The problem of Proto-Balto-Slavic 'aspirates' revisited ONDREJ SEFCIK ABSTRACT: The topic of this paper, as its title suggests, is the fate of the original Indo-European (IE) aspirates in Balto-Slavic, or possibly in its direct precursor, Proto-Balto-Slavic. In contrast with the Indo-European protolanguage, which is generally reconstructed with three modal classes of stops, both the Baltic and the Slavic languages are modeled on the opposition based on the feature l±voicel only, with the opposition based on the feature l±aspirationl not directly attested. Due to this distinction between IE and Balto-Slavic, it is assumed that the original opposition of aspiration was lost at some point during the Proto-Balto-Slavic period. The mechanism of this loss and the question of 'voiceless aspirates' are discussed as well. In the paper it is demonstrated that there is no reason to believe that 'voiceless aspirates' and 'voiced aspirates' ever formed a category of 'aspirates', proportional to the opposition between 'voiceless unaspirates' and 'voiced unaspirates', or to assume that both 'aspirates' ever existed at the same moment. Key words: Indo-European, phonology, aspiration, voice, stops 1. The Indo-European system of modal opposition As it is widely known, the traditional reconstruction of the Indo-European (IE) phonemic system of modal oppositions leads to a triadic or quaternary structure of four modal classes,1 based on two features: l±voicel and l±aspirationl2 (for recent examples see Szemerenyi 1996: 54; Clackson 2007: 40-44; Beekes 2011: 124).3 The given modal classes are thus: 'voiceless unaspirated' stops (schematically 7), 'voiced unaspirated' stops (D), and 'voiced aspirated' stops (Dh). Quaternary models even feature 'voiceless aspirated' stops (schematically marked Th). This paper employs the classical model and its features, since the glottalic theory never gained major acceptance.4 Furthermore, for us the fact that the reconstructed protolanguage originally had two distinctive features is far more important than the specific phonetic values of these features, since the phonetic values could be only guessed at and not objectively determined. All values are thus more of an algebraic nature than a phonetic one. The feature l±voicel has two mutually opposed values: /-voice/ and l+voicel, while the feature l±aspirationl has two opposed values l-aspirationl and l+aspirationl. 1 As modal oppositions we classify oppositions based on the manner of overcoming the obstruction of the second degree (Trubetzkoy 1939: 134 terms it Uberwindungsarteigenschafen, the English translation of 1969: 140-141 terms it in a precisely but overly wordy manner properties based on the manner of overcoming an obstruction), or in more descriptive ways, as oppositions related to the minor modifications of the sound inside the sonority scale. ^ Since we speak of phonemic values of features here (and it is impossible to speak about phonetic values anyway, since we are dealing with reconstructed languages), we prefer to use slash brackets // instead of square brackets [] for both values and features. 3 It should be noted that both Beekes and Clackson tend toward a glottalic interpretation of the Indo-European triad. 4 But later we will see that glottalic solutions may help us to solve some questions in the easiest way. 300 Slovo a slovesnost, 77, 2016 To begin with, we suppose that the value l+voicedl is a marked member of the feature l±voicel and that the value l+aspiratedl is a marked member of another feature l±aspirationl. The first value is linked to the voiced member IDI and the second value is then linked to the aspirated member IDhl. Both values l+voicedl and l+aspiratedl are the positive values of their respective features, since they are obviously manifested with some positive value. On the contrary, values like l-voicedl and l-aspiratedl are the negative values of their respective features, since they are not manifested, but rather they are deducted on a purely abstract basis. Because of their 'negativity', both values l-voicedl and l-aspiratedl are considered to be unmarked values of their features. And logically, both 'positive' values are considered marked values of their features. Note that in this model l+voicedl is an irrelevant feature for Dh, since markedness is sufficiently expressed by the value l+aspiratedl. The positive values of both features are mutually incompatible in the classic triadic model, as they cannot be present together with the same member of opposition, but are mutually compatible in the quaternary model. On the other hand, both negative values are mutually always compatible, for they are present on one member of the opposition, i.e. on the 'voiceless unaspirated' stop (= 7). Because both features are based around one single member T, which shares both unmarked values, we consider them orthogonal. The supposed traditional IE system of modal oppositions can be demonstrated graphically as follows (note that the original system is not attested in any known language): T -► D Figure 1: Traditional IE system of modal oppositions 2. Modal systems in IE daughter languages The Indo-European languages differ in the number of features used (most often l±aspirationl is omitted, as in Balto-Slavic, Iranian, etc.), in the number of members in opposition (new phonemes created), and in the neutral phoneme on which basis of the system is formed (in systems with shifts as in Germanic). All three differences may be mutually bound, as we will demonstrate. Generally, four types of modal systems in specific IE daughter languages are found: Type 1. The original IE triad is preserved in principle. This holds for Ancient Greek and Osco-Umbrian. The /+voice/ quality of *Dh is lost (*Dh > *Th), due to its irrelevance, as described earlier. Even languages that have undergone shifts still represent the old triad, as in Germanic and Armenian (for Armenian, see section 5 below). Slovo a stoves no st, 77, 2016 301 Figure 2: Preserved IE triad Type 2. The original triad is extended to a quaternary system, with a 'voiceless aspirated' stop *Th filling the gap both in opposition to *rand *Dh. This system is attested in Old Indo-Aryan. T -► D A A T T Th -► Dh Figure 3: Extended modal system of OIA Type 3. The original triad is extended by a fourth member as type 2 above, but later the opposition between *D and *Dh is neutralized and a new triadic system is derived. This type is attested in Avestan and in Iranian generally, with the aspirates later becoming fricatives. T -► D Figure 4: Extended and reduced modal system of Iranian Type 4. In the original triad, the opposition between *D and *Dh is leveled. This is valid for Balto-Slavic, but this innovation is shared with (in addition to Iranian) Albanian and the Celtic languages, too.5 A similar process is found even for Anatolian and Tocharian (where, however, all stops eventually merged, with evidence for an earlier system of type 1). A common isogloss (i.e. an areal development) could be assumed probably only for Iranian or Albanian, if it is possible to draw a common isogloss at all. T -► D ft (Dh) Figure 5: Reduced system of Balto-Slavic, Albanian and Celtic 5 At least in the Brythonic languages we can partially trace the former distinction between voiced and aspirated stops, since Welsh has b for original IE *gt and gw for original IE *gifh; Welsh bwyd 'food' vs. Gr. ßiog 'life', but Welsh gweddi 'prayer' vs. Gr. jtoQeco 'I wish' (Kortlandt 1978a: 115). 302 Slovo a slovesnost, 77, 2016 It seems clear that type 3 shares some developments both with type 2 and with type 4. We will demonstrate that the close relationship between types 3 and 4 could, in the case of Balto-Slavic, be even closer than it seems from the lines above. To reach our goal, we will first survey the undisputed reflexes of IE stops in Baltic and Slavic (section 3). We will then focus on the question of the IE 'voiced aspirated' stops (section 4) and on the question of 'voiceless aspirated' in Proto-Balto-Slavic (section 5). 3. Undisputed stops: 'voiceless unaspirateď stops and 'voiced unaspirateď stops Our reconstruction of the Proto-Balto-Slavic modal oppositions will start with the members of the system of modality about which there is no doubt concerning to their original modal class. 'Voiceless unaspirateď stops (*T) are reconstructed for the reflexes (excerpted from Pokorný 1959-1969; Rix 2001, hereafter LIV): *p: OCS petb, Lith. penki, OIA pahca-, Av. panča, Gr. Jtévre, Golh.fimf '5'; OCS pqtb 'way', Pruss. pintis, OIApántháh, Av. panta, 'way', Gr. Jtóvrog 'sea, sea route', Lat. pons 'fascine, bridge', Goth, finpan 'to find (a way)'; *t: OCS trije, Lith. trys, OIA tráyah, Gr. xpeig, Goth, preis, W. tri, '3'; OCS ty, Lith. tu, OIA t(u)vdm, Gr. Dor. xv, Lat. tu, Goth, pu, W. ti 'thou'; *k: OCS krhvb, Lith. kraujas 'blood', OIA kravis- 'raw meat', Gr. Kpéag '(raw) meat', Lat. cruor '(raw) blood', ON hrár 'raw', W. crau 'blood'; *kv: OCS četyre, Lith. keturi, OIA catur-, Gr. Horn, réaaapeg, Att. xéxxapeg, Boeot. néxxapeg, Lat. quattuor, Golh. fidwór, '4'; OCS kbito), čb(to), Lith. kas, OIA káh, Gr. Horn, xéo (gen.), Lat. qui, quod, Goth, has 'who'; *k: OCS shto, Lith. šimtas, OIA šatám, Gr. é-Kaxóv, Lat. centum, Goth, hund '100'; OCS srhdbce, Lith. širdis, Gr. Kapóía, Lat. cor, Goth, hairto 'heart'. From the reflexes as listed here we reconstruct four 'voiceless unaspirateď stops for Proto-Balto-Slavic: *p, *t, *k (for merged IE *k and *k#, some instances of it later palatalized in Slavic) and *s (if already sibilantized). Similarly, 'voiced unaspirateď stops (*D) are reconstructed on the basis of the following reflexes (again excerpted from Pokorný 1959-1969; LIV): *b: OCS boljbjb 'greater', OIA bála- 'strength', Gr. fíéXxepog, fíeXxíoov 'better', Lat. děbilis 'weak';6 *d: OCS desett, Lith. desimtas, OIA dáša, Gr. óÉKa, Lat. decern, Goth, taihun, W. deg '10'; OCS viděti, Lith. pavydéti, OIA avidat '(has) found', véda 'know', Gr. oida 'know', Lat. video, Goth, witan 'see'; *g: OCS stogh 'rick, haystack', Gr. oxéyco 'cover', Lat. tegó 'cover', ON pekja 'cover'; *g Th); B a positional variant of clusters sT (sT > > Th) - here belong the results given by Siebs' Law9 from the original clusters SDh (SDh > STh > (s)Th). 9 As formulated by Siebs (1904: 277-324). 306 Slovo a slovesnost, 77, 2016 Table 2: Reflexes for IE *Th as *Th as *T *IE A tH OIA ratha- 'chariot' YAv. rada 'chariot' Lat. rota 'wheel' Olrish roth 'wheel' OHG rad 'wheel' Lith. ratas 'wheel, pi. (arch.) wagon' *rot-H2-o- OIA pdnthd 'path' OAv. pa do YAv. panta 'path' OCS. potb 'path' Pruss. pintis 'road' Gr. Jiárog 'path' Lat. pons 'bridge' Arm. hown 'ford' *pontH2- OIA mdnthah 'twirling-stick' Lith. mentě 'twirling-stick' OCS metetb 'confuse' R mqtew 'whisk' *mentH2- OIA prthu- 'broad' OAv. paerdthu- 'broad' Gr. JTÁavúg 'broad' Lith. platiis 'broad' OR plota 'roach'10 Ru. plotica, plotvd* pltH2-u- OIA sthd- 'stand' Av. stana- 'stable' Lat. sto 'stand' OCS stati 'stand' Lith. stoti 'stand' *steH2- pH NA11 kH OIA sakhd 'branch, twig' Arm. c'ax 'branch' Ru. soxd 'plough' Cz. socha 'sculpture' SCr. sdha 'forked stick' Goth, hoha 'plough' Lith. Saka 'branch' *kokH2-o- Arm. xap'anem 'hinder' OCS. xapbJQSte (Supr.) 'biting' Ru. xapai 'grab' Cz. chdpat 'seize' Lat. capió 'I take' *kH2ep- Ru., R xuj 'penis' OIA khuddti 'fuck' Lat. cauda 'tail, trail, penis' *kHou-i/kHou-d- B st OIA sthag- 'to cover' Lat. tegó 'I cover' Gr. orsyco 'I cover' OCS stogh 'rick, haystack', ON pekja 'cover' *steg- sp OIA phena- 'foam' Lat. spuma 'foam' OEfdm 'foam' OCS pěny 'foam' Lith. spáiné 'foam' *(s)p(h)oiH-n-(~-m-) OIA sphya- 'chip' Gr. ocpqv 'I yank' OHG span 'chip' *sp(h)e- sk OCS xrab(z>)rz> 'brave' Germ, scharf 'sharp' Latv. skdrbs 'sharp' *(s)k(h)or-bh- OCS xladh R. xdlod R chlod 'cold' Goth, kalds 'cold' Lat. gelidus 'cold' *(s)g(h)el-dh- With the meaning 'broad (and thin) fish'. No sure examples of this cluster. Slovo a sloves no st, 77, 2016 307 It was the influence of de Saussure (1892: 118) and his followers, namely Pedersen (1926: 48) and Kurylowicz (1927: 202-204; 1936: 46-54), that led to rejecting the regular phonemic status of 'voiceless aspirated' in Indo-European and to the statement that the supposed 'voiceless aspirated' are the results of clusters of a stop + laryngeal (*T+H), which is now the prevailing opinion (cf. Shevelov 1964: 35-37; Arumaa 1979: 13-19; Lindeman 1997: 142-148). At least some portion of the 'voiceless aspirated' class are not the product of the assimilation of a TH cluster, but rather a positional variant of a 'voiceless unaspirated' stop in a cluster with the sibilant s (Hiersche provided a very detailed review of the relevant data, see Hiersche 1964; briefly summarized on pages 254-258, though it is debatable whether this secondary aspiration after a sibilant is not limited to Indo-Aryan only). A considerable amount of data is reviewed by Sturtevant (1941: 1-11), Hiersche (1964), Merlingen (1958) and Bičovský (2008), though we do not accept all explanations given therein. In Table 2, onomatopoeia are omitted, as are borrowings from other languages. It should be kept in mind that for some examples of a positional aspiration in clusters s7W other explanations may be possible, since here aspiration could be the result of a lost laryngeal, too. Reviewing the data above, we can see that in both Baltic and Slavic there are no traces of the supposed 'voiceless aspirated' stops in the case of IE *ph and *th. Only kh (either from original k, k or kv) is preserved in Slavic, either from an original cluster *kH or *sk, which is probably only a positional variant after s, similar to the often encountered combinatory aspiration of voiceless stops in OIA, as noted above. Two different origins for Slavic x were already noted by Kozlovskij (1888: 387). In the case of Slavic kh we observe a similar process of spirantization as in Iranian, where the newly formed 'voiceless aspirated' stops are also spirantized (Indo-Iranian kh, th, ph > x, ů, cp). This analogy with Iranian becomes even more interesting if we consider the fact that in Iranian, x is the product of a syntagmatic alternation as well, e.g. in the clusters k+t (č+t), g+t > xt (cf. Av. draoga- ~ druxta; vačo- ~ vdxs). In some Slavic dialects, as in OCS,12 š (as a front variant of x) could be the product of the alternation of a stop before *t-, cf. OCS tekq ~ testi, pekq ~ pešti, ORu. vrbgq ~ vřešti. This alternation is not known in Baltic, cf. Lith. seku ~ sekti, díegiu ~ diegti. An important influence on the status of x in Slavic came from Pedersen's Law (the ruki-mle), which increased the frequency of x/š in Slavic. The ruki-mle added more examples of the spirantized back velar, which merged with the original instances of kh of other origin (cf. Pedersen 1895: 74; Shevelov 1964: 127; Arumaa 1976: 42; Collinge 1985: 143-145; Townsend & Janda 1996: 42-45; Vennemann 1974: 91-97). 12 Note however that in other Slavic dialects, the cluster *kt is realized as c: OCz. teku ~ teci, peku ~ peci, vrhu ~ vrci. Nonetheless, even here we suppose the original cluster with some spirant, secondarily shifted in c, hence IE *kt gave first xt, which was changed into st before a palatalizing environment in OCS and into tj in other Slavic dialects, while before other vowels it was later changed into t only (hence we find on the one hand Ru., Pol, Cz. pot 'sweat' from *pok-t-, but on the other hand OCS pesti, Cz. peci from *pek-t-i, cf. Shevelov 1964: 191). 308 Slovo a slovesnost, 77, 2016 Again, it is worth noting that in Baltic, the ruki-mle is realized to a remarkably lesser degree than in Slavic, and at no time was there a back variant of š as in Slavic. For example, Lith. máišas 'sack', viršús 'hill', širšuô 'hornet', vetušas 'old' (cf. the Common Slavic equivalents méxh, vbrxh, sbršbňb, vethxh) are subject to Pedersen's Law, but this is not true of irregular ausis 'ear', teisus 'silent', sausas 'dry' (cf. the Common Slavic equivalents uxo, tixh, suxh). This could have been a negative influence for a supposed *kh in Baltic, a tendency contrary to that found in Slavic. Slavic x is thus of the following origins (cf. Thummel 1967: 71): 1. from kH clusters; 2. from sK clusters (where AT represents any velar stop), including Siebs' Law; 3. from s as a result of Pedersen's Law / ru&z'-rule; 4. minor sources like onomatopoeia, expressive variants and borrowings. The instances of x arising from these three sources (onomatopoeia and borrowings are left aside) are merged in the following way, as demonstrated in Figure 6: 1. *kH ^> k\ x/š Figure 6: Origins of x in Slavic For Balto-Slavic, the following steps in the development of the 'voiceless aspirates' could be supposed: (1) the existence of 'voiceless aspirated' stops from the clusters kH or sK; (2) the shift of the 'voiceless aspirated' kh to the voiceless spirant x in Slavic but not in Baltic; and finally (3) the dephonemization of the opposition between the 'voiceless unaspirates' and 'voiceless aspirates', which leaves x as a solitary testament to the lost 'voiceless aspirates' (see Table 3). Table 3: Development of the Proto-Balto-Slavic 'voiceless aspirates' Slavic Baltic 1. k-kh t-th p-ph 2. k-x t-th p-ph k-kh t-th p-ph 3. k-x t (<= th) P (*= Ph) k (<= kh) t (<= th) p (*= Ph) The phoneme x in Slavic thus has a solitary position after the merger of the other 'voiceless aspirated' with the 'voiceless unaspirates' (cf. Thummel 1967). Merlingen (1958: 40) points out that a solitary preserved x against the loss of the other 'voiceless aspirated' consonants is a feature that Slavic shares with Armenian, but Armenian data seem to present a more complex situation, since in Armenian almost all Slovo a slovesnost, 77, 2016 309 'voiceless aspirated' consonants are realized in the different form than the IE 'voiceless unaspirates' T (cf. Schmitt 2007: 57-79; Hiersche 1964: 232-253).13 In that regard, Armenian is more likely an example of type 2 (like Indie, but with a shift of modality), not of type 1 (like Greek, Italic or Germanic). 6. Proportionality of aspirates in Proto-Balto-Slavic and Indo-Iranian Considering all the data above, we see that for Proto-Balto-Slavic the following classes must be reconstructed: 'voiceless unaspirates', 'voiced unaspirates', 'voiced aspirates' and 'voiceless aspirates'. This is confirmed by both external parallels outside Balto-Slavic and by internal reconstruction (especially by the results of Winter's Law). From those four, only the first three are inherited from the IE period, while the fourth developed later from various clusters or positional variants of other phonemes. The class of 'voiceless aspirates' probably held a phonemic status only for a short time and during this time kh in Slavic merged with the x created by Pedersen's Law (the ruki-mle). After this merger, the 'voiceless aspirates' in both Slavic (other than x) and Baltic fused with the original 'voiceless unaspirates' T. One might ask if the 'voiceless aspirates' and 'voiceled aspirates' formed a category of 'aspirates', contrasting with the category of 'unaspirates'. In other words, was there a proportional system based on oppositions of voice and aspiration, with categories of 'voiced stops', 'voiceless stops', 'aspirated stops', 'unaspirated stops' and the four classes 'voiceless unaspirated', 'voiceless aspirated', 'voiced aspirated' and 'voiced unaspirated'? This would be a quaternary proportional system similar to those known from Modern Central Indo-Aryan languages (Hindi, Nepali, Assamese, Gujarati, Marathi, Oriya, Bengali), which can be demonstrated graphically as follows (cf. Masica 1991: 100-104,106-107): T -► D t t T T Th -► Dh Figure 7: Indo-Aryan modal system The proportional quaternary system tends to be a very stable one, since in the Central Indo-Aryan languages it has survived for millennia. This is in stark contrast to the situation in Old Iranian (Avestan, Old Persian), which originally inherited the original 13 Generally, IE * T is realized as Arm. t' (and other forms due to its position) in Armenian, similarly to * th (of different origin), cf. Lat. odd but Arm. owt' 'eight' for *T and Gr. atiXig and Arm. awt' 'night camp'. But same results for IE *T and *Th are not a standard in Armenian, since *p is realized in Armenian as zero or h (alongside numerous other realizations, cf. Lat. pater 'father', OIApdhca- 'five' and Arm. hayr 'father' and hing 'five') but *ph as Arm. p' (cf. with Gr. Xaqjuooco, OHG laffan 'to lick' and Arm. lap 'em 'I lick'); IE *k and *bf are realized as Arm. k' (cf. Lat. quam 'how, as' and Arm. k'an 'as') but ld%>h as Arm. x (cf. OIA sakha 'branch, twig', Goth, hoha 'plough', Lith. saka 'branch', Ru. soxd 'plough' and Arm. c'ax 'branch'). It should be kept in mind, however, that examples are scarce and often doubtful. 310 Slovo a slovesnost, 77, 2016 IE triad as well, and developed (or inherited) a new class of 'voiceless aspirates' like Old Indo-Aryan. Later, the original Indo-European class of 'voiced aspirates' was lost in Iranian, similarly as in Balto-Slavic. As demonstrated in Sefcik (2012), the reason for this is that in Indo-Aryan, the system of oppositions of stops became proportional at least in the Middle Indo-Aryan period, but in Iranian and Balto-Slavic this never happened, there never was a category of 'aspirated stops' with a corresponding category of 'unaspirated stops'. The proposed ideal model for covering both Balto-Slavic and Indo-Iranian (Late Eastern Indo-European) is thus quaternary, but without proportionality between the oppositions of 'unaspirated stops' T- D and 'aspirated stops' Th - Dh. The lack of this proportionality is due to the differing phonetic status,14 which resulted from the different origins of Th. The model of the Late Eastern IE retains the original IE triad, but outside of it there exists a solitary opposition of Th and T, not proportional with Dh and D (cf. Figure 8). Figure 8: The modal system of Late IE The relative independence of the T-Th opposition from the old triad is demonstrated by the merger of Dh with D in Balto-Slavic and Iranian: had the system ever been proportional, this could never have happened, as it indeed did not happen in Indo-Aryan. The truth is, we do not have any reasons to suppose that the model shown in Figure 8 ever existed for Balto-Slavic. We suppose that the model is adequate for Indo-Iranian, but here it could be used as a panchronic working model, which we will modify later. Here we can mention again Kortlandt's interpretation of Winter's Law. For him (Kortlandt 1978a: 91-112; 1978b: 447; 1979: 57-63; 1985: 112-124; 1988: 387-396; 2011: 245-250) the IE triad, in the moment when Winter's Law was active, was not the traditional triad *T— *D — *Dh, but rather *T- *'D — *D, which is close to a glottalic interpretation of the IE triad.15 However, if we accept this system for Balto-Slavic, it is clear that there were no phonetic reasons to form a quaternary proportional system, since a value /+aspiration/ was no longer attached to the original IE *Dh. If we consider the place of *Th in Kortlandt's glottalic model, the final form could then be formed as *Th - *T- *'D - *D, without any proportionality between the 'aspirates'. Such an application of Kortlandt's model is even less stable than that described in Figure 8, though both models are inherently unstable and tending to be restructured, which is what indeed happened. 14 Hiersche strongly doubts the aspirated value of Th (Hiersche 1964: 254). 15 It worth of note that Winter himself repeatedly rejected 'glottalic' interpretation of his law, stating that lengthening before voiced stops occurs elsewhere without glottalization present (Winter 1979; Winter 2011). T D Slovo a stoves no st, 77, 2016 311 The different languages in which 'voiceless aspirates' occurred developed different strategies for how to stabilize the system. The following developments occurred: A Indo-Aryan made the system proportional, setting Th as the fourth member of the old triad and hence expanding it into a quaternary system with the categories 'aspirated stops' vs. 'unaspirated stops'. B Iranian spirantized all phonemes of the Th class and, following the merger of Dh with the D class, created a new triad. C Baltic merged D and Dh and, in addition, abandoned any phonemic distinction between T and Th (if there ever was any at all), directly creating a binary system. D Slavic merged D and Dh, abandoning a phonemic distinction between T and Th for all pairs with the exception of k - xls proportional to the pair t - s. Again, if Kortlandt's interpretation of Proto-Balto-Slavic triad is correct, a merger of IE *D and *Dh (Kortlandt's *'D - *D) would be even easier. This leads to the question of a precise phonetic value of Th in Late Eastern IE.16 We know that in Indo-Aryan, the new fourth member was accepted as a 'voiceless aspirate', but we have no proof that in Iranian or Proto-B alto-Slavic it was ever anything other than a voiceless spirant. We could even argue that, instead of spirantization of the 'voiceless aspirate' stops in Iranian and Proto-B alto-Slavic, we could equally say that the same modal class was despirantized in Indo-Aryan and the status of spirant in Iranian and Slavic (and Armenian, too) is the original value of Th. The situation in Baltic could then be equally considered as deaspiration or despirantization - again, if a distinction between T and Th phonemized ever existed in Baltic, as noted above. 7. 'Aspirates' in Balto-Slavic For the development of both 'aspirates' in Balto-Slavic, we propose the following stages (see Table 4): Table 4: Development of Proto-Balto-Slavic obstruents Slavic Baltic 1. im - D - Dh 1 TH 1 s 2. T- 'D-D-Th \ s 3. T — D — Th \ s 4. T-D-Th/x \ s, s/x11 T-D-Th\s, 518 5. T — D 1 x 1 s, s/x T-D 1 s, s 16 An anonymous reviewer of this article offered an interesting solution for the phonemic nature of secondary 'voiceless aspirates' in Proto-Balto-Slavic, considering them not aspirates or spirants, but TH clusters. Such an idea is highly innovative and worth considering pro futuro. 17 The question of whether the original result of the rwfa'-rule in Slavic was x or s, is left aside. 18 For simplicity's sake, we shall not speak here about the results of the development of the IE palatovelars. 312 Slovo a slovesnost, 77, 2016 1. The last stage in the development of Indo-European (Late Eastern IE), with an original triad *T, *D, *Dh, TH clusters, a probably positional aspiration of voiceless stops after a sibilant s7W and one sibilant *s. 2. The post-laryngeal phase, with the original IE triad *T, *D, *Dh later shifted to *T, *'D, *D, newly established *Th (either 'voiceless aspirates' or as 'voiceless spirants') from both original clusters with laryngeal and positional variation, original s preserved. Slavic shares this development of *Th with Indo-Iranian, Armenian and Greek. 3. The Winter's Law phase, lengthening before original 'voiced unaspirated' stops and for lost preconsonantal laryngeals, the merger of both IE *D (BS1. *'D) and *Dh (BS1. *D) - this isogloss is common with Iranian - and thus finally *T, *D, *Th, *s. The situation is then similar to the state attested in Iranian (cf. Figure 4). 4. A 'Pedersen's phase': Slavic merged original kh either from *kH or *sK with x resulting from Pedersen's Law, but the process did not affect the Baltic languages. We suppose that at least the independent existence of x (of any origin) has to share a common period of existence with x from Pedersen's Law, otherwise it is hard to imagine the merger of both xs into one. 5. The final phase is marked by the loss of a distinction between 'voiceless unaspirated' and 'voiceless aspirated' stops, which completes the process. Note that in the proposed model, it is implicitly stated that 'voiced aspirates' and 'voiceless aspirates' never existed at the same stage of development of the Balto-Slavic languages. This is probably the easiest solution for the proposed lack of proportionality in Balto-Slavic between both traditional classes of 'aspirates', though at stage 2 we can model a quaternary model, based on a modified panchronic ideal model as presented in Figure 8. LIST OF ABBREVIATIONS OF LANGUAGES Arm. Armenian Lat. Latin ORu. Old Russian Av. Avestan Lith. Lithuanian Pruss. Prussian Bceot. Boeotian Latv. Latvian P Polish Cz. Czech MIA Middle Indo-Aryan Phryg. Phrygian Dor. Doric OAv. Old Avestan Ru. Russian Germ. German OCS Old Church Slavic/Slavonic SCr. Serbo-Croatian Goth. Gothic OCz. Old Czech Sin. Slovene Gr. Greek OHG Old High German Toch. Tocharian Hit. Hittite OIA Old Indo-Aryan (Vedic) W. Welsh Horn. Homeric ON Old Norse YAv. Young Avestan IE Indo-European OP Old Polish REFERENCES Arumaa, P. (1976): Urslavische Grammatik, 2: Konsonantismus. Heidelberg: Carl Winter Universitätsverlag. Beekes, R. (2011): Comparative Indo-European Linguistics: An Introduction. Amsterdam, Philadelphia: John Benjamins. Bicovsky, J. (2008): Initial *x- in Slavic revisited. Chatressar 2008, 23-45. Brugmann, K. (1922): Kurze vergleichende Grammatik der indogermanischen Sprachen. Berlin, Leipzig: De Gruyter. Slovo a slovesnost, 77, 2016 313 Clackson, J. (2007): Indo-European Linguistics. Cambridge: Cambridge University Press. Collinge, N. E. (1985): The Laws of Indo-European. Amsterdam, Philadelphia: John Benjamins. Derksen, R. (2002): On the reception of Winter's law. Baltistica 37 (1), 5-13. Derksen, R. (2008): Etymological Dictionary of the Slavic Inherited Lexicon. Leiden: Brill. Elbourne, P. (1998): Proto-Indo-European voiceless aspirates. Historische Sprachforschung 111, 1-30. Elbourne, P. (2000): Plain voiceless stop plus laryngeal in Indo-European. Historische Sprachforschung 113, 2-28. Etymologický slovník jazyka staroslověnského. Praha: Academia, 1989-2008 (vol. 1-14); Brno: Tribun EU, 2010-2014 (vol. 15-17). Fraenkel, E. (1962-65): Litauisches Etymologisches Wörterbuch. Heidelberg: Carl Winter Universitätsverlag. HiERSCHE, R. (1964): Untersuchungen zur Frage der Tenues Aspiratae im Indogermanischen. Wiesbaden: Otto Harrassowitz. Jasanoff, J. (2004): Plus ga change...: Lachmann's law in Latin. In: J. H. W Penney (ed.), Indo-European Perspectives: Studies in Honour of Anna Morpurgo Davis. Oxford: Oxford University Press, 405-416. Kortlandt, F. H. H. (1978a): Proto-Indo-European obstruents. Indogermanische Forschungen 83, 107-118. Kortlandt, F. H. H. (1978b): Comment on W. Winter's paper. In: J. Fisiak (ed.), Recent Developments in Historical Phonology. The Hague: Mouton, 447. Kortlandt, F. H. H. (1979): Three problems of Balto-Slavic phonology. Zbornik zafilologiju i lingvistiku 22 (2), 57-63. Kortlandt, F. H. H. (1985): Long vowels in Balto-Slavic. Baltistica 21 (2), 112-124. Kortlandt, F. H. H. (1988): Remarks on Winter's law. In: A. A. Barentsen, B. M. Groen & R. Sprenger (eds.), Dutch Contributions to the Tenth International Congress of Slavists, Sofia (= Studies in Slavic and General Linguistics 11). Amsterdam: Rodopi, 387-396. Kortlandt, F. H. H. (1989): Lachmann's law. In: T. Vennemann (ed.), The New Sound of Indo-European. Berlin: Mouton de Gruyter, 103-105. Kortlandt, F. H. H. (1999): Lachmann's law again: Language change and typological variation. In: E. C. Polome & C. F. Justus (eds.), In Honor ofWinfred P. Lehmann on the Occasion of his 83rd Birthday, 1: Language Change and Typology. Washington: Institute for the Study of Man, 246-248. Kortlandt, F. H. H. (2011): Winter's law again. In: T. Pronk & R. Derksen (eds.), Accent Matters: Papers on Balto-Slavic Accentology. Amsterdam: Rodopi. 245-250. Kozlovskij, I. (1888): Zur Geschichte des slavischen Consonantismus. Archiv für slavische Philologie 11, 383-395. KuRYLOWicz, J. (1927): a indo-européen et h. hittite. In: W. Taszycki & W. Doroszewski (eds.), Sym-bolae grammaticae in honorem Ioannis Rozwadowski, 1. Krakow: Gebethner & Wolff, 95-104. Kurylowicz, J. (1936): Etudes Indo-Europe'ennes. Krakow: Gebethner &Wolff. Lindeman, F. O. (1997): Introduction to the Laryngeal Theory. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Masica, C. P. (1991): The Indo-Aryan Languages. Cambridge: Cambridge University Press. Merlingen, W. (1958): Idg. x. Die Sprache 4, 39-73. Pedersen, H. (1895): Das indogermanisches s im Slavischen. Indogermanische Forschungen 5, 74-87. Pedersen, H. (1926): La cinquieme declinaison latine. K0behavn: A. F. H0st. Pokorný, J. (1959-1969): Indogermanisches etymologisches Wörterbuch. Bern: Francké. Rasmussen, J. E. (1987): On the status of the aspirated tenues and the Indo-European phonation series. Acta Linguistica Hafniensia 20, 81-109. Rix, H. (ed.) (2001): LIV: Lexikon der indogermanischen Verben. Wiesbaden: Reichert. Saussure, F. de (1892): Contribution ä l'histoire des aspirées sourdes. Bulletin de la Société de Lin- guistique 7, 118. Schmitt, R. (2007): Grammatik des Klassisch-Armenischen mit Sprachvergleichenden Erläuterungen. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. 314 Slovo a slovesnost, 77, 2016 Shevelov, G. Y. (1964): A Prehistory of Slavic: The Historical Phonology of Common Slavic. Heidelberg: Carl Winter. Siebs, T. (1904): Anlautstudien. Zeitschrift für vergleichende Sprachforschung (= Kuhns Zeitschrift) 37, 277-324. Smoczyňski, W. (2007): Slownik etymologiczny jezyka litewskiego. Wilno: Uniwersytet Wileňski. Sturtevant, E. H. (1941): The Indo-European voiceless aspirates. Language 17 (1), 1-11. Sukač, R. (2012): Lachmann's law (part 1). Linguistica Brunensia 60 (1-2), 13-36. Szemerenyi, O. (1996): Introduction to Indo-European Linguistics. Oxford: Clarendon. ŠefčIk, O. (2012): On the integration of Old Indo-Aryan 'voiceless aspirated' into the system. In: R. Sukač & O. Šefčík (eds.), Sounds of Indo-European 2. München: Lincom Europa, 271-280. Thümmel, W. (1967): Der phonematische Status des urslavischen x. Münchener Studien zur Sprachwissenschaft 21, 71-91. Townsend, C. E. & Janda, L. A. (1996): Common and Comparative Slavic: Phonology and Inflection. Columbus: Slavica. Trubetzkoy, N. S. (1939): Grundzüge der Phonologie (= Travaux du Cercle linguistique de Prague, 7). Prague: Jednota československých matematiků a fysiků. Trubetzkoy, N. S. (1969): Principles of Phonology. Berkeley: University of California Press. Vasmer, M. (1953-1958): Russisches etymologisches Wörterbuch. Heidelberg: Carl Winter Universitätsverlag. Vennemann, T. (1974): Sanskrit ruki and the concept of a natural class. Linguistics 130, 91-97. Winter, W. (1978): The distribution of short and long vowels in stems of the type Lith. ésti: věsti: městi and OCSjasti: vesti: mesti in Baltic and Slavic languages. In: J. Fisiak (ed.), Recent Developments in Historical Phonology. The Hague: Mouton, 431-46. Winter, W. (1979): Zur 'Überlänge' im Deutschen. In: K. Ezawa & K. Rensch (eds.), Sprache und Sprechen: Festschrift für Eberhard Zwirner zum 80. Geburtstag. Tübingen: Niemeyer, 197-200. Winter, W. (2011): Vowel lengthening before distinctively voiced consonants in Tocharian. Tocharian and Indo-European Studies 12, 221-238. RÉSUMÉ Problém protobaltoslovanských „aspirát" po revizi Otázka počtu modálni ch tříd v indoevropských jazycích neztratila dosud své pálčivosti. Jak známo, zatímco převládající popis pracuje s třemi modálními třídami (neznělých, znělých a znělých aspirát), některé modely stále pracují i se čtvrtou třídou (neznělých aspirát), případně přehodnocují situaci z pohledu různých variant glotální teorie. Baltoslovanské jazyky pak v historickém období mají jen dvě třídy (neznělou a znělou, s tou splynula i historická aspirovaná). Díky vnitřní rekonstrukci (s využitím Winterova zákona) dnes můžeme i pro protobaltoslovanskou fázi vývoje rekonstruovat tři modálni třídy, nicméně stále zůstává otázka neznělých aspirát (autor tohoto příspěvku rozhodně považuje tzv. neznělé aspiráty za sekundární, nikoliv obecně indoevropské), které v slovanštině (ale ne baltštině) v některých případech reprezentuje *x/š. V příspěvku se ukazuje, že není nutné předpokládat, že by neznělé a znělé aspiráty někdy spolu tvořily kategorii aspirovanosti proporční kategorii neaspirovaných, jak je tomu např. v indoárijských jazycích, ba ani časový souběh existence znělých a neznělých aspirát. Department of Linguistics and Baltic Studies, Faculty ofArts, Masaryk University A. Nováka 1, 602 00 Brno, Czech Republic Slovo a sloves no st, 77, 2016 315