Kirčio atitraukimas ir nukėlimas lietuvių kalbos tarmėse

LIETUVIŲ KALBOTYROS KLAUSIMAI, IV LIETUVOS TSR MOKSLŲ AKADEMIJA STRESS RETRACTION AND DISPLACEMENT IN LITHUANIAN DIALECTS V. GRINAVECKIS I. STRESS RETRACTION 1. Terms for stress retraction § 1. At present, Lithuanian dialectology uses the terms “general stress retraction” and “conditional stress retraction” to describe stress retraction in dialects, but neither is precise. General stress retraction is defined as the retraction of stress from a short or long circumflex ending. Thus, in fact, stress retraction is not general: under this rule, stress is not retracted from an acute ending. Conditional stress retraction is defined as the retraction of stress from a short ending to the long penultimate syllable. However, this rule covers only one rule of stress retraction from a short ending. Thus, these terms are insufficient to describe all cases of stress retraction. Since the aforementioned terms are imprecise and insufficient to describe all cases of stress retraction, this article adopts new terms describing each case of stress retraction from an ending. 1. The following terms are used to indicate the dependence of stress retraction on the length and tone of the ending: a) Stress retraction from a short ending is called grave stress retraction. b) Stress retraction from a long circumflex ending is called circumflex stress retraction. c) Stress retraction from an acute ending is called acute stress retraction. 2. The following terms are used to define the rules of stress retraction in individual dialects: a) Stress retraction from a short ending (historically short or historically acute) and from a circumflex ending is called grave-circumflex stress retraction; stress retraction in dialects that retract stress not only from short and circumflex endings but also from acute endings may be called grave, circumflex, and acute stress retraction. b) Stress retraction from a short ending in all cases, that is, onto the next syllable from the end of the word, whether long or short, will be called general grave stress retraction. 117 c) The retraction of stress from a short ending only onto a long syllable farther from the end may be called conditional gravis stress retraction. d) The retraction of stress from a short ending only onto a syllable consisting of a long vowel or the diphthong ie or uo may be called incipient stress retraction. 2. Laws of stress retraction § 2. In certain cases in the northern and northwestern Lithuanian dialects, stress is retracted from the end of a word toward its beginning. Stress retraction is not the same everywhere. The spoken dialectal features of individual localities differ in various respects; therefore, stress retraction can be divided into several groups from at least three perspectives: 1) according to the type of ending from which stress is retracted (from a short ending; from short and long circumflex endings; from short and long circumflex and acute endings); 2) according to which syllable from the end receives the stress (the penultimate; the first (initial), etc.); 3) according to whether, in cases of retraction, all the stress is concentrated in one syllable or distributed over two or more syllables. § 3. The first classification. In the far northwestern part of the stress-retraction area, bounded by the Curonian Lagoon and the Baltic Sea to the west and the Latvian border to the north, we find gravis-circumflex stress retraction: stress is retracted from short endings (from an ancient short vowel, e.g., sūnus ~ sūnùs, daĩlus ~ daí-lùs; from an ancient acute, e.g., bàlta ~ baltà < baltá, màrti ~ martì < martí Joniškis) and from long circumflex endings (e.g., sũnaus ~ sūnaũs, pàčiuos ~ pačiõs Klykoliiai). The aforementioned stress retraction covers the area north and west of a line that starts at the mouth of the Minija and runs via Ž. Naumiestis, Pajūris, Upyna, Vaiguva, Kurtuvėnai, and Kužiai, skirts Šiauliai, then runs via Radviliškis, Baisogala, Upytė, Panevėžys, and Vabalninkas, and meets the Latvian border west of Biržai (cf. APh IV 33). In some northern and western localities within this area, stress is also sometimes retracted from circumflex endings. This is acute stress retraction, which covers the entire northern part of the donininkai area, e.g., árklliẽms ~ arkliáms, gáidẽms (with primary stress on the ending) ~ gaidžiáms (Kojeliai, Klaipėda district); it is also found sporadically along the Latvian border, e.g., nèžinau ~ nežì-nau, bǫvau ~ buvau, gèram ~ gerám, kǫkèm ~ kokiám, laũkam ~ laukám, árkliem ~ arkliám, svèččem ~ svečiám (Akmenė, Klykoliai, Vegeriai, Papilė, Tryškiai, Luokė), gèram ~ gerám, kǫkèm ~ kokiám (Šiupyliai), laũkam ~ laukám, árkliem ~ arkliám, svèččem ~ svečiám (Šimkūnai village, Joniškis district; Vilūnaičiai village, Joniškis district), as well as from within words, e.g., pàmate ~ pamãtė, nèžina ~ nežìno (Šiupyliai), sǫsenėsas ~ susinėsasi, nepàdarisi ~ nepadarýsi (Klykoliai; cf. also A. Jonaitýtė LKK II 114, 115). In the northern part of the Mūša river basin, stress is sometimes retracted from dative endings, e.g., saũsam ~ sausám, viẽnam ~ vienám. Here, however, the dative ending is pronounced with a circumflex intonation: jam̃, taĩ (Uniūnai village, Linkuva district). South and east of the area of gravis-circumflex stress retraction lies the area of gravis stress retraction, where stress is retracted only from short endings. APh IV 33 states that in this area stress is retracted “only onto a preserved long or diphthongal stem.” This statement is not accurate, since in the area mentioned stress is retracted in three different ways: a) From a short ending in all cases, regardless of the length of the preceding syllable. This type of stress retraction is found in the area where the boundary of gravis-circum- 118 flex stress retraction recedes considerably northward, skirting Šiauliai. This area includes the southeastern corner of the dūnininkai region (Kiaunoriai, probably Kiaunorius), Tytuvėnai, Šalōpelkis, Pakražantis, Kražiai, Kušleikiai) and part of the northwestern Highlands area (Šiauliai, Pakapė, Tyruliai, Šiluva, Lyduvėnai). In addition, stress is always retracted from short endings beyond the eastern boundary of gravis-circumflex stress retraction, around Biržai, e.g.: ankšti, ba.l.ti, ta.m.šu; ma.ž.šus, kėlus, vākārė, rutėlės, ni.tikės (Kušleikiai, Kelmė district; see Gtr 103); ainu ~ einu, sōdus ~ sodūs; nāktis ~ naktis, vēdu ~ vedu, nēšu ~ nešu, rāštus ~ raštus, pati ~ pati, rasu ~ rasą, bālu ~ balą, gēru ~ gerą (Vabalninkas); saīs ~ sausą, viēn ~ vieną, sāk ~ šaką, viś ~ visi (Biržai; see I. Lašinskaitė LKK I 189). b) From a short ending only when the preceding syllable is long, e.g.: jāuna, karšti, baūsus, but: mažā, viś, laukai (Troškūnai); girdžu, suprantù, aštūni (Viduklė); laīkus, sveīki (Šimkaičiai; cf. E. Grinaveckienė LKK I 122). c) In the southeasternmost part of the area under discussion (Suginčiai, Karalinava — Molėtai district; Gelvonai — Ukmergė district), stress is retracted from short endings only when the penultimate syllable consists of a long vowel or the diphthongs ie or uo, e.g.: sūnus, pǐ?ksu, stora, lōpus, prōtu, viena, diena, but sausā, laukīs, dalīs, tvarkā (see Gtr 103). The southern and eastern boundary of the area of gravis stress retraction (both general and conditional, as described here: sausā ir šākā, šāk) runs approximately through the following places: Upyna, Nemakščiai, Viduklė, Šimkaičiai, Ariogala, Čekiškė, Kulva, Gelvonai, Giedraičiai, Molėtai, Leliūnai, Vyžuonos, Svėdasai, Kamajai, Kriaunos. § 4. The second classification. a) The accent is not completely retracted from the word ending throughout the entire area of accent retraction described above: in some places, traces of accent can still be heard at the end of the word. This is characteristic of all Samogitian dialects in which accent is retracted, for example: švìsa ~ šviesà, vãpsā ~ vapsā (Laukuva); gálvā ~ galvà, šākā ~ šakà, gáidīs ~ gaidỹs, gėismė ~ giesmė (Grušlaukė); mažā ~ mažà, sausā ~ sausà (Kelmė). In polysyllabic words with retracted accent, every syllable is accented, for example: ãgìnā ~ aguonà, pãvažā ~ pavažà (Laukuva); vàkārė ~ vakarė, pãbō.ngā ~ pabanga ‘end’ (Mažeikiai), cf. Švdt § 3 ff. b) In terms of accent distribution in polysyllabic words with retracted accent, only the northeastern Samogitian corner that retracts accent differs somewhat (Klykoliai, Akmenė, Papilė, Viekšniai), where the ancient accent of short final syllables can no longer be heard, for example: kálba ~ kalbà, ãveles, drūktā, mė̃us (Klykoliai), cf. Švdt § 78. Sometimes the final accent, especially in a short ending, cannot be heard in the Donininkai dialect either: nēšu ~ nešu, vės ~ visà, žiūv ~ žuviu (Dreverna). The old final accent is pronounced much the same way in the northern part of the northwestern Highlander dialect area (cf. A. Jonaitytė LKK II 109, 110); farther south (around Šiauliai, Pakapė, Šiluva, Krakes), the old final accent is better preserved. In both the Samogitian dialects and the Highlander dialects adjoining them, the old (final) accent is secondary, while the retracted accent on the initial syllable is primary. c) Farther east from the Samogitian dialects, traces of the final (old) accent occur in the Highlander dialects along the boundaries of gravis-circumflex and gravis accent retraction. For example, near Panevėžys, in the village of Stetiškiai, which lies by the boundary of gravis-circumflex accent retraction, the final (old) accent in words 119 with retracted accent is pronounced in cases where the final syllable is long: laũką ~ lauką, vaikè ~ vaikai, būvǒ ~ buvau, maišǒ ~ mačiau. In a short ending, no trace of the old accent remains, for example: sausù ~ sausà, šakù ~ šakà. d) Traces of the final (old) accent have also been preserved along the boundaries of conditional accent retraction, for example: laũkus, vaĩkus (Kamajai), tõli, aũkšta (Bikūnai village, Anykščiai district; cf. Gtr 103). These two accents are heard throughout the north of the eastern Highlander and eastern Dzūkian dialect areas (Daugėliškis, Dusetos, Svėdasai, Utena, Zarasai, Linkmenys (see E. Grinaveckienė MADA 2 (9) 176), Rimšė). § 5. The third classification. In some Lithuanian dialects, accent is retracted to the first syllable (in polysyllabic words), and in others to the penultimate syllable (with certain exceptions). In the Samogitian dialects, under the law of gravis-circumflex accent retraction, accent is retracted to the first syllable. Under the law of gravis accent retraction, accent is likewise retracted to the first syllable in the Samogitian dialects (e.g.: nėgàlž ~ negali, nebėtùrù ~ nebeturiu, drūo-bužù ~ drobužiu); however, alongside this, within the Samogitian area of gravis accent retraction, accent is retracted to the syllable to which it shifts from the ending when the word is inflected, for example: aštūni ~ aštuoni (:aštuonis) lapẽlis ~ lapelius (:lapelį), kanapikẽs ~ kanapikės (:kanapikę), bli˙nūkùs ~ blyniukus (: blyniuką), drobužùs ~ drobužius (: drobužį), vuotagù ~ votagų (:votagą), bala.n.džùs ~ balandžius (: balandį); kėtūrì ~ keturi (:keturis), dėdamà ~ dedamà (: dėdamą), vàkarè ~ vakare (:vakarą), neki.m.bama ~ nekimbama (: nekimbama), dèbesù ~ debesį (: debesį) (Kušleikiai, Kelmė district). In the eastern Highlander dialects, accent is retracted to the antepenultimate syllable, for example: pamàčà ~ pamačiau, padàrǒ ~ padariau, pavàž ~ pavažą, šonel ~ šuneliù (Pušalotas); avìž ~ aviža, išìs ~ išiša, pabàig ~ pabaiga, vodèg ~ uodega (Čeniškiai village, Biržai district); malõnus ~ malonūs, nobõdus ~ nuobodūs, septintà ~ septintą, išáusta ~ išaušta (Kupreliskis village, Vabalninkas district); akmenù ~ akmenų, sǫ-draške ~ sudraskei, vakàre ~ vakarai, ketǫr ~ keturi (Pušalotas). Eastern Highlanders retract accent to a syllable farther from the end in words of accent paradigm 3, for example: vuõdeg ~ uodega, vėrgdam ~ verkdama (ibid.); lĩ˙gǫm ~ lygumą, alkàn ~ alkaną, sausùm ~ sausumą, bėgdam ~ bėgdama (Linkuva), cf. Švdt § 4. The northwestern Highlanders, whose settlement area adjoins the Samogitian dialects, have a pattern of accent retraction very similar to that of the Samogitians: accent is retracted to the first syllable (with minor exceptions), for example: pėlenai ~ pelenai, lĩ˙dekà ~ lydeką, bàravìks ~ baravykas, paũkščiuks ~ paukščiukas (Šiupyliai, Šiauliai district); àdara ~ atdarà, ìštìsa ~ ištisà, uõdega ~ uodega, dõvana ~ dovanà; àkmenu ~ akmenų, dõbilai ~ dobilai, pàšaki˙s ~ pasakys (Eglynaukis village, Joniškis district). Forms with retracted accent on the second syllable are also used alongside these, for example: pabaiga (and pàbaiga) ~ pabaigà, sudrāskẽi (and sùdraskei) ~ sudraskei, pasvèikai (and pà sveikai) ~ pasveikai, vandeniũ (and vàndenìnu) ~ vandeniniu (ibid.); maitìnu (and mãitinu) ~ maitinù, dalìnu (and dãlinu) ~ dalinù (Šimkūnai village, Joniškis district). Similar accent variation (retraction to the penultimate and first syllables) occurs in the area from the Samogitian boundary to Joniškis (see Gtr 104; also cf. A. Jonitytė LKK II § 4). Farther east of Joniškis, retracted accent on the penultimate syllable increasingly predominates. 120 3. The history of stress retraction a) Causes of stress retraction § 6. In Lithuanian, the old circumflex vowels at the ends of words have been shortened, e.g.: sausà < * sausą, vàrna < * vórną (see J. Endzelynas GF 22). In the northern Lithuanian dialects, the shortening of circumflex endings has proceeded even further. In the northwestern part of Samogitia, the old circumflex vowels at the ends of words have been shortened so much that they are barely audible, e.g.: sáus(a) ~ sausa (Priekulė), sáusa ~ sausa (Židikai, Leckava, Klykólai); in the northwestern part of the eastern Aukštaitian dialects (in the Mūša river basin), the old circumflex vowels have disappeared altogether, e.g.: saĩs ~ sausà, vàrn ~ vàrna (Pušalotas, Biržai). Since the shortening of vowels in circumflex endings affected word-final position equally whether stressed or unstressed, the further loss and reduction of endings in the northeastern Lithuanian dialects must have been accompanied by another phenomenon—stress retraction from the end of a word to its beginning. The beginning of stress retraction can be understood as follows: as long circumflex vowels in word-final position gradually shortened, and these vowels continued to disappear in the northern dialects, they must, as they disappeared, have lost their stress (:sausà > saĩsa, naujienà > naujienėn), since a stressed sound must be pronounced more, not less, strongly. A word cannot be without stress, so stress must have arisen on syllables farther from the end of the word. Later, stress also began to retract from long circumflex endings: šakòs > šakós, pavažòs < pavažõs (Joniškis). Retraction of stress from long circumflex endings may have been conditioned by analogy (in many words and forms, stress had already been retracted from final syllables); the accentuation of forms within paradigms must have had a particularly strong effect. In addition, circumflex endings also shortened in many areas, especially in the Mūša river basin. It is noteworthy that in many cases the shortening of both stressed and unstressed circumflex endings was the same in the eastern Aukštaitian dialects, e.g.: vuverė ~ voverė, báre ~ bãrė, lùnkàs ~ lankõs, blaižas ~ blaizdos (see J. Šukys LKK II 158, 159), pagalàs ~ pagalõs, pradalgis ~ pradalgỹs (Čeniškiai village, Biržai district), ėžs ~ ežỹs, gėnīs ~ genỹs, ãkma ~ akmuõ (Pušalotas). The second stage of stress retraction formed the present-day area of grave–circumflex stress retraction. b) The process of stress retraction § 7. Of course, not all the cases of stress retraction listed above (§ 2–5) arose at the same time. It must be assumed that stress retraction took place in stages, as may be confirmed by the degrees of stress retraction observable today. a) The first degree of stress retraction, in which stress is retracted from a short ending (an originally short ending or a long circumflex ending) to the long penultimate syllable, while a remnant of the old stress is still audible in the final syllable (ãũkšta, sūnūs, see § 4), most likely reflects the very first and oldest beginnings of stress retraction. In this case, as the shortened circumflex ending gradually weakened, the long penultimate syllable could more easily grow stronger and acquire stress, since long syllables are pronounced more firmly than short ones. This may explain how saũsà came to be pronounced with retracted stress, while šakà still lacks retracted stress (see § 3b). 121 Another variant of the beginnings of stress retraction is the retraction found in certain localities in Molėtai district (Suginčiai, Karalinava), from a short ending to the penultimate syllable, which consists of long vowels and diphthongs ie, uo (see § 3c). This case of stress retraction apparently shows that long vowels and diphthongs ie, uo (which, in their processes of change in the dialects, generally do not differ from vowels), by forming a closer phonetic unit, are pronounced more firmly and for longer than the diphthongs a, e, i, u + l, m, n, r and ai, ei, au, ui, whose components do not form such a close unit. The latter retraction of stress is probably not a primary stage in the movement of stress toward the beginning of the word throughout the area of stress retraction, but a phenomenon characteristic only of the southern eastern Aukštaitian dialects. This is suggested by the absence of analogous cases of stress retraction in other dialects (Samogitian, northwestern and central Aukštaitian). b) The further development of stress retraction undoubtedly formed conditional grave-accent retraction (saũsa, but šakà), and later universal grave-accent retraction (šakà, šakù, šak). It is possible that in some localities these two stages of stress retraction (saũsa and šakà) constituted a single phase, that is, in both cases stress was retracted at once. However, the existence of both stages of stress retraction in the eastern Aukštaitian dialects (saũsa šakà — Anykščiai and saũs šak — Biržai) and the northwestern Aukštaitian dialects (saũsa šakà — Šiauliai, saũsa šakà — Šimkaičiai) suggests that at least in part of the dialect region (roughly in the areas where these stages are found today), such stages of stress retraction existed. c) Once established, retracted stress moved from short endings to the beginning of the word, and also affected stress on circumflex endings. Analogy contributed most to the development of this stage of stress retraction, and shortening of endings contributed in part as well (see § 6). d) In the northernmost and westernmost dialects (in some localities), stress is sometimes also retracted to the beginning of the word from acute endings and from within the word (see § 3). This is, of course, a still more recent step on the way toward the formation of fixed stress on the first syllable. § 8. None of the stages of stress retraction mentioned above arose suddenly, all at once. At every stage, the process of stress retraction proceeded gradually, over a long period: final stress gradually weakened, while retracted stress (at the beginning of the word) grew stronger in parallel. This is clearly shown first of all by the Samogitian and Samogitian-adjacent Aukštaitian dialects, where the old stress persists (though already greatly weakened), while retracted stress is also pronounced: vìna ~ vienà, màžà ~ mažà, láukà ~ laukai (see also § 5, 7a). We find the same in the Aukštaitian area, except that here traces of the old (final) stress have been preserved not throughout the whole area, but only along the boundaries of grave (§ 4c, d) and grave–circumflex stress retraction. It is noteworthy that in the eastern Aukštaitian dialects beyond the boundary of grave-stress retraction, when the ending is short, the long vowels and diphthongs ie, uo in the penultimate syllable have not been shortened (as they normally are when unstressed), e.g.: tvorà, sūnùs, storà, vienà, juodà, but tva.rà:s ~ tvorõs, su.nauis, sta.rà:s ~ storõs, ve.nà:s ~ vienõs, ja.dà:s ~ juodõs (Utena); see also Kirtis 116. Here, word-initial (retracted) stress is not yet perceptible, but the failure of the diphthongs ie, uo to become monophthongs and the narrowing of the vowels ē, ā (their development into ė, o) show that these sounds were not shortened, just like stressed sounds (cf. plõnas, vėjas, viẽnas, juõdas — Utena); and if so, then the beginnings of the stress-retraction process must already be present here as well. We find the same phenomenon in the Veliuona dialect, where conditional grave-stress retraction- 122 along the boundary: thin, thick, one, black, but thin, thick, one, black (Vilkija district, Saulėtekis village). After stress was retracted from a long circumflex ending, traces of the old final stress remained in only one place in the eastern Aukštaitian dialects—in the village of Stetiškiai, near Panevėžys (see § 4c). § 9. The process of stress retraction did not begin at the same time in all the places where stress is now retracted. The gradual weakening of stress retraction from the north and northwest toward the south and east alone suggests that stress retraction began first in the northwestern (coastal Samogitian and northern dounininkai) and northern eastern Aukštaitian (the Mūša river basin) parts of the dialect area, and then gradually spread south and east from these locations. a) The following facts show that stress began to be retracted first in the coastal Samogitian dialects. In Lithuanian generally, if stress occurs on the syllable before the old stressed syllable, the new stress on a long syllable always has a circumflex or a medial (which, in terms of pitch movement, is circumflex) intonation. This is the case in the Samogitian dialects (except the coastal variety) and in the northern Aukštaitian dialects that retract stress, for example: bàltà ~ baltà, bàltos ~ baltõs, laũkų ~ laukų (Joniškis), gìva ~ gyvà, jaũna ~ jaũna (Obeliai). In the coastal Samogitian dialect, however, the situation is different: if stress is retracted onto a circumflex root, it receives a sustained (standard-language circumflex) intonation, for example: ùžì ~ užì, žõsės ~ žąsis (Kartena), or a medial intonation (which is simply a shorter sustained intonation), for example: laũkà ~ laukai, ná.ujuõs ~ naujõs (Salantai); if stress is retracted onto an acute root, it receives a falling (standard-language acute) intonation, for example: bá.ltà ~ baltà, bá.ltũos ~ baltõs (Grūšlaukė); pé.rmõs ~ pirmõs, viéžš ~ vėžys (Salantai). This duality in the intonation of retracted stress in coastal Samogitian shows that the coastal Samogitians must have retracted stress at a time when they still pronounced two different intonations in pre-stress syllables: circumflex and acute. We therefore conclude that the coastal Samogitians (like the Latvians) retracted stress before the intonations of unstressed pre-stress syllables had merged, whereas the other Samogitians did so only after that merger. Thus, the coastal Samogitians retracted stress earlier, and all the others later. When the time came for the acute intonation in pre-stress syllables to turn into circumflex, the long pre-stress syllables of coastal Samogitian were already stressed by retraction and therefore could no longer change their intonation, since only the intonations of unstressed syllables had merged: circumflex became generalized in pre-stress syllables, and acute in post-stress syllables, for example: kalnài, but kálnus, vẽžėlės, but vẽžgėlės (cf. K. Būga LKŽ I XIX). That dounininkai retracted stress earlier than dūnininkai is also indicated by the greater lengthening of vowels and diphthongs under retracted stress. For example, in dūnininkai, the intonation of retracted stress in disyllabic words whose two syllables are long is medial, and the length of vowels and diphthongs does not differ from their length when unstressed. For instance, in the first syllables of the words láukù ~ laukų, dǫntù ~ dantų, the diphthongs au, ǫn do not differ in length from the same diphthongs in unstressed syllables, e.g.: laukåms ~ laukåms, dǫntìms ~ dantìms; in dounininkai, by contrast, the first component of the diphthongs is half-long in this case: lá.ukàm ~ laukų, dǫ.ntùm ~ dantų, but laukåms ~ laukåms, dǫntìms ~ dantìms (Salantai). 123 On the basis of the statements set out above, we may conclude that stress retraction must have begun in the northwestern part of the Samogitian dialect area and then spread southward. This conclusion is further supported by the following data from the Samogitian dialects: 1. Retracted stress is pronounced more forcefully by dounininkai than by dūnininkai. 2. In the dūnininkai dialect area, retracted stress weakens from north to south until it disappears completely at the southern edge of the dialect area. 3. In dounininkai, especially in the northern varieties, word-final vowels bearing ancient stress are pronounced more weakly than in standard Lithuanian, and the ancient (secondary-final) stress is less audible, for example: gražǫs ~ gražūs, tǫ? ~ turiù (Salantai). b) In the eastern Aukštaitian dialects, stress retraction probably began first in the Mūša river basin, since the old short endings and old acute endings have completely disappeared here (e.g.: sūns ~ sūnùs, ràs ~ rasà). This indicates that acute endings (and, along with them, the old short endings) began to weaken and disappear earlier in this area; stress must have shifted from these endings to syllables farther from the end. We should assume that the gradual weakening of final-syllable shortening (see Gtr 101) and of stress retraction (see § 7) from the Mūša river basin southward and eastward shows how these phenomena spread from a single center. § 10. From everything stated in § 9, we may conclude that there were two initial centers of stress retraction: in Samogitian, in the northwestern dounininkai area; in Aukštaitian, in the Mūša river basin. Most likely, the areas of stress retraction in coastal Samogitian and the Mūša river basin were not contiguous when this phenomenon first began. Only later did Samogitian stress retraction, moving eastward, meet the wave of eastern Aukštaitian stress retraction moving westward. Yet the gap between the two areas of stress retraction has not been completely closed; this gap is clearly indicated by the narrowing of the area of circumflex-stress retraction in the Šiauliai corridor (see § 3a). The fact that stress retraction in Samogitian and Aukštaitian originated independently of one another is also shown by the differences between their stress-retraction rules: 1. Samogitians always retract stress to the first syllable, whereas Aukštaitians retract it to the second (with a few exceptions; see § 5). 2. To this day, Samogitians have also preserved traces of the old (final) stress (sáusà ~ sausà, mažà ~ mažà, vã-kã ~ vaikai); moreover, in polysyllabic words with retracted stress, stress is distributed across all syllables (blėzdž'n.gà ~ blezdingà) (see § 4). Traces of final stress in Aukštaitian occur only along the margins of conditional and general stress retraction (tõli ~ toli, in Bikūnai village, Anykščiai district; sākė ~ sakė, in Stetiškiai village, Panevėžys district). Thus, depending on the differing nature of stress retraction, we distinguish Samogitian and Aukštaitian stress retraction (see also Gtr 103, 104). The Samogitian and Aukštaitian stress retraction patterns meet in the area of present-day northwestern and central Aukštaitian dialects (Šiupyliai—Joniškis). Therefore, in this area we find features of both stress retraction systems (Samogitian and Aukštaitian), that is, in some cases stress is retracted in two different ways: both to the first syllable and to the penultimate syllable, e.g.: àštuoni and aštuõni, nežinau and nežinau, mãitinu and maitinu, dàlinu and dalinù (Šimkūnai village, Joniškis district); pàbaiga and pabaigà (Eglynlaukis village, Joniškis district); stress may also be distributed over several syllables of a word or concentrated on just one syllable, e.g.: atèìn and ateìn ~ ateina, pàti || pàti~ pati (Šiu- 124 pyliai, Šiauliai district), dárbininkus ‖ dárbininkus (Šimkūnai village, Joniškis district); compare also A. Jonaitītė LKK II 109, 110. However, in the area mentioned, the Samogitian system of stress retraction predominates over the Eastern Aukštaitian system, although in the Akmenė and Kuršėnai districts, on the edge of the Samogitian area, we also find features of the Aukštaitian stress retraction system that have spread there; for example, sometimes only one retracted stress occurs here (ketųru ~ keturių, vėsuokė ~ visokie, Klykoliai), and at times stress is retracted only to the penultimate syllable, e.g.: palėkau ~ palikau, nebėmatau ~ nebematau (Klykoliai). c) The antiquity of stress retraction § 11. The acute accent on a retracted stress in the acute-accented syllable of the Coastal Samogitian dialect already shows that stress was retracted in that area before the accentual leveling of pretonic syllables (see § 9). This fact might suggest that stress retraction began earlier in the Samogitian dialect than in Aukštaitian, where it began later (since in Aukštaitian, retracted stress in acute-accented syllables occurs only with the circumflex (or mid) accent; thus, stress was retracted after the accentual leveling of pretonic syllables). However, the accents of pretonic syllables may have leveled later in Samogitian, since Samogitian has generally preserved clearer (more perceptible) accents than Aukštaitian, especially Eastern Aukštaitian. Thus, stress retraction may have begun at the same time in the Samogitian and Aukštaitian dialects; in Coastal Samogitian, it may have occurred while the acute pretonic accent had not yet changed into a circumflex, whereas in Eastern Aukštaitian (the Mūša river basin) the accents of pretonic syllables may already have been leveled. Therefore, stress retraction in Aukštaitian need not be more recent than in Samogitian. § 12. Stress retraction is undoubtedly more recent than the shortening of acute-accented endings. While acute-accented endings were long, stress could not have been retracted, if for no other reason than that even today stress is not retracted from acute-accented endings (except in areas of the far northwestern part of the stress retraction zone; see § 3); and the process of stress retraction depends on the loss of acute-accented endings that had already shortened (see § 7). On the basis of endings in Slavic loanwords—neuter nouns—K. Būga established that acute-accented endings in Lithuanian shortened around the 13th century (K. Būga LKŽ I XXX—XXXIV). Stress retraction must have begun somewhat later. § 13. a) In Latvian, endings corresponding to acute-accented endings in Lithuanian shortened earlier than in Lithuanian (K. Būga LKŽ I XXXIV). In J. Endzelīns’s view, stress in Latvian was first retracted from short endings (see J. Endzelīns Gr 32—33). Thus, at one time Latvian too had only gravis stress retraction; later, of course, it gradually proceeded to complete stress retraction, that is, to establishing fixed stress on the first syllable. If the process of stress retraction in Latvian was the same as (or similar to) that in Lithuanian dialects, then its causes, too, were apparently the same, namely the shortening of endings. Since endings in Latvian shortened earlier, stress retraction must also have begun earlier there, and at present the establishment of stress on the first syllable has been completed in Latvian, although even in Latvian there are still 125 secondary stresses, which, as in the Samogitian dialect, are probably traces of stress shifting from the end of a word to its beginning (see § 30). Stress retraction, as a result of the shortening of endings, spread across the Latvian-speaking area (of course, not all at once) and must have extended into the Lithuanian-speaking area as well. This is how the history of stress retraction in the eastern Baltic languages can be understood. b) The common roots of stress retraction in Latvian and the northern Lithuanian dialects are also indicated by the fact that this linguistic phenomenon forms a single, continuous area. Some believe that the eastern corner of the Lithuanian-speaking area along the Latvian border does not have stress retraction, which precludes linking its retraction with Latvian stress (see APh IV 33). In fact, however, the eastern corner of the Lithuanian-speaking area (Kamajai, Zarasai, Dusetos, Rimšė, Apsas) also has stress retraction, although it is very weak there (see § 4 d). The further spread of stress retraction in this area (Eastern Aukštaitian and Eastern Dzūkian) was, it seems, impeded by another phonetic process—the shortening of long unstressed sounds, which must have preempted stress retraction: it is harder for stress to shift onto shortened, that is, more weakly pronounced, pretonic syllables (cf. § 8). In our river basin, where long unstressed vowels have shortened, sounds bearing retracted stress have not shortened: dõvan ~ dovanà, giẽsme, ~ giesmẽ, but dovančėla ~ dovanėlė, gedõt ~ giedóti (Pušalotas). In Biržai, where stress is not retracted from long endings, the forms pronounced are viẽn ~ vienà, but venõs ~ vienõs. II. ACCENT SHIFT 1. Rules of accent shift § 14. In Samogitian dialects that retract the accent, words bearing the primary (old) accent also have, in certain cases, a secondary accent called the secondary-shifted accent, which falls on post-tonic syllables. § 15. The secondary-shifted accent occurs in the following cases: 1. In a long syllable (occurring after the primary (old) accent), e.g.: audė ~ audèi, brėzgėls ~ brizgelas ‘brizgìlas’, bùom ~ bùtų, bimbāls ~ bimbalas, dedžiojo ~ didžiojoje, ēržēls ~ erželas ‘eržìlas’, ēsēls ~ ėselas ‘asilas’, galvuožerses ~ galvoviršis ‘viršugalvis’, ēšlēi ~ išleja ‘išlieja’, kēpāls ~ kėpalas, lāpē ~ lapei, lādite ~ láidyti, lēpōnt ~ lipant, mōnì ~ manìe ‘man’, mēdie ~ mėdėje ‘medyje’, nuvažiuōnt ~ nuvažiuojant, vuobōls ~ obuolas ‘obuolys’, pāta-line ~ patalynė, parvažāēven ~ parvažiuojavę, ri'tioji ~ rytójuo, sēnēms ~ seniams, sūnū ~ sūnuo ‘sūnui’, sprāgēls ~ sprāgelas ‘spragìlas’, skrōždieles ~ skruzdėlės, šēršūnc ~ širšūnas, šōni ~ šūnie ‘šuniui’, trēzdesimtē ~ trisdešimti ‘trisdešimts’, vākārs ~ vakaras, vėžōnt ~ vežant, vīrāms ~ vyrams, vēršiū ~ veršiuo ‘veršiui’, vēštā ~ vištai, vėjēne ~ vėjėję ‘vėjyje’, viedārs ~ vėdaras, žōnsēnc ~ žąsinas, žing-dāms ~ žįsdamas. The examples given each contain only one long syllable occurring after the primary (old) accent. If a word has many long syllables after the primary (old) accent, 126 they all bear a secondary accent, e.g.: bånduodãržie ~ bandódaržėję ‘bandodaržyje’, darbúodavomuos ~ darbuodavomós ‘darbuodavomės’, júodalksnes ~ júodalksnias, káustidavom ~ káustydavoma ‘kaustydavome’, mókytúojos ~ mokytojus ‘mokytojas’, pùšimpėlkė ~ Pušynpelkėje (a marshy forest in Šiauduva village, Šilalė district), si̇̀nuotãrpė ~ sienõtarpėje ‘sienotarpėje’, tórpdėrvùonės ~ tórpdirvonis ‘dirvonas, ant kurio mina, džiovina durpes’, žiópčiúodams ~ žiópčiodamas. 2. On a(e) < Baltic ã in words that are compounds or have a postposition attached: a) In the nominative singular forms of pronominal adjectives, e.g.: bál-tãsės ~ báltasis, dūorãsės ~ dõrasis, gẽltuònãsės ~ geltonasis, gẽrãsės ~ gẽrasis, grẽitãsės ~ grẽitasis, káiltãsės ~ káltasis, krẽivãsės ~ krẽivasis, mãžãsės ~ mã-žasis, mẽnkãsės ~ mẽnkasis, sėnãsės ~ sėnasis, áiškãsės ~ áiškiasis, ãštrẽsės ~ ãštriasis, dáilẽsės ~ dailiasis, gẽlẽsės ~ giliasis, grãudẽsės ~ grãudžiasis, jáukẽ-sės ~ jáukiasis, línksmẽsės ~ línksmiasis (lìːn.k.smõs ~ linksmius ‘linksmas’, plã-tẽsės ~ plãčiasis, žõngẽsės ~ žãngiasis (žõː.n.gõs ~ žangus ‘kas sparčiai eina’), aukštẽsnẽsės ~ aukštẽsniasis, dẽdẽsnẽsės ~ didẽsnasis, gerẽsnẽsės ~ gerẽsnasis; naujẽlẽsnẽsės ~ naujẽlẽsnasis, plònẽlẽsnẽsės ~ plonilẽsnasis; aukščẽusẽsės ~ aukščiausiasis, gražẽusẽsės ~ gražiausiasis, jaunẽusẽsės ~ jauniausiasis, lengvẽudẽ-sės ~ lengviausiasis. b) In reflexive 3rd-person verb forms, e.g.: áudãs ~ áudasi ‘audžiasi’, bãrãs ~ bãrasi, dẽrbãs ~ dirbasi, džáugãs ~ džiau̇gasi ‘džiaugiasi’; sùkãs ~ sukias ‘sukis’, nẽškẽs ~ nẽškias. c) In the verb suffixes -av-, -dav- (in the past simple and frequentative tenses), e.g.: júodava ~ júodavo, plūdùrava ~ plūdùravo, šùkava ~ šūkavo, úogava ~ úogavo; dainiúodava ~ dainiuodavo, švẽlpaudava ~ švilpaudavo, táuškindava ~ táuškindavo, vartýdava ~ vartydavo. If a long syllable follows the a of this suffix, the a is unstressed; in this case, the final long syllable of the word bears the secondary accent, e.g.: šùkavóm ~ šúkavoma ‘šūkavome’, úogavóm ~ úo-gavoma ‘uogavome’, lìbdavìot ~ liep­davote. In the case described in § 2b, not all Samogitian dialects that retract the accent have a shifted accent; in the northwestern donininkai dialects, there is no accent shift in this case. § 16. The secondary-shifted accent is usually placed on short vowels in long words, most often in compounds in which three or more syllables follow the primary (old) accent, e.g.: ėkẽtvẽrbãlės ~ eketvirbalis, pãkùpẽtė ~ pãkupetė ‘pakugė’, pãvãkãrė ~ pãvakare, sùrmaĩšẽlės ~ sùrmaišelis, òšpakã-lės ~ užpakalis (see § 21c). In this case, the secondary-shifted accent has arisen on the second syllable after the primary (old) accent. If this syllable is long, there is no secondary-shifted accent on either the first or the third syllable after the primary (old) accent, e.g.: pòžbaprùotės ~ pùsbėprotis. § 17. As all the examples given show, the primary (old) accent always precedes the shifted accent(s). If the primary (old) accent is at the end of a word, that word never has a shifted accent, e.g.: alsavãu ~ alsavau ‘kvėpavau’, prìvežẽms ~ prìvežimas. § 18. In some western Aukštaitian dialects there is also a secondary accent 127 cases reminiscent of the Samogitian secondary retracted stress, e.g.: užžvakar ~ užva- kar, párvažiāva ~ párvažiavo ‘drove across’ (Turbarkas); Fr. Kuršaitis Gr 64, 65 also gives examples of analogous secondary stress, e.g.: pérvaryti, péralkes, pér- gabénti. 2. Intonations of retracted stress § 19. In the Dūnininkai and southern Donininkai dialects (south of Priekulė), syllables bearing retracted stress always have a broken intonation, except in a few cases where the intonation is short; see Švdt § 7–8, 68. § 20. In the Dounininkai and northern Donininkai dialects (north of Priekulė), the broken intonation of retracted stress is pronounced: a) in every long final syllable with retracted stress, e.g.: làpà ~ lapai, vìrà ~ vyrai; kìelė ~ kėlei, làukāu ~ laukiau; làkėnàu ~ lãkinau, làkėnuot ~ lãkinote, pàvàžā ~ pāvažai, gėrūjuo || gėrōjuo ~ gėruoju ‘for the good one’, gerėjė ~ gerėjai; vàkārs ~vãkaras, jōvāls ~ jōvalas, spràgėls ~ sprãgilas (Alsėdžiai, Seda, Telšiai...); b) in roots that are inherently acute (usually, see § 21 b), e.g.: párkõštė ~ párkãsti, párpjāutė ~ párpiauti, párlāužtė ~ párlaužti (Kretinga; see J. Aleksandravičius VVPI MD VIII 131, 132), pòžbrùolis ~ pūšbrolis (Grūšlaukė); c) in roots that are inherently circumflex, e.g.: nùtrėnkė ~ nùtrenkė: trefkė, nùšlàužė ~ nùšliaužė: šliaužė, kélvėitė ~ kélvietė: vietą (Kretinga); cf. § 20 b above; d) in suffixes containing inherently acute long vowels or diphthongs, e.g.: gáudėtė ~ gáudyti, jūdìntė ~ jūdinti, šāukūtė ~ šāukoti ‘to shout’; e) in suffixes containing inherently circumflex long vowels and diphthongs, e.g.: pašūtėlis ~ pasiūtėlis. § 21. In the Dounininkai and northern Donininkai dialects (north of Priekulė), syllables bearing retracted stress are pronounced with sustained intonation in the following cases (see B. Rokaitė, LKK, IV, 144). a) Syllables that are inherently circumflex: 1. word roots: àtvàrsla ~ ãtvarslai: verčia, vārto, pālàukis ~ pāliaukis: liaũkos, parsegōnda ~ pársigando: išsigāndo, pàgàikštis ~ pãgaikštis; 2. suffixes: rùpūžė ~ rūpūžė: kepūžė ‘shaggy hat’ (see LKŽD 390), sālontus ~ Sālantus: Šlavañtas (a lake in Prienai District; see LKŽD 374), ōrkštinė~urkštinė; 3. endings: līgūoje ~ lýgiojo < ligjājā, līgōji ~ lýgįjį < ligjanjīn, līgūoje ~ lýgiąją < ligjanjān, līgūoses ~ lýgiosios < ligjāšjās (plural dative). b) Syllables that are inherently acute: 1. roots of individual words: pālōnge ~ pālangė: lāngas, pālūpis ~ pālūpis: lūpa, pāmùosis ~ pānosis: nōsis, pòspūrė ~ pūspūrė: pūras, pašūostė ~ pāšostė: sōstas; 2. individual endings: dérbontējē ~ those who work, cf. verptējē ~ spinners (Kretinga, see J. Aleksandravičius VVPI MD VIII 138, 139); 3. individual suffixes: peìškūne ~ Paeškūnė (a village in Skuodas District): Eiškūnas (a river in Skuodas District). 128 c) Syllables consisting of the vowels a, e, ė, o < a, e, i, u: 1. vowel a < ã: mažàsis ~ mažasis, pérmãsis ~ pirmasis, padėrptãsis ~ padirbtasis; dùksãva ~ dùksavo, mãinekãva ~ mãinikavo, ãpnãris ~ ãpnasris ‘bridle’, ekė̇tverbalìs ~ ekėtvirbalis, pãrsãkė ~ pársakė ‘told again’, pãvãkãrė ~ pãvakare, ǒšpãkãlis ~ užpakalis; 2. vowel e < ã: dèrbonėtės ~ dirbančiasis (Kretinga, see J. Aleksandravičius VVPI MD VIII 138); 3. vowel e < ė: pàoudėgė ~ pàuodėgė, pòsėsėrė ~ pusseserė, sùrmãisėlis ~ sùrmaisėlis; 4. vowel e < i: dėdervėnė ~ dėdervinė, šòlėnė ~ šulinė; jọdėna ~ judino, lākėna ~ lā kino, párėlėje ~ páriljo ‘soaked through’, cf. also § 23. 5. vowel o < u: ãpōšė ~ ãpūšė. 6. In the Donininkai dialect, especially farther south, the vowels a, e, ę, o, when bearing retracted stress, are pronounced with a short intonation, e.g.: karūmėnė~karūmenė, vėšvanāges~vištvanagis (Priekulė), cf. § 16, 19, but: párėlėje~páriljo, kėmpėnė~kėmpinė (Budrikai, Klaipėda District). § 22. a) In the Dounininkai dialects, the intonation of inherently falling-accent syllables may in many cases have changed to circumflex (§ 21 b) through metatony or analogy. b) Syllables bearing a circumflex intonation under retracted stress can occur only in non-final syllables; the intonation of long final syllables bearing retracted stress is always broken in all Samogitian dialects that have retracted stress, even when the syllable is inherently circumflex, e.g.: lápą ~ lápai, vīrā ~ vīrai; gėrōjūō ~ gėruoju ‘to the good one’ (Alsėdžiai). Apparently, retracted stress in the final syllable is somewhat later in origin, having arisen only after the intonations of post-stressed syllables had become levelled, unless this is the result of analogy. 3. History of stress retraction a) Causes of stress retraction § 23. Retracted stress in the Samogitian dialects apparently arose for several reasons. a) Stress retraction is undoubtedly related to stress withdrawal. When pretonic syllables began to receive stress as a result of stress withdrawal—that is, when the pronunciation of pretonic syllables was strengthened—post-tonic syllables likewise began to be pronounced more strongly by analogy, that is, they acquired retracted stress. The fact that the emergence of retracted stress is related to the emergence of stress withdrawal is shown by the almost complete overlap between the areas of stress withdrawal and retraction (see § 14) and by their having arisen at the same time (see § 27–29). b) The second cause of the emergence of retracted stress was probably the same as that of stress withdrawal: shortening of the word ending. In the Samogitian dialects, a law of compensation in the strength and duration of vowel and diphthong pronunciation was operative, according to which, as the ending shortened, the pronunciation of preceding vowels or diphthongs (those farther from the end of the word) grew stronger; that is, they received stress and, in connection with this, lengthened, e.g.: pãvakãrė <*pàvakãrė, òšpãkãlis <*ùžpakãlis, būdāva <*būdāva. However, in the forms būdavom ~ būdavome, būdavot ~ būdavote, the vowel a in the syllable -da- is no longer 9. Kalbotyros klausimai, IV 129 lengthened, because in this case the compensatory strengthening has affected the long pre-final syllable, which has become final as a result of the shortening of the word ending. § 24. The importance of ending-shortening for the emergence of retracted stress is clearly demonstrated by the inflectional paradigms of nouns and verbs (see B. Rokaitė LKK IV 145), where, for example, the forms, låkėno ∼ låkino < *lakinóo, låkini < *lakinė do not have retracted stress, whereas låkėna ∼ låkino < *lakinā already does. This is understandable, since acute-accented endings must have shortened before stress withdrawal (see § 12), while stress retraction must have taken place somewhat later than stress withdrawal. Thus the shortening (not disappearance) of acute-accented endings could not have played a role in the emergence of retracted stress; retracted stress is apparently related to the change and shortening of circumflex-accented endings. This explains why, among the forms mentioned, only the one that had a circumflex-accented vowel at the end has retracted stress. Compare also the accentuation of noun paradigms (examples from Grušlaukė): Singular Nom. šõlėnė ∼ šūlinė pɔ̀seserė ∼ pusseserė Gen. šõlėnės ∼ šūlinės pɔ̀seserės ∼ pusseserės Dat. šõlėnė ∼ šūlinei pɔ̀seserė ∼ pusseserei Acc. šõlėnę ∼ šūlinę pɔ̀seserę ∼ pusseserę Instr. šõlene ∼ šūline pɔ̀seserė ∼ pussesere Loc. šõlėnie ∼ šūlinėje pãvakārė ∼ pãvakarėje Plural V. šõlėnės ∼ šūlinės pɔ̀seserės ∼ pusseserės K. šõlėnų ∼ šūlinių pɔ̀sesėrų ∼ pusseserių N. šõlėnėms ∼ šūlinėms pɔ̀sesėrėms ∼ pusseserėms G. šõlėnes ∼ šūlines pɔ̀seseres ∼ pusseseres Įn. šõlėnėms ∼ šūlinėmis pɔ̀sesėrėms ∼ pusseserėmis Vt. šõlėnės ∼ šūlinėse § 25. The influence of ending shortening on the emergence of retracted stress is also indicated by the stronger pronunciation of retracted stress in the final syllable when a word has several retracted stresses, e.g.: vålkūotīs ∼ valkioties <*valkiotiesi, pūoãukštė ∼ poaukščiai (see Švdt, § 30, 68 b). This is also understandable: the final syllable in the dialect took over the strength of those word-final sounds that were shortened (reduced) or lost. Incidentally, compensatory vowel lengthening is best evidenced by a, e < ai, ei: as the second component of the diphthong disappeared, it lengthened the first, as in the standard language ą, ę, į, ų < an, en, in, un, where we likewise find compensatory lengthening. The longer pronunciation of vowels a, e < ai, ei than that of other sounds had already been noted by K. Jaunius; see Pocc роб 24. Since these diphthongs have merged into single vowels in word-final position in the Samogitian dialects, where they are used quite frequently, this may also have contributed considerably to the emergence of retracted stress in final syllables. Here one must also bear in mind that in the Samogitian dialect a vowel derived from a diphthong tends not only to retain its original (diphthongal) length, but also to concentrate within itself the full strength of the diphthong’s articulation. In this way, secondary stress arises on a post-stress syllable. 130 § 26. Retracted stress in medial syllables must have arisen by analogy with secondary retraction of stress. This is best demonstrated by the fact that in the dounininkai dialect (and partly on the margins of the northern dūnininkai area), short syllables situated between a syllable with primary—historical stress and a long syllable with secondary—retracted stress also bear retracted stress (weak), e.g.: pãvažã ~ pãvažai, ašarã ~ ašarai, ãtnėšė ~ ãtnėšei, pãrvėžė ~ pãrvėžei ‘you transported’ (Grūšlaukė); cf. also B. Rokaitė, LKK IV 143. The fact that retracted stress is not preserved ancient stress, for example in compound or prefixed words, is shown by its having arisen in word syllables that were formerly unstressed, e.g.: pãvakare ~ pãvakarė, plãutkėpenės ~ plãutkepenės, pùseserė ~ pùsseserė, pãbudėgė ~ pãuodėgė. In all cases, retracted stress on the short vowels a, e, ė, o has arisen in the penultimate syllable, which confirms the view that shortening or loss of the final syllable (final vowels and diphthongs) caused the penultimate syllable to become stronger, that is, led to the emergence of retracted stress; cf. § 23. b) The antiquity of stress retraction § 27. In the dūnininkai dialects, syllables bearing retracted stress are always pronounced with a broken tone, e.g.: ląpã ~ ląpai, pỹlė ~ pỹlei, gẽrasės ~ gẽrasis, jùkās ~ juõkasi ‘laughs’ (see Švdt, § 6–9, 68). In the dounininkai and northern donininkai dialects, syllables bearing retracted stress have either a broken or a falling-rising tone, e.g.: a) vìrūo ~ výruo ‘to the man’, párbėktę ~ párbėgti ‘to run across’, vãsarã ~ vãsarai; b) párlėke ~ párlėkė ‘flew across’ (Kretinga), Pãlėiknis ~ Pãlieknis (a place name in the Endriejavas environs, Rietavas district). The examples show that in the dounininkai dialect the tone of retracted stress depends on the syllable’s original tone. If the tone of a syllable bearing retracted stress was originally rising, it has often remained so to this day (broken in the dialect); cf. párbėktę ~ párbėgti and bėktę ~ bėkti. If it was falling-rising, it remained falling-rising (sustained in the dialect); cf. párėte ~ párėiti, párkėrštę ~ párkirsti, párlėkte ~ párlėkti and lėkte ~ lėkti (Kretinga; see J. Aleksandravičius VVPI MD VIII 131, 132); see also § 19. Post-stress long endings have also preserved the ancient tones, if they are not in the final syllable, e.g.: bãltõje ~ bãltojo, bãltõjų ~ bãltãjų, bãltõ-sės ~ bãltosios (Grūšlaukė); see § 21 a. However, even in the dounininkai dialect there are many cases in which syllables bearing retracted stress have a broken tone in place of the old falling-rising tone, e.g.: párlėke ~ párlėke, but pràlėke ~ pràlėke, nãsvėka ~ nẽsveika (Kretinga; see J. Aleksandravičius VVPI MD VIII 131, 132; cf. also § 20 a). The broken tone may have arisen in place of the old falling-rising tone through analogy, or in individual cases the retracted stress itself may be more recent. Retracted stress and the broken tone in final word syllables (in all cases; see § 21 b) are apparently more recent. Their origin most likely depends on other long- 131 … the retracted stress of post-tonic syllables, as well as stress retraction (see § 22 a) and the monophthongization of final diphthongs (see § 24). Thus, the retracted stress (like retracted stress, see § 9 a) in the Dounininkai and northern Donininkai dialects arose at a time when the intonations of post-tonic syllables were still distinguished and had not yet merged. The intonation of a post-tonic syllable that received stress could no longer change, since in Lithuanian only the intonations of unstressed syllables were merging. In the Dūnininkai dialects, retracted stress must have arisen later—after the merger of the intonations of post-tonic syllables, when all long post-tonic syllables already had a circumflex intonation; it is therefore understandable that, upon receiving retracted stress, a post-tonic syllable had to retain its acute (broken) intonation. § 28. The difference in the intonations of retracted stress in the Dounininkai and Dūnininkai dialects shows that stress retraction began earlier in the northern part of the Žemaitian dialect area and later in the southern part. Since stress retraction must also have begun earlier in the north (this is likewise shown by the intonations of the pretonic syllables under retracted stress; see § 20 a), we must assume that stress retraction and stress shift proceeded in parallel. However, stress shift, if understood as having arisen (at least in part) as a result of stress retraction (see § 23 a), must have followed stress retraction, that is, it occurred somewhat later. § 29. a) One may assume that retracted stress, like stress retraction, first arose in words whose original primary stress fell on short syllables. This is shown by the fact that in this case the secondary-shifted stress is stronger, and the vowels and diphthongs bearing it are more lengthened than in words where the original stress falls on long syllables, e.g.: lėpàm ~ lipama ‘we climb’, dėdėsės ~ didžiasis, rėzdàms ~ rįsdamas; but nèšõnì ~ nešant, áugàm ~ augama ‘we grow’, dūobėls ~ dóbilas (Dūnininkai), cf. § 21. b) The Dounininkai and Donininkai pronounce words with secondary-shifted stress, in which the original stress falls on a, e, as follows: nàktėi ~ nãktie ‘for the night’, vẽstėis ~ vėstis (Kretinga), làpà ~ lãpai, žẽmė ~ žẽmei (Normantai, Klaipėda district), whereas among the Dūnininkai the same words are pronounced nãkti ~ nãktie, vẽstis ~ vėsties (Laukuva). Thus, in the Dounininkai forms the a, e bearing original primary stress are short, whereas among the Dūnininkai they are long. On this basis, one may assume that secondary-shifted stress among the Dounininkai arose before short stressed a, e were lengthened. Weakened by the emergence of secondary-shifted stress, the original primary stress could no longer lengthen these vowels. Secondary-shifted stress among the Dūnininkai must have arisen after a, e had lengthened. c) The vowels ė, o (~ i, u) bearing original primary stress have lengthened in both the Dounininkai and Dūnininkai (e.g.: kė.ta ~ kitą, šò.va ~ siūvo), but in words with secondary-shifted stress these vowels have remained short (e.g., among the Dūnininkai stėprė ~ stipriai, sòkrė ~ sūkriai; in Kretinga: lėpàm ~ lipama ‘we climb’, bòvòm ~ būvoma ‘we were’). Thus, among both the Dounininkai and Dūnininkai, secondary-shifted stress must have arisen before the lengthening of ė, o (cf. § 22a). § 30. Latvian also has secondary stress, which in many respects resembles the secondary-shifted and secondary-retracted stress of the Žemaitian dialects. As in the Žemaitian dialects, syllables bearing Latvian secondary stress have in many cases retained the old intonation, e.g.: dẽrības: 132 derýbos, ciemĩns: kaimýnas, vasarãjis: vasarójas, vãcietis: vókietis, zinãms: žìno-mas, žinóti, aklė̃: aklėjì, aklùos: aklúosius (see J. Endzelīns Gr. 41 ff.), but in Latvian, apparently because of the system, both secondary stress and its intonations have mostly been leveled and made uniform (see J. Endzelīns Gr. 32, 33). Thus, the ancient intonations of secondary stress preserved in Latvian and in the Žemaitian dialects provide further evidence that first-syllable stress arose by the same process both in Latvian and in the northern Lithuanian dialects. Moreover, the greater stabilization of the system of secondary stress (and intonations) in Latvian indicates that the present-day Latvian stress system is older than that of the northern Lithuanian dialects. This further supports the view that the wave of stress retraction (along with stress shift), as a result of the shortening of acute (and circumflex) endings (the long Latvian endings), spread into the northern Lithuanian dialects from the Latvian-speaking area. Evidently, contacts between the dialects of ancient Žemaitia and those of the Mūša river basin and Latvian dialects were stronger than elsewhere, which made it easier for the first centers of stress retraction to arise here within the Lithuanian-speaking area. Abbreviations APh — Archivum Philologicum, Kaunas, 1930—1938. K. Būga LKŽ — K. Būga, Lietuvių kalbos žodynas, I sąsiuvinis, Kaunas, 1924. K. Būga RR — K. Būga, Selected Writings, I, compiled by Z. Zinkevičius, edited by V. Mažiulis, Vilnius, 1958; II, compiled by Z. Zinkevičius, Vilnius, 1959. J. Endzelynas GF — J. Endzelynas, Baltų kalbų garsai ir formos, Vilnius, 1957. J. Endzelynas Gr — J. Endzelīns, Latviešu valodas gramatika, Rīgā, 1951. Gtr — Vl. Grinaveckis, Shortening of Endings in Lithuanian Dialects. — “Research Papers of the Vilnius State Pedagogical Institute. Lithuanian Language and Literature,” VIII, 97—105, Vilnius, 1959. K. Jaunius — Rossian dialect — Dialectal features of the Lithuanian language in the Raseiniai district (by K. Jaunius). — K. Gukovsky, Description of the Raseiniai district, pp. 20—25, Kovno, 1893. Stress — V. Grinaveckis, Stress in the Northwestern Dūnininkai Dialects. — “Lithuanian Linguistics Questions,” I, 109—118, Vilnius, 1957. F. Kuršaitis Gr — Grammatik der litauischen Sprache von Dr. Friedrich Kurschat, Halle, 1876. LKK — Questions of Lithuanian Linguistics, I, 1957; II, 1959; III, 1960. LKŽD — Pr. Skardžius, Lietuvių kalbos žodžių daryba, Vilnius, 1941. MADA — Proceedings of the Academy of Sciences of the Lithuanian SSR. Series A. švdt — Northwestern Dūnininkai dialects. Švdt — Vl. Grinaveckis, Phonetic Features of the Northwestern Dūnininkai Dialects and Their Development. — “Research Papers of the Vilnius State Pedagogical Institute,” XI, 41—118, Vilnius, 1960. VVPI MD — Research Papers of the Vilnius State Pedagogical Institute. Places from which the material is presented, and the contributors and recorders of the material Alsėdžiai, Plungė District, 19581— G. Sėlenytė. Bikūnai village, Anykščiai District, 1960—O. Žakytė. 1 The number beside a place name indicates the date when the material was recorded or contributed. The examples given in the article from places not mentioned in this list were recorded on site by the author of the article. 133 Čeniškiai village, Biržai District, 1958 — V. Drevinskaitė. Eglynlaukis village, Joniškis District, 1960 — A. Lauciūtė. Endriejavas, Rietavas District, 1956 — E. Martinkutė. Grūslaukė, Kretinga District, 1960 — A. Vaitkutė. Ylakiai, Seda District, 1958 — E. Malukaitė. Jokampolė village, Kelmė District, 1960 — N. Tamutytė. Kamajai, Rokiškis District, 1955 — D. Gargasaítė. Karalinava village, Molėtai District, 1958 — A. Pivoras. Kartena, Kretinga District, 1959 — E. Tamošaitytė. Kulų I village, Skuodas District, 1960 — I. Žutaitė. Pušalotas (Dikoniai village), Joniškėlis District, 1955 — O. Sučvaitė. Saulėtekis village, Vilkija District, 1959 — G. Mozurkaitė. Seda, 1958, 1959 — R. Mickevičiūtė, V. Majausskaitė. Skliaustės village, Seda District, 1958 — D. Jonikaitė. Stetiškės village, Panevėžys District, 1959 — M. Taruškaitė. Suginčiai village, Molėtai District, 1959 — A. Sugintaitė. Šakiai village, Molėtai District, 1958 — E. Lukošiūnaitė. Šimkūnai village, Joniškis District, 1958 — O. Riniūtė. Šiupyliai (Norvaišiai village), Šiauliai District, 1955 — Alb. Vaitkutė. Tyruliai, Radviliškis District, 1960 — V. Galinauskaitė. Uniūnai village, Linkuva District, 1959 — S. Dauparaitė. Židikai, Seda District, 1959 — B. Malukaitė. II The material provided from various locations in Kelmė District (Adomaičiai, Kiaunoriai, Kražiai, Kušleikiai, Pakražantis, Salopelkis, Tytuvėnai, etc.) was recorded in 1959 by a dialectological expedition of the Vilnius State Pedagogical Institute, led by senior lecturers J. Aleksandravičius and V. Grinaveckis and involving 20 students (B. Rokaitė, Ž. Urbavičiūtė, E. Tamošaitytė, I. Ermanytė, P. Katinaite, J. Keliolytė, A. Stunžėnaitė, etc.). Material from Akmenė District (Klykoliai, Papilė, Viekšniai) was also collected in 1959 by a VPI dialectological expedition, in which, besides the author of this article, B. Rokaitė, E. Martinkutė, I. Ermanytė, J. Ilkevičiūtė, A. Verbickaitė, P. Katinaite, and others also took part. RETRACTION AND SHIFT OF STRESS IN THE DIALECTS OF THE LITHUANIAN LANGUAGE В. З. ГРИНАВЕЦКИС Abstract 1. Stress Retraction Stress retraction from the end of a word to its beginning covers the entire northern and northwestern territory of the Lithuanian language. However, the laws of stress retraction are not the same throughout this territory. In the far north, near the Latvian border, in some localities (Akmenė, Joniškis, Žeimelis), stress always shifts from the end of a word to its initial syllables: that is, from a short ending, e.g., vien~vieną 'one'; from a long circumflex ending, e.g., laukė~laukai 'fields'; and from a long acute ending, e.g., vaikam~vaikám 'to the children'. Such a 134 stress retraction occurs in the northwestern part of the Lithuanian-speaking territory and in the vicinity of the city of Klaipėda. Southeast of the territory indicated, stress shifts to the initial syllables only from short (including historically acute) and circumflex endings. This is the law of retraction of grave-circumflex stress. The southeastern boundary of this law’s distribution runs from the coast of the Curonian Lagoon, at the point where the Minija River flows into it, approximately through the following places: Žemaičių Naumiestis, Pajūris, Upyna, Vaiguva, Kurtuvėnai, Kužiai, Šiauliai (to the north), Radviliškis, Baisogala, Upytė, Panevėžys (to the southeast), Vabalninkas, and west of the city of Biržai reaches the Latvian border. South and east of this line, stress is retracted only from short (and historically acute) endings, but this retraction is still variable. In the vicinity of the city of Šiauliai and to its south (Pakapė, Kiaunoriai, Kelmė, Kražiai, Pakražantis, Tytuvėnai, Šiluva), as well as near the city of Biržai, stress is always retracted from a short ending, e.g.: saũsa šakà~sausa šakà 'a dry branch' (Kelmė), saũs šak 'the same' (Biržai). Throughout the rest of the territory where the law of stress retraction from short endings applies, stress is retracted only if the penultimate syllable is long, e.g.: saũsa 'dry', but šakà 'branch' (Viduklė, Čekiškė, Jonava, Ukmergė...). The southeastern boundary of stress retraction from short endings runs approximately through the following places: Upyna, Nemakščiai, Viduklė, Šimkaičiai, Ariogala, Čekiškė, Kulva, Gelvonai, Giedraičiai, Molėtai, Leliūnai, Vyžuonos, Svėdasai, Kamajai, Kriaunos. In the Samogitian dialects, when stress is retracted, voice intensity is distributed across all the syllables of the word, e.g.: dẽinā~dienà 'day', vakãrė~vakarė 'in the evening'. The same distribution of stress across all syllables also occurs in the Aukštaitian dialects adjoining the Samogitian ones, as well as in the areas where the boundary between retraction of acute and acute-circumflex stress runs. The residual stress at the end of a word when stress is retracted to its beginning attests to the gradual nature of the movement of stress from the end of a word to its beginning. In the Samogitian dialects stress is retracted to the first syllable, whereas in the Aukštaitian dialects it is retracted to the penultimate syllable (with minor deviations). One of the most important factors that led to the disappearance of stress at the end of a word is the reduction of ancient endings with strong initial stress, e.g.: *sausã>sausà>sūsà (in the Samogitian dialects) >saũs (the Mūša River basin). It is therefore understandable that the process of stress disappearing at the end of a word began in short (and historically acute) endings. Stress retraction began earliest in the northwestern Samogitian dialects and in the Mūša River basin in the Aukštaitian dialects, and spread from there across the rest of the present-day territory where stress retraction occurs. Thus, at its outset stress retraction had two centers (the northwest of the Samogitian dialects and the Mūša River basin), which apparently did not come into contact with one another; only later did the distribution areas of stress retraction in the Samogitian and Aukštaitian dialects join up. 135 Stress retraction could have arisen only after the shortening of ancient acute endings, that is, after the 13th century. In Latvian, acute endings shortened earlier than in Lithuanian. In J. Endzelīns’s opinion, in Latvian stress was first retracted from the end to the beginning in words whose endings were short or acute by origin (see Gr. 33—32). It follows that Latvian, too, once had retraction only of grave stress, and that only in its subsequent development (through gradual analogical generalization, as also in Lithuanian dialects) did a genuine stress system take shape in Latvian, that is, fixed stress on the first syllable. If the process of retracting stress to the first syllable was the same in Latvian and in Lithuanian dialects, then the causes of retraction must also have been the same—reduction of acute endings (in Latvian, of all long endings). Since long endings shortened earlier in Latvian, stress retraction must also have occurred earlier than in Lithuanian dialects; and since no linguistic phenomenon arises at once throughout an entire language territory but spreads from one or several centers, the wave of stress retraction, as a result of ending reduction, undoubtedly came to the northwest and north of the Lithuanian dialects from Latvia, and then spread over a wider territory. The ancient centers of stress retraction in Lithuanian territory mentioned above apparently had closer contacts at that time with ancient Latvian dialects. 2. Stress shift In addition to retracted stress, Samogitian dialects are characterized by shifted stress, that is, stress that arose in post-stress syllables, e.g.: kàtė~kàtei ‘to the cat’. All long (and sometimes short) post-stress syllables are pronounced with shifted stress. The cause of the emergence of shifted stress was apparently stress retraction. Since after stress retraction many words were already pronounced with stress on every syllable, the force of the voice must also have been redistributed in words unaffected by the stress-retraction law. A second cause of stress shift is the reduction of long circumflex endings. This is demonstrated by the following: in declension or conjugation paradigms, shifted stress occurs in forms whose final syllable was long and circumflex, e.g.: šọ́lėne (< *šūlinē)~šùlinė ‘well’ (nominative singular), šọ́lėnęs (*šūlinēs)~šùlinės (genitive singular), šọ́lėnę (< *šūlinēn)~šùlinę (accusative singular), šọ́lėnių (< *šūlinjū)~šùlinių (genitive plural), but šọ́lėne (< *šūlinēn)~šùline (instrumental singular), šọ́lėnęs (< *šūlinēs)~šùlines (accusative plural). The antiquity of stress shift is evidenced by the intonations with which syllables bearing this stress are pronounced. If a post-stress syllable was pronounced with acute stress, then even today in many cases it is pronounced with broken intonation, which derives from the acute (e.g.: pàrkõste~pàrkąsti ‘to have a snack’, cf. 136 kãsti ‘to bite’) and conversely: if a post-stress syllable was pronounced with circumflex intonation, then in most cases even now, under shifted stress, it is pronounced with strong-median intonation (e.g.: pārsēgōnda ~ pārsigando ‘was frightened’, cf. išsigāndo ‘the same’, lygio-je~lygiojo <*ligjā̃ ‘of the even’), which derives from the circumflex. It follows that shifted stress, like retracted stress, arose before acute intonation became generalized in post-stress syllables and circumflex intonation in pre-stress syllables (as in Latvian). This distinction between intonations in syllables with shifted stress occurs only in northern Samogitian dialects; farther south, the intonation of shifted stress is always broken. This indicates that stress shift began earlier in the north of the Samogitian dialect area and later in the south. Latvian, too, has secondary stresses, which in most cases have also preserved the ancient intonations, e.g.: akliē: Lithuanian aklieji ‘the blind’, ciemīns: Lithuanian kaimynas ‘neighbor’. This is further evidence that the process of forming the accentual system (with stress at the beginning of the word) was similar or comparable in Latvian and in the northern dialects of Lithuanian. Stress Retraction and Shift in Lithuanian Dialects V. GRINAVECKIS Summary I. Retraction of stress Stress retraction from the end of the word to the beginning covers the entire northwestern territory of the Lithuanian language. However, the rules of this retraction are not the same throughout the territory. In the far north, near the Latvian border, in the vicinity of the towns of Akmenė, Joniškis, and Žeimelis, stress is always shifted from the end of the word to the beginning, that is, from a short ending, as in vien ~ vienà ‘alone, one’, from a circumflex ending, as in laukė ~ laukai ‘the fields’, and from an acute-stressed ending, as in vaikam~vaikám ‘to the children’. This retraction is found in the northwestern part of the Lithuanian-language territory and in the vicinity of Klaipėda. In the southeastern direction of the territory mentioned above, stress shifts to the first syllables only from a short ending (and, historically, also from an acute and a circumflex). This is the rule of retraction of gravis-circumflex stress. 137 The southeastern boundary of the distribution of this rule, from the coast of the Curonian Lagoon, where the Minija River flows into it, runs approximately through the following areas: Žemaičių Naumiestis, Pajūris, Upyna, Vaiguva, Kurtuvėnai, Kužiai, Šiauliai (north), Radviliškis, Baisogala, Upytė, Panevėžys (southeast), Vabalninkas; west of the town of Biržai it reaches the Latvian border. South and east of the line mentioned, stress is retracted only from a short ending (and historically from an acute ending), but this retraction can take two forms. In the vicinity of the town of Šiauliai, to its south (Pაკapė, Kiaunoriai, Kelmė, Kražiai, Pakražantis, Tytuvėnai, Šiluva), as well as around the town of Biržai, stress is always retracted from a short ending, e.g. saũsà šakà ~ sausà šakà ‘dry branch’ (Kelmė), saũs šak ‘ibid.’ (Biržai). In the rest of the territory subject to the retraction rule from a short ending, stress is retracted only when the penultimate syllable is long, e.g. saũša ‘dry’, but šakà ‘the branch’ (Viduklė, Čekiškė, Jonava, Ukmergė…). The southeastern boundary of retraction from a short ending runs approximately through the following areas: Upyna, Nemakščiai, Viduklė, Šimkaičiai, Ariogala, Čekiškė, Kulva, Gelvonai, Giedraičiai, Molėtai, Leliūnai, Vyžuonos, Svėdasai, Kamajai, Kriaunos. In the Lower Lithuanian (Samogitian) dialects, in cases of stress retraction, vocal force is distributed across all the syllables of a word, e.g. dẽinà ~ dienà ‘the day’, vàkarè ~ vakarè ‘in the evening’. We also find this distribution of stress in Upper Lithuanian (Aukštaitian) dialects that border the Lower Lithuanian dialects. It is also found in the areas through which the boundary of retraction of acute and gravis-circumflex stress runs. The remnants of stress at the end of the word during the process of stress retraction to the beginning of the word show that stress retraction from the end to the beginning of the word takes place gradually. In the Lower Lithuanian dialects, stress is retracted to the first syllable, but in the Upper Lithuanian dialects, to the penultimate syllable (with a few exceptions). One of the most important factors causing the disappearance of stress from the end of the word is the reduction of archaic acute endings, e.g. *sausá > sausà > sãus² (in the Lower Lithuanian dialects) > saũs (in the basin of the Mūša River). It is therefore understandable that the process of stress disappearing from the end of the word began with short endings (and, historically, with acute endings). Stress retraction began somewhat earlier in the northwestern Lower Lithuanian dialects, and from there spread across the rest of the territory where stress retraction occurs. Thus, stress retraction initially had two centers (the northeastern Lower Lithuanian dialects and the Mūša River basin), which apparently did not come into contact with one another; only later, as their territories expanded, did they unite the Lower and Upper Lithuanian patterns of stress retraction. 138 The retraction of stress could only have appeared after the shortening of the archaic acute endings, i.e. after the 13th century. The shortening of long endings took place somewhat earlier in Latvian than in Lithuanian. According to J. Endzelin, stress in Latvian was retracted primarily from the end of the word to its beginning in words whose endings were short (Gr. 33—32). It follows that Latvian, too, once had only the retraction of a grave-accented stress, and only in its further development (in the process of analogical generalization, as also in the dialects of Lithuanian) did the present-day stress system, i.e. stress on the first syllable, develop in Latvian. If the process of retraction of stress to the first syllable was the same in Latvian and in the Lithuanian dialects, the causes of this retraction should also have been the same—the reduction of (long) endings. Since long endings were shortened earlier, stress retraction should also have taken place much earlier than it did in the Lithuanian dialects. Since no linguistic phenomenon can arise suddenly throughout an entire territory, but spreads from one or several centers, the retraction of stress, as a result of the reduction of acute-stressed endings, undoubtedly came from Latvia into the northwestern territories and the northern dialects of Lithuanian, and subsequently spread over a larger territory. The centers of the emergence of stress retraction in the territory of Lithuanian mentioned above apparently maintained close contact with the archaic Latvian dialects. II. Stress Shift In addition to retracted stress, stress shift is characteristic of the Lower Lithuanian dialects, i.e. stress that arises after a stressed syllable, for example: Ràte ~ kãtei, ‘to the cat’ (D.).¹ With shifted stress, all long (and sometimes short) syllables occurring after a stressed syllable are pronounced with stress. The cause of the emergence of shifted stress was apparently stress retraction. Since after stress retraction many words were pronounced with stress on every syllable, the force of the voice should also have been distributed among the syllables of words unaffected by stress retraction. The second cause of stress shift is the reduction of long circumflex endings. This is demonstrated as follows: in the paradigm of declensions or conjugations, shifted stress occurs in those forms whose final syllable was long and circumflex, for example šòłẽnę < *šùlinē ~ šùline ‘well, N. Sing.’ šòłẽnes < *šùlinēs ~ šùlines ‘G. Sing.’, šòłẽnę < *šùlinēn ~ šùline ‘Akk. Sing.’ šòłẽnu < *šùlinjun ~ šùliniu ‘G. Pl.’, but šò-łẽnę < *šùlinēn ~ šùline ‘with the well’, šò.łẽnes < ~ šùlinēns ~ šùlines ‘Akk. Pl.’ The intonation with which words bearing this stress are pronounced testifies to the age of stress shift. When the post-stressed 139 syllable was pronounced with acute stress, even today this syllable is pronounced with broken tone, which derives from an acute intonation (e.g.: pãrkõste ~ pãrkašti ‘to bite through’, compare kãšti ‘to bite’); conversely, if the post-stressed syllable was pronounced with circumflex stress, then even now this syllable is usually pronounced with shifted stress and sustained tone (e.g.: pãrsẽgõnda ~ pársigan̄do ‘was frightened’, compare išsigãndo ‘id.’ lĩgõje ~ lygiojo < *lĩgjãjã ‘of the same…’), which derives from a circumflex. It follows that both shifted and retracted stress arose before acute intonation became generalized on post-stressed syllables and circumflex intonation on pre-stressed syllables. Such a difference in intonation on syllables with shifted stress occurs only in the northern Lower Lithuanian dialects; farther south, shifted stress always has a broken tone. This indicates that stress shift began earlier in the Lower Lithuanian dialects than in the south. Latvian also has secondary stress, which has usually preserved the archaic intonation, for example aklīē: Lithuanian aklieji ‘the blind’, ciemīns; Lithuanian kaimymas ‘neighbor’. This again attests that the process of forming the accentuation system (with stress at the beginning of the word) was the same (or similar) in Latvian and in the northern dialects of Lithuanian.