubykh and PAK vowels and a uniformly correspond to each other. We should mention some particular regularities that seem to modify the reflexes in individual cases:
    1) The vowel * yields a in PAK and Ubykh in monosyllabic roots that are included in the enklinomena (unaccented) type in PAT; the old * is preserved in roots of the orthotonic (accented) type and after labialized consonants;
    2) The vowel *a is usually preserved everywhere. Only in the initial syllable *ma- in bisyllabic roots do we observe the development a > in Ubykh and in PAT - also only in roots with the initial unaccented (minus) syllable.
    The situation with vocalic correspondences is somewhat complicated by the evident presence of the */a ablaut in PWC (in PAT there are clear traces of this ablaut in nominal roots, and in PAK it is very productive in verbal stems). This often leads to a violation of correspondences even between close dialects.
    On the prehistory of West-Caucasian bivocalism see above (page 73); its development from a richer original system (as a result of transferring qualitative features onto preceding consonants) seems quite clear. This allows us to explain the extreme richness of PWC consonantism as well as the four-way opposition of consonantal qualitative features. However, it is hardly reasonable to reconstruct qualitative oppositions of vowels as late as on the PWC level, as does A. I. Abdokov; we should rather date the process of destruction of qualitative (and quantitative) vocalic oppositions from the period that immediately preceded the division of the PWC unity.


    1.10.3. Root structure and prosody.

    In PWC, as well as in modern West-Caucasian languages, the bulk of roots had the structure (C)CV; longer roots with the structure CVCV or CVCVCV were more rare. The prevalence of monosyllabic roots in PWC is historically explained by the fall of syllables with resonant and laryngeal consonants (see above, page 85).
    PWC undoubtedly had a tonal accent structure. In the modern Ubykh and Abkhaz languages the original tonal system has been transformed into a system with a dynamic mobile accent, wherein the positioning rules are determined by information about the belonging of each syllable of the given word-form to one of the two accent classes (which are respectively marked as "+" and "-" by V. A. Dybo). Both Abkhaz and Ubykh have a rule according to which dynamic accent in the word is placed on the last syllable in the sequence of "plus" syllables, and in the case when the word has no "plus" syllables - on the last syllable in the sequence of "minus" syllables. See the description and an attempt of the reconstruction of the original system (it is suggested that the morphonologic "+" and "-" characteristics reflect PWC tones) in the works [Dybo 1977, 1989].
    Some irregularities in accent correspondences between Ubykh and Abkhaz are probably explained by the presence of a third tone in PWC, which cannot be discovered by synchronous morphonological analysis of the Ubykh and Abkhaz systems. While working with speakers of the Tapant dialect of the Abaza language, we discovered that it still preserves tonal oppositions (already lost in Abkhaz), that are, however, already combined with a developed system of dynamic accent. The differences between the Tapant and Abkhaz systems also serve as an argument in favour of reconstructing one more tonal feature in PAT, whose postulation would allow us to explain many cases of irregular Abkhaz-Ubykh accent correspondences. The solution of this problem now depends on a careful field examination of the Abaza data.
    In the modern Adygh languages the accent distinctions seem to be absent. However, S. L. Nikolayev was able to reconstruct a distinctive accent in bisyllabic nominal PAK roots, the place of which, in most cases, corresponds to the place of accent in related Abkhaz and Ubykh forms (see below).
    The reconstruction of the PWC accent system, we hope, will be completed in the nearest future, after which it will be possible to attempt a comparison of prosodic systems in West-Caucasian and East-Caucasian languages.


    1.10.4. From PAT to the modern Abkhaz and Abaza dialects.

    All modern Abkhaz and Abaza dialects are very close to each other, and therefore we will limit ourselves to the briefest information about their comparative phonology (we use the data of the best described systems: the Bzyb and Abzhui dialects of Abkhaz and the Tapant dialect of Abaza).
    For PAT we reconstruct the following system of consonants (a similar system - with some minor differences - is suggested in the works of K.V. Lomtatidze, see [Lomtatidze 1976] et al.):

Labials             p       b       ©       f             m       w
Dentals             t       d       ®                   r,l    n
Labialized dentals         tº       dº       ®º
Hissing             c       ½       ˆ       s       z
Hissing-hushing (palatalized)             c´       ½´       ˆ´       s´      
Labialized hissing         cº       ½º       ˆº       sº      
Hushing             ‰       ¾       Š       «       ¼
Palatalized hushing         ‰´       ¾´       Š´       «´       ¼´    j
Labialized hushing         ‰º       ¾º       Šº       «º       ¼º
Velars                k       g       ™       x      
Palatalized velars         k´       g´       ™´       x´       ‘´
Labialized velars         kº       gº       ™º       xº       ‘º
Uvulars             q              ª
Palatalized uvulars                       ª´
Labialized uvulars         qº              ªº
Emphatic laryngeals                           ‚       “      
Labialized emphatic laryngeals                              “º       º


    We establish the following correspondences between modern languages and dialects:

PAT    Bzyb       Abzh       Tap
*p    p          p          p
*b    b          b          b
*©    ©          ©          ©
*f    f          f          f
*w    w          w          w
*m    m          m          m
*t    t          t          t
*d    d          d          d
*®    ®          ®          ®
*r    r          r          r
*l    l          l          l
*n    n          n          n
*tº    tº [tp]    tº [tp]    ‰º
*dº    dº [db]    dº [db]    ¾º
*®º    ®º [®©]    ®º [®©]    Šº
*c    c          c          c
*½    ½          ½          ½
*ˆ    ˆ          ˆ          ˆ
*s    s          s          s
*z    z          z          z
*c´    c´          c          c
*½´    ½´          ½(¬z)    ½(¬z)
*ˆ´    ˆ´          ˆ          ˆ
*s´    s´          s          s
*z´    z´          z          z
*cº    c´¹          ‰¹       ‰º
*½º    ½´¹          ¾¹       ¼º
*ˆº    ˆ´¹          й       Šº
*sº    s´¹          «¹       «º
*zº    z´¹          ¼¹       ¼º
*‰    ‰          ‰         
*¾    ¾          ¾          ¾
*Š    Š          Š          Š
*«    «          «          «
*¼    ¼          ¼          ¼
*‰´    ‰´          ‰´          ‰´
*¾´    ¾´          ¾´          ¾´
*Š´    Š´          Š´          Š´
*«´    «´          «´          «´
*¼´    ¼´          ¼´          ¼´
*j    j          j          j
*‰º    f          f          c
*¾º    v          v          ½
*Šº    ©          ©          ˆ
*«º    «¹          «¹       «º
*¼º    ¼¹          ¼¹       ¼º
*k    k          k          k
*g    g          g          g
*™    ™          ™         
*x    x          »          »
*‘    ’          ’         
*k´    k´          k´         
*g´    g´          g´         
*™´    ™´          ™´          ™´
*x´    »´          »´          »´
*‘´    ’´          ’´          ’´
*kº    kº          kº      
*gº    gº          gº      
*™º    ™º          ™º       ™º
*xº    xº          »º       »º
*‘º    ’º          ’º       ’º
*q    »          »          q
*ª    ª          ª          ª
*ª´    ª´          ª´          ª´
*qº    »º          »º      
*ªº    ªº          ªº       ªº
*‚    ª          ª         
*“    “          “         
*    0(ƒ)       0(ƒ)      
*“º    “º          “º       “º
*º    º          º       º



    Comments.
    1. Phonemes, that are reconstructed as hissing labialized and hushing labialized, could be interpreted as hissing-hushing labialized and hissing labialized respectively. Historically, however, only the former have developed from hissing labialized consonants (the rare PAT hushing labialized consonants go back only to PWC laterals, see above), and we prefer the interpretation suggested above.
    2. PAT had no distinction between velar and uvular fricatives (it has arisen only in the Bzyb dialect as a result of the fricativization q > »). In modern Abkhaz and Abaza dialects these phonemes have a uvular articulation. This also concerns the Bzyb dialect with its opposition x - » : the phoneme that we denote as x also has a uvular (or, at least, back velar) place of articulation, and differs from the uvular » primarily by the character of friction (flat friction slot by x as opposed to round friction slot by »). In PAT we could also reconstruct uvular rather than velar fricatives; historically, however, these phonemes go back to PWC velars (and laterals, see above), while old uvular fricatives in PAT have undergone laryngealization. Therefore, for PAT we prefer to reconstruct velar fricatives (though it is certainly only one of the possibilities).
    3. The PAT voiced laryngeal disappears in Abkhaz, leaving behind a compensatory lengthening of the vowel (both and a give ƒ in this case). The correlated labialized laryngeal, preserved in Abaza, in Abkhaz is pronounced as a specific "emphatic-palatalized" w (/w´I/). For the sake of uniformity we denote this specific Abkhaz phoneme as º, using the same transcription as in Abaza.
    As for vowels, we reconstruct the same bivocalic system with the vowels and a for PAT, as for PWC in general. In Abkhaz as well as in Abaza there is a rule according to which the vowel is dropped in unaccented position (if the accent is being shifted to this unaccented syllable, the vowel is restored). As a result of -reductions in long words new clusters can arise, consisting of a large number of consonants; the morphonological analysis, however, clearly shows that all such sequences are recent and that in PAT we should reconstruct dropped vowels. Both in Abkhaz and in Abaza there are some contexts in which the unaccented does not disappear; however, since such contexts are rather few, and the operating rules are rather complicated, we will not dwell on them here.
    In some words Abkhaz dialects have clusters of identical consonants, never divided by . This is mostly expressive vocabulary, but it may in principle point to the appearance in Abkhaz of a new class of tense consonants (geminates) (the old PWC tense consonants have been lost in PAT, see above).
    As a result of the complete simplification of consonant clusters, the PAT root has acquired the structure CV (or CVCV, more rarely CVCVCV). In modern languages this structure is preserved on the morphonological level, but, as a result of -reduction, on the phonological level there are already many roots with new consonant clusters.
    For a short characteristics of the Abkhaz accent system see above, page 193; for more detailed information see the works [Äûáî 1977, 1989].


    1.10.5. From PAK to modern Adyghe languages.

    The PAK reconstruction was made by A. Kuipers (see [Kuipers 1963]; later Soviet works - Áàëêàðîâ 1970 and Êóìàõîâ 1981 - have little added to our knowledge of PAK). There are the following minor differences (suggested by S. L. Nikolayev) between his reconstruction and ours:
    1) On the basis of (rare) correspondences "Ad. ‰´: Kab. s´" and "Ad. ¾´: Kab. z´", we reconstruct the PAK hissing-hushing affricates *c´ and *½´, which are parallel to the hissing-hushing fricatives *s´, *z´ (the latter two are present in the reconstruction of A. Kuipers, too). Consequently, the correspondence "Ad. s´·: Kab. s´·" (according to A. Kuipers reflecting PAK *s´·) is interpreted by us as the result of a uniform fricativization of the original PAK glottalized hissing-hushing affricate *ˆ´.
    2) On basis of the correspondence "Ad. ‘: Kab. ¼" (also very rare) we reconstruct a fourth PAK lateral phoneme *ž (on its origin see above). Consequently, instead of the lateral glottalized fricative *›· (according to A. Kuipers) we reconstruct a PAK lateral glottalized affricate *.
    3) On basis of the correspondences "Ad. “: Kab. »" and "Ad. ‘: Kab. z´", we reconstruct two palatalized back fricatives *»´ and *‘´, not present in the reconstruction of A. Kuipers (they are also very rare).
    4) Instead of the fricative *s·º, we reconstruct the affricate *ˆº; thus, glottalized fricatives are totally eliminated from the PAK system (their secondary character is also clearly seen in comparison with the evidence of other West-Caucasian languages, see above).
    5) On the other hand, instead of the glottalized affricate ª of A. Kuipers we reconstruct the emphatic laryngeal *‚ (preserved with the same articulation in modern Adygh languages); as we have seen above, though this laryngeal developed from an earlier uvular consonant, it did not develop directly from *ª. The fact that this consonant is rendered as ª in Ubykh loanwords, is irrelevant in this case (because of the lack of emphatic laryngeals in Ubykh).
    6) Instead of the initial h, reconstructed by A. Kuipers on the basis of system considerations, we prefer to reconstruct a zero beginning in PAK.
    7) In bisyllabic roots of PAK we reconstruct two possible places of accent with the following development of structures in descendant languages:

PAK       Ad       Kab
*Ca´Ca    CƒC       CƒCa
*Ca´C    CaC       CaC
*C´Ca    CC       CCa
*C´C    CC       CC
*CaCa´    CƒCa       CƒCa
*CaC´    CaC       CaC
*CCa´    CCa       CCa
*CC´    CC       CC


    In Adyghe (Temirgoy) there is a rule, according to which every unaccented final vowel is reduced; in Kabardian there is a rule according to which every final is reduced, but final a is preserved. It is possible that we will be able to reconstruct relevant accent oppositions in monosyllabic morphemes as well (by analysing composita); however, this has not yet been done.
    Let us now give a short table of correspondences between the Adyghe languages and dialects (for more detailed information see the works [Kuipers 1963, Kumakhov 1981]). We will give the data of the best described systems (Bzhedug, Temirgoy and Kabardian):

AK          Bzhed       Tem       Kab
*p          p          p          p
*p:          p:          p          b
*b          b          b          b
*©          ©          ©          ©
*w          w          w          w
*m          m          m          m
*t          t          t          t
*t:          t:          t          d
*d          d          d          d
*®          ®          ®          ®
*r          r          r          r
*n          n          n          n
*c          c          c          c
*c:          c:          c          ½
*½          ½          ½          ½
*ˆ          ˆ          ˆ          ˆ
*s          s          s          s
*z          z          z          z
*cº          c´º          c´º          f
*c:º       c´:º          c´º          v
*½º          z´º          z´º          v
*ˆº          s·´º          s·´º         
*sº          s´º          s´º          f
*zº          z´º          z´º          v
*c´(c´:)       ‰´(‰´:)       ‰´         
*½´          ¾´          ¾´         
*ˆ´          s·´          s·´          s·´
*s´          s´          s´         
*z´          z´          z´         
*‰          «          «          «
*‰:          ‰:          ‰          ¼
*Š          Š          Š          s·´
*«          «          «         
*«:          «:          «         
*¼          ¼          ¼         
*‰´          «´          «´          «
*‰´:          ‰´:          ‰´          ¼
*¾´          ¾´          ¾´          ¼
*Š´          Š´          Š´          s·´
*«´          «´          «´         
*«:´          «´:          «´         
*¼´          ¼´          ¼´         
*ž          ‘          ‘          ¼
*          ›·          ›·          ›·
*›          ›          ›         
*L          L          L          L
*k´          ‰´          ‰´         
*k´:          ‰´:          ‰´          ¾
*g´          ¾´          ¾´          ¾
*™´          Š´          Š´          Š
*x          x          x          x
*‘          ‘          ‘         
*kº          kº          kº         
*k:º       k:º          kº         
*gº          gº          gº         
*™º          ™º          ™º          ™º
*xº          f          f         
*‘´          ‘          ‘         
*q          q          q-,q:       q
*q:          q:          q:          ª
*»          »          »          »
*’          ’          ’         
*»´          “          “          »
*qº          qº          qº-,q:º   
*q:º       q:º          q:º       ªº
*»º          »º          »º          »º
*’º          ’º          ’º          ’º
*‚          ‚          ‚         
*“          “          “