Shortening, Lengthening, and Reconstruction: Notes on Historical Slavic Accentology

Introduction1

0.

Frederik Kortlandt has, in one of his articles (Kortlandt 2016: 478–479), criticized a few random issues out of my recent monograph (Kapović 2015), while misinterpreting some of my stances. I responded to his criticism inKapović 2017a and he responded back in Kortlandt 2018. In this article, I will respond to him once again, while taking the chance to discuss some issues in historical Slavic accentology, since I have to admit I do not actually believe that there “is simply no viable alternative to the theory of Slavic accentuation that [Kortlandt] proposed 45 years ago” (Kortlandt 2018: 295). However, I would like to thank my co-discussant for providing me with an opportunity to present and further elaborate some of my views. I hope that other scholars will find our discussion useful and interesting, despite occasional heavy words and heated argumentation.

Before we start with the actual issues, a few notes on the modus operandi and rhetorical strategies of the other discussant are in order. Kortlandt’s response is written in his usual style. He often does not really criticize other scholars’ views with arguments and discussion – he just dismisses them out of hand. He also tends to ignore the data that do not fit into his theories and rarely discusses anything outside of his framework. Likewise, instead of actual discussion he has a tendency to denounce certain ideas as a product of other scholars’ ignorance or as “outdated”, even in case of ideas held by most scholars outside of his own school of thought.

In his article, Kortlandt constantly accuses me of my “lack of a chronological perspective”, which seems to be a code for “not accepting Kortlandt’s ideas on relative chronology”. I can indeed confess that I have neither tried to present my complete version of a prosodic relative chronology from Proto-Indo-European to Slavic, nor to criticize the totality of Kortlandt’s ideas on relative chronology of historical Slavic accentuation – that would take a whole monograph, and while it could perhaps be interesting and useful, it is nonetheless my view that there are much more important issues still to be solved in historical Slavic accentology than criticizing views of just one scholar. My aims were always quite modest – I merely attempted to show that some of Kortlandt’s theories, parts of his grand relative chronology scheme, simply do not work, are not convincing, or do not actually explain the data. Before trying to fit one’s ideas into a neat chronological perspective, one must be sure that the theories actually explain the facts – the problem with at least some of Kortlandt’s ideas is that they do not or that they do it much more poorly than the alternative explanations. All the chronological perspectives in the world cannot fix disregarding of data and implausible analogies. While relative chronology is indeed important, it is not everything and the data and obvious explanations should not be twisted in order for them to nicely fit a preimagined wider hypothesis. One can take the example of the Moscow Accentological School (MAS), which does not deal with relative chronology at all. Not because it is not important but because the main aim of MAS is paradigmatical reconstruction of morphologic and derivational categories. MAS scholars compile enormous amounts of data and work with full sets of words (e.g. of various types of verbs, derivatives, etc.), trying to reconstruct Proto-Slavic accentual paradigms for all of them. That is hardly less valuable than operating with a few chosen words (repeated in paper after paper) in order to try to make an elaborate relative chronology.

The reflex of the Proto-Slavic short neo-acute in Kajkavian

1.

When discussing Kajkavian forms like õsmi ‘eighth’ and rešẽta ‘sieves’ (Kortlandt 2016: 475), it is a shame that Kortlandt did not take into account new analyses from my book (Kapović 2015), because it might have been useful for him to acknowledge the lengthening of the short neo-acute in the positions Ivšić (1936: 72) has missed – most importantly loc^sg stõlu ‘table’ < *stòlu2 (Kapović 2015: 379, 384–386) and also derivatives like sẽlce ‘hamlet’ < *sèlьce, in opposition to no lenghtening in forms like smȍkva ‘fig’ < *smòkъv- (ibid. 390–396).3 Had he done so, perhaps he would not need to reconstruct such a complex and unnecessary system like the supposed Proto-Slavic distinction of *è/ò : *ẽ/õ : *ⁱè/^uò (in my opinion, it is enough to reconstruct *è and *ò only) or to make rather dubious claims that Kajkavian forms like nȍsim ‘I carry’ (a. p. B) are “evidently analogical” (Kortlandt 2016: 476). That is not only not “evident” but is very unlikely: it would entail a staggering analogical development in the Kajkavian accentual paradigm B where the whole present nȍsim – nȍsiš – nȍsi – nȍsimo – nȍsite – nȍse (cf.Kapović 2018: 230–231), in all six persons in both singular and plural, would have a secondary accent! To make things worse, this supposedly innovative accent (nȍs- in all forms) would be due to analogy to the sole original but not attested 1^sg *nošȕ (and perhaps to other non-present forms like the imperative nosȉte! and the like) (cf. alsoKapović 2017b:611). How likely is it that a single form would influence five other forms in such a way (or eight if one considers the now extinct dual as well)? To make things even worse, this original *nošȕ (and other forms like the imperative), did not even have nȍs- but only unaccented nos-. So one would have to assume that the original *nošȕ – **nõsiš – **nõsi – **nõsimo – **nõsite – **nõse somehow magically changes to nȍsim – nȍsiš – nȍsi – nȍsimo – nȍsite – nȍse without any remnants of the “original” forms and through a very suspicious process (**nõs- > nȍs- by analogy to *nos-). How can this be “evidently analogical”? Unfortunately, Kortlandt does not even discuss this “evident analogy”, thus leaving a false impression on an uninitiated reader that this supposed analogy is indeed easy and reasonable. Kortlandt very often avoids discussing the details and problematic issues. He usually leaves out discussion and argumentation even of his own hypotheses (together with counterexamples and most data that do not fit his theories), which results in obfuscation of the problem and leaves many uninitiated readers unable to judge for themselves if a certain hypothesis is convincing or not. I point to a similar case inKapović 2017a: 395 , the fact that Kortlandt believes that the present tense forms 1^pl *nòsimo – 2^pl *nòsite in Slavic a. p. b are actually supposedly analogical to 2^sg *nòsišь – 3^sg *nòsitь – 3^pl *nòsętь, but he never puts it in so many words, nor does he explain why there are no traces anywhere of the supposedly original **nosîmo > **nosìmo – **nosîte > **nosìte (one finds only forms like ložĩ – ložĩmo/ložīmȍ ‘we make fire < lay’ in a. p. b₂ in some dialects, but those have a neo-acute, cf.Kapović 2017b: 395). To get back to Kajkavian nȍsiš ‘you carry’, there is nothing “evidently analogical” about it – it is a perfectly expected reflex of Slavic *` in Kajkavian in initial/medial syllables in a position before a non-contractional length (cf.Kapović 2015: 380–381). The same reflex is also seen in Kajkavian forms like mȍreš ‘you can’ < *mòžešь or nȍšen ‘carried’ < *nòšenъ (both with a short following syllable). The difference between Kajkavian nȍsiš and gen^pl nõvih ‘new’ is in the contractional origin of -i- in the latter form (*nòvyjixъ),4 where the contractional length was originally probably super-long at the time of the lengthening of the short neo-acute in Kaj­kavian (and Slovene),5 cf.Kapović 2015: 388–389).

Pretonic and posttonic length in Slavic

2.

In my last article (Kapović 2017a: 385–386), I pointed out numerous problems with Kortlandt’s unorthodox theory that the length in a. p. c forms like Neo-Štokavian rúka ‘arm’ is secondary because it was supposedly shortened originally:

The generalization of brevity in a. p. c of e.g. ā-stems in West Slavic (Czech ruka – acc^sg ruku) would entail only the generalization of the original accent in forms like *rǫ̑kǫ, for which there are ample typological parallels (cf. the footnotes inKapović 2017a: 383), or the generalization of a short vowel, which is in any case expected in the great majority of cases – in 17 of 21 (Kapović 2015: 429,2017a: 383). Thus in Czech (and West Slavic in general), it is a simple question of a complete generalization of *rŭk- instead of the older alternation of *rŭk- (17x) and *rūk- (4x), or a simple complete generalization of *rȗk- (which originally occurred in 9 of 21 case forms) in all the cases. In Štokavian/Čakavian/Kajkavian, Kortlandt’s supposed reintroduction of length in a. p. C is much more complicated and unconvincing because it would have entailed the change of **rŭkȁ to *rūkȁ by analogy to *rȗku (thus a lengthening of a short unaccented vowel by analogy to a long accented vowel), while forms like dat^pl **rŭkȁm would for some reason remain unaffected (because there was no complete generalization).7Kortlandt’s (2018: 290) solution that the length is not restored “in polysyllabic word forms” is an attempted description, not an explanation. First of all, why was the supposed analogy always and everywhere perfect in disyllabic forms, while it did not affect polysyllabic forms?8 Secondly, Kortlandt’s (not really explanatory) description is imprecise because it would be a stretch to claim that Čakavian dat^pl rŭkȁm, loc^pl rŭkȁh (which obviously have the same kind of shortening as Neo-Štokavian rùkama)9 are “polysyllabic” and that would entail counting the final yers in order for this to work. Additionally, Kortlandt’s disyllabic/polysyllabic terminology is confusing and imprecise in one more way – aren’t forms like Čakavian planīnȁ ‘mountain’ or lovīmȍ ‘we hunt’, which one would expect to behave like Čakavian rūkȁ, also polysyllabic? And isn’t gen^sg rŭkẽ also disyllabic?

Kortlandt’s (2018: 290) response to these critiques is surprising. Though one would expect him to defend his theory and try to offer some kind of justification for the raised objections or take the opportunity to provide some details for his theory, he does no such thing. This is what he has to say. First of all, there is the mantra of my supposed “lack of chronological perspective”, though the problem is not in chronological perspective but in Kortlandt’s unconvincing massive reintroduction of pretonic length in a. p. C. Then he reiterates once more where the supposed restoration of length in a. p. C comes in his relative chronology (though that is not the issue) and says that I am disregarding the difference between a. p. b and c (which makes no sense at all). And that’s it – no real explanation is provided.

Another problematic issue is the one of e-presents in a. p. c (Kapović 2017a: 386–387). According to my theory, both the West Slavic and Western South Slavic attested paradigms are very easy to understand. In Czech, the whole paradigm shows a short root (třesu – třeseš – třese – třeseme – třesete – třesou) because this is what you would expect in the majority of original forms (1^sg, 1/2/3^pl) (ibid. 386). The oldest attested paradigm type in Western South Slavic is preserved in some Čakavian dialects: (trēsȅn) – trēsȅš – trēsȅ – trĕsemȍ – trĕsetȅ – trĕsũ.10 This is exactly what one would expect (cf.Kapović 2015: 417–419). However, in Kortlandt’s framework, the archaic Čakavian paradigm is again difficult to explain.Kortlandt (2018: 290) states that the restoration of length in trēsȅš – trēsȅ but not in trĕsemȍ – trĕsetȅ “is a consequence of the fact that pretonic length was limited to the first pretonic syllable”. If this has something to do with pretonic length in word forms which experienced Dybo’s law, where there was indeed no pretonic length outside of the first pretonic syllable, how is then one to explain forms like Štokavian zábava ‘fun, party’ and národ ‘people’ with Dybo’s law (as perKortlandt 2018: 289) but no length in forms like rùkama or dialectal loc^pl rŭkȁh in a. p. c? If one could find length in new forms (created by Dybo’s law according to Kortlandt) like zábava < zābȁva and národ < nārȍd in the system, why was there no restoration of length in the same positions in a. p. c (rùkama and rŭkȁh)?

Of course, when talking about the present tense paradigm Kortlandt does not mention the inconvenient fact that the primary source (i.e. one of the present tense forms) for the supposed restoration of length in trēsȅš – trēsȅ can only be the early disappearing and unattested old 1^sg *trȇsu,11 which makes the whole thing suspicious (again, as in the case of rūkȁ, one would have to assume the analogy of a pretonic vowel to an accented one). Kortlandt conveniently leaves out one more crucial point: according to his doctrine one would expect not trēsȅš – trēsȅ but rather **trĕsẽš – **trĕsẽ,12 where one would presumably expect the short pretonic vowel as in 3^pl trĕsũ. The length would thus have to be restored from the unattested and early eliminated 1^sg *trȇsu to 2^sg **trĕsẽš – 3^sg **trĕsẽ but not to 3^pl trĕsũ. Though one could obviously claim that pretonic length was reintroduced in the singular only, it would be strange that the forms with the short root in 2/3^sg are not attested anywhere (unlike 3^pl trĕsũ). Likewise, would it not be strange that no dialect generalized the secondary **trĕsẽm by analogy to the supposedly original **trĕsẽš – **trĕsẽ? In any case, even if one is to accept Kortlandt’s analogies as plausible, we still remain with a number of forms in pretonic a. p. c syllables that have to be explained by analogies instead of regular phonetic developments. If one has to choose which development is secondary and which is original in a. p. c – the shortened pretonic syllables in West Slavic or the long pretonic syllables in Western South Slavic, the latter seems a better bet, since the required generalizations are much simpler in West Slavic (see above) and remnants of the old pretonic length in a. p. c do exist there (Czech třásti, gen^sg devíti), while the required restoration of pretonic length in Western South Slavic is very messy and problematic and there are no traces whatsoever of the supposedly original forms like **rŭkȁ with shortened pretonic length in a. p. C.

In his works, Kortlandt has maintained the claim that there are no traces of pretonic length in a. p. c in West Slavic. Curiously, he usually ignored the length that occurs in West Slavic a. p. c infinitives of e-verbs like Slavic *orstı̍ ‘to grow’ (cf. Štokavian rásti) > Czech růsti, Slovak rásť, Polish róść; Slavic *tręstı̍ ‘to shake’ (cf. Štokavian trésti, Slovene trẹ́sti) > Czech třásti, Slovak, triasť, Polish trząść, Slovincian třḯsc, etc. (for more such infinitives in West Slavic cf.Stang 1957: 153,Kapović 2015: 431). While he had previously skipped Czech třásti with no real explanation (Kortlandt 2011: 264), I am glad to see that he has now (Kortlandt 2018: 290) accepted my suggestion (Kapović 2015: 431–432,2017: 384) that from his perspective this length might be explained as analogical to the l-participle (Czech třásl, Slovak triasol, Polish trząszl). However, as I have already pointed (ibid.), this kind of explanation might work in the case of verbs with stems ending in a consonant (like *tręs-tı̍, *orst-tı̍), which had b-forms in l-participles (*trę͂slъ) but not in the case of verbs with stems ending in a vowel13 (like *klętı̍ ‘to swear’, *mertı̍ ‘to die’), which had regular c-forms in l-participles (*klę̑lъ, *mь̑rlъ) (cf.Kapović 2018: 171–172 for the apparent – > + metatony in l-participles of verbs with stems ending in a consonant). Obviously, Kortlandt cannot accept that because it would undermine his theory of phonetic shortening in pretonic syllables in a. p. c14 since in the case of Czech klíti ‘to swear/curse’ – klel and mříti ‘to die’ – mrel there is no ad hoc solution to explain away the length in the infinitives through an analogy to the length in the l-participle. So how does he deal with these crucial word forms?

Kortlandt (ibid.) says: “The retraction of the stress in Czech klíti ‘to swear’ and mříti ‘to die’ was much earlier (stage 4.4)”, thus introducing a retraction without any additional explanation or argumentation on why this supposed change would have occurred except in order to be an ad hoc “solution” for problems with his theory of pretonic length. It is unclear how such a retraction could have taken place in infinitives like *klętı̍, *mertı̍, but not in forms like *rǫka̍ ‘arm’ or loc^sg *žьlčı̍ ‘bile’ (especially considering the fact that the infinitive ending *-t-i has the exact same historical origin as loc^sg *-i of i-stem nouns).

In a. p. c, the end-stress in *klętı̍ and *mertı̍ is expected (cf.Дыбо 1981: 233, 235),15 just like in *tręstı̍, and directly attested in Štokavian kléti, mrijéti and older Russian (Stang 1957: 152) yмретú, взятú ‘to take’.16 Kortlandt does claim some kind of relation of this supposed retraction with “S/Cr. vȉti ‘to twist’, grȉsti ‘to bite’, sjȅći ‘to cut’” (ibid.), but this makes no sense since these forms underwent Hirt’s law, a known and almost universally accepted phonetic law (though details of it are disputed), which occurs in word forms with acute pretonic syllables (as traditionally conceived), while forms like *klętı̍, *mertı̍, *ętı̍ ‘to take’ do not have acute roots (as seen by their reflexes). While not bothering to elucidate on his position, Kortlandt repeats the mantra of my supposed “lack of chronological perspective” (i.e. not accepting Kortlandt’s theories). However, the problem is that all West Slavic infinitive forms with length are completely expected in traditional theory (cf. alsoStang 1957: 153), while in Kortlandt’s theory forms like Czech klíti, mříti cannot be explained in the same way as třásti and he offers no acceptable solution for the problem. Thus, one must assume that these infinitives indeed preserve the old pretonic length of a. p. c (which was ousted in most other forms in a. p. c by analogy to word forms where it was not phonetically expected). Kortlandt’s additional and implausible ad hoc sound laws in order to preserve his general doctrine are unnecessary and superfluous.

FollowingStang (1957: 88), I maintain that Slavic *dȅvętь ‘nine’, *dȅsętь ‘ten’ (a. p. c) had the archaic and original end-stress in gen/loc^sg *devętı̍ and *desętı̍ (attested directly in Russian gen/loc^sg девятú, деcятú), and that the pretonic length of these original forms is directly attested in Czech gen/loc^sg devíti, desíti (cf. alsoCarlton 1991: 209), which would be one of the few traces of the original preserved pretonic length in a. p. c in West Slavic (in most other forms, this length was analogically removed). The old end-stress of gen^sg *devętı̍ and *desętı̍, disappearing analogically in nominal i-stems like gen^sg *kȍkoši ‘hen’ (cf.Stang 1957: 87–88, alsoДыбо 1981: 25, 28, 30), is confirmed by the Lithuanian end-stress in gen^sg -iẽs in mobile nouns (a. p. 3/4). However, these straightforward reflexes of Czech devíti, desíti are opposed byKortlandt (2018: 290) because they are not in accord with his theory that pretonic length in Slavic phonetically shortens.

I am glad to see that Kortlandt (ibid.) has accepted my suggestion that Štokavian dȅvet, dȅset beside dȅvēt ‘nine’, dȅsēt ‘ten’ are a result of allegro shortening (together with possible analogy to sȅdam ‘seven’, ȍsam ‘eight’), because his own explanation of the shortening was hardly convincing.17 However, there is no need to assume allegro shortening for West Slavic (Czech deset, Slovak desať, Polish dziesięć) as Kortlandt does, since posttonic length is clearly phonetically shortened in West Slavic in a. p. c (see below). We find a similar problem with the way Kortlandt (ibid.) explains Czech gen^sg (also dat/loc/instr^sg)18 devíti, desíti, by positing that the length is preserved from the barytone forms – again, the view that West Slavic preserves posttonic length in a. p. c is patently wrong (as shown below). Thus, Czech devíti, desíti cannot be explained from *dȅvęti, *dȅsęti (the forms expected in dat^sg). In any case, it would be strange, even if old posttonic length was indeed preserved in West Slavic (which clearly it is not), that it would disappear in Czech nom/acc^sg devět, deset but not in oblique devíti, desíti.19

Kortlandt (e.g. 2011: 30) believes that, unlike pretonic length (which is supposedly always shortened), posttonic length (in disyllabic words like *msęcь ‘month’, *ȍbvolkъ ‘cloud’, not counting final yers)20 is originally preserved both in West and South Slavic, in both a. p. a (in words like *msęcь) and a. p. c (in words like *ȍbvolkъ). To explain the actual reflexes, Kortlandt (ibid.) has to assume vast analogies in both South and West Slavic a. p. C. Thus, Štokavian old pretonic length (which should be shortened according to Kortlandt) in loc^sg obláku (ARj) would supposedly be analogical to posttonic length in ȍblāk ‘cloud’ – gen^sg ȍblāka, etc., while Czech short second syllable in oblak ‘cloud’ (which should be long according to Kortlandt) would be analogical to original end-stressed forms like loc^sg oblaku21 (where the pretonic length should shorten according to Kortlandt). However, in my view both Štokavian gȍlūb – gen^sg gȍlūba and Czech holub – gen^sg holuba have completely regular phonetic reflexes of posttonic length in most cases22 and no analogies are needed. While posttonic length is consistently preserved in Štokavian/Čakavian no matter the original accentual paradigm23 (cf. the examples like mjȅsēc, gȍlūb inKapović 2015: 504–506, 509),24 West Slavic shows clear difference in the treatment of posttonic length in a. p. a and a. p. c. In a. p. c, the length is always shortened (cf. e.g. Czech25 čeleď ‘family’, čelist ‘jaw’, deset, devět, holub ‘pigeon’, kolovrat ‘spinning wheel’, oblak ‘cloud’, oblast ‘region’, obruč ‘hoop’, rozum ‘sense’, tetrev ‘capercaillie’, žalud ‘acorn’ + předivo ‘spinning material’, pečivo ‘pastry’)26 – a few cases with length are easily explained as secondary (cf. e.g. Czech jeřáb ‘partridge’, ovád ‘horsefly’).27 However, in a. p. a one finds numerous examples of preserved posttonic length (cf. e.g. Czech měsíc ‘month’, zajíc ‘hare’, pavouk ‘spider’, peníz ‘penny’, tisíc ‘thousand’, participles like řezán ‘cut’, jestřáb ‘hawk’, personal names like Branimír, Slovak pavúz ‘shaft for pressing hay in a carriage’) but also many of them with shortening (Czech havran ‘raven’, labuť ‘swan’, paměť ‘memory’, kapraď ‘fern’, vítěz ‘winner’).28 Thus,Kortlandt (2018: 290–291) is not correct when saying that “posttonic long vowels were consistently preserved in accent paradigm (a)” (also, the fact that the length is preserved in a. p. a does not automatically mean that it was originally the same with a. p. c, as he claims, and that only later the length was lost in a. p. c in West Slavic due to analogy to end-stressed forms). They are indeed preserved in many words, but hardly consistently. This clear difference of the reflexes of posttonic length in a. p. a (length preserved in slightly more than half of the words/forms) and a. p. c (length always shortened except in a few usually clearly secondary words) would point to the regular phonetic reflexes originally being different in a. p. a and a. p. c in West Slavic.

My view is that posttonic length was originally phonetically shortened in a. p. c in West Slavic (but not in the posttonic length preserving area of South Slavic, i.e. in Štokavian/Čakavian). This not only explains the material with no need for analogies, but also provides a unitary theory for the reflexes of both a. p. c polysyllabic root forms like Czech oblast ‘region’ < *ȍbvolstь and a. p. c monosyllabic root forms like Czech vlast ‘country’ < *vȏlstь (cf. Štokavian ȍblāst ‘region’, vlȃst ‘rule, reign’), that would have to be completely separated if one is to accept Kortlandt’s theory.29 The rule in West Slavic would be simple – the length is always shortened in accentless word forms, as a. p. c forms with initial circumflex are usually described. All forms with an automatic phonetic initial circumflex in phonologically accentless words (thus both the *vȏlstь and *ȍbvolstь type) would have been shortened – regardless of whether original long vowels (as *-ol- in *volstь) were under phonetic initial circumflex stress (as in *vȏlstь) or not (as in *ȍbvolstь).30

How to explain the almost haphazard reflexes (e.g. Czech peníz but havran) of posttonic length in a. p. a forms in West Slavic? One may obviously consider the analogical influence of polysyllabic forms (where the length would be expectedly shortened)31 in declension, like Czech (secondary) dat/loc^sg havranovi, instr^sg havranem, (secondary) nom^pl havranové, gen^pl havranů, dat^pl havranům, loc^pl havranech, or in derivation, like Czech pamětný ‘memorial’, pamětlivý ‘mindful’, pamětník ‘witness’, etc. However, such an otherwise possible explanation perhaps does not look too convincing in light of a rather good agreement within West Slavic of the forms with a short suffix (Czech/Slovak havran ~ Polish gawron; Czech/Slovak labuť ~ Polish łabędź; Czech paměť ~ Slovak pamäť ~ Polish pamięć; Czech kapraď ~ Slovak papraď ~ Polish paproć). Shortening by analogy to longer forms would probably result in more variety in West Slavic reflexes, i.e. not all West Slavic languages would have the short reflex in exactly the same words. Thus one should perhaps entertain a Moscow Accentological School type of explanation via the valences of the suffixes. If one is willing to accept valences as a phonetic reality at the time of this shortening,32 an explanation of the length being preserved only in dominant suffixes becomes possible (cf. the detailed analysis of the suffixes inKapović 2015: 508). This would then be in agreement with the shortening of length in recessive suffixes in a. p. c as well.

Kortlandt (2018: 291) attributes the lack of length in some a. p. a forms (like Czech labuť ‘swan’) to an early shift of a. p. a to a. p. c.33 This is ad hoc but not impossible. However, one should note that that explanation also works better in my framework because I posit the phonetic shortening of posttonic syllables in a. p. c, while he needs not only the shift from a. p. a to a. p. c but also the subsequent analogy of posttonic long vowels to the shortened pretonic vowels (in other cases). Similar goes for Czech pekař ‘baker’ and rybář, for whichKortlandt (2011: 266) says I dismiss them “without discussion”. If the first one is originally a. p. c (which would be strange because although *pek- is indeed a recessive root, *-arь is nonetheless a dominant suffix,34 though Štokavian pȅkār would at least formally account for such a reconstruction)35 and the other a. p. b, the forms are self-explanatory in my model (the shortening of posttonic length in a. p. c and the preservation in a. p. a in West Slavic) with no analogies needed. In Kortlandt’s framework both work as well, but pekař again needs an analogy to forms with -ař- in pretonic position. As always, Kortlandt’s theories need much more analogies, though he tries to present the picture otherwise. In any case, Kortlandt’s theory ignores a clear difference in the phonetic treatment of posttonic length in a. p. a, with most but not all words preserving the length (some 8 words or types of forms36 with length and 5 with a shortened suffix), and a. p. c, with almost all words showing a shortened suffix (14 a. p. c words with the shortening and 3 with length, most easily explained as secondary),37 while also disregarding the connection of words such as Czech vlast and oblast and the same kind of shortening occurring in all-recessive (“phonologically accentless”) word forms. Again, his theory is substandard in explaining the material.

According to Kortlandt’s theory, a. p. b verbs like *xvaliti ‘to praise’ or *skakati ‘to jump’ should preserve the original length of the root in all positions, because the accent was still on the root (*xvãliti, *skãkati) at the time of the supposed general shortening of pretonic length (in a. p. c). The length is indeed what we see in major Slavic languages in these positions, cf. e.g. Štokavian hváliti, Czech chváliti and Štokavian skákati, Czech skákati. However, as I have warned repeatedly (most recently inKapović 2017a: 387–388), this does not work for the original reflexes in West Slavic and the adduced infinitives have actually reintroduced the secondary length by analogy to the present tense forms like Štokavian 2^sg hvȃlīš, skȃčēš (dialectal and older hvãlīš, skãčēš), Czech chválíš, skáčeš. The original shortened root in a. p. b is seen in Old Polish infinitives like sędzić ‘to judge’, przystępić ‘to approach’ and żędać ‘to demand’38 but not in the present forms 2^sg sządisz and przystąpisz.39 Crucially, the problem for Kortlandt’s theory is Slovincian, which, unlike other Slavic languages with preserved quantitative distinctions or remnants of it, has regular short vowels in a. p. b infinitives (but a long one not only in the present but in the l-participles as well),40 e.g. Slovincian sʒĕc ‘to judge’ – present 2^sg sʒïš – l-participle sʒĕl, stpjĭc ‘to step in’ – present 2^sg stpjïš – l-participle stpjĕl (cf. the complete list of such verbs inДыбо 2000: 91–92). The same is true of Slovincian a-je-verbs like kpac ‘to bathe’ – present 1^sg kpją (cf. the list inStang 1957: 42 and alsoStankiewicz 1993: 315–316, 318 for both). The shortened root in the imperative (which behaves like the infinitive as it also has a dominant *-i-) of a. p. b can be found in the Czech (Moravian) Hanakian dialect:41 imperative 2^sg młať! – 2^pl młaťte! but present 2^sg młátíš and l-participle młátił (the infinitive młátiť has a secondary length probably by analogy to the original supine).42 Further evidence for such a distinction in the treatment of infinitive/imperative (shortened root) and l-participle (preserved length in the root) can possibly be found in the Middle Bulgarian manuscript Apostle from the 14^th century, where the stress seems to have moved only to the dominant *-i-:43 infinitive сѫдúти, imperative cѫдúте! but l-participle сдилъ.

Instead of addressing the Slovincian data (or for that matter Hanakian or Middle Bulgarian),Kortlandt (2018: 291) chooses to address the three Old Polish verbs in detail, as if the problem was only those three verbs and not the whole system in Slovincian (supported by other data). Old Polish examples only corroborate that Slovincian system is indeed archaic and that West Slavic in general originally had the same quantitative alternation. In his discussion of the Old Polish sędzić – sządisz and przystępić – przystąpisz, Kortlandt says (ibid.) that the “short root vowel (…) offers a serious problem for the theory that these verbs belong to accent paradigm (b)”. There is no point in a detailed critique of Kortlandt’s rather elaborate ad hoc schemes, using which he tries to explain these forms that are very problematic from the point of view of his theory, since there is absolutely no reason why these two verbs should be considered anything else than a normal a. p. b. Neither Old Polish nor anything else in Slavic merits such wild theories to explain these two verbs – the only reason not to reconstruct the usual *sǫdı̋ti (a. p. b) and *stǫpı̋ti (a. p. b) is that their reflexes in Old Polish do not fit well with Kortlandt’s doctrine on length in Slavic. These two verbs belong to a normal a. p. b without any doubt and have always been reconstructed as such – cf.Дыбо 2000: 441–442 (for Slavic in general),Kapović 2011b: 210–211 (for Croatian dialects) and *sǫ͂dъ ‘court of law’ (a. p. b).

Still, I would like to address a methodological point concerning Kortlandt’s struggle to explain Old Polish sędzić – sządisz and przystępić – przystąpisz because they are a good example of his modus operandi. When he cites (ibid.) the data from the Čakavian dialect of Kukljica, it is interesting to note that he does not mention where it is taken from. He mentions neitherBenić 2011 norBenić 2014, where the data can be found. This is not only scholarly inappropriate, since he is not giving proper respect to the fieldworker but is also methodologically incorrect since it does not allow the reader to check the data himself and assess it in the context. Unfortunately, this kind of erasure of other accentologists is quite in line with Kortlandt’s general suppressing of most modern non-Leiden scholars in most of his works (outside of his usually very scathing reviews of new accentological works that are not written within the framework of the Leiden Accentological School), which I have already mentioned before.44 It is methodologically incorrect to look at one or a couple of forms without context, as Kortlandt tends to do, like in the question of the Kukljica forms such as budȋn – prebȗdin ‘I wake up’ (Benić 2011: 7). Kortlandt completely ignores the known process in Čakavian (and Štokavian) through which the old long a. p. c verbs first yield a mixed a. p. C-Cː, and then later shift to either a completely shortened a. p. C (rarely) or to an innovative a. p. Bː (Kapović 2011b: 228–231,2015: 477–488). The forms in question are probably connected to the poluotmetnosť phenomenon as well (the prefixal accent, cf. Дыбо, Замятина and Николаев 1993: 43–51,Kapović 2018: 218), however the historical process of the a. p. C(-Cː) > Bː cannot be left out of the picture in a serious treatment of the problem (cf. especiallyKapović 2015: 487–488).

To get back to the topic,Kortlandt (2018: 291) does not completely ignore the Slovincian data. He does mention two Slovincian a-je-verbs and explains the short root in the infinitive through the original short ablaut in them (cf. OCS pьsati ‘to write’, dъxati ‘to breathe’). However, he conveniently ignores the rest of the examples where no such explanation is possible (like the cognates of OCS lizati ‘to lick’, skakati ‘to jump’, *kǫpati ‘to bathe’), just like he ignores the i-verbs in general. He does, on the other hand, falsely attribute (Kortlandt 2018:291) the claim that these different stem formations have something to do with dominant and recessive suffixes. They do not and I have never said that they do: it is inappropriate of Kortlandt to discuss a-je-verbs like *pьsa̋ti – *pĩšešь only and then make reference to my paper where I discuss i-verbs (i-verbs that he otherwise completely ignores). By the way, the explanation of the West Slavic (and Middle Bulgarian) forms through the earlier progressive accentual shift to the dominant acute syllables (as explained inKapović 2017a: 291) is, of course, not mine but of the Moscow Accentological School (Дыбо, Замятина and Николаев 1993: 7–9,Дыбо 2000: 92–93), that Kortlandt has not commented on for a quarter of a century now, even though the West Slavic data goes against his whole theory on the development of pretonic length in Slavic.

To conclude, Kortlandt’s theory is again inadequate when dealing with real data. Unfortunately, the way he seems to deal with obvious shortcomings of his doctrine is to focus on irrelevant details (coming up with complex and unnecessary hypotheses out of thin air in the process) like the three Old Polish verbs, present the problematic data partially (as in the case of a-je-verbs in Slovincian), and ignore most of it (the Slovincian i-verbs). However, serious historical accentology cannot be based on just a few chosen forms. One has to look at the whole system.

The genitive plural

3.

FollowingДыбо (2000: 21), I interpret the lengthening in gen^pl (in Western South Slavic and West Slavic – though the situation in the latter is not as clear) of all accentual paradigms – cf. dialectal Croatian (Štokavian/Čakavian/Kajkavian) gen^pl krȃv ‘cows’ (from krȁva, a. p. A), gen^pl žẽn ‘women’ (from ženȁ, a. p. B), gen^pl võd ‘waters’ (from vodȁ, a. p. C) – as caused by the old ending *- < post-Proto-Indo-European *-ōm (cf. more details with references inKapović 2017a: 389–390). The advantage of such a theory is that it not only explains the reflexes in all three accentual paradigms phonetically and in a simple manner (a long ending that has a tendency to drop, at least in certain conditions, lengthens the preceding syllable), but also simply and convincingly explains otherwise aberrant endings – Neo-Štokavian -ā (Old Štokavian -ã) and Slovene -á (cf.Kapović 2015: 537–540).

Kortlandt’s explanation, on the other hand, is not phonetically regular for all accentual paradigms and involves analogies, some of which are completely unbelievable. In his view (e.g.Kortlandt 2011: 54), only võd (C) is phonetically regular. Gen^pl žẽn (B) is analogical to võd, which is not impossible although it introduces an unnecessary analogy, but the real problem is krȃv (A) (the same in Slovene, cf. also Czech krav from kráva) where the supposed analogy, which Kortlandt does not really explain, looks very strange. It is difficult to grasp how krȃv can be analogical to võd. Likewise, the old long *- can explain length in posttonic syllables of a. p. a forms like gen^pl jȁgōd ‘strawberries’ phonetically (again, a long ending with a tendency to disappear in certain conditions lengthens the preceding syllable), while in Kortlandt’s theory those also have to be analogical.

Additionally, what is problematic is his explanation that the supposedly only phonetic lengthening, the one in võd (C), is due not to the old *- but to the retraction of the accent from the old *vod – Kortlandt believes that retraction of stress from a weakening yer causes regular lengthening of a preceding short vowel. While I agree that such a retraction causes regular phonetic lengthening in the case of a resonant (or at least *m and *v) preceding the final yer (cf. the lengthening in the dat^pl – Old Štokavian sinovõm ‘sons’, Slovene možẹ́m ‘husbands’,45 Old Kajkavian (Pergošić) lughoom ‘groves’46 and the lengthening in gen^pl like Old Štokavian/Čakavian/Kajkavian sinõv ‘sons’, Slovene sinv)47 I do not find the cases with non-resonants convincing. Cf. the lack of the supposed retractional lengthening in forms like Štokavian jȅst ‘is’ < *jest (cf. also Czech jest), (secondary) Štokavian aorist rèkoh ‘I said’ < *rekox (cf. also Old Czech řečech ‘I said’,48 vedech ‘I knew’), and Štokavian 2^sg tréseš ‘you shake’ < *tręseš (cf. also Czech třeseš). While one could imagine that Štokavian jȅst is analogical to jèsi ‘thou are’ (however, Czech jest can hardly be analogical to jsi and other forms in js-) and rèkoh to rèkosmo ‘we said’, rèkoste ‘you said’ (where no lengthening is possible), the same is not true for forms like tréseš. Though forms like trésēš (older/dialectal trēsẽš, cf. also Slovak trasieš) do indeed exist (cf. Kapović 2015: 367–370), they can be easily explained as analogical to the 2^sg lovĩš ‘you hunt’ type, and the final -ẽš, -ẽ can hardly be original since in that case, as we have already mentioned in the paper, one would expect a shortened root vowel in Western South Slavic (i.e. **trĕsẽš, **trĕsẽ), which is not attested anywhere. Thus, the case for a general retractional lengthening from a falling yer seems to be weak and, in any case, it is hardly necessary to explain the lengthening in genitive plural forms.

Kortlandt (2018: 292) first says that “Kapović still adheres to the outdated view that the Proto-Indo-European gen.pl. ending was *‑ōm, for which there is no evidence”. However, Kortlandt actually misrepresents my position. I take the short *-om (not *-ōm) to be the original Proto-Indo-European gen^pl ending49 and consider the long *-ōm a post-Proto-Indo-European contractional development in o-stems (*-o-om)50 and eh₂-stems (*-eh₂-om), which is then generalized in certain languages. This is hardly an unusual position – for an overview of different reconstructions and problems concerning the Proto-Indo-European gen^pl cf. e.g.Olander 2015: 255–257, 261–265.51 Still, negating the existence of Proto-Indo-European *-ōm with harsh rhetorics52 is hardly a proper way to go when evaluating Slavic evidence for the reconstruction of *-. This long *- seems to have yielded the same results as the plain *-ъ in most Slavic dialects (e.g. in East Slavic), but has left accentual (the mentioned lengthening of the krȃv, žẽn, võd type), and more rarely morphological (Old Štokavian -ã, Slovene -á) traces in Western South Slavic and the Czech-Slovak area of West Slavic.53

As concerns the Štokavian gen^pl ending -ā,Kortlandt (2018: 292) simply says “The S/Cr. ending ‑ā does not continue an original long jer but was introduced on the analogy of the loc.pl. ending of the i‑ and u‑stems”, adding references that elucidate his opinion on the matter. It seems that Kortlandt believes that Old Štokavian -ã (-ьь in 14^th century Old Serbian texts54) and Slovene -á, though they look as the same ending, are not actually genetically connected in any way.Kortlandt (1978: 286) apparently followsOblak (1890: 439–440) for the explanation of Slovene -á andJohnson (1972: 349–358) for the explanation of Štokavian -ā – neither of those explanations are either convicing or simple.

Oblak (ibid.) believes that the (standard) Slovene gen^pl variant ending -á (gorá together with gr ‘mountains’) somehow originates in analogy to dat^pl -am, loc^pl -ah, instr^pl -ami. It is completely unclear how a new ending can spring right out of thin our air by analogy to a the first syllable of other endings. While this explanation may seem formally satisfying at first glance, it is difficult to imagine how this kind of analogy would actually work. What is more, this analogy does not explain the accent of Slovene -á. If this is an analogy to dat^pl -ȁm, loc^pl -ȁh, instr^pl -ȃmi, why does -á have a completely different accent? How is that motivated? Now, an easy solution would be to say that -á has the accent by analogy to the i-stem gen^pl -í, but where is the motivation for that? Why would gen^pl gorá be made by analogy to gen^pl kostí ‘bones’? Where is e.g. the structural similarity of gen^pl gorá – dat^pl gorȁm – loc^pl gorȁh – instr^pl gorȃmi and gen^pl kostí – dat^pl kostm – loc^pl kosth – instr^pl kostmí? Both the endings and accents in dat/loc/instr^pl of ā- and i-stems are different. Neither Oblak nor Kortlandt provide answers to these questions. Now, one of the reasons why Oblak (ibid.) thinks that the it is “schwerlich zu glauben” in the genetic relation of Štokavian -ā and Slovene -á is the fact that the Slovene ending is attested only from the 17^th century. This is indeed a valid point – however, if one is to take that Slovene -á derives only from *- (thus in a. p. c), only in certain positions (originally in trisyllabic forms – see below) and only in some dialects (in others, *- may have phonetically just disappeared like *- and *-ъ), this becomes far less suspicious. In any case, if one is to believe that Slovene -á is somehow secondary (in spite of the fact that Oblak’s hypothesis looks very unconvincing and does not take accent into account at all), then one should also believe the following:

Can all of that be just a coincidence? Certainly, but not very likely.

Johnson’s (1972: 349–358) scenario for a secondary origin of Štokavian -ā is very similar to Oblak’s Slovene scenario and has similar problems. His basic claim (ibid. 356) is that Štokavian new gen^pl *žen-ā (dat^pl *ženam – loc^pl *ženah – instr^pl *ženami) somehow analogically mirros i-stem gen^pl *kost-ī (dat^pl *kostim – loc^pl *kostih – instr^pl *kostimi56). Again, the analogy in ā-stems internally is very abstract and it is hard to picture it and the analogy to the less frequent i-stems is also doubtful. Again, neither Johnson (ibid.) norKortlandt (1978: 286) explain the accentual details (i.e. how one gets from the older *glãv to younger *glāvã, etc.), which are highly problematic. If one derives the final -ã from the old *-, the form glāvã is more or less a direct reflex and easy to understand (disregarding here the length of the root and the disyllabic form – see below). But if one starts from the original glãv and adds a secondary -ā of whatever origin, it is not so easy to explain the actually attested glāvã > glávā. Yes, glāvã could perhaps be analogical to the i-stem gen^pl vlāstĩ ‘governments’ or gen^pl slūgũ ‘servants’ (though slŭgũ is older) and perhaps instr^sg glāvõm, but that kind of analogy is hardly simple and it is strange that no dialect shows **glãvā > **glȃvā from a supposedly direct combination of glãv plus the secondary -ā. Now, if one wants to put the ā-stems in correlation to the i-stems as Johnson did,57 gen^pl glávā ‘heads’ (Cː) can theoretically be, as already said, analogical to the i-stem vlástī ‘governments’ (Cː), while trávā ‘grass [pl]’ (Bː) can be analogical to glávā (Cː). However, vódā ‘waters’ (C) is not in accord with i-stem kòstī ‘bones’ (C) and krȃvā ‘cows’ (A) is also different from i-stem smȑtī ‘deaths’ (A) (there is no root lengthening in the i-stem a. p. A). Thus, the accent in long a. p. Cː should be analogical to i-stems (*glāvã in analogy to *vlāstĩ) in order to get the actually attested forms, but in short a. p. C the long root of *vōdã must not conform to the i-stem *kostĩ (the same in a. p. A) in order for the forms to work. All this is perhaps not impossible but does not look very convincing.58

The major problem with Johnson’s hypothesis is that, even if one takes his -a by analogy to -am/-ah/-ami suggestion as possible, Štokavian -ā would be an original *a, not a yer. Now this contradicts not only Old Serbian -ьь but also Montenegrin dialectal -ǣ(h), where [æ] points to *ь/ъ and not to *a. Kortlandt (1978:297) does not mention -ьь and tries to relativize -ǣ(h) by pointing to some secondary forms where æ occurs for old *a (cf.Милетић 1940: 225–239 for the historical origin of æ in the Montenegrin dialect of Crmnica). However, certain secondary forms in some Montenegrin dialects hardly dispute the fact that /æ/ generally derives from *ъ/ь. Korltandt (ibid.) explains -ǣ(h) by claiming “the vowel timbre of the gen. pl. ending must be derived from the original loc. pl. endings of the i- and u-stems, which contained a jer”, which is one more complication in his explanation.59

In my view (cf.Kapović 2015: 539), the original *- in gen^pl (of o- and ā-stems),60 stemming from the older *-ōm (< Proto-Indo-European *-o-om and *-eh₂-om), behaved as the short *-ъ in some Slavic areas (e.g. in East Slavic). However, in South Slavic (with traces mostly in Western South Slavic)61 and West Slavic the ending *- lengthens preceding syllables (the situation is less clear in West Slavic than in South Slavic). The ending *- itself always disappears in West (and East) Slavic, just like the short *-ъ, and the same may easily be true for Kajkavian and Čakavian (and perhaps even for some Štokavian dialects) in South Slavic. In Štokavian and Slovene the old *- disappears in most cases, but not in all of them. It is originally preserved when under accent (thus in a. p. c only) in trisyllabic forms,62 where it yields *-. This variant ending can then disappear in some dialects, linger on as remnant in others (as a variant in Standard Slovene, e.g. gen^pl gorá together with gr ‘mountains’), while in some it may eventually secondarily and gradually spread to become the main ending in gen^pl of o- and ā-stems (as is the case in most Štokavian dialects). The original reflexes must have been something like the following63 (nouns with the originally short root are adduced for a. p. b and c to indicate the lengthening):

Kortlandt (1978: 286) claims that Štokavian -ā cannot be old because “Final -ъ was lost in the Serbo-Slovenian dialectal area as early as the tenth century, while the gen. pl. ending -ā appeared in Serbo-Croat in the 14th and in Slovene in the 17th century.” The argument with the loss of the final short *-ъ is hardly compelling since the very essence of the theory we have laid out here is that the long final *- had a different reflex in some positions (when stressed in a. p. c in trisyllabic forms) in some (Štokavian) dialects. The time of attestation of the ending in Štokavian and Slovene is also hardly a problem, since the material in general is very scanty – so much so that some (e.g.Svane 1958: 80) believe that attempts at explaining the ending -ā are doomed to fail from the very start. If one is to originally start with a complex gen^pl ending distribution (*krȃv – *žẽn – *planin) in some (Štokavian) dialects only, why would it be strange that -ā is attested from the 14^th century (the same goes for Slovene mutatis mutandis)? It is simple enough – the ending -ā was attested when it started to spread (both internally in the system – at the expense of -Ø – and through diffusion between different dialects). Kortlandt’s (ibid.) point that “the rise of the medial syllable in SCr. sestárā, otácā would remain unexplained if -ā continued the Proto-Slavic ending -ъ” is valid (cf. the same objection inOblak 1890: 439), but only if one is to assume that *- yielded -ā generally. If one is to take that -ā is the phonetic reflex only in some positions,68 one can just assume the original forms such as gen^pl *metl ‘brooms’ < *metъl with the same development as in *võd < *vod.69 The ending -ā in forms like metálā would then be secondary, just like in jȁgōdā, vódā, etc.

To conclude, Kortlandt’s doctrine presupposes a number of analogies (some of which, like gen^pl krȃv supposedly by analogy to võd, or gen^pl vódā supposedly by a very abstract and strange analogy, seem completely implausible) and bases itself on dubious claim of the general retractional lengthening in Slavic (which is not really supported by the data, especially by the 2^sg trēsȅš type), while my theory explains the root lengthening in the genitive plural phonetically and additionally provides a rather simple phonetic interpretation of the Štokavian/Slovene -ā as well.70

The accentual type bȏg/kȍkōt

4.

As I have shown clearly and in detail (Kapović 2015: 231–233, 627,2017a: 391–394), in Western South Slavic the old short circumflex is lengthened in monosyllabic roots (plus a final yer), thus *bȍgъ ‘god’ > Štokavian/Čakavian/Kajkavian bȏg, Slovene bg. The same lengthening occurs in polysyllabic forms in Štokavian/Čakavian with an initial circumflex in a. p. c, where the unaccented last syllable of the root is also lengthened (Štokavian/Čakavian kȍkōt ‘rooster’) – this lengthening cannot be seen in Kajkavian/Slovene due to the early disappearance of posttonic length in them and/or progressive shift of the circumflex (cf. Slovene kokt). Though the change is subsequently somewhat obfuscated by later changes (like preresonant lengthening in Čakavian or the secondary kȁmēn ‘rock’ type in Neo-Štokavian),71 it is clear that the lengthening in the kȍkōt type cannot be separated from the lengthening in the bȏg type – both occur in a. p. c only, in forms with initial circumflex only, and regardless of the final consonant or declension (cf. the i-stem type mlȁdōst ‘youth’). The lengthening in the Štokavian/Čakavian kȍkōt type is in line with the general lengthening tendency in a. p. c forms with initial accent, not only with (Western South Slavic) *bȍgъ > bȏg and the already mentioned Slovene (and marginally Kajkavian and Čakavian) *kȍkotъ > kokọ̑t type, but also with the (Western South Slavic) *sto > stȏ ‘hundred’ type.72

That the lengthening in the kȍkōt type is the same as the one in the bȏg type is also clear when one considers Štokavian polysyllabic forms like pȍmōć ‘help’ (gen^sg pȍmoći) with monosyllabic forms like mȏć ‘power’ (gen^sg mȍći) – acc^{sg ­} pȍ ̮mōć ‘for the power’.73 The lengthening in the polysyllabic pȍmōć and the prepositional phrase pȍ ̮mōć is obviously exactly the same phenomenon. Otherwise, one should have to, disregarding the Occam’s razor rule to an utmost extreme, interpret the length in pȍ ̮mōć as analogical to the form mȏć (which is unnecessary), while the length in pȍmōć would have to be, unconvincingly, something completely different.

Kortlandt pushes the lengthening of *bȍgъ further back into the history and fails to see the obvious connection of this type to the kȍkōt type lengthening.Kortlandt (2018: 292) explains that when “Dybo’s law shifted the stress to the following syllable (…) yielding long falling vowels in opposition to short and long rising vowels in non-initial syllables, the tonal opposition on short vowels became limited to monosyllables, e.g. *bȍgъ versus *kònjь. This anomalous distribution was resolved by lengthening short falling vowels in monosyllables (…), resulting in the same opposition between short and long rising versus long falling vowels that existed in non-initial syllables (...)”. The problem with this explanation is that it obliges one to accept certain questionable views. The first one is that it is not at all clear that all *` (and *˜) not preceding a yer shifted to the right, e.g. in forms like 2^sg *mòžešь ‘you can’, *nòsišь ‘you carry’, definite adjective *nòvъjь ‘new’, etc. (cf.Kapović 2017a:390). The second one is that Kortlandt’s theory only works if one accepts that the result of the rightward shift onto a long non-acute vowel yields a new long falling accent (e.g. **nosîšь), which is bogus (see the next section of this paper). The supposed Common Slavic *bȍgъ > **bȏgъ is a nice symmetrical explanation from the point of view of Kortlandt’s doctrine, but it is set upon questionable or clearly false presumptions and does not change anything substantial since the reflexes in all Slavic languages remain the same (*bȍgъ and **bȏgъ would yield exactly the same reflex in all languages). Thus, there is no point in reconstructing **bȏgъ instead of the obvious *bȍgъ. There was no lengthening in Common Slavic due to Dybo’s shift in order to set the same kind of the supposed tonal distinctions in initial as in medial syllables. The *bȍgъ > bȏg and *kȍkotъ > kȍkōt lengthening was a later Western South Slavic innovation due to a simple compensatory lengthening caused by the fall of final yers, which occurred in all “phonologically accentless” forms in a. p. c74 with an original final yer (except in instr^sg forms like bȍgom in a. p. C, where it probably disappeared early by analogy to popȍm ‘priest’ in a. p. B and rȁtom ‘war’ in a. p. A). Kortlandt (ibid.) has nothing to say on the lengthening in the kȍkōt type except that it is supposedly “a more recent development of analogical origin that did not reach all S/Cr. dialects and has nothing to do with the lengthening in bȏg.” He adds nothing to prove that it is “more recent”, nothing to prove that it is “of analogical origin” (analogical to what exactly?), he remains vague about the change supposedly not reaching “all S/Cr. dialects” (though it is clear, as I have shown, that it is a trait of all Štokavian/Čakavian dialects,75 while it cannot be present in Slovene/Kajkavian, which have no posttonic length and/or have progressive shift of the circumflex), and he offers no argumentation on why the length in kȍkōt supposedly has nothing to do with the length in bȏg. However, an ad hoc claim that a phenomenon is “more recent” and “of analogical origin”, without any argumentation to back it up, is hardly valid in a serious linguistic discussion. Pages of careful detailing of the kȍkōt lengthening in Štokavian and Čakavian and discussing of data from different local dialects cannot just be dismissed without a shred of evidence – even if such claims would not produce strange assumptions like the one that pȍmōć and pȍ ̮mōć (also mȏć) have completely different origins of the length in the second syllable.76

However, what is astonishing is the following claim (Kortlandt 2018: 292): “The length in S/Cr. gȍspōd ‘lord’, kȍkōt ‘rooster’, kȍkōš ‘hen’, mlȁdōst ‘youth’, bȍlēst ‘illness’, gȍvōr ‘speech’, kȍrēn ‘root’, plȁmēn ‘flame’, jȁblān ‘poplar’ beside gȍspod, kȍkot, kȍkoš, mlȁdost, bȍlest, gȍvor, kȍren, plȁmen, jȁblan (…)”. This is simply factually wrong on a very basic level. First of all, there is no “beside” – there are no old variants with these words. As I have shown (Kapović 2015: 231–233,2017a: 392–394) the lengthening of the unstressed syllable in forms ending in a yer in a. p. c is completely regular and expected in Štokavian and Čakavian (as already said, Slovene/Kajkavian lost posttonic length early and is thus irrelevant in this regard). There are no variants like gȍspŏd, kȍkŏt, kȍkŏš, mlȁdŏst, bȍlĕst, gȍvŏr, kȍrĕn, plȁmĕn, jȁblăn anywhere in Štokavian/Čakavian, except in dialects without posttonic length (or with late phonetic loss of posttonic length in some positions)77 or in very infrequent cases of clear later analogies. The famous Čakavian dialect of Vrgada (Jurišić 1966,1973) is, for instance, a very rare and exceptional example of a dialect with a late analogical loss of length in o-stems like gȍvor ‘speech’. Kortlandt seems to believe that this is widespread, but that is not the case. The Vrgada forms are exceptions and, more so, easily explainable exceptions. The Vrgada forms like gȍvŏr could be simply interpreted as analogical to all other cases (like gen^sg gȍvora),78 but were probably more influenced by the original a. p. A pattern.79 That this is a late and simple analogy in Vrgada is proven by i-stems, where the expected lengthening in the nom/acc^sg is found in all cases like kȍkōš ‘hen’ – gen^sg kȍkoši (cf. the analysis inKapović 2017a: 393–394), which is not accidental but is connected to the preservation of the original mobile a. p. C pattern in i-stems and the early loss of old polysyllabic a. p. a i-stems.80 All Štokavian dialects (including literary Neo-Štokavian) that preserve posttonic length regularly have forms like kȍkōt – gen^sg kȍkota in o-stems (*kȍkotъ, a. p. c)81 and kȍkōš – gen^sg kȍkoši in i-stems (*kȍkošь, a. p. c),82 cf. the lack of length in Standard Neo-Štokavian forms like pȁžen ‘taken care of’ < *pa̋ženъ (a. p. a) or nȍšen ‘carried’ < *nòšenъ (a. p. a), which point to no lengthening outside of a. p. c (i.e. outside of forms with an initial old circumflex).83 As my careful analysis of Čakavian has shown, the kȍkōš – gen^sg kȍkoši type in i-stems is found in all Čakavian dialects that preserve posttonic length phonetically. Except for Vrgada, which has lost all gȍvōr type lengths analogically, all other Čakavian dialects show the length in o-stems as well (I have provided numerous examples of that inKapović 2010: 88,2015: 232⁸⁴²,2017a: 393), but the data is not easy to find in the sources because most Čakavian dialects have phonetically lost posttonic length, while those that did not often have preresonant lengthening in posttonic syllables, which makes forms like gȍvōr irrelevant. This leaves only kȍkōt ‘rooster; sea robin’, gȍspōd ‘Lord’, and trȍskōt ‘some kind of weed/grass (e.g. knotgrass)’84 of the more known o-stems not ending in a resonant to show the original a. p. C lengthening. However, gospod seems to be missing or is at least not attested in many Čakavian dialects (this is not an everyday word even in the standard language), while kokot85 and troskot86 are often attested in dialects that have not preserved distinctive posttonic length. Less frequent a. p. C forms not ending in a resonant like grȍhōt ‘loud laughter’ or trȍpōt87 are only rarely attested because dictionaries of Čakavian local dialects often, quite irritatingly, prefer to attest Romance loanwords (or “unusual” words of Slavic origin, like those not present in the standard language) instead of the “normal” inherited Slavic lexicon.88

To conclude the point, there are no relevant old short variants “gȍspod, kȍkot, kȍkoš, mlȁdost, bȍlest, gȍvor, kȍren, plȁmen, jȁblan” that Kortlandt adduces (and thus wrongly informs the uninitiated reader) – in Štokavian, these are always clear later local developments and in Čakavian they are, as we far as we know, limited to the dialect of Vrgada due to a specific innovative process in the o-stems (while the Vrgada i-stems preserve the expected a. p. C mobile pattern together with the expected quantitative alternation). In Štokavian, there are even such cases as in the dialect of Dubrovnik, where one can without any doubt show that in the past the old a. p. C quantitative alternation was present, but that it was subsequently analogically lost in some words – e.g. modern Dubrovnik kȍkŏt ‘rooster’ now has the short second syllable by analogy to the oblique cases like gen^sg kȍkota, but in older Dubrovnik dialect the old and expected kȍkōt – gen^sg kȍkota is attested, as written down by not one but four earlier scholars of the dialect: della Bella, Rešetar, and Bojanić and Trivunac (cf. the analysis and the references for this and other words in Ligorio and Kapović 2011: 343).89

Furthermore, the examples Kortlandt adduces are very different among themselves and cannot be listed together. The first two (gospod, kokot) are o-stems (just like plamen), where the usual quantitative pattern (gȍspōd – gȍspoda) is attested almost everywhere (with rare specific and local later generalizations of brevity or length). The next three are i-stems (kokoš, mladost ‘youth’, bolest ‘illness’), which show the expected a. p. C alternation (kȍkōš – kȍkoši) everywhere in both Štokavian and Čakavian. The form ‘root’ has the old variant suffixes (*-enъ/-enь and *-ěnъ), where koren would be the type like gȍvōr (short suffix lengthened only in nom/acc^sg)90 and korijen would belong to the ȍblāk ‘cloud’ – ȍblāka type (the old long suffix with length in all cases). The last word (jablan ‘poplar’) does not belong to this group at all, as it is originally not only a word with a long suffix, like ȍblāk (cf. the usual Štokavian jȁblān – gen^sg jȁblāna), but also an original a. p. a word (*a̋bolnь),91 unlike all the other words in the list which are originally a. p. c. The inclusion of jablan in this list (which Kortlandt obviously put there under the influence of my mentioning the word,92 though in a different context),93 again points to Kortlandt’s inadequate approach.

To conclude this section, Kortlandt, as elsewhere, unfortunately avoids argumentation, he does not seem to have a complete grasp of the material, and ignores obvious points. Some of the implicit outcomes of what he claims, like the one that the length in Štokavian pȍmōć and pȍ ̮mōć is of different origin, seem almost bizarre.

The *obőrna and *čьrnĩna type accent and retractions of contractional neo-circumflex

5.

Slavic prefixed derivatives like *naròdъ ‘people’, *obőrna ‘defense’, *sъdőrvъ ‘healthy’ and o-compounds like *bosonògъ ‘barefoot’, *golobőrdъ ‘barefaced’ seem to point to a frequent or even (near-)generalized fixed accent (a. p. a) on the root following the prefix or the compositional *-o- (*` on short roots, * ̋ on long roots), even when the root does not originally have the fixed accent (cf. *rȍdъ ‘kin’, *bornı̋ti ‘to defend’, *dȇrvo ‘tree’, *noga̍ ‘foot’, *borda̍ ‘beard’). The accent in these formations obviously cannot be analyzed by means of valences (or by means of acute and non-acute syllables) as is usually possible in Slavic, which would points to a later generalization of accent in this type (cf.Kapović 2017a: 396). One could typologically compare the secondary spread of the acute to the spread of non-etymological length in some modern Neo-Štokavian dialects in prefix-derivatives like pónos ‘pride’, prólaz ‘passage’, ómot ‘wrap’ (instead of the older pònos, pròlaz, òmot),94 before the -ńa suffix (cf. the secondary length in vóžńa ‘drive’, nóšńa ‘attire’ compared to vòziti ‘to drive’, nòsiti ‘to carry’ but the expected length in kúpńa ‘buying’ ~ kúpiti ‘to buy’),95 or to the generalization of secondary length in the root in -je derivatives in some modern Neo-Štokavian dialects, as in grȏbļe ‘graveyard’ (but grȍb ‘grave’, older/dialectal grȍbļe), grȏžđe ‘grapes’ (older/dialectal96 grȍžđe), or nárūčje ‘armful’ (cf. the expected root length in lȋšće ‘leaves’, prízēmļe ‘ground floor’).97 A tendency of a secondary spread of a certain type of accent in a specific derivational type is hardly unusual – in such cases, a specific derivational type is “strengthened” through a specific generalized accentual type.

Kortlandt (2018: 293) tries to solve this problem by assuming a generalized original accent on the prefixes (e.g. *òborna) and the connecting *-o- (e.g. **bosònogъ). The generalization of the accent on the prefix (i.e. the generalization of dominant prefixes) may indeed be a possible origin in at least some of the adduced forms. However, this does not solve the unexpected and non-etymological old acute in the long roots (as in *obőrna), except in Kortlandt’s doctrine.Kortlandt (e.g. 2011: 322, 340) believes that the accent shift via Dybo’s law results in a new long falling accent on the originally long vowel, e.g. *òborna > **obôrna. This is the first problem because, as I have pointed out, this medial long falling accent (originating in Dybo’s law) is a mirage, as is clear from examples like the North Čakavian type črnĩna ‘blackness’, ravnĩca ‘plane’ (Kajkavian kraļĩca ‘queen’), dvorĩšće ‘courtyard’, popĩć ‘little priest’, the accentual development of Slavic types like Slovene volár ‘ox-keeper’, Old Štokavian (Posavina) sestrĩn ‘sister’s’, etc. (cf.Kapović 2017a: 395 with further references). Examples like these prove that the new accent, resulting after the progressive shift of *` and *˜ (Dybo’s law), is the neo-acute (e.g. *čьnina > *čьrnĩna). The other problem with Kortlandt’s theory is that it presupposes that this supposed new long falling accent retracts (via “Stang’s law”) from final syllables (not counting final yers) but shortens in medial ones (Kortlandt 2011: 8, 322,2018: 293), which then accounts for examples like *nòsišь ‘you carry’ but **obòrna (which would have the same reflexes as *obőrna). However, as I have already pointed out (Kapović 2017a:395), this would account for numerous alleged alternations and variations that are simply not attested anywhere. In a. p. b presents, one finds only reflexes of the initial accents like 2^sg *nòsišь, though one would expect such an accent in that form and 3^sg *nòsitь ‘s/he carries’, 3^pl *nòsętь ‘they carry’ but not in 1^pl *nòsimo/ъ/e ‘we carry’ and 2^pl *nòsite ‘you carry’, where Kortlandt’s theory envisages the reflexes of **nosìmo, **nosìte (i.e. the same as **nosı̋mo, **nosı̋te in traditional reconstruction) that simply appear nowhere in the present tense. In the *povőrtъ ‘(re)turn’ type, Kortlandt’s theory would provide the expected reflexes in oblique cases like the gen^sg *povőrta, but not in the frequent nom/acc^sg where his supposed **povôrtъ should retract the accent. Again, no such forms are found.98 Thus, to summarize, while the generalization of the original accent on the prefixes might work, Kortlandt’s views on the results of the progressive shift and the supposed retraction and shortening of the assumed medial long falling accent are simply untenable.

Kortlandt (2018: 293) begins his discussion of these topics with incorrect imputations. He says I did not explain the origin of the non-etymological old acute in forms like *sъgőrda, which I did,99 and that I do not “explain the Slovene neo-circumflex” in forms like zgrȃda ‘building’, though I actually dealt with the problem extensively and consider the neo-circumflex as the phonetic and expected reflex before the dominant length as in *sъgőrdā̟ (cf.Kapović 2015: 317–325). The same applies to Kortlandt’s (ibid.) claim that I do not give an explanation of the accentuation of forms like golòbrad or zlòduh ‘evil spirit’. As already said, I believe that the accent like *golobőrdъ and *zъlodűxъ is secondary, originating in a tendency to generalize the original *vьlkodla̋kъ type accent. The expected original accent would be *gòlobordъ > (post-Dybo) *golòbordъ and *zloduxъ > (post-Dybo) *zъlòduxъ due to *gòlъ and *zlъ having dominant roots (i.e. a. p. b) – see below for the valence theory, according to which the stress is assigned to the first dominant syllable in a word. In most cases and in some types in general, the new accentual type was generalized (as in golòbrad and zlòduh),100 however there are indeed cases where the old and new accent both exist. Cf. the original accent in Štokavian kȍlovrāt ‘spinning wheel’ (*kȍlovortъ – cf. *kȍlo ‘wheel’ and *vȏrtъ ‘neck’, a. p. c)101 but the innovative, generalized accent type in Russian коловopóт ‘brace, drill’ (*kolovőrtъ). Cf. also perhaps Neo-Štokavian zȁpād ‘west’ which agrees with pȃd ‘fall’ (C) together with the usual type západ < *zapa̋dъ.102

The generalization of the acute on the long roots in prefix derivatives is, when the situation is looked at carefully, not really so hard to understand. If one starts with generalized dominant (i.e. accented) nominal/adjectival prefixes (which is what Kortlandt himself does),103 one would originally expect pre-Dybo forms like:

*zãstava – *òborna – *prĩroda

[~ *sta̋viti ‘to put’ (a) – *bornı̋ti ‘to defend’ (b) – *rȍdъ ‘harvest, crop’ (c)]

The accent on the syllable due to Dybo’s law depended on the characteristics of the syllable which got the new accent – i.e. on whether it was short or long and whether it was acute or non-acute. If it was short, *` appeared (cf. *gòtovъ > *gotòvъ ‘ready’). If it was non-acute and long, * ͂ was the result (cf. *blina > *bělĩna ‘whiteness’). If it the was acute (and long), it was * ̋ (cf. *bõrniti > *bornı̋ti). Applying those rules, one would expect post-Dybo forms:

*zasta̋va – *obõrna – *priròda

Since the opposition of acute and non-acute was possible only on long vowels, all short roots had the same accent (*`). Thus, the was characterized by the opposition of only *` on short vowels to both * ̋ and * ͂ on long vowels.104 What occurred was the simplification of the system in that type of forms to a simple opposition of *` on short to * ̋ on all long syllables. Thus, we got the innovative:

*zasta̋va – *obőrna – *priròda

This is then reflected in Slavic languages and we get: Russian застáва ‘outpost’, оборóна ‘defense’,105 прирóдa ‘nature’, Slovene zastȃva ‘flag’, obrȃna, etc.106 The same kind of generalization occurred in other prefix derivatives like *naròdъ, *priròdьnъ, *zaslűžьnъ ‘deserving’ (cf. *zaslűga), *povőrtъkъ ‘return’ (cf. *povőrtъ ‘return’), etc. Thus, almost all nominal/adjectival derivatives of this type generalized the innovative synchronic rule that the first syllable after the prefix (and after *-o- in compounds) is always stressed and the accent is either *` if the syllable (root) is short or * ̋ if the syllable (root) is long. Just like the original opposition of dominant and recessive morphemes is gone (in both the prefixes and the nominal/adjectival roots), the original opposition of long acute and long non-acute roots disappears as well.107

Kortlandt (2018: 293) goes on to claim that I reconstruct “Slovak pýta < *pȳtȃ < *pyta̋je without explaining the long vowel and the difference between Čakavian pĩtā ‘asks’ and kopȃ ‘digs’, Bulgarian píta versus kopáe, Old Polish kopaje.” However, I have indeed tackled this problem extensively in my book (cf.Kapović 2015: 341–354), much more so, at least concerning the Western South Slavic material and history, than Kortlandt himself ever has. I even explained for the first time the difference between Čakavian North and Štokavian/Čakavian South when it comes to forms like *kopȃ(mo) (ibid. 351–354) and why we find both 3^sg kȍpā and kopȃ ‘digs’ in Čakavian (or kȍpā and kòpā, and nȅ ̮znā and nè ̮znā ‘doesn’t know’ in Neo-Štokavian), the answer being in the different rules for retraction to preceding short vowels in Štokavian/Čakavian South and Čakavian North.108 I also thoroughly discussed a later similar and connected retraction in Štokavian/Čakavian in masculine nom^sg of l-participles in -a- (*kopȁo > *kopȏ > kȍpō ‘dug’) and in the a-imperfects of the Montenegrin dialect of Pljevlja dȑžāg ‘I held’ < *držȃh < *držȁah (cf. the details and references in Kapović 2018: 261–267).109 Kortlandt’s (ibid.) approach to data seems to be quite lax when quoting Čakavian as having pĩtā ‘asks’ and kopȃ ‘digs’, which is true only for pĩtā – as already said, in the kopa type Čakavian has both kopȃ and kȍpā (depending on the dialect), sometimes with a generalized one or the other type in all verbs and sometimes with a combination of both depending on the verb (cf.Kapović 2015: 344–345).

When talking about my objection on his formulation of “Stang’s law” and the lack of non-retracted forms in present 1/2^pl, he (Kortlandt 2018: 293) adduces the following: “Carpathian (Ublja) byváuu, bývaš, bývať, byváⁱeme, byváⁱete, byvá^uuť”. However, this is neither what I was talking about, nor does it prove what Kortlandt wants it to prove. My point (see above) was that there are absolutely no traces anywhere of the supposed phonetic forms like 1^pl **nosîmo > **nosìmo ‘we carry’ (where Kortlandt’s supposed neo-circumflex should be shortened but not retracted from the medial syllable). In his Carpathian forms, the situation is quite simple. There is no imaginary “restoration of the thematic vowel in *‑à(e)me, *‑à(e)te on the analogy of *kopà(j)e‑” (ibid.) – forms that look uncontracted (like 2^pl byváⁱete) indeed had no contraction. Forms with contraction show also the retraction (2^sg bývaš, 3^sg bývať), while forms without contraction have no retraction (1^sg byváuu, 1^pl byváⁱeme, 2^pl byváⁱete, 3^pl byvá^uuť). One can compare that to Neo-Štokavian where the only present form without contraction also has no retraction (3^pl pítajū < pītȁjū ‘they ask’), while all other forms have contraction and thus also the retracted accent (e.g. 2^pl pȋtāte < pĩtāte). Again, this has nothing to do with the complete lack of attestation for the supposedly original **nosîte, which Kortlandt simply ignores.

To my arguments against “Stang’s law” (Kapović 2017a:390,391) Kortlandt sadly has nothing to say but to apodictically claim that “they have adequately been refuted in the literature”, while citing works of his own and a paper by Vermeer (a Dutch scholar working in Kortlandt’s framework). Once more, simply stating that something is the case does not actually prove it. Obviously, I am well aware of the usual (and Kortlandt’s) arguments for “Stang’s law”, but, as already argued, I simply do not find them convincing when positing the emergence of a neo-circumflex from Dybo’s law (as in the supposed 1^pl *nòsimo > **nosîmo). The only real neo-circumflexes that can retract (which one can call Stang’s law, though Ivšić was actually the first to explain it)110 are those of contractional origin (like in the mentioned 3^sg *ptatь < *pytâtь < *pyta̋jetь).111 Other retractions are local (like gen^pl lȍpāt(ā) in Štokavian/Čakavian) or later.

Kortlandt (2018: 293) completely misses my point when discussing forms like Čakavian črnĩna ‘blackness’, ravnĩca ‘plane’, dvorĩšće ‘courtyard’, etc. (see above). He says that “Vowel length in derivational suffixes is mostly generalized, e.g. S/Cr. ‑at, ‑av, ‑ica, ‑ina versus ‑ār, ‑īk, ‑īn, ‑īna (cf. Dybo 1968)”. The problem is that he does not see that I have adduced forms that clearly show that there are suffixes that have not generalized length in modern dialects112 (that accentologists have up to now disregarded – including Dybo, cited above). Cf. e.g.113 Vrgada Čakavian114 planīnȁ ‘mountain’115 (cf. the secondarily shortened planìna in Neo-Štokavian) but travȉna ‘grass [pejorative]’ (Neo-Štokavian tràvina),116 Neo-Štokavian nom^pl kolíca ‘cart’117 but glàvica ‘little head’,118 Neo-Štokavian zìdić or zìdīć ‘little wall’ (depending on the dialect – both variants, -ȉć and -ĩć, occur in Čakavian and Kajkavian/Slovene as well),119 Posavina Old Štokavian120 sestrĩn ‘sister’s’ but Neo-Štokavian sèstrin,121 and Senj Čakavian122 dvorĩšće ‘courtyard’ but toporȉšće ‘axe handle’ (cf. Neo-Štokavian both -ište and -īšte depending on the dialect).123 I call these “the Hirt suffixes” (cf.Kapović 2015: 184–193 for detailed treatment with references) because I believe that the non-acute/acute and recessive/dominant variants in them originally stem from Hirt’s law. Originally, one should start with a suffix like the nom^sg *-ina,124 where the first syllable (*-in-) would be acute and recessive (–). Before the secondary dominant (+) ending *-a, the suffix would metatonize to secondary dominant (+) while remaining acute – i.e. in non-valence terms, the acute syllable *-in- would attract the accent from the originally stressed ending *-a̋. This metatony/retraction is what one calls Hirt’s law. However, in cases like the acc^sg the ending *-ǫ was recessive (in non-valence terms, it was unaccented), where there would be no Hirt’s law (because the ending was not dominant/accented). Thus, one would expect the original nom^sg *dolı̋na ‘valley’ but acc^sg *dȍlinǫ with the mixed a. p. a-c. This would then, as is often the case with Hirt’s law,125 lead to generalization and one gets variant suffixes (and regular a. p. a and a. p. c): one dominant/acute (*dolı̋na – *dolı̋nǫ, a. p. a, with the new acc^sg analogical to the nom^sg) and the other recessive/non-acute (*dolina̋ – *dȍlinǫ, a. p. c, with the new nom^sg made in accordance to the acc^sg). The first type *dolı̋na (a. p. a) can be seen in the Neo-Štokavian variant dòlina and Russian долúнa. The second type *dolina̍ (a. p. c) can be seen in the Neo-Štokavian variant dolìna.126 These acute/dominant and non-acute/recessive variants generalize depending on dialect/language, different word-forms, and semantics. Sometimes, only one variant is preserved in a dialect/language, sometimes both are. In any case, both variant suffixes can follow a. p. a, b, and c roots, e.g. *sta̋rъ (a) ‘old’, *čьr͂nъ (b) ‘black’, *žȋvъ (c) ‘alive’ – cf. Neo-Štokavian stȁrina ‘antiquity’, secondary crnìna ‘blackness’ (the original accent is seen in North Čakavian type črnĩna, which has usually been generalized),127 and živìna (older acc^sg žȉvinu) ‘cattle’. What is relevant for our present discussion is that in cases of such non-acute/recessive suffixes when they are added to the dominant non-acute (a. p. b) roots we see that the original *čьr͂nina (before Dybo’s law, i.e. before the progressive shift of dominant circumflexes) yields *čьrnĩna with the neo-acute (as attested in numerous, already adduced, forms in Čakavian, Old Štokavian, and Kajkavian). This in turn proves that Kortlandt’s supposed **čьrnîna (with the alleged long falling medial tone after Dybo’s law) > **čьrnìna (with the alleged shortening of this long falling tone) is an illusion. This then means that his explanation of prefix forms like *obőrna, however simple, cannot be correct.

Thus,Kortlandt’s (2018: 293) explanation that “[v]owel length in derivational suffixes is mostly generalized” unfortunately means nothing. His claim (ibid.) that the “Original differences have been preserved e.g. in dvòrište (b) ‘yard’ versus blȁtīšte (a) ‘mud-pit’ and Czech pekař (c) ‘baker’ versus rybář (a) ‘fisherman’” also have nothing with to do with my argument. I have already discussed Czech pekař and rybář (see above) but this has nothing to do with the neo-acute as the result of Dybo’s law in the case of non-acute suffixes. As for Štokavian dialect preserving the opposition of the blȁtīšte – dvòrište type,128 the situation is not as simple in mjȅštāni ‘locals’ – sèļani ‘villagers’ type (which is the only widespread and Standard Neo-Štokavian example of the preserved length in an originally acute suffix),129 because the -ište suffix has both -ȉšte and -ĩšte variant (see above). Thus Neo-Štokavian -īšte can be both the reflex of the unaccented acute posttonic length (corresponding to the accented -ȉšte) or the reflex of the unaccented non-acute posttonic length (corresponding to the accented -ĩšte). In any case, the origin of the long -īšte in words like blȁtīšte is irrelevant for our topic here and the form dvòrište in no way disproves the forms in -ĩšte that also exist, which cannot be secondary because there is no phonetic or analogic way by which -ȉšte could yield -ĩšte (one simply must reconstruct both types of accent in this suffix, as in the other mentioned suffixes).

Again, as previously, Kortlandt unfortunately misinterprets my claims, seems not to understand certain obvious and simple examples, does not even try to explain the lacunae in his theory, ignores the important problems, and simply repeats the forms he always repeats (as Czech pekař and rybář), which are simply not pertinent at all to the subject being discussed.

The reflex of *ò in Slovak and Czech monosyllables

6.

Concerning the reflex of *ò in Slovak/Czech monosyllables,Kortlandt (2018: 293–294) sticks to his view that the length in o-stem monosyllables like Slovak kôň, Czech kůň ‘horse’ “did not arise phonetically but was adopted from the case forms where the accent had been retracted in accordance with Stang’s law before the loss of weak jers, the shortening of long falling vowels in initial syllables, the loss of distinctive tone, and the fixation of the stress on the initial syllable”, while I maintain “the traditional view that Czech ů, Slovak ô is the phonetic reflex of *ò in monosyllables.” (which is true, though I stress that Czech monosyllables are less convincing due to the later tendency of generally lengthening o > ů before voiced final segments).

To be more precise,Kortlandt (2011: 345–346) thinks that the long reflexes in nom/acc^sg (as Slovak kôň, Czech kůň) is analogical to loc^sg, gen^pl, loc^pl, instr^pl, which according to him all had *ò in the first syllable:130 loc^sg *kòńi,131 gen^pl *kòńь,132 loc^pl *kòńixъ, instr^pl *kòńi.133 His explanation for this analogy (the transfer of the original length from loc^sg, gen^pl, loc^pl, instr^pl to nom/acc^sg, where it is supposedly not phonetically expected) is that the length was generalized in all forms with an initial accent – i.e. initial-stress forms like nom/acc^sg *kòńь, instr^pl *kòńi had length (whether it was original or not), while end-stressed forms like gen^sg *końa̍, dat^sg *końu̍ had a short root. Kortlandt (ibid.) then says that in the oblique cases the length was later lost “before the new long case endings in gen.pl. -ôv,134 -í and loc.pl. -iech, -ích”135 and “the paradigm could be further regularized by generalization of the short root vowel (…)”.

If one accepts Kortlandt’s reconstruction of a. p. b forms and its reflexes, which are hardly certain and doubtfree (see the footnotes above), his scenario would not be impossible, but accepting it would presume accepting a lot of unnecessary analogies, none of which have any real function except being a part of Kortlandt’s wider ideas on historical development of Slavic accentuation. WhenKortlandt (2018: 294) says, yet again, that my “treatment again lacks a chronological perspective”, it is clear that means that I am unwilling to accept his reconstructions and his vision of relative chronology, i.e. his completely unnecessary and complex analogies and generalizations. Is it imaginable that Czech originally had *skot, which was then changed to the attested Old Czech skót ‘cattle’ (by analogy to some of the oblique cases), which later again changed to the modern Czech skot? Yes, it is perhaps imaginable, but it is neither economical nor necessary. Leaving Kortlandt’s elaborate relative chronology aside, it is an unnecessary complication to assume that the only form where the length is today attested (nom/acc^sg forms like Slovak kôš ‘basket’) is actually secondary, while none of the forms that supposedly originally had the length now do not have it (cf. Slovak loc^sg koši, gen^pl košov, loc^pl košoch, instr^pl košmi – however, the last three have innovative endings). On the other hand, should one start with the expected length in nom/acc^sg, the modern paradigm is quite straightforward even if one has to explain the root in forms like loc^pl košmi (with a secondary ending) as analogical (though that is hardly certain – see above).

An important reason why Kortlandt assumes all these subsequent analogies is his idea that the reflex of Slavic *` and * ̋ merge in Czech (i.e. yield brevity in monosyllables and length in disyllables in front of a short vowel). However, while it is not impossible that *` and * ̋ indeed merge in Czech, that is hardly the likeliest option. Kortlandt’s theory that the old acute phonetically yields short vowels in monosyllables (cf. *ma̋kъ > Czech mák ‘poppy’) is not very persuasive (cf.Kapović 2017a: 397) and the general merger of the two Common Slavic prosodemes is very much in question taking into account paradigmatic reflexes such as Czech 2^sg můžeš ‘you can’ < *mòžešь but javor ‘maple’ < *a̋vorъ (for details cf. Kapović 2019).

Kortlandt (2018: 294) ends his short comment on the problem with a statement that the traditional (and my) view “does not explain the short vowel in Czech osm, Slovak osem ‘eight’ < *òsmь”. However, this is an oversimplification of a complex issue. There are a lot of counterexamples for both the possibility that *ò yields length and that *ò yields brevity in Czech/Slovak. Short reflexes in osm/ /osem can be easily explained in a variety of ways. AsVerweij (1994: 515), who works in Kortlandt’s framework, notes, there are generally no long reflexes in initial *o- (this may have even been a separate phonetic law). It could also be a an analogy to sedm/sedem ‘seven’ (monosyllables never show length from *è) or analogy to osmý ‘eighth’ in Czech (however, Slovak ordinal ôsmy has the long reflex which must be secondary).136 In any case, osm/osem is hardly a crucial or only example which would prove a point. For more details on my take on the reflexes of *` in general in West Slavic cf. Kapović 2019.

The valence theory and the Proto-Indo-European origin of Balto-Slavic accentuation

7.

Kortlandt (2018: 295) ends his paper with the remark that “[a]ttempts to solve classic problems in terms of dominance patterns have resulted in complete failure”. By this, he means to say that the Moscow Accentological School approach to the issue of the origin of Balto-Slavic accentuation, which sees it as the most archaic reflex of the original Proto-Indo-European tonal system,137 is wrong and that one should stick to the dominant norm of Western Balto-Slavic scholarship in the last half of century, which basically means deriving Balto-Slavic complex paradigmatical accent from a simple “Graeco-Aryan” Proto-Indo-European accentual system (i.e. Proto-Indo-European accentual system reconstructed primarily on the basis of Ancient Greek and Vedic).138 However, it is quite the opposite – mainstream Western Balto-Slavic historical linguistics has been sleeping on the problem and is still desperately clinging on to implausible sound laws and a Vedic-centric Proto-Indo-European reconstruction, which has been abandoned long ago in other spheres of Indo-European linguistics. I will point to a couple of problems with the usual way of deriving Balto-Slavic accentuation from a simple Proto-Indo-European accent, reconstructed primarily on the basis of Vedic and Greek.

When comparing Balto-Slavic accentuation with Vedic/Greek accentuation, what is immediately clear is that the Balto-Slavic system (even if one does not accept the valence theory) is much more complex. In Vedic/Greek accentual system, the accent can be mobile in a very limited way: usually only in athematic nouns with monosyllabic roots like Vedic pt – gen^sg padás and Greek πούς – gen^sg πoδóς (which is then reconstructed as PIE *pṓds – *pedós ‘foot’). However, in Vedic not all athematic root-nouns have mobile accent139 (cf. Vedic śv ‘dog’ – gen^sg śúnas),140 while in Greek it is automatic there141 (cf. Greek κύων ‘dog’ – gen^sg κυνóς), which is often disregarded. In Greek, the verb (except for participles, which behave like nominal forms) usually has the predictable propenultimate or penultimate stress, while the accent in Vedic verb is connected to ablaut, which must be secondary (and not original as the Indo-Europeanist mainstream holds). Unlike both Vedic and Greek, the original Balto-Slavic accentual system has non-trivial and unpredictable142 accentual classes in both nouns and verbs (and elsewhere) – all words belong either to an immobile or mobile class. There is no way to predict which word will belong to which class – neither phonological, morphological, nor semantic criteria have any say in this.143 In Balto-Slavic, the whole system shows the opposition of immobile and mobile stress, while Vedic and Greek show only traces of that – this must mean that the more complex system, the Balto-Slavic one, where there is no morphological limitation to the mentioned opposition, is older, in spite of the traditional belief that the classical languages must be the most archaic.

Besides tradition, the implicit reason to assume that Vedic and Greek must have a more archaic accentuation than Balto-Slavic seems to be the timeframe of the attestation of these Indo-European branches. Since both Indo-Iranian and (Mycenaean) Greek were attested already in the second half of the second millennium BCE and Balto-Slavic is attested only much later (Slavic from the 9^th century and Baltic from the 14^th century), the implicit reasoning is that Vedic and Greek must have a more archaic accentual system (though Vedic was written down much later and the Greek tradition of writing of the accents begins only in 3^rd-2^nd century BCE). However, this is not necessarily so. If Balto-Slavic accentuation in the 1^st millennium BCE was such as we can reconstruct it – why would it necessarily be more innovative than Vedic or Greek just because it was accidentally not attested in writing? We do know that Balto-Slavic languages are very archaic in other regards – e.g. many of them preserve seven (of eight Proto-Indo-European) cases, while Ancient Greek preserved just five cases (Modern Greek only four)144 and Modern Indo-Aryan languages have lost all but two of the Old Indo-Aryan eight cases already a thousand years ago.145 And if Modern Greek has more or less preserved the free stress of Ancient Greek up till today, just like Modern Pashto (and some other modern Indo-Iranian languages) preserved reflexes of a free stress system similar to what is attested in the 2^nd millennium BCE Vedic, why would it be strange that some modern Balto-Slavic languages preserve archaic traits of the 1^st millennium BCE Proto-Balto-Slavic? Given the otherwise famous archaic nature of Lithuanian, that would hardly be impossible.146

The mainstream Western Indo-European and Balto-Slavic accentology usually presumes that the mobility in Balto-Slavic ā-stems (and o-stems and polysyllabic athematic stems) is somehow secondary (in comparison to Vedic and Greek columnar oxytone accent in ā-stems). However, there is a big problem, usually ignored, concerning this. Cf. the ā-stem ‘head’ in Proto-Slavic (*golv) and Lithuanian (galvà) with the accentuation of a polysyllabic (vanden-) athematic stem in Lithuanian (vanduõ ‘water’) and an athematic stem (ἀρήν ‘lamb’) with monosyllabic root (ἀρν-) in Greek:

Now, the usual approach is to assume that Balto-Slavic forms as the ones in the table have a secondary mobile accent that is due to some innovations, usually retractions of some sort – cf. recently e.g. Olander 2009147 and Jasanoff 2017148 for this type of approach. The most usual and traditionally accepted retraction is some kind of “Pedersen’s law”, which interpretes the initial accent in mobile stems like Lithuanian acc^sg dùkterį ‘daughter’ (Slavic *dkťerь) as retracted from the supposedly original PIE *d^hugh₂térm̥ (reconstructed on the account of Vedic acc^sg duhitáram and Greek acc^sg ϑυγατέρα).149 The other approach may be to explain the mobility in thematic stems or polysyllabic forms like dukt – acc^sg dùkterį as somehow analogical to the athematic root-nouns (as Kortlandt does), though that would be very strange since thematic stems were much more productive than the athematic ones, which tended to weaken and eventually even disappear altogether in most later Indo-European languages. In any case, if the Balto-Slavic mobility in most of the nominal stems is due to some kind of innovative retraction or some similar process (something like “Pedersen’s law” or Olander’s “mobility law”), how is it possible that what one gets is exactly the same kind of mobility as seen in Vedic and Greek root-nouns? As the table above clearly shows, acc^sg/nom^pl/acc^pl is barytonic everywhere, while nom^sg/gen^sg/gen^pl is oxytonic. Is this just a coincidence?150 Is it possible that some kind of innovative sound law in Balto-Slavic (“Pedersen’s law” or some law similar to it) would provide the same type of mobility that was supposedly originally already there in athematic root-nouns?151 Is it possible that Balto-Slavic also secondarily developed the same kind of immobile/mobile opposition in verbs as well? What were these strange retractions (or even stranger analogies) that produced a Balto-Slavic split of the verbal system to immobile and mobile stems, completely paralel to the accentual split in nouns, and at the same time completely unknown in the traditionally reconstructed “Graeco-Aryan” Proto-Indo-European accentual system? If all this were true, that would be one enormous coincidence and one very unusual development. Why wouldn’t one rather assume that the opposition of immobile and mobile stress was originally present in all stems and that it was later lost almost everywhere in Vedic and Greek except in the athematic root-nouns, with the thematic and polysyllabic athematic stems generalizing the columnar oxytone accent (e.g. Greek ϑεᾱ́ ‘goddess’ – acc^sg ϑεᾱ́ν)? A typologically similar development can be seen in many contemporary Štokavian/Čakavian/ /Kajkavian dialects, which have (similarly to Vedic and Greek so many centuries ago) lost the mobility in a. p. C except in nouns with monosyllabic roots – cf. Modern Štokavian loc^sg zúbu (from zȗb ‘tooth’) but loc^sg ȍbrūču (from ȍbrūč ‘hoop’) instead of the older loc^sg obrúču, or acc^sg glȃvu (from gláva ‘head’) but acc^sg sramòtu (from sramòta ‘shame’) instead of the older acc^sg srȁmotu. The same can occur in dialects without the accent retraction – cf. Brač Čakavian (Šimunović 2009: 35–37) mobility in monosyllabic vrȏg ‘devil’ – gen^{sg +}vrȏga152 – loc^sg vrōgȕ – nom^pl vrȏzi – gen^pl vragv – dat/loc/instr^pl vrōzȉma(n) (thus in most monosyllabic a. p. C stems) but lack of it in disyllabic gȍlub ‘pigeon’ – gen^sg gȍluba – loc^{sg +}gȍlubu – nom^pl gȍlubi – (gen^{pl +}gȍlubih) – dat/loc/instr^{pl +}gȍlubima (thus in most disyllabic old a. p. C stems).153

According to the valence theory,154 the attested accentual system in Balto-Slavic (prior to the operation of de Saussure’s and Dybo’s law) can be analyzed by assuming that all syllables (or morphemes) were either dominant (+, probably high tone) or recessive (–, probably low tone),155 somewhat similar to modern Japanese. The stress is attributed to the first dominant syllable in a word (e.g. *kő̟rva̟ ‘cow’, *vo̠da̟̋ ‘water’). If all syllables are recessive, the word is phonologically unstressed, which means it gets an automatic initial circumflex (acc^sg *vȍ̠dǫ̠ ‘water’), which can then shift further to the left if it gets a preceding recessive clitic (*nȃ̠ ̮vo̠dǫ̠ ‘onto the water’). Dominant morphemes are a. p. a and b roots (*kő̟rv-, *sè̟str- ‘sister’) and accent-“attracting” endings in a. p. c (nom^sg *vo̠da̟̋) and suffixes (*vo̠dı̟̋ca̟ ‘little water’). Recessive morphemes are a. p. c roots (*vo̠d- ‘water’), together with endings (*-ǫ̠ in acc^sg *vȍ̠dǫ̠) and suffixes that do not “attract” the accent in a. p. c (*s̑no̠vъ̠ ‘son’s’).

What the valence theory is especially good at explaining is the accentuation in Slavic derivation,156 which the proponents of deriving the Balto-Slavic accentuation from the “Graeco-Aryan” Proto-Indo-European system usually completely ignore. If one takes that all morphemes/syllables in Balto-Slavic were either dominant/high or recessive/low, the accentuation in derivation is very simple to explain – the accentuation in a derivative will simply depend on the valence of the root, the suffix, and the ending. E.g. if a root is dominant (like *kő̟rv-), all forms made with this root will be initial-stressed (and the accentual characteristics of suffixes and endings will not matter): dat^sg *kő̟rvě̠, instr^pl *kő̟rva̟mi̟, *kő̟rvi̟ca̟ ‘little cow’. If a root is recessive (like *vo̠d-), the accentuation of the forms made with this root will depend on the accentual characteristics of suffixes and endings: dat^sg *vȍ̠dě̠, instr^sg *vo̠do̠j, instr^pl *vo̠da̟̋mi̟, *vo̠dı̟̋ca̟.

However, if one does not believe in valence theory and derives Balto-Slavic accentual paradigms (which are much more complex than those found in either Vedic or Greek) from the “Graeco-Aryan” Proto-Indo-European system via specific unconvincing retractions and analogies, the accentuation in derivatives is practically impossible to explain.157 It is not problematic in a. p. a and a. p. b – there, one can simply claim that derivatives preserve the same immobile root-accent as in basic forms, e.g. *kőrvica like *kőrva, *sèstrica like *sèstra.158 Where the situation becomes problematic is with roots belonging to a. p. c. How can one explain that derivatives of words that have mobile accentuation can have any kind of accentual type (except, logically, having a dominant root-stress)?159 It is very difficult to explain phonetically the accentuation of such forms as *s̑nъ̠ ‘son’ – *s̑no̠vъ̠ – fem. *sno̠va̟̋ – *sno̠vь̟̀cь̠ ‘nephew’ – *snъ̟̀kъ̠ ‘sonny’ or *go̠lva̟̋ ‘head’ – *nȃ̠ ̮go̠lvǫ̠ ‘on the head’ – *go̠lvı̟̋ca̟ ‘little head’ – *go̠lva̟̋tъ̠ ‘with a big head’ – *go̠lvã̟ŕь̠ ‘chief’ – *gȏ̠lvь̠nъ̠ ‘main’ – fem. *go̠lvь̠na̟̋ – def. *go̠lvь̠nъ̠jь̟̍ – def. neut. *gȏ̠lvь̠no̠je̠ if one does not take into account the valence theory.160 What is more, it is implausible that such derivation rules, obviously governed by underlying (high and low) tones, can be derived from a much simpler “Graeco-Aryan” Proto-Indo-European system as is usually reconstructed.

In any case, while the valence theory and the new approach to Proto-Indo-European accentuation is definitely worth at least a serious consideration, it is not difficult to see why such a theory cannot easily become an Indo-Europeanist mainstream. Balto-Slavic is traditionally held as unimportant for the reconstruction of Proto-Indo-European accent and the reconstruction of Proto-Indo-European accentuation on the basis of Vedic and Greek (with a touch of Germanic via Grimm’s and Verner’s law) is well established and extremely simple. Making Balto-Slavic accentuation perhaps the main cornerstone of reconstructing Proto-Indo-European accentuation is not practical because of its complexity, which is such that even many Balto-Slavic specialists are not too comfortable with it.161 It will be interesting to see what the future brings concerning the reconstruction of Proto-Indo-European accentuation.

Conclusion

8.

It is understandable that Frederik Kortlandt does not want to give up on or substantially modify the doctrine he has been working on for almost half a century. This, however, does not mean that it is not full of lacunae and that it can explain the real data in a satisfactory manner. But it is Kortlandt’s modus operandi that is most disturbing: working with a rather limited set of examples repeated tirelessly (e.g. Czech pekař and rybář) without taking into account the whole context and system, obfuscating his own theories and making it very difficult for uninitiated readers to properly assess them themselves (e.g. concerning Stang’s law), constantly ignoring important data that do not fit his doctrine (e.g. Slovincian data for pretonic length), avoiding detailed argumentation (e.g. concerning Štokavian/Slovene gen^pl -ā), having an inadequate grasp of the basic material (e.g. concerning the kȍkōt type length in Štokavian/Čakavian), and suppressing recent research done outside of his own school of thought. Kortlandt appears to be at pains to defend his clearly problematic doctrine, while unfortunately not being able to produce anything new in its defense except to repeat what he wrote decades ago. Bold rhetorics and fierce conviction may go a long way, but they cannot replace careful argumentation, in-depth knowledge, overview of the data, and honest scholarly discussion.