A critical lexicostatistical examination of Ancient and Modern Greek and Tsakonian Текст научной статьи по специальности «Языкознание и литературоведение»

Applied Linguistics

UDC: 811.14 DOI: 10.33910/2687-0215-2019-1-1-18-68

A critical lexicostatistical examination of Ancient and Modern Greek and Tsakonian1

N. Nicholas®1

1 Independent researcher, 440 Collins Str., Melbourne VIC 3166, Australia

Abstract. This article provides a lexicostatistical comparison of Ancient and Modern Greek Swadesh-100 vocabulary with data from the three recorded dialects of Tsakonian: Southern, Northern, and Propontis. Propontis Tsakonian (now extinct) has undergone the most influence from Modern Greek; Northern Tsakonian is known to have undergone more influence than Southern. Tsakonian is renowned for its Doric heritage, and there are some startling archaisms in its core vocabulary; but its lexicon overall takes Early Modern Greek rather than Doric or even Attic Greek as its departure point. Tsakonian phonology is distinctive compared to Modern Greek, which helps identify loanwords readily; the phonological developments that led from Ancient to Modern Greek, and from Ancient and Modern Greek to Tsakonian, are discussed in some detail. The etymologies of the Tsakonian forms in the Swadesh-100 vocabulary are also discussed in detail. There is a high number of cognates between Modern Greek and Tsakonian, that observe Tsakonian phonology, as well as a significant number of clear loanwords from Modern Greek that do not. Previous lexicostatistical studies on Tsakonian are examined, including the necessity for sound etymological analysis, and the challenges in identifying the primary term for a wordlist item; but also the fragility of interpreting the same etymological data, depending on one's default assumptions about the relation between the two variants.

Keywords: Ancient Greek, Modern Greek, Tsakonian, etymology, lexicostatistics.

T0ov Ou^rcepxo nepvo (1870-1946), to ^aaxopZiKO xZai vor|xe veoypa^^aKixZe, e/Ki^ou

1. Introduction

In the development of lexicostatistical and glottochronological theory, a prominent role has been given to the evidence from Greek, as a language with a long written tradition. Until now, however, the lexicostatistical data used for Greek has not been published; nor has it been subject to the kind of scrutiny exemplified by Bergland and Vogt's (1962) classic study.

While glottochronology is not generally considered reliable, lexicostatistics is still used to give a rough guide of how closely related languages are, especially in the absence of more reliable data (e. g. in Amerindian, Australian, and Papuan linguistics). So long as it is taken as only a rough guide, it can shed some light. In this study, I use lexicostatistics to give such a rough guide on how closely related Tsakonian is to Standard Modern Greek. Judgements on how cognate forms are involve more subtle judgement than is usually done in lexicostatistical studies: I expose those difficulties in arriving at such judgements here, to foreground how problematic they are, but also to illustrate the ways in which more mainstream forms of Greek have influenced Tsakonian at various stages.

Tsakonian has three dialects: Northern, Southern, and Propontis Tsakonian — formerly spoken by a colony in North-Eastern Turkey, and heavily influenced by Thracian Greek, a dialect much closer to Standard Modern Greek. There is thus a range of data to draw on for a lexicosta-tistical investigation.

1 This paper was first written in 1996, and has circulated as a draft. My thanks to Maxim Kisilier and Nikos Liosis for helping finalise it for publication.

2. Tsakonian

Tsakonian is a now moribund language, spoken by perhaps 1000 people in the Peloponnese, and formerly in two villages in North-Eastern Turkey. Tsakonian is not mutually intelligible with Standard Modern Greek, and deviates from Greek noticeably in its morphology. After some debate earlier this century, it has become accepted that Tsakonian is descended from Doric, the Ancient Greek dialect associated with Sparta, and not Attic-Ionic, the dialect which was to give rise to all other Modern Greek dialects, via Hellenistic Koine. Indeed, the word Tsakonian is often derived from Laconian, Laconia being the region in which Sparta was situated. (Caratzas (1976) has more convincingly concluded that it derives from SiàKovoi, and refers to the military office often given to Tsakonians.) The association was already made by the Byzantines; thus, the Byzantine historian George Pachymeres (I 309) refers to "many of the Laconians, who are also called Tsakonians in corrupted form" (a^oi xe nÀeîaxoi sk iœv AaKœvœv, ouç Kaî TÇàKœvaç napa^Geipovxeç ëÀeyov...) A good deal of the lexical evidence for the association with Doric is adduced in this paper; one of the major questions such a study could help resolve — and a question which does not seem to have been raised in the literature until now — is to what extent Tsakonian derives from Doric proper, as opposed to a Doric-coloured Hellenistic Koine, as was the case with Italiot Greek.2

Tsakonian can be divided into three dialects. Southern Tsakonian (Pernot 1934) includes the villages of Melana, Prastos, Tiros, the smaller villages of Pramatefti and Sapunakeika, and the town of Leonidio (vernacular: Lenidi; Tsakonian: AgieLidi). Northern Tsakonian (Costa-kis 1951) includes the villages of Sitena and Kastanitsa, and formerly the town of Prastos. Standard Modern Greek made its presence known much more strongly in these villages than in Southern Tsakonia, since primary schools opened up in the area much earlier, and the Kas-tanitsiots were very mobile as a result of their trade as travelling whitewashers. In addition, Northern Tsakonia was not as geographically isolated:

Kastanitsa seul a un accès plus facile vers certaines régions de l'intérieur, ce qui me paraît expliquer et le caractère plus moderne du tsakonien qu'on y parle et le fait que celui-ci n'a pas pratiqué certains changements qu'on observe à Lénidi; les noms de famille indiquent aussi qu'il y a eu là un afflux de gens du dehors. (Pernot 1934, 139) (The settlers in Northern Tsakonia (13th-14th century ad) appear to have been Epirots: Costakis 1951, 59.)

As a result, Northern Tsakonian is closer to Standard Modern Greek than Southern Tsakonian is, both lexically and phonologically, and the language has retreated much more rapidly; indeed, Haralambopoulos (1980, 7) reports he was unable to find consultants who could offer more than isolated words as far back as 1971. The date on which the two metropolitan dialects diverged is unknown, but Turkish traveller Evliya Çelebi, who visited the area in 1668, gives both a northern and a southern form for 'house' (Costakis 1951, 153): tanjala (xav xaéÀa tan 'tsela) 'the house' (Northern) vs. tanja (xav xaea tan 'tsea) 'the house' (Southern)—the elision of /l/ before back vowels is the main isogloss separating the two dialects. Pernot (1934, 503) prefers to see in this the singular / plural contrast evident in the glosses: /tha 'teele/ 'in the houses', [than 'dzea] 'in the house'. However, he believes the data was given to Çelebi by a Kastanitsiot, and accepts that the dialects had diverged by that time. Tzitzilis (in prep, cited in Liosis 2007, 37) believes Çelebi's data represents a distinct, now extinct dialect of Tsakonian, which he calls Papadianika after the nearby village, and which Liosis calls Western Tsakonian.

2 In that regard, cf. Pernot (1934, 214) on the etymology of Ar|phi 'relativiser': "La forme ancienne nfj [pe:i; Attic-Ionic] d'où on le [phi] tire est plus indiquée [Attic-Ionic /e:i/ > Modern Greek /i/]. Comme on ne trouve pas ici le dorien nâ [pâ:i > *pha], il y faut voir une preuve de plus de la pénétration en Laconie de la koine ancienne."

The third dialect, Propontis Tsakonian, was formerly spoken in the villages of Vatka (Musatsa) and Havoutsi (now Havutga and Misakga), on the mouth of the Gonen (Aesepus) River, in the Propontis (Sea of Marmara) in North-Eastern Turkey. The inhabitants of these villages were subject to the 1922 population exchanges between Turkey and Greece, although they had already ceased speaking Tsakonian as their primary language in 1914, when they were internally exiled with other ethnic Greeks at the outbreak of World War I (Costakis 1986, X). Although Propontis Tsakonian is grammatically more conservative than metropolitan Tsakonian, morphologically and lexically it has been influenced significantly by Thracian, the dialect of Greeks in the surrounding area, which is much closer to Standard Modern Greek. The dialect has now died out.

The date on which the Tsakonians emigrated to the Propontis is under dispute. Koukoules (1924) argued for a 13th century ad emigration on the basis of historical evidence; the Byzantine historian George Pachymeres says explicitly that Emperor Michael VIII resettled Tsakonians in Constantinople. Costakis (1951, 151-155) argues for a much later date, possibly the time of the Orloff revolt (1770-1780), given what he considers to be the good preservation of archaic features of Tsakonian. He also adduces Turkish traveller Evliya Qelebi's information that Tsakonian was spoken far south of Leonidio in 1668, in a region since settled by Arvanites (confirmed by Scutt 1912-13, 139) — and which includes a village called Vatika. As Scutt (1912-13, 138) discusses, the 14th century Chronicle of Morea includes in Tsakonia villages further south from the modern boundary, like Geraki, but not Vatika or Monemvasia. Tsakonian is spoken in the district of Cynuria, which is now administratively part of Arcadia, but was historically northeastern Laconia. If Tsakonian was spoken as far south as Monemvasia, Papadianika (where ^elebi visited) and Vatika, as the historical evidence hints, then it was originally spoken throughout the eastern half of Laconia.

Fixing a date on the divergence of Tsakonian from Standard Modern Greek is even more problematic. If we accept the received knowledge that Tsakonian is directly descended from Doric, without any Koine admixture, the divergence should be dated at the break-off of Doric from Proto-Greek — some time in the second millennium вс. By 1st century вс, the Ancient Greek dialects were dying out in the face of Hellenistic Koine; Laconian Doric experienced a revival in inscriptions up to 2nd century ad (where it is called Neo-Laconian), with varying degrees of success, but researchers have concluded the revival was largely artificial (Panayiotou 1993).

Phonologically, the divergence must be dated at around 800, the date by which almost all phonological changes and the most substantial morphological changes between Ancient and Modern Greek had been realised. Tsakonian phonology deviates significantly from Standard Modern Greek; nevertheless, most of the phonological changeovers from Ancient to Modern Greek have also taken place in Tsakonian. This does not imply that the languages could not have diverged before 800. Greek voiced and voiceless aspirated stops, for example, had started leniting to voiced and voiceless fricatives in 1st century ad. So with /g, ph, th, kh/ already lenited to /у, f, 0, x/, the remaining stops /d/ and /b/ could have readily lenited to /6/ and /v/ independently in Middle Greek and Proto-Tsakonian after 500, under typological pressure. The major phonological archaism of Tsakonian (with the exception of the Doricism /th/ ^/s/) is и /u/ ^/u/, as opposed to its development in Middle Greek, /y/ ^/i/ (known to have taken place by 11th century ad); Tsakonian shares this feature with other archaic dialects of Greek — notably Old Athenian and Maniot — but it is not universally applied: ш 'io, for example, is the Tsakonian reflex ofAncient Greek ибюр hddo:r (cf. yuvaixa gunaika > youvaixa yu'neka.) Furthermore, as seen below, neither /th/ ^/s/ nor /u/ ^/u/ are particularly widespread processes.

Grammatically, the situation is even more complex. Some innovations of Middle Greek are absent in Tsakonian: the complementiser лшдpos, datable to 2th-6th century ad (Nicholas, 1996);

the negator 5sv den, first used as a reanalysis of ouSsv ouden 'nothing' in 1st century bc; the analogical remodelling of u^sig humeis 'you.pl' into eaeig e'sis to obviate phonological merger with he.meis 'we' (6th century ad, according to Palmer 1980, 184.) The use of a participial periphrasis for the present tense is a feature of Koine abandoned in Standard Modern Greek (where it seems to have died out around 1400), but retained in Tsakonian. Other Middle Greek innovations, however, are present both in Standard Modern Greek and Tsakonian: the use of a locative as a relativiser (t^phi; cf. MG noupu; first instance 5th century ad); the use of a volitive future (0a da; volitive futures appear around 1st century ad, although they become the predominant expression of future tense only around 15th century ad); infinitive loss (!0th-16th century ad).

The historical endpoint for the divergence of Tsakonian from Standard Modern Greek seems to be the Slavic invasions of the Peloponnese in the eighth centuty. The first reference to Tsako-nians is considered to be Constantine Porphyrogenitus' De Arte Imperiando (224) (ca. 950), in which he mentions that

the inhabitants of the district of Maina are not from the breed of the aforementioned Slavs, but are of the older Greeks, who are to this day called Hellenes (pagans) by the locals for being pagans in time past and worshippers of idols, like the Hellenes of old, and were baptised and became Christians during the reign of the late Basil (867-886). ("Iaxeov oxi oi xoC Kaaxpou x^c; Ma'ivn? oiK^xopec; ouk eialv ano x^c; yeveac; xrnv npopp^evxwv xkmpwv, a^' ¿k xrnv naAaioxeprnv Trn^airnv, oi Kai |expi xoC vCv napa xrnv evxomwv "E^nve; npoaayopeiovxai Sia xo ev xotc; nponaAaiotc; xpovoi; eiSwAoMxpac; eivai Kai npoaKuvrpcag xrnv e'tSmAwv Kaxa xou; naAaiou; "E^.r|vas, o'ixiveg enl x^; PaaiAeia; xoC aoiSi|ou BaaiAeiou PanxiCT0evxe; xpiaxiavol yeyovaaiv.)

Mani is a distinct region of the Peloponnese, which has remained isolated and distinct from its neighbours in modern times; but scholars routinely assume the passage also refers to the even more linguistically isolated Tsakonians. The fact that the Tsakonians remained pagan for so long is indicative of long-time isolation from mainstream Greek society.

In later Byzantium, Tsakonians were frequently conscripted into the Byzantine army, and were known of in Constantinople; hence the routine references to xZaKwveg or xZeKwvec; as border guards. The first explicit allusion we have to Tsakonian not being intelligible by Standard Modern Greek speakers comes in the 15th century satire Mazaris' Sojourn in Hades (quoted in Costakis 1951, 26):

I thought I would myself turn into a barbarian, just like the Laconians have become barbarians, and are now called Tsakonians. (AeSoiKa [...] ilva PapPapw0w Kai auxo;, wanep apa PePapPapwvxai ye oi AaKwve; Kai vCv KeK^vxai TaaKwve;)

As Pernot (1934, 240) points out, the words Mazaris goes on to quote as 'barbaric' are actually from Mani — the region we have already seen Porphyrogenitus refer to, which scholars refer to Tsakonia as well; that seems to indicate once again that the term 'Tsakonian' was formerly used with a much broader denotation than nowadays.

But while there is much evidence for a long-lasting separation of Tsakonians from Standard Modern Greek, it is just as true that Tsakonian has undergone gradual and constant contact with Standard Modern Greek for an extended period — earlier than the modern period, since the inception of the Greek state, when Tsakonian started dying out. (Already by the end of the 19th century, Tsakonians sung their folk songs in Standard Modern Greek.) Since, in particular, Tsakonian was spoken much further to the south of Leonidio in times past, there would have been extensive opportunities for contact. This contact is apparent even in the core vocabulary of Tsakonian.

3. Historical Phonology 3.1. Modern Greek

The phonological developments in Modern Greek (often concealed by historical orthography) can be summarised as follows:

• Aspirated consonants have been lenited to unvoiced fricatives: /ph, th, kh/ ^ /f, 0, x/ (ph5:s ^ fos 'light', thdnatos ^ 'danatos 'death', khara ^ xa'ra 'joy').

• Voiced consonants have been lenited to voiced fricatives: /b, d, g/ ^ /v, 5, y/ (barus ^ va'ris 'heavy', dekhomai ^ 'dexome 'to accept', gumnos ^ yim'nos 'naked'). This process failed to take place in clusters, particularly with nasals; e. g. /mb/ ^ /mb/ (embaino: ^ 'mbeno 'to enter').

• The phoneme /h/, occurring only before word-initial vowels, has been lost (ho:ra ^ 'ora 'hour', hopou ^ 'opu 'where'), as has the voicelessness of initial /r/ ^ /r/ (rizda ^ 'riza 'root').

• Clusters have dissimilated in manner, so that stop-stop clusters have lenited to fricative-stop (ktizdo: ^ 'xtizo 'build', pteron ^ fte'ro 'wing'), and fricative-fricative clusters have likewise dissimilated to fricative-stop (with the exception of /sf/) (pHhano: ^ fdano ^ 'ftano 'arrive', khthes ^ xdes ^ xtes 'yesterday', skhizdo: ^ 'sxizo ^ 'skizo 'split', but sphendone: ^ sfen'dona 'slingshot').

• The cluster /zd/ has been lenited to /z/ (ktizdo: ^ 'xtizo 'build').

• Pitch accent has become stress accent.

• Contrastive vowel length has been lost.

• Long mid vowels were phonemically distinct from short; they had also become quantitatively differentiated, becoming raised. Of these, /e:/ regularly became raised to /i/ (agape: ^ a'yapi 'love'), although there are counterexamples (ne:ron ^ nero 'water', kse.ros ^ kse'ros 'dry'). /o:/ became /o/ regularly (o:mos ^ 'omos 'shoulder'), although there are many examples of it also raising, to /u/ (klo:bos ^ kluvi 'cage', trago:ido: ^ trayu'do 'sing', sapo:nion ^ sapuni 'soap').

As there are also sporadic instances of /o/ ^ /u/ in Modern Greek (kombion ^ kum'bi 'button', ropho: ^ rufo 'suck'), the /o:/ ^ /u/ development is considered by most linguists a secondary phenomenon; this is how it is discussed, for instance, by Hatzidakis (1975 [1892], 105). It should be noted, nonetheless, that in MG this raising occurs rather more frequently for Ancient /o:/ than it does for /o/.

• Of the ancient diphthongs, those whose first vowel was long had the second vowel drop out: /a:i, e:i, o:i/ ^ /a, e, o/ (ha:ide:s ^ 'adis 'Hades', zdo:ion ^ 'zo(o) 'animal').

• Diphthongs whose second vowel was /u/ had it fortitioned to a labiodental fricative, assimilating in voicing to the following segment: /au, eu/ ^ /af, av, ef, ev/ (auge: ^ av'yi 'dawn', eutukhe:s ^ efti'xis 'happy'). The diphthong /ou/ had already been monophthongised and raised to /u/ in late Attic — a value it has retained since.

• Diphthongs whose second vowel was /i/ were monophthongised and raised: /ai, ei, oi, yi/ ^ /e, i, y, y/ (haima ^ 'ema 'blood', peina ^ 'pina 'hunger').

• Ancient Greek /u/ fronted into /y/ in Attic, although it remained /u/ in the other dialects of Ancient Greek, including Doric. By Roman times, it had been joined by the reflexes of the diphthongs /oi/ and /yi/. Eventually, /y/ was unrounded to /i/ (e. g. hupnos ^ hypnos ^ 'ipnos 'sleep', moira ^ myra ^ 'mira 'fate'). This change came about rather late in Greek; 11th century ad is the conventional date (Lauritzen 2009 gives evidence it had taken place by 1030), although toponymic loanwords (Korfoi ^ Old French Corfu, Oinoe ^

Turkish Unye) suggest the old pronunciation remained widespread as late as 13th century ad.

• Word-initial unstressed vowels have dropped out (embaino: ^ 'mbeno 'enter', oligos ^ 'liyos 'few'); such vowels present in the contemporary language are usually reimportations from Puristic Greek, although the elision was less frequent for disyllabic and trisyllabic words, for low and high rather than mid vowels, or where it would lead to unacceptable initial clusters (e. g. argos ^ ar'ya 'late', akouo: ^ a'kuo 'hear').

• Unstressed prevocalic /i/ has become a yod (diafora ^ [Sjafo'ra] 'difference', opsaria ^ ['psarja] 'fishes'). As an extension, stressed prevocalic /i/ and /e/ also become yods (paidia ^ [pe'Sja] 'children', eleutheria ^ lefte'rja] 'freedom', ennea ^ [e'pa] 'nine', me.lea: ^ [mi'Aa] 'apple tree'). This process has not taken place in Tsakonian; it has also not taken place in Maniot.

3.2. Tsakonian3

For the most part, Tsakonian vocalism and consonantism are the same as those of Modern Greek. There are a plethora of phonological processes listed in the introductions to Pernot's and Costakis' works, not all of which are regular. In the following, the major processes are listed, including those which constitute archaisms with respect to Modern Greek.

3.2.1. Doricisms

The following processes in Tsakonian have been claimed to be Doricisms — that is, to continue characteristics which separated Doric from other Ancient Greek dialects. As should become clear in the discussion, subjective factors have been at work in claiming Tsakonian to be directly descended from Doric; Pernot's critical discussion of these claims was crucial in forestalling their overly enthusiastic acceptance.

1. /0/ /s/ (*dy'yatir ST 'sati PT 'sae 'daughter', *'deros NT 'sere ST 'sen 'harvest', *di'lazo ST si'lindu 'to suckle', *kri'di ^ ST PT 'krisa 'barley'). The rendering of Ancient Greek 0 /th/ as Doric a was first attested at the beginning of 4th century bc (Palmer 1980, 120); e. g. adAaaaa salassa is written for 0dAaaaa thalassa, MG 'dalasa ST 'daasa 'sea' (Hatzidakis 1989 [1905], 365).

There has been controversy over the phonetic value of this a. The obvious process is for /th/ to have lenited to /0/ earlier in Doric than in the other dialects of Greek, which would have been orthographically ill-equipped to denote the new sound. Hatzidakis' counterargument, that the Laconians would not represent a new sound with the same letter as a sound /s/ still extant in their dialect, but would have invented a new letter, is unconvincing. So is the argument that words later borrowed into Tsakonian from MG retain their /0/. However, the (phonologically more plausible) /0/ interpretation of Laconian a requires the process /th/> /0/> /s/ to have been completed in Proto-Tsakonian before it started borrowing MG words with /0/. Note that initial /0/ is frequently retained in Tk in words which cannot have been borrowed from MG — e. g. ST 0uou 'diu 'to slaughter' < AG 0uw thuo: 'to offer a burnt sacrifice; to kill' (MG a^dZw 'sfazo).

3 The following abbreviations are used: M — Melana; T — Tiros; P — Prastos; K — Kastanitsa; S — Sitena; V — Vatka; H — Havoutsi; NT — Northern Tsakonian; ST — Southern Tsakonian; PT — Propontis Tsakonian; MT — Metropolitan (Peloponnesian) Tsakonian; Tk — Tsakonian; Â — Âkerblad (1785-88 or 1796-97); Ç — Çelebi (1668); Vi — Villoison published word list (1785-86); D — Doric, typically as given in Hesychius; C — Costakis' (1986) etymology; MG — Standard Modern Greek; EMG — Early Modern (Late Byzantine) Greek; AG — General Ancient Greek (including Attic and Doric); IG — (Attic-)Ionic Greek; IE — Proto-Indo-European. Text references are given as follows: Diet — texts given in appendix to Costakis (1986); Samp — Costakis (1980); Har — Haralambopoulos (1981); Thus — Houpis (1990); Mal — Houpis (1993); Kar — Karaliotis (1969); Mak — Makris (1952). If a word is not given a location, it is generally used in the region; thus, no location in Northern Tsakonian corresponds to S K, in Southern Tsakonian to M P T, and in Propontis Tsakonian to V H.

So as Scutt (1912-13, 152) finds, "the small number of these words makes it unlikely that the change 0 > a was ever a regular law of the dialect, and the only explanation is that at the time when Laconian used a for 0, a number of such words spread to the Tsakonian district, and have since become reduced under the influence of Modern Greek to the few quoted above." After a fashion, Pernot (1934, 132-134) concurs: the number of instances is much smaller than claimed by Deffner, and the only convincing examples precede front vowels, making this an instance of palatalisation (see below). Pernot (1934, 138) concludes that this should be eliminated from the number of Doricisms in Tsakonian "certainement pour une partie des exemples, sinon pour la totalité."

2. /s/ ■ /r/ word-finally (*pos ■pur 'how', *'tinos ■ 'tsuner 'whose?'). Rhotacism was a prominent feature of Neo-Laconian (Pernot 1934, 17). In Modern Tsakonian, word-final /r/ persists only as a euphonic liaison phoneme; in the unmarked case, it drops out. This seems to explain why word-final /s/ has dropped out in Tsakonian.

3. /s:/ ■ /a/. One of the most salient features of Doric was that it used long /a/ where other Greek dialects used long /e/; for example, Doric used gâ: for Ionic gê: 'earth'. This Doricism undergoes what Pernot (1934, 18) calls "persistance et extension"; e. g. ST a 'mati PT a 'mai corresponds to Ionic hê: mê:tê:r and MG i mi'tera 'the mother'. The domain where Doric /a/ is most prominent is in noun morphology: while MG uses both /i/< /s:/ and /a/ as feminine noun endings, Tsakonian uses only /a/—to the extent that Tsakonians hypercorrect MG -/a/ nouns to -/i/ (e. g. u'ri for MG u'ra Tsakonian nu'ra 'tail') (Pernot 1934, 42). This /a/ is less widespread in other parts of the word.

4. /z/ ■ /nd/ in verb endings (*fo'nazon ■ ST fo'pandu PT fo'nazo 'to shout', *dil'azon ■ ST si'lindu 'to suckle'). This also occurs in the word *'riza ■ ST 'Jinda PT 'riza 'root'. Scutt (1912-13, 156) believes that this change, without parallel in Modern Greek dialects, is a further development of Doric */zd/ ■ /dd/; see also Pernot (1934, 106, 122), who postulates the intermediate stage /00/. The phenomenon does not apply to PT.

3.2.2. Archaisms

The following processes in Tsakonian appear to be more archaic than their equivalents

in Standard Modern Greek.

1. The Tsakonian reflex ofAncient Greek /u/ is not /i/, but /ju/ after a coronal, and /u/ otherwise (*'lykos ■ *'ljukos ■ ST '¿uko 'wolf', *ty'ros ■ tju'ros ■ ST [cu're] 'cheese', *'kyon ■ ST NT 'kue 'dog', *'mya ■ ST PT 'muza 'fly'). In this way, Tsakonian keeps Middle Greek /y/ distinct from /i/; this also occurs in the Old Athenian-Maniot group of Modern Greek dialects, and sporadically in Standard Modern Greek itself (*physkê: ■ fuska 'bubble' (cf. Tsakonian fukha 'belly'), *kyllos ■ ku'los 'maimed', *kytion ■ ku'ti 'box', *mystaks ■ mu'staki 'moustache') — although most instances can be considered later developments after a labial or velar, rather than survivals (Hatzidakis 1990 [1907], 295). It does not seem this tendency regularly extends to the other Ancient Greek phoneme realised as /y/ in Middle Greek—/oi/: the regular reflex for this is /i/, as in MG (koitê: ■ *'kyti ■ ST 'teita 'bed', poio: ■ *py'o ■ ST NT piu 'to do'), although examples with /(j)u/ are extant (*khoiros ■ *'xyros ■ ST NT PT 'xjure 'pig'; Costakis (1951, 37) calls this example "unique in Tsakonian.")

The palatalisation of 'xjure is inconsistent with the rule for /u/, and also occurs for other instances where /ju/ is preceded by a non-coronal — both of /oi/ (heptakoilion ■ *epta'kyli ■ ST fia'teuli 'fecund (of vine)') and /u/ (kurios ■ *'kyrios ■ *'kjuri ■ ST NT 'teuri 'father', gurizdo:n ■ *yy'rizon ■ *yju'rizu ■ ST [ju'rizu] PT [ju'rizo] 'to turn'). Scutt (1912-13, 146) considers this evidence of two original phonemes: Tsakonian retains the Doric pronunciation of u as /u/, as is evident in Hesychius, notwithstanding his vacillation between (inscriptional?) u and (transcribed?)

ou: the proto-form for a regular word like 'Auku was likelier *'lukos than *'lykos, with the palatalisation caused by the consonant rather than any frontness in the vowel. The reflex of /oi/ and the instances in 'teuri and [ju'rizu], on the other hand, where the /y/-sound palatalises a velar, mean that the proto-form here was a front vowel, which dates from later in the language. This is either Middle Greek /y/ or Modern Greek /i/; the former is likelier, although Scutt insists on calling it a i sound.

After extensive discussion, Pernot (1934, 115) concludes that the words where /u/ reflects an old Doric /u/ (like *kuan'a ^ ST NT ku'vane 'black') "ne dépassent pas la dizaine et probablement ne l'atteignent pas." He finds that the development of modern /u, ju/ is not a survival of Doric (unlike Scutt), but a later development from mediaeval /y/, which also affected /i/ in certain contexts (* 'strive)n ^ ST 'fufu NT 'stfufu PT 'stufo 'to twist', *olis'denon ^ ST Au'tfenu 'to slip', *ksifion ^ ST tfu'Oi 'awl', *'otinos ^ ST 'otsune NT 'otsunu 'whose') — although what that context is is not clear.

Instances displaying /i/, the Standard Modern Greek reflex of Ancient Greek /u/, are not at all infrequent. The instances are dismissed by both Scutt (1912-13, 144) and Hatzidakis as loans from MG — but this does not ring true, as these words include core vocabulary (* 'ydor ST NT 'io 'water', *y'os ST NT PT i'ze 'son', *'ypnos ST 'ipre PT 'ipne 'sleep', *'9yon ^ ST 'dio 'to slaughter'). Indeed, in the case of 'io and 'dio, and other instances like *'ollyme ^ ST olin'dumene 'to get lost; to howl', *ta'nyzo n ^ ST ta'pindu NT ta'pizu PT ta'pizo 'to stretch', *'drymbi ^ ST 'fipji NT 'Qripji 'ripe olive' (Pernot 1934, 110), the words had probably died out from MG by the time the transition /y/ ^ /i/ was complete after 11th century ad; the Modern word for 'water', ne'ro, first appears as early as 2nd century bc. In MG, in fact, /y/ has gone to /u/ in the reflex of *'drymbi: 'drumba. In Tsakonian, /y/ ^ /i/ occurs regularly in the prefix syn- and word-initially; Pernot (1934, 111) finds only two exceptions, *ye'tia 'raininess' ^ ST [ju'cia] 'wet weather' and *'yli 'matter' ^ ST NT 'uAi PT 'luni 'silt' (Deffner's etymology, uncontested by Pernot; C derives it from *i'lys 'mud'), which appears to be a loan in its -i ending (Pernot 1934, 364), and which is found in that form in other Greek dialects (Cythera) (Hatzidakis 1990 [1907], 278).

2. Word-initial /r/ becomes /JV (*'ravon ^ ST fafu 'to sew', *'riza ^ ST 'finda 'root'). The easiest way to explain this development seems to be by appeal to the Ancient Greek voiceless allophone of /r/, [r ], which only occurred word-initially. Pernot (1934, 100) prefers to treat this as a modern phenomenon, related to initial /tr/> /JV (see § 3.2.3.1a): he posits initial /r/ had /t/ preposed as a dissimilatory epenthesis after preceding /n/. For example, he derives 'finda as *tin 'rinda ^ tin 'trinda ^ tin 'tfinda ^ tin 'finda ^ ti 'finda 'the root'. This phenomenon does not apply to PT.

3. Classical /o:/ goes regularly to /u/ rather than /o/ in Tsakonian (* poma ^ ST 'puma 'plug', *'xora ^ ST 'xura 'field', *'yrafon ^ ST NT 'yrafu 'to write'). This would indicate that Proto-Tsakonian kept /o:/ > */o / and /o/ apart, as Hatzidakis (1989 [1905], 635 and elsewhere) has claimed. The counterargument raised for Standard Modern Greek also applies for Tsakonian, however: there are instances where Classical short / lax / o/ goes to /u/, and these include four words on the Swadesh-100 list: 'podas ^ ST NT 'pua 'foot', 'yona ^ ST 'yuna 'knee', 'stoma ^ ST NT 'thuma 'mouth', and 'onoma ^ ST NT 'onuma 'name'. This counterargument is particularly strong in Scutt (1912-13, 145), who asserts that, where /o:/ ^ /u/ is not morphologically predictable (primarily in verb endings), it has arisen "perhaps under the influence of neighbouring sounds [...] the evidence for [Hatzidakis'] theory can scarcely be said to be satisfactory." Nonetheless, a (near)-minimal pair like *oron ^ ST NT o'ru 'to see', *'ora ^ ST NT 'ura 'hour' corroborates Hatzidakis' surmise.

Hatzidakis (cited in Costakis 1951, 37) considered /o/ ■ /u/ to have been triggered first by adjacent velars or labials, and then extended to liquids and nasals. (This also holds for PT.) For /o:/ ■ /u/ not to be an instance of the same phenomenon, there must be an instance of /o:/ ■ /u/ between two coronal obstruents — the environment Hatzidakis ruled out for /o/ ■ /u/. *ateleio:tos ■ ST a'te faute NT ate'faute 'unfinished' seems to be such an instance. However, the environment forcing /o/ ■ /u/ is rather broad: velars, labials, liquids and nasals can either precede or follow (*'yona ■ ST 'yuna 'knee', *'ofis ■ ST 'udi 'snake', *'poda ■ ST NT 'pua 'foot', *'olos ■ NT PT 'ule 'all'); and the rule is by no means exceptionless (*'yonos ■ ST 'yone 'spawn', *yomfari ■ ST ofari 'boulder', *poros ■ ST NT PT 'pore 'door', *'olos ■ ST PT 'ole 'all').

While the phenomenon applies to PT (*ton ■ dun 'their', *fo'nazi ■ fu'paz 'he shouts', *loyion ■ o'yju 'kinds.gen'), it is much less widespread. In particular, the participial ending used as the citation form for verbs in Tk (AG -o:n) is PT /o/, rather than ST NT /u/. Since in addition Thracian, the MG dialect PT was in contact with, raises unstressed /o/ to /u/, the evidence for the persistence of the phenomenon is weak. 4. There are a few Tsakonian words where /v/ appears to correspond to the Ancient Greek digamma /w/, which had already disappeared from Attic-Ionic; e. g. pa^voç wamnos 'lamb' > *wanne > 'vane, SapeÀoç (SaPeÀoç in Hesychius) dawelos 'torch' > dave'le.

3.2.3. Innovations

The following processes in Tsakonian are innovations with respect to Middle Greek, and have no precedent or equivalent in Standard Modern Greek.

1. One of the most characteristic features of Tsakonian is its avoidance of clusters; these are uniformly replaced by affricates or aspirated stops.

a. Thus, clusters involving a dental and a liquid (/ôr, 0r, tr/) go to /tJV (*'dryas ■ ST NT 'tjua 'oak', *'adropos ■ ST NT 'atfopo 'man', *'trayos ■ ST 'tfao 'he-goat'). Word-initially, /0r/ is further lenited to /JV (*'drinaks ■ ST 'finaka 'pitchfork', *'9refon ■ ST fefu 'to fatten'); this also occurs at times for /tr/ (*'tremon ■ ST 'femu 'to tremble'). This phenomenon does not apply to PT.

b. Clusters involving a sibilant followed by a stop or fricative go to the homorganic aspirated stop (*'spiron ■ ST phiru 'to sow', *is'tos ■ ST NT 'ithe 'sail', *sta'fida ■ PT tha'fea ST *thiai'dia > kHai'dia [chiai'0ia] 'raisin', *epi'asdi ■ ST ekj'athe [e'cathe] 'I seized', *as'kos ■ ST NT a'kho 'sack', *'isxon ■ ST 'ikhu 'to hold'). However, there are instances where /sx/ goes to /sk/, as in Modern Greek (*is'xada ■ ST NT as'ka PT ska'ia 'dry fig'); and /sf/ does not go to /ph/ at all (*sfi'ri ■ ST sfi'ji ST NT PT sfi'ri 'hammer') (Pernot 1934, 128; Pernot thinks *'sfondylos ■ ST phondile 'vertebrum' is a recent importation, and derives from *'spondylos). There is evidence to suggest this phenomenon dates from Proto-Tsakonian: AG àaKoç askos Hesychius àKKop akkor ■ ST NT a'kho 'sack'.

c. Clusters involving a nasal followed by a fricative go to the homorganic aspirated stop (*omfa'los ■ ST PT apha'le 'navel', *yron'9ia ■ ST yroWa ST PT yru'thea 'punch', *'ryyxos ■ ST 'fukho 'nose').

d. The cluster /ks/ goes to the affricate /ts/ (*kse'ros ■ ST NT PT tse're 'dry', *fy'lakso ■ STfi'atsu 'I guard.suBj').

e. The stop-stop cluster /kt/ goes to the aspirate /th/ (*'daktylo ■ ST NT 'datHle 'finger', *'nykta ■ ST NT 'putha PT 'nutha 'night'), although /pt/ goes to /ft/ like MG, instead of /th/ (opto: ■ ST NT ftenu 'to cook' (MG mayi'revo), ptoê.sis ■ ST ftoisi 'fear' (MG fovos),pémptê: ■ ST PT pefia 'Thursday' (MG 'pemti).) Similarly, the fricative-fricative cluster /x0/ goes to /th/ (*dex'do ■ ST de'thu 'I accept.suBj', *vrex'do ■ ST vreVu

'I get wet.suBj'), although /f0/ goes to /ft/ like MG, instead of /th/ (*fdo'no ^ STfto'nu 'to envy' (MG zi'levo) — but *mazefdume ^ PT mazu'thume ST mazu'dume 'we get together.suBj')

f. Clusters whose second element is /l/ frequently lenite it to /r/ (*pla'ty ^ STpra'kju PT pla'ti, pwa'ti 'wide', *'kleftis ^ ST NT 'krefta 'thief', *'ylosa ^ ST NT 'yrusa 'tongue', *ax'lades ^ ST ax'rae PT 'xrae, xra'ie 'pears'). This also occurs for clusters whose second element is /n/ (*'ypnos ^ ST 'ipre 'sleep', *'texni ^ ST 'texra 'craft'). This phenomenon is severely restricted in PT.

g. Clusters involving /r/ and a stop or /5/ go to the homorganic nasal + voiced stop cluster (*skorpi'os 'scorpion ^ khom'bio ST 'spider' PT 'mole', *'artos ^ ST NT 'ande 'bread', *'arka ST NT PT 'ayga 'I took', *por'di STphun'da 'fart').

h. There is evidence suggesting that AG /kk/> Tk /kh/: AG sakkos ^ ST 'sakho 'sack', AG kokkos ST NT PT 'kokho 'fava bean', AG lakkos ST PT 'akho 'hole', Latin accumbo MG akum'bo NT akum'bixu ST khuy'gixu 'to touch, to lean on' (Costakis 1951, 50, 61-62). This phenomenon may even date from Proto-Tsakonian: Costakis cites Phrynichus (2nd century ad), who says that the Dorians pronounced sakkos with two /k/'s, while the Athenians used one.

2. Word-final /o/ is raised after a coronal, becoming /e/ (* 'onos ^ ST NT 'one 'donkey', *'xyros ^ ST NT PT 'xjure 'pig', *yraftos ^ ST yrafte 'writing'). Coronals here include front vowels (*'xreos ^ ST 'xrie 'debt'). This change is widely seen, as it involves the major masculine and neuter citation-form endings for nominals. In PT, it is extended to labial nasals (* 'dromos ^ ST 'dromo PT 'drome 'path', *'kosmos ^ ST 'kosmo PT 'kosme 'world', *'polemos ^ ST NT PT 'polemo PT poleme 'war, effort'), and sporadically to other labials and velars (* pyryos ^ ST 'kiryo NT piryo PT pirye 'tower').

3. Various palatalisations, involving consonants preceding front vowels, and in particular /i/, go further in Tsakonian than in Standard Modern Greek. In fact, Katsanis (1989) draws an intriguing correspondence between Arumanian and Tsakonian: palatalisation proceeded much further in both languages than in dialects of Greek proper, because the latter were much more subject to conservative pressure from Standard Modern Greek than were Tsakonian (a distinct Hellenic language) and Arumanian (a Romance language). Thus, Standard Modern Greek has palatal allophones of velars before front vowels (/k, qg, x, y/ ^ [c, jij, g, j]), and a palatal allophone of the alveolar liquids (/l, n/) before non-syllabic /i/. In Tsakonian, the following palatalisations take place:

a. The palatal unvoiced stop preceding front vowels is further fronted to an alveopalatal affricate4 (*ke'ros [ce'ros] ST NT PT tee're 'weather', *'kipos ['cipos] ST NT 'teipo 'garden'). Its voiced counterpart fronts even further, to become an alveolar affricate (*ay'gizon [ap'jizon] ^ ST an'dzixu ST NT an'dzizu PT an'dzizo 'to touch', *stray'gizon [strap'jizon] ^ ST stran'dzixu PT stran'dzizo 'to strain'). Occasionally, the unvoiced stop also fronts further, to become alveolar (*kefa'li ^ ST tsufa NT tsufala 'head', *'kipos ^ ST 'teipo, 'tsipo 'garden', Turkish kayik ^ MG ka'iki ^ ST ka'itei, ka'itsi 'boat'). Pernot (1934, 72) notes that, with the exception of 'tsufa, > ts is characteristic of recent loans, and Costakis himself used /te/ in such words; consistent with this,

4 There is some controversy and imprecision in the literature about how to describe both this phoneme and its more front counterpart, and I use the alvelopalatal vs the alveolar for convenience; Haralambopoulos (1980) analyses the phonemes I call /te/ and /ts/ as /ts/ and /tsh/ respectively. (Hence the minimal pair *keros > tee're 'weather', *kseros > tsere 'dry'; in Haralambopoulos' transcription, tse're 'weather', tshere 'dry'.) Liosis (2007, 342, 699) finds that [te] is a rare realization of the /te/ phoneme only found among the very oldest female ST speakers; the more common realisations he found were [dz] ~ [ts]. (For fluent speakers, he found 13 % of instances of the phoneme were realised as [te], 47 % as [ts], and 40 % as [dz].) By contrast, he found the phoneme I call /ts/, and Haralambopoulos calls /tsh/, was realised as [tsh] 78 % of the time by fluent speakers, and as [ts] 22 % of the time).

C only records words such as ka'itei with /te/. So it would seem /k/> /ts/ is a late and restricted development, which has exceptionally become generalised in the case of tsufa. In PT, > ts is more common (*bu'kia 4 PT bu'tsia 'mouthful', *ky'lia 4 PT tsi'lia 'belly', *kyp'rinos 4 H tsip'rine V kip'rini 'carp').

b. The typological gap vacated by palatalised velars is filled by both palatalised labials and dentals, which become palatals before /i/ (*pi'yadi 4 ST [ci'yaôi] 'well', *pisa 4 ST ['chisa] 'tar', *kum'bi 4 ST [kup'ji] 'button', *ty'ros 4 ST [cu're] NT [ci're] 'cheese', *dia'mandi 4 ST NT [ja'majiji] 'diamond'). (Following Haralambopoulos' (1980, 53) analysis, these palatal stops are henceforth treated phonologically as velars preceding a yod: [cu're] = /kju're/.) In NT, these palatalisations occur but are severely restricted (*pi'anon 4 ST NT kjanu ['canu] 'to get', *'ty'liyon 4 ST NT ki'lixu [ci'lixu] 'to wrap', *praymateftis 4 ST NTpramakefki [pramakefci] 'salesman', but *pi'non 4 NTpi'nu ST ki'nu ['cinu] 'to be hungry', *a'yapi 4 NT a'yapi ST a'yaki [a'yaci] 'love', *ti'mon 4 NT ti'mu ST ki'mu [ci'mu] 'to honour', *ku'ti 4 NT ku'ti ST ku'ki [ku'ci] 'box') (Costakis 1951, 55). This phenomenon does not take place at all in PT; the instances given by Costakis (1951, 163) of /d/> /g/ do not involve palatalisation (*ma'do 4 'maygo 'to pluck', *ma'dumenos 4 may'gomne 'plucked')

Palatalised dentals are also realised as dental affricates (*'oti 4 ST NT PT 'otsi 'coMPLEMENTiSER.', *op'tilos 4 *op'tsilos 4 ST NT epsi'le PT psi'le 'eye'). According to Pernot (1934, 74), /ti/> [ci] is older than /ti/> /tsi/, and there is no reason to think that /ti/> [ci]> [tsi]. The transition /ti/> [ci] went through the intermediate step [t-i], still extant as a realisation in Pernot's time, although not as frequent as [ci]; Pernot found that [ti] was frequently backed: [tji]. The labial seems to have gone through the same transition (Pernot 1934, 77): /pi/> */pti/> */ti/ > [t-i]> [ci]. Pernot found that like /ti/, /pi/ was realised as both [t-i] and [ci], so that there was no difference in realisation between *'tima 4 ST ['t-ima, 'cima] 'honour.iMP !' and *'pima 4 ST ['t-ima, 'cima] 'drinking', although he had the impression that "pi se gutturalise plus aisement que ti." Indeed, Pernot even heard instances where backed /p/ was no longer palatal: *ndro'pi 4 ST [dro'ki] 'shame'.

c. Palatalised alveolar liquids become palatal; unlike Standard Modern Greek, palatal liquids enjoy phonemic status. Such palatalisation also occurs before syllabic /i/, unlike in MG, where it is restricted to non-vocalic /i/ (*a'niyon [a'niyon] 4 ST NT a'pindu PT a'piyo/ 'to open' (MG [a'niyo]), *ili'azon [i'Àazon] 4 ST '¿azu PT '¿azo 'to sun' (MG ['Àazo]).) (As Costakis (1951, 33) notes, this is also characteristic of the Standard-Greek speaking villages to the south of Leonidio, such as Tsitalia,5 and indeed of Peloponnesian Greek in general; for example, /'yianis/ ['janis] 'John' is pronounced as ['japis] in Tsitalia — although, perplexingly, as ['janis] in Tsakonian).

4. Intervocalic voiced fricatives — particularly /5/ and /y/—frequently drop out (*'podas 4 ST NT 'pua PT pa 'foot', *evôo'maôa 4 ST vdi'ma NT PT vdi'mada 'week', *dy'yatir 4 ST 'sati NT fati PT 'sae 'daughter', *'trayos 4 ST 'tfao 'he-goat'). This process is characteristic of other Modern Greek dialects, such as Cypriot.

5. Intervocalic hiatus or yod is fortitioned to /z/ (* 'mya 4 ST PT 'muza 'fly', *y'os 4 ST NT PT i'ze 'son') or /v/ (AG khaos 'chaos, chasm' 4 *'xaos 4 ST NT 'xavo 'chasm', *y'datu 4 y'atu 4 ST NT i'vatu 'water.GEN', AG kyanos 'blue' 4 ST NT ku'vane 'black'). There seems to be no clear conditioning factor at work: cf. 'muza with ku'vane. This development was already

5 Tsitalia is a Standard-Greek-speaking village immediately to the south of Leonidio; it is the village Costakis' father came from. According to Liosis (2007, 41) Albanian placenames stop between Tsitalia and Leonidio, indicating that it was the start of a historically mixed Albanian / Tsakonian-speaking district.

in place by 1668; Qelebi records izemi 'my son'. This contrasts with MG, which inserts /y/ instead (ae:r, aeros ^ a'eras ^ a'yeras 'air', akouo: ^ a'kuo ^ a'kuyo 'to hear', ao:ros ^ 'ayoros ^ 'ayuros 'unripe').

6. Infrequently, stops become aspirated in Tsakonian; this is a characteristic of South-Eastern Greek dialects, as well (Pernot 1934, 131). In Peloponnesian Tsakonian, this only occurs for /k/ and /p/, predominantly word-initially (*'kambi ^ ST PT 'khamba 'caterpillar', *por'di ^ ST phun'da PT phor'da 'fart', *pondi'kos ^ ST phopji'ko NT pondi'ko PT pondo'ke Hphondi'ko 'mouse'). In PT, this aspiration is further extended to /t/ (Mirambel 1960, 58) (*dyya'teres ^ sa'there 'daughters').

7. There is a tendency for /e/ to be raised to /i/ word-finally (*'edare^ ST NT eda'ri 'now', *ame'ros *ame're ST ame'ri 'day.GEN', * 'kalos * 'kale ST NT 'kali 'stick', * 'sates ST NT 'satsi 'this year'), although the conditioning is not obvious; and in the word-final groups /'ea/ and /'eo/ (*e'lea ST NT e'lia PT e'faia 'olive', *'xreos ST 'xrie 'debt'). The raising of /e/ word-finally does not apply to PT.

8. As seen above, /s/ + stop combinations go to aspirated stops. When an /s/ + stop combination results in Tsakonian as a secondary development — e. g. through an interceding vowel dropping out, or as a more recent loan into the language—/s/ becomes /J7 (Mediaeval Greek *'laspi ^ ST 'lajpi PT 'ajpi, o'ajpi 'mud', *'sa tano ^ ST 'Jtanu 'above, in the mountains').

3.2.4. Local Innovations

The following processes in Tsakonian are regionally restricted to Southern Tsakonian, rather than Northern. In all these instances, Northern Tsakonian patterns consistently with Standard Modern Greek; this has led Costakis to conclude that Northern Tsakonian 'corrected' the phonological innovations, undoing them under Standard Modern Greek influence. Liosis (2007, 167) however rejects this as an instance of internalised bias against Northern Tsakonian, which retreated much earlier than Southern, and believes they are simply NT archaisms — which is indeed the more economical explanation.

1. /l/ disappears before back vowels (*av'dela ^ ST av'dea NT av'dela PT av'dela, av'dewa, av'daa 'leech', *'loyos ^ ST 'oyo NT 'loyo PT 'loyo, 'oyo 'word', *'luzon ^ ST 'ukhu NT 'lukhu PT 'luzo, 'uzo 'to bathe'). This change may have been in place by 1668, as discussed above, and the data from Propontis Tsakonian shows it was also in place before the colonisation (Costakis 1951, 56); it is, as Hatzidakis (1989 [1905], 365) pointed out, more recent than the fronting of /o/ to /e/ after coronals (*ka'los ST NT PT ka'le 'good', *'olos ST NT PT 'ole 'all'). Costakis' (1951, 57) argument that it is Northern Tsakonian that innovates in reintroducing /l/ is based on the existence of hypercorrect insertion of /l/ (which also took place in Leonidio, the urban centre of Modern Tsakonia) (* 'axyra ^ NT 'laxira ST 'axure 'hay', *ayp'nia ^ NT aip'nila ST PT aip'nia 'sleeplessness', *'ayano ^ Leonidio 'layane ST NT PT 'ayane 'beard of wheat').

This process also takes place in PT; but while it is regular in Vatka, it is not attested in Havoutsi — which, being on the coast, was subject to much stronger influence from MG (*'ala ^ ST a NT H 'ala V 'awa, 'aa 'other.FEM', *la'yos ST V a'yo NT H la'yo 'hare', *los'tos ST os'te NT H los'te H lus'te V us'te 'crowbar'). Vatka Tsakonian also presents other consonants in place of the lenited /l/; Costakis' (1951, 161-162) discussion implies they are epenthetic, but they are clearly intermediate stages in the lenition (awkwardly described by Costakis), which has thus not progressed as thoroughly as in ST. Thus, velarised */!/ is realised in Vatka as 0, /o/ (*'lampa ^ o'ampa 'lamp'), "a sound resembling more or less /u/" (clearly /w/; Makris (1951, 201) describes it as "corresponding almost to the digamma (p)", which in Ancient Greek represented /w/, and is a common reflex cross-linguistically of /!/) (*'ylosa ^ yu'osa ['ywosa] 'tongue'),

"a sound comparable to /uv/, with /v/ barely audible" (possibly fricated /w/) (*xalakos 4 xauva'ko [xawa'ko] 'to be ruined'), "a clear /v/" (by fortition of /w/) (*ke'fala 4 tse'fava 'head'), /vo/ (combining two lenited forms (*ka'la 4 kavo'a 'well.ADv'), and /vu/ (possibly /vw/ or /vw/, as an intermediate stage between /w/ and /v/) (*a'naskela 4 a'nasteevua [a'nasteevwa] 'supine').

2. Palatalised /m/ assimilates to /p/ (*mia [mja] 4 ST pa PT mœ 'one', *'mia 4 ST pia NT PT 'mia 'one', *ma'mi 4 ST ma'pi NT PT ma'mi 'midwife'). This is consistent with the behaviour of palatalised /n/ in Tsakonian.

3. Unlike Standard Modern Greek, the palatalisation of liquids also extends to /r/, whose palatalised variant is realised as /3/ in women's Southern Tsakonian, although Northern Tsakonian and modern men's Southern Tsakonian realise it as /r/ (Haralambopoulos 1980, 35) (e. g. *ry'aki 4 'jatei, ri'aki 'stream', *'ryzi 4 'jizi, 'rizi 'rice'). In xix ad, this palatalised variant seems to have been more like a fricative trill [r ] (Scutt 1912-13, 151: "it becomes something very similar to a Czechish r, often sounding like Z."); by the early twentieth century, this realisation had substantially died out (Pernot 1934, 97). In Northern Tsakonian, when /r/ is palatalised, Costakis (1951, 33) distinguishes between old people's [3], and the younger / male realisations [rz] and [r3], which he considers assimilations towards MG /ri/—a changeover apparently complete by Haralambopoulos' time. While Liosis (2007, 351) did observe variation between [3] and [r3], he did not find it correlated with gender, and indeed he found [r3] is now extremely rare: 3 % for fluent speakers, exclusively in NT, against 43 % for [3] and 50 % for the MG [r].

4. There are several instances where NT /i/ corresponds to ST /u, ju/ (* 'artyma 4 NT PT 'artima ST 'artuma 'cheese', *y'mon 4 NT 'pimu ST 'pumu 'you.PL', *vy'zi 4 NT PT vi'zi ST PT vu'zi 'teat') and vice versa (*try'yon 4 NT tfu'yu ST tji'yu PT tri'yo 'to harvest', *syyy'rizon 4 NT suyiur'jizu ST siyi'jizu PT siyi'rizo 'to tidy', *ky'lon 4 NT ku'lizu ST akha'liu PT tei'lo, tei'o 'to roll'). In this instance, one dialect is conservative while the other patterns with Standard Modern Greek; neither dialect is consistently more conservative, which points to either different loans in different dialects, or lexical diffusion.

5. NT uses /JV "more frequently" (Costakis 1951, 32) than ST. Altough Costakis gives no conditioning factors, the innovation specific to NT seems to be a palatalisation of /s/, which also extends to the clusters /ks/ (which goes to /ts/ regularly) and /ps/ (*kra'si 4 NT kra'fi ST PT kra'si 'wine'; *'ksila 4 NT 'kfila PT 'ksila, 'ksia, 'ksioa (ST 'kaúa) 'pieces of wood', *'taksi 4 NT 'takfi ST PT 'taksi 'order; menstrual period', *ek'sinda 4 NT ek'finda ST ek'sinda PT 'ksinda 'sixty'; *psy'xi 4 NT 'pfuxa ST PT 'psuxa 'soul'). A palatalisation interpretation is reinforced by Costakis' (1951, 37) observation that, where /*u/ > /y/ > /ju/ after a coronal, "/j/ is not heard after /JV (/kf, pJV) or /3/" (* karyo 4 NT 'kaju, 'karju, 'karu, ST 'kaji, ST PT kari 'walnut'; *'syko 4 NT fukho ST PT 'suko 'fig', *ok'symilo 4 NT 'tfumale ST tfu'mati 'plum', *psy'xi 4 NT 'pfuxa ST PT 'psuxa 'soul'). Strangely, C makes no mention of this process, nor does it list many of the variant pronunciations given in Costakis (1951). This process also applies to cases where Tk /s/ is a reflex of AG /th/ (*ka'dimenos 4 NT ka'fimene ST ka'simene 'seated'). In the case of NT fati/ ST 'sati 'daughter', we may have a secondary development from the etymon thugáte:r > dy'yatir: *dy'yatir> *sy'yatir> *sju'yatir (normal development after coronal)> *sju'ati (morphological reanalysis) > NT *fu'ati> 'fati; in ST the palatalisation did not take place, so *sju'ati> *'sjati> 'sati. Costakis (1951, 32) attributes this development to the neighbouring non-Tsakonian villages, "where /s/ is pronounced thick [as /J/]"; one would presume that the phenomenon in those villages is also an instance of palatalisation.

6. PT has lost the Peloponnesian Tsakonian contrast between /u/ and /ju/ (*'lykos 4 ST 'áuko H 'luko H V 'uko 'wolf', *'endyma 4 ST 'opjuma PT 'onduma 'clothing', *'stylos 4 ST ['chule] PT 'thule 'post'). /ju/ is present only in the stem *'yyro 4 *'yjuro 4 ST NT PT ['jure] 'around'.

7. PT regularly merges /ia/ into /œ/ or /œa/, when /i/ is unstressed (i. e. a yod) (*mya'lo ■ mœ'le 'brain', *fii'ano ■ ftœno 'to make'; *vori'ades ■ vorœ'aôe 'north winds', *kami'a ■ kamœ'a 'any.fem'). In mainland Tsakonian, this applies only to the proper name *milti'adis ■ mil'tœôi 'Miltiades', which is of recent importation.

That a form has undergone characteristic Tsakonian phonological processes is not in itself sufficient guarantee that it is not a loan. Some phonological processes — particularly those involved in morphology and the palatalisation of /k/ — are universal, and apply to words which are definitely loans. Examples given by Pernot (1934, 43) include Turkish inat ■ MG /i'nati/ ■ [i'naci, i'natsi] 'spite', MG /irino'ôikis/ [irino'ôicis] ■ ^ino'ôici, 3ino'ôitei] 'justice of the peace', and MG /yramati'ki/ ■ [yramaci'tei] 'grammar'. Phonological assimilation can be undertaken quite systematically in Tsakonian; as Pernot (1934, 43) comments, "il faut tenir compte du fait que tous les villageois ont dans l'esprit un système de correspondances phonétiques et morphologiques, une sorte de sentiment linguistique inné, qui leur permet d'ordinaire une grande précision."

One final point which should be raised is the likelihood of non-Greek influence on Tsakonian. As Pernot (1934, 141), himself formerly a sceptic on the Doric origins of Tsakonian, admits, "je touche ici à une point délicat" — Hesseling's speculations on non-Greek admixtures in Tsakonian drew the indignant response of a Leonidio town meeting, and as Pernot (1934, 144) grumbles, "ce n'est pas ainsi que la question pourra être éclaircie."

The Doric heritage of Tsakonian seems now to be beyond dispute. Yet it is also true that this Doric heritage has been overstated, for reasons of the prestige which comes with the unique status of Tsakonian as heir to Doric. One need only contrast Pernot's scrupulous caution with the enthusiasm of both Deffner before him, and Costakis after him. Yet Costakis' dictum (pers. comm.) that "the foreign influences on Tsakonian stopped at the shepherd's hut door," rings true: while livestock colours are largely borrowed from Albanian, for example (Costakis 1975-76), the names of kitchen utensils are native. The influence of Slavonic on Tsakonian is similarly small (Kisilier 2017, 114-115 lists 13 loanwords, and teepa 'skin, membrane' discussed in this article is likely a fourteenth) — although Tsakonia has bordered on first Slavonic-speaking, and then Albanian-speaking territory for a long period. (Liosis 2007 nonetheless has extensive documentation on the possibility of Albanian-Tsakonian contact, with a formerly mixed population immediately south of present-day Tsakonia. Kisilier 2017, 115-116 also adduces a longer list of Albanian loanwords.)

Pernot's (1934, 143-144) speculations on foreign (Slavonic) influence on Tsakonian are limited to morphology; they appear somewhat impressionistic, and in any case are beyond the scope of a study on historical phonology. One phonological feature Pernot does find worth mentioning is the extent of palatalisation in Tsakonian — both how widely it is spread (including, for example, all liquids), and the phenomena /ti/> [tsi] and /pi/> *[ptJi], which are not characteristic of Greek. Pernot's (1934, 141) impression is that

une influence extérieure qui n'est pas celle de la langue commune, s'est exercée à un moment, donné, qu'elle n'a eu phonétiquement qu'une influence limitée et que le dialecte est en train de réagir, dans la mesure où il le peut.

4. Historical Lexical Issues

Two major processes involved in the transition from Ancient to Modern Greek vocabulary are the replacement of nouns by -ion diminutives (followed by the dropping of the /on/ neuter suffix after /i/) (poûs ■ podion ■ podi 'foot', onuks ■ onukhion ■ 'nixi 'nail'), and the morphological regularisation of nominals by generalising their oblique stems to the nominative (gunê:,

gunaikós ^ yi'neka, yi'nekas 'woman', ané.r, andrós ^ 'andras, 'andra 'man'). Both these processes date from Middle Greek; the use of diminutives is already widespread in the New Testament, while the morphological simplification is noticable in papyri after 3rd century ad.

The citation forms given for verbs vary. In Ancient and Modern Greek, the form given is the first person indicative present; the infinitive has not survived into Modern Greek. In Tsakonian, the indicative present tense is formed with an auxiliary and a participle; so the active present participle is the citation form. The Tsakonian participle corresponds to the Ancient Greek o:n ending; as the /n/ has dropped out, the reflex /u/ is usually the same as that for the indicative present, whose first person ending is o: (Standard Modern Greek o, Metropolitan Tsakonian *u).

The effect of Puristic on MG has been profound; ancient revivals have displaced vocabulary even in the Swadesh-100 list (e. g. 'skin'; Kassian 2018 elicits the archaic fli'os for 'bark'). In order to account for this, I have attempted to provide the pre-Puristic vernacular term where possible.

5. Swadesh-100 List

In the following, I draw mainly on Costakis' (1986) comprehensive dictionary to compile the basic word lists. As this is a unidirectional dictionary, in which the entries for different dialect equivalents are not always cross-linked, I have not been successful in tracing all entries for Northern Tsakonian; I have supplemented this with enquiries to Dimitris Houpis. As comparison with Costakis (1951) and Pernot (1934) makes clear, the dictionary does not list all phonetic variation present in the language — although it seems fairly reliable for Southern Tsakonian. To determine the unmarked form out of the available options, I have used textual frequency in the Tsakonian texts at my disposal: for all three dialects, the dialect texts collected and published by Costakis (1981; 1986 — predominantly Southern); in addition, for Southern Tsakonian, the texts in Haralambopoulos (1980); for Northern Tsakonian (Kastanitsa), Houpis' texts (1990; 1993) — although the fact that Houpis is literate and late means his texts should be looked at with caution, notwithstanding the fact that Houpis checked his language use with old Kas-tanitsiots; and for Propontis Tsakonian (Havoutsi), Karaliotis' (1969) autobiography, written at the behest of Costakis. I have occasionally cited the word list gathered in Kisilier's (2017) fieldwork, where it supplements other findings.

Our major source in tracing the transition from Doric to Tsakonian, after the Neo-Laconian inscriptions, is Hesychius' 5th century ad dialect dictionary. This is a highly heterogeneous source containing words ranging from the Homeric to comtemporary Latin loanwords, with no indication of provenance other than a one-word regional epithet, like 'Cypriot' or 'Laconian'. Pernot (1934, 103-105) spends some time discussing how reliable a witness Hesychius is; while there are some inconsistencies, particularly in transcription, he takes Hesychius' Laconian forms present in Tsakonian but unattested in Neo-Laconian as reasonably belonging to 4th century ad Proto-Tsakonian. More recently, in Tsakonian proper, we have 35 Tsakonian words recorded by Evliya Qelebi in 1668 (cited in Costakis 1951), the 55 words recorded by Jean-Baptiste-Gaspard d'Ansse de Villoison in his 1785-86 trip to Greece and published in 1788 (Famerie 2007), with a further 100 words recorded in his travel journal and recently published (Famerie 2006, not seen for this study); and the 52 words and phrases recorded by Johan David Ákerblad, in one of his trips to Greece, either 1785-1788, or 1796-1797 (Manolessou and Pantelidis 2018).

The villages drawn upon as reference points for the three dialects are those about whose language we know most: Melana (Costakis' birthplace) for Southern Tsakonian, Kastanitsa for Northern Tsakonian, and Havoutsi for Propontis Tsakonian. The orthography used is that of Costakis (1986). Comparative text counts between two synonyms in a Tsakonian dialect are only given when neither is cognate with the Standard Modern Greek form, to determine whether there is a match or not. To determine the unmarked stems in Classical Greek, text counts

were used: primarily in the corpus of Plato, secondarily in Aristophanes and Thucydides (using the Thesaurus Linguae Graecae); Classical words were checked against Woodhouse 1979 [1932] and Liddell, Scott et al. (1940).

Word MG NT ST PT

1.1 eyrn e'yo eÇoû e'zu eÇoû(ve) e'zu(ne) eyrn e'yo, eyrnva e'yona, (e)yrnve (e)'yone

 esou. MT e'zu < eyiœ [e'jo] < A èyœ ego: (Pernot 1934, 203): first the intervocalic fricative is lenited to a yod or dropped (§3.2.3.4), and then it is refortitioned to /z/ (§3.2.3.5), with final /d:/ raised to /u/ (§3.2.2.3). e'zune, e'yoneby analogy with e'kjune, e'tune'thou'. All forms cognate, although PT may have borrowed the MG form.

2. thou

eau e si

eKiou e'kju

eKiou(ve) ekju(ne)

(e)xou (e)'tu, (e)xouve (e)'tune, exouva e'tuna

A etjou. MG e'si < IG au sy < IE *tu 'thou.NOM', assibilated by *tw- 'thou.OBL'. Tk e'kju, e'tu < D tu tu: < IE *tu (C; Palmer 1980, 288; Hesychius is somewhat vague in this respect). For the palatalisation /tj/ > /kj/ [c] see §3.2.3.3b. Pernot (1934, 187) gives the form as [e't-iu]; at the time Pernot collected his data, there was still fluidity between [t-i] and [c], which seems now to have been resolved in favour of [c]. Costakis (1951, 191) cites e'tu as an instance of PT being more conservative than MT; however, the loss of palatal /ju/ is a PT innovation (§3.2.4.6).

Pernot (1934, 204) explains the endings of ST 'you' as being completely remodelled after those for 'I' (*eti'u after *eyi'u), rather than pointing back to Hesychius' touv^ touns: 'you', which does not explain the palatalisation of /t/. The palatalisation before reflexes of EMG /y/ (§3.2.2.1), Pernot had decided, occurs only to those instances where Doric /u/ had not survived — and it seems from Hesychius that Doric /u/ has survived here. (This seems to be confirmed by PT e'tu, which does not palatalise.) He conceded nevertheless that ("semble-t-il") the form is likely Doric in origin, and (e)'kju is the regular Tk reflex of D tu: (tu: > *tju> *kju). The ne suffix is also present as an emphatic in the Modern Greek accusative of 'thou' (e'sena, e'senane), whence it has been generalised in Tsakonian to the nominative. The initial /e/ of e'si developed in Greek by analogy with ego: 'I' around 1st century ad (Palmer 1980, 184). All forms cognate.

3. we

e^eiç e mis

e^ei e mi, e|ieiv e miß, e^eive e'mine

evei(ve) e'ßi(ne)

e^ei e mi, e|ieiv e miß, e^eive e'mine

 eni. ST e'ßi follows from palatalisation of /m/ (§3.2.4.2). Costakis (1951, 84) also gives e'mu for NT. Tsakonian has not held on to D à^eç ha:mes (Palmer 1980, 288), but has adopted IG ^eîç he.meîs > MG e^eiç e'mis. The D form survives in M K oblique va^ou 'namu 'us', which Pernot (1934, 202) derives from D à^œv ha:mo:n 'of us', with the initial /n/ metanalysed.

Anagnostopoulos (cited in Pernot 1934, 203) and Scutt (1912-13, 163) attempt to assert the Doric status of e'ßi by adducing D e^io hemio and è^iœ hemio:, used by the ancient comedian Rhinthon (C). The simplest explanation nevertheless seems to be that this is an early MG (or IG) loan — with which Pernot (1934, 204) concurs: "La non persistance de a'ßi = ^eîç (cf. 'namu) rend possible, mais non certain, un emprunt à la langue commune." All forms borrowed (very early).

4. this auxôç aftos évxev'n 'endeni évxepn 'enderi xep^ te'ri

5. that eKeivoç e'kinos ex^ve e'tine éxper| 'etrei x^ve 'tine

Á etineri 'that.MASc', eteneni 'that.FEM' (see below). The correspondences of demonstrative pronouns are complex. MG has a three-way distinction: xoúxog 'tutos (proximal), auxóg aftos (unmarked), eKeivog e'kinos (distal). In Pernot (1934, 191), ST is said to also have a three-way distinction: evxev'n 'endeni or évxepn 'enderi (proximal), exev'n 'eteni (medium distance), ex^ve e'tine (distal—"ce dernier indiquant un grand éloignement dans le temps out dans l'espace"). (This is not counting the emphatic variants ending in -ori and -eri.) Pernot reports that Costakis (his language consultant) "saisit mal une opposition de xoúxog et de auxóg", and thus glossed the three Tsakonian demonstratives as auxóg ебю (aftos here), auxóg eKeí (aftos there), and eKeivog. In C, however, 'eteni is listed simply as a variant of 'endeni. Furthermore, C refers to e'tine as "indicating no clear differentiation of distance" — in other words, it is claimed to be unmarked, not distal. This is illustrated in Pernot's (1934, 43) report that Tsakonians speaking Standard Modern Greek say yia xsivo це ^GeAeg yia 'kino me 'ideles instead of yi' avió fe ^deXeq yi afto me 'ideles for 'that's why you wanted me'; this is because the unmarked demonstrative, used in both languages for clausal referents, is in Tsakonian the distal rather than the medial (cf. Tsakonian ['ja ci m 'esa '0eu] 'ibid.', where ('e)ci is the neuter distal).6

While NT has a huge panoply of demonstratives (11 nominative masculine forms), Costakis (1951, 86) claims that the "same forms are used in NT whether for a great or small distance, while in ST the distinction is clear" (i. e. the distance distinctions are effaced in NT.) PT is similarly complex, with 6 demonstratives. To help make some sense of all this, I tabulate the demonstratives according to the etymologies given them by Pernot (1934, 206-211); the demonstratives fall in three groups:

Group A (ende-): EMG évi xog 'eni tos 'it is he' > Tsitalia evxog 'endos 'there he is' (Pernot 1934, 208). Pernot gives the traditional derivation xog tos < auxóg aftos 'that one.NOM'; but Joseph (1994) has proposed that this is in fact an analogical development from accusative xov ton < auxóv afton.

Deffner (cited in Scutt 1912-13, 164) sees the en element as derived from AG ^v e:n 'behold!', rather than évi 'eni 'it is'. Scutt (1912-13, 164) finds these etymologies improbable, and prefers a connection with Hesychius av5a ánda 'that' (Cypriot), avxexoCg antetoüs 'that year' (Laconian); Pernot (1934, 207) comments that "il y a loin de la, sous tous les rapports, au tsakonien 'enderi, 'endai". C prefers an etymological connection with ex^ve e'tine; as a result, he spells the pronoun endings with n, although in Costakis (1951) he used i.

Tzitzilis (in prep, cited in Liosis 2007, 415) rejects the proposed grammaticalisation of 'eni tos 'it is he', and prefers to see proximal 'endeni and medial 'eteni as cognate, with et- > "emphatic" ett- > eth- > e(n)d-, given that eth- and e(n)d- pronouns are attested in Greek dialect. The forms he cites starting in ett- (Italy, Cappadocia) are more readily derived from aftos > eftos by assimilation. The claim of et- > e(n)d- is based on Cappadocian 'etos ~ 'ettos ~ 'edos as recorded in the Historical Dictionary of Modern Greek (-th- variants also occur in Chios and Southern Italy): the likeliest account for that variation is analogical extension from clitic pronoun do, which in turn resulted from metanalysis of -n to (a commonplace development in Greek dialect). (A similar development would not be feasible in Tsakonian, which lacks that specific clitic.)

Tk evxev'n 'endeni, PT evxeva 'endena, 'vxevá nde'na, 'vxev'n nde'ni < évxove 'endone, acc.

of evxog. The unpalatalised /n/ in 'endeni indicates the form was originally *'endene, so that

6 Liosis (2007, 413) retorts that Tsakonian 'eci is the neuter of both medial 'eteni and distal e'tine. In addition, he notes that ekinos is unmarked as to distance in the Peloponnesian dialects of Modern Greek as well. He prefers Tzitzilis' (in prep) etymological argument identifying proximal 'endeni and medial 'eteni as cognate, and so being in effect a single proximal pronoun opposed to e'tine; I admit to finding that derivation arbitrary. He does concede that Pernot's observation hints that there was a three-way distinction in Tsakonian, which has since died out under Standard Modern Greek influence; Liosis has found all distinctions effaced between demonstrative pronouns, and notes that Costakis (1951, 86) had already made the same observation for NT.

palatalisation was not triggered (Pernot 1934, 208). Unlike other variants of Greek, prenasalisation of voiced stops in Tsakonian is still the usual option, as confirmed in Liosis (2007, 344); so 'endeni, which is occasionally realized as 'edeni.

MT evxepn 'enderi, M vxep^ nde'ri, PT xep^ te'ri < *evxe 'ende (< 'endos) + opi o'ri 'look!' This means that evxepi was originally an emphatic form, and the emphatic forms evxevopi 'endeno'ri, exevopi 'eteno'ri, ex^vepi 'etine'ri maintain the same suffix. Pernot (1934, 192) reports that ori is the most frequent emphatic pronominal suffix, and that for his native language consultant (Costakis) it constitutes an independent word. In fact, the -ori suffix has been semantically bleached in the emphatics as well, giving the renewed forms evxevopopi 'endenoro'ri and exivepopi 'etinero'ri (Pernot 1934, 193). Costakis (1951, 177) derives the -ri suffix from Tsakonian rhotacism and the Ancient emphatic clitic -i, which is well attested for pronouns in Ancient Greek (e. g. outos-i'this very one'); thus, évxoç-i 'endosi > evxea-i 'endesi> evxep-i 'enderi. But there is an attested tendency for renewal of -ori, and the posited -i suffix seems artificial (Modern Greek affixes -e to pronouns instead, and only to the accusative); so Pernot's derivation is much more believable.7

NT evxeyn 'endeyi, evxen 'endei. Given the NT emphatic form evxevoyi endeno'yi (oyi o'yi 'here'), one can easily surmise that this is yet another blanched emphatic: 'endeyi < * 'ende + o'yi. 'endei would then follow by lenition of /y/ []]. Tzitzilis (in prep, cited in Liosis 2007, 415) instead sees here the Ancient formant -ge/-gi, which he claims is also used in Southern Italian and Heptanesian Greek.

Group B (ete-): EMG eSe xoç 'ede tos 'look! he' > Dialectal Greek e xoç 'e tos 'there he is' (attested, for instance, in Crete and Chios); eSe 'ede> EMG e! e 'interjection' + iSe i'de 'look!' Tzitzilis (in prep, cited in Liosis 2007, 415) considers this derivation "unsustainable", and derives it instead from autos > aftos 'that' > *eftos > 'etos. MT exev'n 'eteni < exove 'e tone, acc. of 'e tos.

NT exepn 'eteri, PT xop^ to'ri, xep^ te'ri < *exe 'ete (< 'e tos) + opi o'ri 'look!' Pernot (1934, 209) expresses surprise that Costakis' paradigm does not contain this term, by analogy with 'enderi.

Group C (eti-): D x^voç tê:nos 'that one' (cognate with MG exeivoç e'kinos; cf. Aeolian K^voç kê:nos, AG (è)Keîvoç (e)keînos) has been proposed as an etymon (Hatzidakis 1989 [1905], 94, 365, Scutt 1912-13, 165; defended in Liosis 2007, 416). Costakis does not question the Doric etymology of e'tine (presumably because of PT 'tine)—hence its spelling with an n in C (x^voç > ex^ve). But this proposal is emphatically rejected by Pernot (1934, 209-210) ("s'il existe des lois phonétiques, le x prouve précisément qu'il ne s'agit pas du dorien x^voç"), as /t/ here has undergone neither of the two palatalisations characteristic of Tsakonian, to [c] or [ts]. (tê:nos is not to be confused with MG e'kinos, whose reflex would have been *e'teine.)

As Scutt points out, the neuter of e'tine is indeed eKnvi 'ekini; this would suggest *'etini, confirming tê:nos as the etymon — but this does not explain the unpalatalised masculine and feminine forms. Scutt believes that the /t/ failed to palatalise because /i/ had dropped out (etine'ri > *etne'ri), and that in modern ST the reappearing /i/ is epenthetic (cf. *kap'nos > kapi'ne 'smoke'); the neuter did not have the /i/ drop out, he maintains, "owing to the number of syllables and position of the accent". But as Pernot retorts, /i/ drops out of etine'ri

7 In doing so contra Costakis and Liosis, I take Pernot's approach of considering modern etymologies in Tsakonian more plausible than ancient accounts, where they involve elements that have not otherwise survived in Greek, ceteris paribus. Liosis (2007, 415) argues for a clitic -i based on the older feminine genitives en'dari < endas + i , et'rari < etinas + i (s# > r by §3.2.1.2), which analogically extended to the feminine nominatives en'dari, et'rari, where a final -s would not be justified. But if an ancient clitic -i can extend from fem.gen to fem.nom, then a modern suffixed -o'ri can just as readily (and with fewer moving parts) extend from masc.nom and fem.nom to fem.gen.

in rapid speech, and at any rate the stressed /i/ of e'tine could hardly be epenthetic. In the end, Pernot cannot offer an alternative etymology; but drawing on the fact that, for Costakis, the declension paradigm of e'tine is contaminated in the neuter by that of 'eteni, he suspects that there might have been influence working between several forms: in particular, that the /t/ of e'tine is related to a'tos or e'tos, dialectal variants of aftos.

Tk8 ex^ve e'tine, ST ex^ e'ti, PT x^ve 'tine, vx^ve 'ndine. Of these, e'ti is used as an animate or inanimate pronoun where MG would drop the pronoun; this even includes dummy pronouns, as in m et e'catße 'vrexo 'but it started raining' (Pernot 1934, 192). Pernot (1934, 210) does not believe enough is known to establish whether e'ti is a truncation of e'tine or not.

ST exnvepi, exnvepi 'etiner'i, 'etiner'i is listed by Scutt as a variant of e'tine; there is no reason to doubt 'etiner'i < e'tine + o'ri 'look!' In fact, the etymology Pernot (1934, 193) gives for it is e'tina 'that.FEM' + o'ri > e'tina + ri > 'etiner'i 'that.MASc'. However, Scutt derives the form as tê:nos> 'tiner (§3.2.1.2, §3.2.3.2), with the 'i ending by analogy to other Ancient Greek emphatic pronouns, such as oûxoai houtosi 'this very one', aûxn'i' haute.i 'that very one'. Costakis gave a similar derivation for 'enderi, and both derivations are defended by Liosis (2007, 416). (For ex^vevt 'etine'ßi, see below.)

PT ex^vet etine'i (Costakis 1951, 176; not mentioned in C); presumably from e'tine + o'yi 'here', just as with 'endei above.

PT x^vxe 'tinde, xnxa ti'ta. No etymology has been offered, but it is tempting to see in the second syllable a reflex of the definite article / relativiser to to; alternatively, this could be an old analogical remodelling of (e)'tine after *'ende and *'ete. At any rate, these demonstratives are not common in PT texts.

NT éxpevï 'etreni < *'etneni < ex^vevt 'etine'ßi (emphatic demonstrative in ST, though not in NT); this is derived by Liosis (2007, 417) from e'tine + the accusative ending -ne, as already seen in nde'ni < 'endone.9). The unpalatalised /n/ is a problem, but may be an analogical formation after 'eteni and 'endeni. =The etr- demonstratives are listed under evxev'n 'endeni in C; éxpeyn 'etreyi, expen 'etrei and expepn 'etreri are analogous to évxeyn 'endeyi, evxen 'endei and evxepn 'enderi. The neuter of NT 'etreri in Scutt (1912-13, 164) is 'ekini, which is also the neuter of ST 'etiner'i; this indicates that the two are cognate. For our table, the conclusion is as follows. Of the Tk demonstratives, ende- and ete-, and possibly eti- demonstratives are related to the unmarked MG demonstrative (and third person pronoun) aftos; there is a less compelling case for eti- demonstratives being cognate with the distal MG demonstrative e'kinos, via D tê:nos. Whether to treat them as matches depends on the choice of MG terms corresponding to 'this' and 'that'. All three dialects of Tk have a distal eti- demonstrative: PT 'tine is glossed as e'kinos, and the NT phrase srpsq xip a^epe 'etrei tir a'mere as exeiveç xiç n^épeç "those days" (Costakis 1951, 88) strongly suggests a similar role for 'etrei, which seems to be the main NT cognate for e'tine. All these forms are taken to correspond to MG e'kinos.

For 'this', 'endeni has been selected for ST; 'eteni seems to be a much more infrequent variant. The NT phrase evxepou oyi m 'enderu o'yi 'with these people here' (Costakis 1951, 88) shows 'enderi (glossed as aftos) to be equivalent to 'this'. The main term other than 'tine used in PT texts is te'ri and that is the form quoted here for 'this'; Makris (1952, 205), it should be noted, glosses te'ri as 'tutos. (Costakis never uses MG 'tutos 'this' as a gloss in his work,

8 C claims e'tine does not occur in NT, counter to Costakis (1951).

9 In an earlier draft of this paper, I derived 'etine'ßi from e'tine + o'ri, with assimilation ('etineßi is a variant of 'etineri); "The unpalatalised /n/ is a problem, but may be an analogical formation after 'eteni and 'endeni" Liosis (2007, 417) rejects this account as convoluted, and I accept his counterproposal: instead of 'eteni providing a model for analogical adjustment, it provides the simpler derivation.

and this is very probably ST influence on his MG idiolect, as pointed out by Pernot.) Since the distal scale appears to be weakened in NT and PT, and 'eteri is infrequent in ST, the MG equivalent selected for 'this' here is the unmarked form aftos, while the MG distal e'kinos is chosen to render 'that'. As a result, all 'this' forms are cognate, and no 'that' forms are cognate (following Pernot in rejecting e'tine < ts:nos).

6. who

noioç pjos

noíe(p(e)) pie(r(e)), ne(p(e)) pe(r(e))

noíe(p) pie(r), ne(p) 'pe(r), noi(p) pi(r(e))

noío(p) 'pio(r), XCTÍ(p(e)) 'tsi(r(e))

Already in IG, noîoçpoîos 'what sort of' had started displacing xiç tis 'who'. Pernot (1934, 215) points out that the regular process /pi/ > /ki/ (§3.2.3.3.b) did not take place with this word (cf. D ngpâ:i> Tk Kia kja 'where', cognate with poîos); "Ceci me paraît supposer que le mot est entré en tsakonien à une époque où cette évolution phonétique ne se produisait plus". In other words, this is probably an MG loan (Scutt 1912-13, 167 concurs), displacing the earlier usage of xaip tsir < D xip tir 'who' (cf. IG xiç tis 'who') (§3.2.1.2, §3.2.3.3.b), which survives in PT: xaip eip tsir ir 'who is it?' (C)—Costakis (1951, 191) mentions 'tsire as a PT archaism. tsi is clearly the unmarked form for PT: there are 8 instances in the PT corpus, while the 2 instances of pio are both adjectival. (Liosis 2007, 427 notes that this distinction between reflexes of nominal tis and adjectival poîos is characteristic of Asia Minor Greek — i. e. it resulted from contact, and incidentally helped preserve the archaic tsir.) C gives 'what, who' as the gloss for tsi throughout Tsakonian, but there are no grounds to believe tsi means 'who' for MT, and C gives masculine and feminine (i. e. animate) forms for tsi only for PT.

/pi/ > /ki/ must have taken place quite early: MT uses noiou 'piu for 'make; do', and noiœ pi'o seems to have been displaced by Kavœ 'kano in MG by 16th century ad; so the loan of 'pie may also have occured early. All forms borrowed.

7. what

tí tí

Ti ti

TCTl tsi

TCTi tsi, tcte tse

Apsi. For /ti/> /tsi/ (§3.2.3.3.b), see above. All forms cognate.

not ou(k) ou(k) 8ev den

o o

o o

tctœ tsa

Sev den

Q wmehu 'there isn't any' (NT o^' e^ou om 'exu 'I don't have', where om < o or u + e^i 'emi 'am'; in ST, this would be op 'exu). According to C, MT o < AG o o variant of ou(k) ou(k) (EMG u(k) 'not'). Pernot (1934, 295) more plausibly derives o from the Tsakonian process /u$e/> /o/ (also present in many dialects of Greek: Andriotis 1956): u 'ipi > 'upi 'they are not', u 'eni > 'oni 'he is not'. The exigiencies of the Tk conjugational system have /e/ as the usual vowel before which a negator occurs, so this was generalised at the expense of /u/. MG den < AG ou5ev ouden 'nothing' has penetrated into MT, although Costakis (1951, 125) explicitly refers to it as a loan from MG, and it does not appear to have become the unmarked alternative in Tk texts. AG ou(k) did not survive after EMG outside Pontic (khi < u'ki < uk). NT, ST forms non-cognate; PT form borrowed.

9. all

nàç pâs

ô^oç 'olos

o^e 'ole, oû^e 'ule

o^e 'ole

MG 'olos < AG oÀoç holos 'whole'. All forms cognate.

10. many

no^oí polloí

no^oí po'li

no^oú po'lu, npeCT oú pre'fu, náCTou pafu, náCT oi pafi

npeCTCToí pre'si, náCTe 'pase

o^e 'ole, oû^e 'ule

noó po'o

ST pre'si < AG nepiCTCToi perissoi 'too many'; MT 'pafu, 'pase < AG nag pas 'all', with the declension class regularised to singular *pasos > pase, plural pasi > pafi. Houpis (1990, 100) names the distinction between NT pafa and ST pre'sa (neut.pl) as one of the main differences between NT and ST. Pernot (1934, 178) explicitly refers to ST pre'si and 'pase as synonyms; the texts, however, clearly show pre'si to be indeed the unmarked variant in ST: 18 instances of an allomorph of pre'si in Samp, Diet and Har, as against 2 of 'pase. V po'o < MG po'li, which (as noAupo'li) is indeed the PT singular of 'many' ('a lot'). According to Costakis (1951, 78),po'lu is the usual, suppletive form for the masculine plural of NT pre'se 'too many; many' rather than prefu, although the feminine and neuter plural and the singular use forms ofpre'se. ST form non-cognate; NT and PT form borrowed.

11. one

' hén

Q ena. All forms cognate.

12. two | Suo duo | Suo 'bio | Suou 'diu, Su' di | Suou 'diu, Su' di | vxuo 'dio, vto do

Q dwya; Pernot (1934, 502) believes Qelebi or a copyist has conflated the ending of 'two' with that of the next word in the list, trwya 'three' ('tria). No NT form is listed in C, but 'two' can be found in NT texts (di in Diet §9 ii; 'diu in Samp §26.) Of the two forms, 'diu is emphatic (Pernot 1934, 183); the unmarked form is di. There are only two instances in PT of AG /d/ > /d/: duo > 'dio, and enduma > 'oduma 'clothes' (Costakis 1951, 163). All forms cognate.

13. big

цеуад megas

^eyà^oç me'Yalos

ата é a'tfe

ата é a'tfe

^eyá^e me'Yale

MG me'yalos is a morphological levelling of AG mégas, megâlou. C places a'tfe only in S, but it is used abundantly in Houpis' (K) writings. Kisilier (2017, 125) records a'tse for Melana. Tk a'tfe < AG àSpoç hadros 'stout, fully grown' (C); /dr/> /tf/ by §3.2.3.1.a. NT, ST forms non-cognate; PT form borrowed.

14. long

So^i^óg dolikhás, цакpóç makrás

ца^ид ma'kris

цак^ие mak'jie

цак^и mak'ju

ца^^е) ma'kri(e)

While dolikhos is a good Homeric word, by Classical times makros predominates: in Plato there are 7 instances of the dolikh stem and 5 of the adjective itself, while there are 94 instances of the adjective makros, 33 of its comparative and superlative degrees, and 11 other instances of the stem in compounds. This pattern persists in other Attic authors: there are 35 instances of the adjective makros in Aristophanes, and 31 in Thucydides, but none in either of dolikhos.

MG ma'kris < EMG mak'rys is morphologically remodelled afterpakhus >pa'xis 'fat', batus > va'dis 'deep' (Andriotis 1990 [1983]). ST mak'ju < *mak'rju < *mak'ry is consistent with EMG mak'rys. NT ^aKpZjue DH. NT, PT, ST forms cognate.

15. small

[i^póg mi:krás

H^póg mi'kros

цпта^ mi'tsi, vrea^ßi'tsi S

(е)цгш^ (e)mi'tsi

H^pé mi'kre, цтгс mi'tse

Tk mi'tsi < EMG ^ixaoç mi'tsos 'small' (cf. Cypriot ^na^g mi'tsis, Cretan ^lkloç mi'teos) < D ^iKKog mikkos, according to Andriotis (1974 §4048; the etymology goes back to Hatzidakis); it still seems safe to treat mikkos as cognate to AG ^iKpog mikros. Pernot (1934, 177), on the other hand, considers this implausible, since it would necessitate the palatal stem ending spreading from the feminine to the masculine and neuter forms; he favours viewing mi'tsos as originally a babytalk form ("mi'tsos en grec commun est un caritatif et vient, je crois, de mik'ros par une imitation

de la phonétique enfantine.") Pernot also claims mi'tsi is a loan, since the native equivalent of MG /ts/ is /tf/ (EMG aygy'lista 'small crooked thing'> MG 'glitsa, ST aygra'itfa 'shepherd's stick'); but the evidence for Pernot's claim is weak, and at any rate /ts/ is not a cluster of any great antiquity in Greek. Of course, if the baby-talk form was independently innovated in Tk, then the sound correspondence need not obtain, but it is worth noting that standard MG does not have palatoalveolars like Tk. All forms (for want of better evidence) considered cognate.

16. woman

yuv^ guns:

yuvaira yi'neka

youvaiKa

yu'neka

youvaiKa

yu'neka

youvaiKa

yu'neka

Vi youvaiKa yu'neka. All forms cognate.

17. man àv^p ans:r dvxpaç 'andras axa rnno 'atfopo d(v)9prnno 'a(n)dropo, dvxprnno 'andropo avxpoi 'andri (pl), (dGprnno 'adropo)

18. person avGpœnoç anthropos dvGpœnoç 'andropos dGprnno 'adropo, d(v)9pouno 'a(n)drupo d(v)9prnno 'a(n)dropo, dvxprnno 'andropo d9p(m/e)no 'adr(o/e)po, d(9/S)epno 'a(d/d)erpo

MT 'atfopo< AG avGpœnoç ânthro:pos (MG 'anGropos); AG /thr/> Tk /tf/ follows by rule §3.2.3.1.a. Although the adjective avxÇiKo andji'ko 'male, man's'< AG àvSpiKoç andrikos exists in ST, the original word for 'man', preserved in MG, has been displaced by the word for 'person'; this development is paralleled in Cypriot. The two reflexes of anthropos in MT, 'adropo and 'atfopo, have become semantically differentiated in ST, with 'adropo meaning 'person' and 'atfopo meaning 'man, husband'. While 'adropo is used in ST to denote 'man' (enepdxae eva aGpœno xaai Su' youvaixae epe'ratee 'ena 'adropo tee di yu'netee 'one man and two women went past': C), this appears to be a marginal development; the unmarked term for 'man' in ST is 'atfopo. According to Pernot (1934, 163), 'adropo is an MG loan:

Le mot dvGpwnoç est curieux à plusieurs régards. Il se présente en tsakonien sous une double forme: 'atfopo = dvxpaç, 'mari' et 'homme, opposé à femme', 'adropo, 'homme en général, individu'. Pour le sense de dvGpwnoç 'mari' cf. Matthieu 19, 10. L'aspect phonétique de cette deuxième forme ne laisse aucun doute sur son origin: elle provient de la langue commune. Il est évident que les Tsakoniens n'ont eu, à un moment donné, qu'une seule forme, 'atfopo, pour ces différents sens; cf. en fr. 'homme' opposé à 'femme' et 'mon homme' pour dire 'mon mari'. L'introduction de gr. com. d(v)9pœnoç a amené une distinction sémantique.

In the NT texts I have, there is one instance of 'man'—'atfopo in Samp §29 (K): nôzs va fioXai oi arâoinoi! pote na 'moloi i a'tjipi 'when were the men going to come!' Houpis uses 'a(n)drupo for 'person' and 'atfopo for 'husband' (Thus 60), but never uses the word for 'man'. Though a sample size of one is slim, one could conclude that in NT as in ST, 'adropo has only occupied the first stage of the cline person > man > husband. Since in PT 'adropo is known to also mean 'husband', it seems reasonable that it would also mean 'man', given the same cline.

Recent data points to 'atfopo being used for both 'person' and 'man': Vyatkina (2015, cited in Kassian 2018) records 'atfopo for 'person' in Prastos and Melana, and 'adropo only in Tyros, while Kisilier (2017, 129) records 'atfopo for 'person' in Prastos, Tiros, Melana, and Vaskina, and 'andropo in Tyros and Kastanitsa. If Pernot was correct about the semantic distinction made between the MG and the indigenous form, it was a distinction that did not persist; the indigenous form has picked up both meanings again latterly (as was likely the case before Tk borrowed the form from MG).

In the PT texts I have access to, there is one instance of dvxpoi 'andri 'men' (Diet §11 ii (V) a^a a' PaÀeKa^' oûo, e^eiv', oi âvrpoi, oaoi 'xa ex' Çœa 'ama s va'lekam 'uo, e'min, i 'andri, 'osi

ta et 'zoa 'when we had loaded them all up — us, the men, whoever had animals'; there is no other instance of 'man'), and one of avSpag 'andras ('andras?) 'her husband' (Mak p. 212 a YouvaiKa yupixg va nayaiv aneg axav nopxa, Ppexg xov avdpag, nexg xivepL. a yu'neka yi'rits na pay en apes stan 'porta, vrets ton 'andras, pets tine'ri 'the woman went back inside the door, found her husband, and told him...'). Thus: for 'man', PT seems to have borrowed the MG term; while it may well also use 'adropo, there is no textual evidence of this. NT and ST are not cognate. For 'person', PT, NT and ST all seem to have borrowed the MG term.

19. fish

i/Suç ikhthus

■dpi 'psari

■dp(Ç)i 'psar(j)i

^d(p/Ç )i psa(r/j)i

■41dpi psari

'psari is derived from the dim. o^apiov opsarion of o^ov opson 'snack'. Kisilier (2017, 128) records the variants 'psariyi (Melana), 'pfadji (Melana), 'psara (K). All forms cognate.

20. bird

opviç ornis

nou^i pu'li

nou^i pu'ti

nou^i pu'tä

nou^i pu'tä

pu'li is the dim. (pu'lion) of a Latin borrowing (pullus). All forms borrowed, given that the MG word is borrowed.

21. dog

< kuo:t

ctkû^oç 'skilos

. 'kue

Koûe 'kue, axâû^e 'steile

axâû^e 'steile, cxcuÀe 'stsile

MT 'kue < AG kûmv kûo:n 'dog'; after considerable discussion, Pernot (1934, 112-113) cannot justify the /e/ ending of 'kue, as the expected reflex would be *'kuu > *ku (cf. géro:n > 'yeru 'old man'). The form as it stands suggests a morphological remodelling to *'kuos > *'kuer, as Hatzidakis (1990 [1907], 585) had suggested; but there is no preceding coronal or front vowel to justify raising /o/ to /e/.

MG 'skilos< AG aKÛÀaÇ skulaks 'puppy'. Pernot (1934, 380) identifies Leonidio 'stile (palatalising /k/) as meaning 'whelp', and describes it as "d'importation récente." Words derived from 'steile are attested in C only for PT; their MT counterparts are derived from 'kue, e. g. H axâuÀeÛYœ stei'levyo 'to become angry as a dog' vs. ST Kouvé^ou ku'neygu 'to become angry as a dog' (M, T), 'to idle' (P). (The verb is unknown in NT: Pernot 1934, 354). Furthermore, there are 12 instances of 'kue in the main ST texts — Samp, Diet, and Har — with no instances of 'steile; so 'kue remains the basic term for 'dog' in MT. All the same, the MG term has made its presence known, and indeed the only forms for 'female dog' Costakis knew were based on it ('tila, 'steia), while Deffner's Kouvapa ku'nara 'female dog, peevish woman' was to Costakis only an augmentative. NT, ST forms non-cognate; PT borrowed from MG.

22. louse

90eip phtheir ■eipa psira

■ eipa 'psira

■ eipa 'psira

■eipa 'psira

For 'psira instead of the expected reflex 'ftira, Hatzidakis (1989 [1905], 172) sees the influence of psilos 'flea'; 'ftira is attested in Cyprus, Southern Italy and the Pontus. All forms cognate.

23. tree

SévSpov déndron

Sévxpo 'öendro

SévxÇ i 'dendji, SévxÇou 'dendju

SevxpiKÖ dendri'ko, SevxÇ iko dendji'ko

Sévxpe 'dendre, Sevxpé den'dre

ST dendji'ko < AG 5ev5piKov dendrikon 'tree-like (neut)'; cf. EMG oviKov oni'kon 'don-key'< AG oviKov onikon 'donkey-like (neut)'. Kisilier (2017, 126) adds 'dendri for Melana (which would be the MG equivalent of 'dendji). All forms cognate.

24. seed

CTnôpoç spóros, CTnép^a spérma

anôpoç 'sporos

anope 'spore

ánópi 'spori, n oúpe 'phure, n pd^a 'phrama

anope 'spore

Of the two AG words, sporos refers primarily to the act of sowing; the two terms, however, are clearly cognate. In Plato, sperma is clearly the dominant term: there are 27 instances of the word, as against 1 of sporos.

ST 'phure is cognate with MG 'sporos < AG anopog sporos; the process /sp/ > /ph/ word-initially is regular in Tsakonian, and there are only a couple of cases where a word-initial /sp/ can be ruled out as being a later development — most notably the word for ash, ano'i'a spo'ia (but also Ao'i'apho'ia in Leonidio). Tentatively, the NT and PT forms should be treated therefore as loans, displacing earlier phure (cf. the cognate word 'sow': PT ansipw 'spiro, NT ansipou 'spiru, ST neipou phiru.) ST phrama 'seed; sperm; silkworm egg' < *'sparma (cf. AG anep^a sperma 'seed; sperm'). Kisilier (2017, 128) adds phi'ratsi (Melana) as a cognate ofphrama; but this is clearly the Tk counterpart of MG spi'raki (dim) 'pimple; seed' < AG spyros.

It is impossible to tell which of the three ST forms is basic; the evidence of 'phiru strongly suggests 'spori is a recent loan — and indeed, the examples given in C suggest the primary meaning of 'spori in ST is 'the act or season of sowing', with its cognate phure used to mean 'seed', 'shoot', or 'pimple'. There is only one instance of any of these words in the available ST texts, 'phrama, in the meaning 'silkworm egg' (Samp §17). Haralambopoulos (1980) cites 'phrama frequently in his phonetic study; but that is because it forms a neat minimal pair with npa^a prama 'thing'. At any rate, all three ST words are cognate with 'sporos; since 'spori is unlikely to be the ST basic term, ST is considered not to have borrowed its term.

25. leaf

I 9ÄM.0V phúllon I 9ÄM.0 'filo I 9ÄM.E 'file I 9ÚX1 'fili

'file

The unpalatalised /l/ in ST 'fili points to *file, preventing the palatalisation. Kisilier (2017, 127) records the MG-looking forms filo (Prastos), fio (Melana) alongside 'fili (including in K). All forms cognate.

26. root

piÇa rízda

pí^a 'riza

â ÍVTa 'Jinda

â ÍVTa 'Jinda

pí^a 'riza

ST 'finda < 'riza; for consonants, see §3.2.1.4, §3.2.2.2. NT aivxa DH (without diacritics). Kisilier (2017, 127) records 'rize (Melana), 'riza (Kastanitsa) alongside 'finda.

27. bark

9^oiôç phloiás

9^oúSa 'fluda

9^oúSa 'fluda, 9poú^a frula

9poúa frua

9^oúSa 'fluda, neTaÍ pe'tsi

MG 'fluda < ^AouSiov 'fludion< phloudion, dim. of ^Aotig phlous 'reed'. PT pe'tsi 'bark'< 'skin'. PT normally drops /l/ before back vowels. Kisilier (2017, 127) records only fruda in Vaskina, and 'fluda in Prastos, Vaskina and Kastanitsa; as often is the case, the forms he records as currently in use are far closer to MG.

28. skin

Sép|ia dérma

Sep|ia 'derma

neTQÍ petsi, Taéna 'tsepa

neTQÍpe'tsi, TO|iàÇ i to'maji, Tâ éna 'teepa, Taéna 'tsepa

nETâi 'petsi

Tk pe'tsi < EMG nexaiv pe'tsin, dim. of Italian pezzo. Both pe'tsi and to'mari are also extant in MG; the latter as 'hide' (a meaning also suggested by C's examples: to ro^afy evi Gevxa aKia^a to to'maji 'eni 'denda 'akisma 'the hide needs salting'; 0a vxi ^naAou to ro^aty da ndi 'mbalu

to to'maji 'I'll skin you alive!'). It seemspe'tsi was the normal MG word, before it was displaced by the reintroduction of 'derma < AG 5ep^a derma through Puristic Greek.

'teepa is given in C with the second meaning 'skin', and Deffner (1923) defines it as 'human skin'. Deffner gives the fanciful derivation < AG skepe: 'covering', but Pernot (1934, 376) objects that that would give 'stsepa or 's.epa. The primary meaning 'milk skin', and the metaphorical meaning 'shame', Pernot concludes, make it clear that these are merely cognate to MG 'tsipa 'milk skin, (fig.) shame' < Bulgarian tsipa 'membrane, film', with 'human skin' a secondary development within Tsakonian. Kisilier (2017, 127) records 'derma in Prastos and 'zerma in K, and 'petsi in Melana; he further records 'maji in Tiros and Vaskina, and speculates it is derived from MG mu'lari 'mule' (though it is tempting to see a reanalysis of to'maji as to 'maji, with to the definite article). All forms borrowed, given that both 'tsipa and pe'tsi are themselves loanwords in MG.

29. flesh

Kpeag kreas

Kpeag 'kreas

Kpie 'krie

Kpie 'krie

Kpea 'krea V, Kpie 'krie

In MG 'kreas means 'meat', while 'sarka (perpetuated through ecclesiastical usage) means 'flesh' as opposed to 'soul'; both are used for human flesh. The 'meat' definition has been assumed. NT 'krie unlisted in C, but given as declension paradigm in Costakis (1951, 75). 'krie has been morphologically remodelled from *'kreas (a restricted declension paradigm) to *kreo (/o/ to /e/ by §3.2.3.2, /e/ to /i/ by §3.2.3.7). There are only two instances of the word in the PT corpus (Samp §35); both are 'krie. Since furthermore C claims 'krea is limited to Vatka, the PT form is considered here to be 'krie. All forms cognate.

30. blood a!|ia haima ai|ia 'ema ai|ia 'ema ai|ia 'ema ai|ia 'ema

31. bone QCTxeov osteon K6KKaAo 'kokalo KOKaAe 'kokale KOKaAe 'kokale KOKaAe 'kokale

MG 'kokalo < AG K6KKaAog kokkalos 'pine cone'. Kisilier (2017, 127) records ko'kali for K. All forms cognate.

32. grease

Ainog lipos, dAoi^ aloiphe:

Ainog 'lipos, 'ksiygi

^ivx^i 'ksindzi

^ivx^i 'ksindzi

Ainog 'lipos

Of the two terms given for AG, aloiphe: is a nominalisation of dAei^w aleiph: 'to smear', and refers generally to anything smeared, including oinment, unguent, and varnish. Of the two terms given for MG, 'lipos corresponds more generally to 'fat', and is probably a Puristic reimportation into MG, while 'ksiygi refers specifically to cooking fat / grease — although MT 'ksindzi has no such restriction: OAiou <;ivrZi ski evxev'n o (3ouAe n' eGuKa^e 'ofau 'ksindzi 'eki 'endeni o 'vule ph e'dikame 'that rooster we slaughtered was all fat'. MG 'ksiygi < Latin axungia; the presence of initial /ks/ in ST 'ksindzi and the absence of /u/ strongly suggests this is a loan from MG, and not an independent reflex of axungia, despite the normal palatalisation of the velar. NT ^ivx^i DH. Kisilier (2017, 126) records only the more MG-like 'ksiygi, 'ksigi.

33. egg

mov o:ron

a^yo av 'yo

aP(ou/i)y6 av(u/i) 'yo

a^ouyo avu'yo

aP(ou)y6 av(u) 'yo

A augo, corresponding to MG auogo (mistakenly swapped with the Tk form?) Kisilier (2017, 129) records MG av'yo for NT, and avu'yo for ST. According to Hatzidakis (1990 [1907], 322), t av'ya 'the eggs'< *ta'va< *tau'a< *ta o'a< AG ta o:ia. Tk forms involve epenthesis on a form which does not fit with Tk phonology, and the dissimilation *tav'ya (which is not fully explained by Hatzidakis) is unlikely to have been independently arrived at in Tk, particularly as the hiatus consonants there are /z/ and /v/ (§3.2.3.5). There is no u'vo stem in Tk which might point to an independent survival of o:ion; the form ouyia 'uyia ['uja] 'selvedge', present in both Tk

and MG, has been traced to o:ia by Deffner and Andriotis (citing Koraes); the Turkish oya 'fringe' is more obvious, as Pernot (1934, 364) points out, although oya itself has been traced by Koukoules (cited in Andriotis 1990 [1983]) back to EMG ouia 'uia. At any rate, 'uyia too is a MG phenomenon, and whether it has anything to do with o:ion or not, it does not represent an independent Tk survival. All forms borrowed from MG.

34. horn

Kepag keras

Kepaxo 'kerato

xa Epaxe 'teerate

xa epa(xe) 'teera(te), xaoxave 'tfoxane

xa epaxe 'teerate

MG 'kerato < AG Kepag, Kepaxog keras, keratos. Kisilier (2017, 128) adds 'tfepi (Vaskina), etymology unknown. According to Deffner, 'tfoxane < AG xpw^avov tro.ksanon 'twig', xpau^ava trauksana 'dry chips'; ST 'tfoxane means both 'horn' and 'twig', and Pernot (1934, 380) mentions a similar polysemy for French le bois. It is difficult to tell whether 'tfoxane or 'teera(te) is the older form in ST. 'teera preserves the AG nominative Kepag keras, and displays the normal palatalisation of /k/, while 'teerate, like MG 'kerato, has been morphologically regularised: 'teerate < *'kerato; but this does not mean 'teerate is a loan from MG. 'teera has been semantically expanded to include 'carob' (parallel to MG ^uAoKepaxo ksilo'kerato 'wood-horn = carob') and a legume disease. Pernot (1934, 71) glosses the plural 'teerate as MG xapounia 'carobs'; this implies strongly that 'carob' is the primary meaning of ST 'teerate.

'tfoxane also means 'twig' — and the etymology suggests this to be the original meaning. Both terms show up in Melana proverbs: 'teera in ^ae vi eKiou va ^naAepe xo raspa faepi e'kju na 'mbalere to 'teera 'you should eat it, so you can get the horn out', and 'tfoxane in Ka xaai raoxavs a' e^noixaepe ka tee 'tfoxane s em'biteere 'you did it well and horns (i. e. not at all well)'. 'teera(te) has given rise to xaepaxa teera'ta 'cuckold' (cf. Italian cornuto), also present in MG (Kepaxag kera'tas), and the words for 'carob tree', 'carob leaf', and 'small carob'; 'tfoxane has given rise to xao^avi tfo'xapi 'small horn', and xaoxa(o)via tfoxa(o)'nia 'big-horned'. The fact that 'tfoxane has been more lexically productive than 'teera(te) suggests that it is the ST basic term; furthermore, there are two instances of 'tfoxane in ST texts (Samp §7, §10), but none of 'teera(te). We have no evidence to suggest 'tfoxane is also used in NT. ST form not cognate; NT, PT forms cognate.

35. tail

oupa oura:

oupau ra

voupa nu ra

voupa nu ra

vopa no ra

Tk nu'ra < EMG nu'ra < AG oupa oura:, by metanalysis. This process can be seen at work in Pernot's (1934, 25) transcription of a story by Costakis: there are 6 instances of 'tail'; one is the nominative a nura, but the other five are accusatives with a preceding /n/: tan ura 'the tail', mitcan ura 'small tail'. NT voupa DH. Kisilier (2017, 129) records MG 'ura alongside nu'ra; he also records 'nurle in Prastos. NT, PT, ST forms cognate.

36. feather

nxepov pterin ^xepo fte'ro

All forms cognate.

37. hair

0pi§ thriks, Ko^n kome:

|ia^.ia ma'lja, xpi/a 'trixa

^xepe fte're

xa i/a 'tfixa

^xepe fte're

xa i/a 'tfixa

^xepe fte're

xpi/a 'trixa, |ia^ia ma'Aia

Of the AG terms, thriks refers to both single hairs and (in the plural) hair on the head; kome: refers only to the hair on the head, and is described by Woodhouse (1979 [1932]) as occurring in Plato but rare in other Attic prose. Text counts are ambivalent on this: there are 5 plural instances of thriks in Aristophanes against 7 of kome:, but 11 plural instances of thriks in Plato against 3 of kome:. MG 'trixa < AG 0pi^, xpi^og thriks, trikhos.

Kisilier (2017, 126) records 'tsixe, 'tsiixa for Melana. MG ma'lja, plural of ^a^i ma'li 'hair, wool' < ^a^iov ma'lion < mallion, dim. of ^a^og mallos 'flock of wool'. Of the two MG words, ma'lja refers to the hair on one's head (French chevelure), while 'trixa refers to individual hairs. While derivatives of the singular of ma'lja, ^a^i ma'li, exist in Tk (^a^idZpu ma'faazu 'grow hair', ^a^iape mafaa're 'hairy'), ma'li according to C exists only in PT, and even there predominantly refers to wool, according to C. (Kisilier 2017, 112 also reports it in K, under recent MG influence.) The examples given in C for 'tfixa indicate that it has been extended to cover chevelure: Ko^e tou ralxa vti, n' e^dKai xda' xoup e^ou vti 'kopse tu 'tfixa ndi, ph e'zakai tas tur ep'su ndi 'cut your hairs (=hair), which has gone into your eyes'; tou raixs vti a' dyKai Ta nouAia tu 'tfixe ndi s 'aygai tapu'faia 'the birds have taken your hairs ('trixa) away' (said to a bald man, glossed in MG as ma'lja). See also the example given in the definition in C of nAe^iSa ple'ksida 'braid of hair': V nAe^a '^a Ta pa^ia nAe^ou5e plexa ma ta ma'faia mplek'sude, M evi npeya tou raixs (pl) ^i nAe^i5e 'epi preya tu 'tfixe miplek'side 'I am braiding my hair into tresses'.

The conclusion seems to be that 'tfixa covers the function of both ma'lja and 'trixa in MT, with ma'faia either reintroduced into PT from MG, or more likely constituting a PT archaism. This is because of the archaic stress in ma'faia < EMG ^a^ia ma'lia, as opposed to MG [ma'Aa]; the retention of word-final /'ia/ is characteristic of other archaic dialects, such as Maniot and Old Athenian, but not of Thracian, the MG dialect PT was in contact with. In the context of the wordlist, the chevelure definition seems to be what is intended; therefore NT and ST must count as non-cognate, and PT as cognate, although not borrowed.

38. head

Ke^aA^ kephale:

Ke^aAi ke'fali

Tuou^aAa tsu'fala

TCTou^a tsufa

tct e^aaa tee'fala, tct e^aai tee'faÄi

PT tee'fala < MG Ke^aAa kefala, augmentative of ke'fali; the final vowel and gender change implies kefala is the etymon for all variants of 'head' other than teefafai (C). ST tsufa < *tsufaa < NT tsu'fala, as ST drops /l/ before back vowels. For /ts/ instead of /te/ before a front vowel, see discussion on rule §3.2.3.3.a; Deffner recorded the form as ST T^ou^a teufa NT T^ou^aAa teufala, and Pernot (1934, 114) recorded tsjufa in Tiros; Kisilier (2017, 126) records tsufa, tfufa in Tiros. Pernot (1934, 112) believes this is related to the occasional development in Tk of AG /i/> /u/, comparing it to Southern Italian Greek eiofali /tjo'fali/. All forms cognate.

39. ear

oug ous

a^Ti afti

aßo(u)Tava av(u/o)'tana

aßo(u)Tava av(u/o)'tana

(a)^Ti (a)fti

MT avu'tana< a avut'ana 'the ear' < *a vua'tana < *ua'tani < *'uata + vipi 'it'< D ornTa ouata 'ears' (C). The derivation C gives is somewhat stretched (particularly with the suffixing ofpi), but the relation between avu'tana and ouata seems plausible enough. MG afti < AG WTiov o:tion, dim. of oig ous 'ear', through ta o:tia 'the ears' > tautia > taftia > ta aftia [taftja] (Hatzidakis 1990 [1907], 321-322); the word is cognate with D ouata (cf. AG plural WTa 5:ta.) All forms cognate.

40. eye

o^BaA^og ophthalmos, ö|i|ia

|iaTi 'mati

^iAe psi'le, sxp iAe epfi'le

^iAe psi'le

^iAe psi'le

Of the two AG forms, ophthalmos is barely predominant over omma: there are 72 instances of ophthalmos in Plato as against 58 of omma, and 82 of the ophthalm stem against 60 of the omma stem. Adding Aristophanes and Thucydides, this becomes 93 instances of the word ophthalmos (and 112 of the stem) against 70 of the word omma (72 of the stem). The two forms are ultimately cognate, both derived from PIE *okw 'see'; omma< op (reflex of *okw) + ma 'nominaliser', and ophthalmos< PIE *okuph- with 'expressive aspiration' of the expected /pt/ to /phth/ (so Pokorny).

Q epsile; the form i^e epfi'le is also given for NT in Scutt (1913-14, 27) and Costakis' (1951, 32, 45) NT grammar, and is recorded for K by Kisilier (2017, 126), but is not recorded in C. The palatalisation is consistent with NT (§3.2.4.5). MG 'mati< AG o^axiov ommation, dim. of omma. Tkpsi'le < D onxiAog optilos 'eye'. As Pernot (1934, 74) notes, the form optilos 'escaped' the older palatalisation /ti/> [t-ii]> [ci], undergoing instead the later palatalisation /ti/> /tsi/. The initial vowel is dropped in the citation form through hiatus: nominative o psi'le, accusative ton epsi'le (Pernot 1934, 171). In C, initial /e/ is only listed for accusative forms, where it is for the most part optional. D optilos is cognate with omma (and indeed ophthalmos), being derived from *oku-. All forms cognate.

41. nose | puy/og mykhos | |iuTr| 'miti | ctouko 'fiukho | ctouko 'fiukho | |iuTr| 'miti

No AG etymon for 'miti < EMG 'mytis (AG *mutis) has been convincingly identified, but it appears to be cognate to such words as ^ukt^p mukte:r 'nostril, nose, snout' and muksa 'mucus'. ST 'fiukho< AG puyxo? ruykhos 'snout, beak'. Consonants by §3.2.2.2, §3.2.3.1.c. NT ctouko DH (no diacritics indicated). Kisilier (2017, 128) gives the variant 'sukho for Melana. ST form non-cognate; PT form borrowed.

42. mouth

CTTO^a stoma

CTTO^a 'stoma

T ou|ia 'thuma

T ou|ia 'thuma

CTTO^a 'stoma, T 6|ia 'thoma

There are two possible derivations for Tk 'thuma. The first is from AG stoma 'mouth'; /st/> /th/ is normal for Tk, but one would have to explain /o/ > /u/, which occurs sporadically in Tk. The second is D xou^a touma 'mouth' (C), presumably cognate with stoma 'mouth'. This does explain /u/; now it is the change /t/> /th/ which would have to be explained. As Mirambel (1960, 58) points out, /t/> /th/ is characteristic only of PT, and not of MT (§3.2.3.6); so stoma is the least problematic etymon. (It is possible, as Pernot (1934, 127) hints, that Hesychius' touma is in fact Proto-Tsakonian 'thuma, in which the aspiration of /st/ has already taken place.) All forms cognate.

43. tooth | oSoug odous | Sovxi 'dondi | ovxa 'onda | ovxa 'onda | Sovxa 'donda

MG 'dondi < oSovxiov o'dondion < odontion, dim. of odous, odontos. The MT forms presumably by *o'donta > *o'onda > 'onda. All forms cognate.

44. tongue |y^waaa gloissa |y^rnaaa 'ylosa |ypouaaa 'yrusa |ypouaaa 'yrusa |y^rnaaa 'ylosa

NT 'yrusa unattested in C but used abundantly by Houpis (in the meaning 'language', at least.) All forms cognate.

45. claw | ovu§ onuks | vu/i 'nixi | vu/i pixi | vu/i pixi | vu/i pixi

MG 'nixi < ovu^iov o'nyxion < onukhion, dim. of onuks. The failure of /u/ to retain its archaic pronunciation in Tsakonian (*puxi) in a context where this normally occurs (cf. AG nuks, nuktos> putha 'night') renders these forms suspect.

46. foot |nougpous |noSi 'podi |noua pua |noua pua |napa, napha

MG 'podi < no5iov podion < podion, dim. of noug, no5ogpous, podos. ST 'pua < *poda < EMG no5ag podas; intervocalic /5/ regularly drops out in Tk. C lists the dim. noui pui for NT (cognate with MG 'podi, although the latter has lost its dimuntive force), but does not list 'pua; however, 'pua is used by Houpis (Thus 24). /o/> /u/ is conditioned by the preceding labial. All forms cognate.

47. knee

yovu gonu

yOvaxo 'yonato yovaxe 'yonate yoûva 'yuna

yovaxe 'yonate

ST 'yuna is a back-formation from 'yunata 'knees'; this back-formation is also present in other dialects of Greek (Hatzidakis 1990 [1907], 15). /o/> /u/ is conditioned by the preceding velar. Kisilier (2017, 127) records only 'yonato for Prastos and yona'te for Melana and Kastanitsa. All forms cognate.

48. hand | xeip kheir | /épi 'xeri | xépa 'xera

MG uses a diminutive. All forms cognate.

49. belly

yaax^p gaste:r, KoiAia koilia:

KoiAid ki'lja

^oûKa 'fukha

Xepa xera

^oûKa fukha

Xepa xera

^oûKa fukha

Of the AG terms, gaste:r is etymologically primary, while koilia literally means 'hollow'. The terms are fairly evenly matched in Attic: 7 instances of koilia in Plato against 5 ofgaste:r; but including Aristophanes and Thucydides, 18 of koilia against 26 ofgaste:r. (Counting stems, there are 19 instances of the koili stem — unlike MG, koil itself cannot mean 'belly'—against 43 of gaste:r.)

Q fwqa. Kisilier (2017, 126) records 'fuka for K. Tk 'fukha < D ^uaKa phuska: 'belly; and the large intestine'; IG ^uaK^phuske: 'large intestine; blister'. (Liddell-Scott-Jones Lexicon considers the use of D phuska: to mean 'pot-belly' "probably a nick-name.") Note that ^ouaKa fuska, a reflex ofphuske:, is extant in both Tk and MG with the meaning 'bubble; cyst; blister'. Presumably, the latter form became differentiated from the earlier within Tk as was the case with 'atfopo 'husband' and 'adropo 'man; person'. No forms cognate.

50. neck

xpd/nAoç trâkhe:los

Aai^Oç le'mos

Aai^O le'mo

Aai^O le'mo

MG le'mos < AG Aai^oç laimos 'throat, gullet'. All forms cognate.

51. breasts

ax^9oç sté:thos, liaaxOç mastos, axépvov stérnon, kOAkoç kolpos

ax^9oç 'stidos, ßu^id vi'zja

ax^9r| 'stiOi, ßu^ia vi'zia

ax^9n 'stiOi, ßou^ia vu'zia

Aai^O le'mo

aax^9ia a'stiOia, ax^9a 'stiOa, ß(o)uZia v(i/u)'zia

Of the AG terms, ste:thos and sternon refer to the breast or chest, mastos specifically to a woman's breast, and kolpos to a bosom or lap. Of the two MG terms, 'stidos corresponds to 'breast', while vi'zja specifically refers to a woman's or female animal's breasts; while it is now marked as vulgar in MG, it appears to be the basic vernacular term corresponding to 'breasts'.

Kisilier (2017, 126) adds ma'stari for Melana. Of the Tk terms, 'stidi seems to be a recent loan. Its normal reflex would have been *['chi0e], although the Tk singular ctt^Bi 'stidi is a backformation from the EMG plural ctt^B^ 'stidi. Nevertheless, the plural is characteristically MG, not Tk, and Pernot (1934, 173) concludes on that basis that the form is "probablement un emprunt." Given that vi'zja has been chosen as the basic term, all forms are cognate. The NT and PT forms display AG /u/ > MG /i/; however, Tk itself is inconsistent in retaining AG /u/, so the NT form need not be a loan. NT 'stidi: Thus 65 (referring to man's breast).

52. heart

KapSia kardia KapS id kar'dja KapS ia kar'dia KapS ia kar'dia KapS ia kar'dia

Scutt (1913-14, 27) gives the form Kop5ia kor'dia once, but four instances of KapSia kar'dia. Normally AG /rd/ goes to /nd/ in Tk (§3.2.3.1.g); furthermore, /8/ does palatalise into /v/ in Tk, although there are not many examples of this. On the other hand, the failure of -ia to reduce to yod indicates this is not a recent loan. All forms cognate.

53. liver

^nap M:par

CTUKrnxi si'koti ctkwki 'Jkoki

ctkwki 'Jkoki

CTUKrnxi si'koti

The post-classical forms are derived from the collocation ^nap auKwxov he:par suko.ton 'figged liver', in the same way as the equivalent Latin collocation iecur ficatum gave rise to the Romance words for 'liver'. The ST form is the normal Tk reflex of si'koti (§3.2.3.8); the PT form seems to have been reborrowed from MG. NT ctkwki DH; no diacritics indicated, though the pronucia-tion 'skoki may be possible under MG influence. Kisilier (2017, 128) gives fi'koki for Prastos and Kastanitsa, si'koki for Melana, 'skoki for Tiros and Vaskina. All forms cognate.

54. drink

nívw píno

nívw 'pino

nívou pinu

'kinu

nívw 'pino

The ST form is the normal Tk reflex of 'pino; the palatalisation of /pi/ does not seem to have taken place in NT, so the NT and PT forms have not necessarily been reborrowed from MG. On the other hand, -inu verbs in Tsakonian are generally recent loans; Pernot (1934, 272) suspects the Tsakonian form was originally 'piu, through analogical remodelling which had eliminated the -inu verb class, and that 'kinu is a more recent, phonologically assimilated importation from MG. All forms cognate.

55. eat

éüBíia esthío:

xprnrn 'troo

xcrou 'tju

xcrou 'tju

xpm(y/x)m 'tro(y/x)o

Aphare 'that you eat' (suppletive aorist, Tk faere, MGfas). MG 'tro(y)o < AG xp«Y« tro:go: 'to chew'; the verb is already used in its new meaning in the Gospel of Matthew. ST /tJV is the normal MT reflex of /tr/ (§3.2.3.1.a); however, the process does not seem to operate in PT, which consistently retains AG /tr/. Since Q records 'three' (MG xpia, ST xaia) as trwya, this is probably an archaism, rather than a loan from MG. NT 'tfu e. g. Houpis (1992, 12): Eivi xaiouvxa aave; 'ini 'tfunda sa'ne 'do they eat hay?' Kisilier (2017, 126) records the variant 'tfu for Tiros, Vaskina and Kastanitsa. All forms cognate.

56. bite

Sdrcvia dákno:

SayKwvw da'gono

SayKwvou da'gonu S

Kaxa ívou ka'teinu

SayKwvw da'gono V, yaxa ívou ya'teinu

NT, V da'gono is explicitly annotated by C as an MG loan. Kisilier (2017, 127) records ta'tsinu for K. According to Pernot (1934, 350), the metathesis /tsak/ > /kats/ is commonplace; as an etymology, Pernot (1934, 252) suggests MT Kaxaouvou 'break' < SMG xaaKMv« *'smash' < possibly EMG xaaKiov 'pen-knife'. He derives Deffner's ka'tsinu by metathesis from xaaKivwv *tsa'kinon, a variant of MG xaaKi^M tsa'kizo 'break'. However, the /te/ of ST PT point to *ka'kinon not *ka'tsino n—"ce qui complique le question." What seems to have taken place is that the metathesis form *ka'tsinu (which may indeed still have been extant in Deffner's time) was influenced by its still current etymon, which would have been palatalised to *tsa'teinu. Alternatively, * ka'tsinu may have been reanalysed to *ka'kinu [ka'teinu] as a sort of assimilation, given that /ts/ is not a very common etymological phoneme in Tk.

Elsewhere, Pernot (1934, 350) suggests ka'teinu < dakno: via consonant metathesis; that account is not much more convincing.

Liosis (pers. comm.), in his contribution to the Cognacy in Basic Lexicon project at the Max Planck Institute, Jena, has tentatively proposed *Kaxa5dKvw *kata-dak-no: 'to bite down' > *ka-dak-ino (remodeled present) > *ka-ak-inu (§3.2.3.4) > *kakinu > kateinu.

Since H is the reference dialect for PT and da'gono is unattested there, the PT form is non-cognate; the NT form is borrowed from MG. If we accept Liosis' etymology over Pernot's more problematic derivations, ST is cognate.

57. see

opm oro:

ßAenm 'vlepo opou o'ru

opou o ru

Bwpw do'ro

Q wri 'see!'. A ora 'I see', nandioru (= na ndi o'ru) 'that I may see you', orade 'you saw' (for o'rare). MG 'vlepo < AG pAenw blepo: 'to watch'. ST o'ru< AG op« oro: 'see'; this form survives in PT only in the participle opaKw ora'ko. PT do'ro < dialectal Greek do'ro 'see' (unmarked in e. g. Cretan, but absent in MG.) NT o'ru: Thus 36. No forms cognate.

58. hear

aKouw akoüo:

i'kuo

viou piu

viou piu

aKou(y)rn a'ku(\)o

According to C, MT 'piu < AG vo« noo: 'perceive'. But this verb already has the reflex ST voou no'u, PT voyw no'yo 'feel, understand, remember', which is cognate with MG vow no'o (high register) and voydw no'yao (colloquial), and which is a much better semantic and phonetic match for AG noo:. (no'u may well be an MG loan.) Pernot (1934, 277) points out a suitable cognate for 'piu would be MG voiwGw ni'odo 'feel' < voiwvw ni'ono (morphologically remodelled) < evvoia 'enia 'care, meaning'; however, the obvious reflexes in Tsakonian would have been *ni'azu > *'pandu. (The verb ending du in Tk, as discussed for 'burn', is problematic.) What Pernot considers likeliest is that ni'ono > *'pukhu, e'puka 'I feel / hear, I felt / heard' was influenced by its semantic match o'ru, o'raka 'I see, I saw', giving the aorist e'paka, which corresponds to the present tense piu (the aorist suffix jaka corresponds to the present 'iu.) In ST, a'kuo is present in the imperative dKou 'aku, dKo 'ako. NT, ST forms non-cognate; PT form borrowed.

59. know

olSa oida, yiyvwuKW gigno:sko:, eniCTTa^ai

^epw 'ksero

§epou 'kseru

§epou 'kseru

^epw 'ksero

Of the AG terms, oida is originally the perfective of the verb *e'i8M/pi8M eido: ~ wido: 'to see', which survived into Classical Greek only in the aorist e!5ov eidon (so still in MG: eiSa 'ida.) This form clearly predominates in AG: for the first person present indicative, Plato uses oida 139 times, epistamai 30 times, and gigno:sko: 10 times; Aristophanes uses oida 28 times, epistamai 6 times, and gigno:sko: once. EMG n^eupw i'ksevro (< ?AG e^etipw ekseuro: 'find out') > MG 'ksero.

While 'kseru is clearly the unmarked form for 'know' in Tk (26 instances in Samp, Dict and Har, and the reduced form o^a 'oksa for 'I don't know' in ST), it is just as clearly borrowed from MG, since /ts/ is the native reflex of AG /ks/. NT ^epou DH. All forms borrowed from MG.

60. sleep

KaBeuSm katheüdo:, Koi^m^ai koimo:mai, euSm heudo:, unvm^ai hupno:mai

Koi^a^ai ki'mame

(KaCT %eve kafmene), npayia^ou pra'yiazu

Kiou^ou 'kjufu, KaCTiou ka'siu

KaCT^Km ka'siko, KaCTim ka'sio, nAayiaZm 'azo

Ä tjupse 'sleep!' (= 'kjupse). Of the AG terms, katheudo: is clearly predominant. Plato uses katheudo: 32 times against koim3:mai 11 times and heudo: 6 times, while Aristophanes uses katheudo: 32 times, koimo:mai twice and heudo: 4 times. MG ki'mame< AG Koi^m^ai koimo:mai.

ST 'kjufu 'sleep; lie' < AG KunTm küpto: 'to stoop'; the verb also turns up in Corsican Maniot (Pernot 1934, 285). Tk ka'si(k)u 'I sit; I sleep (present)' < KaCT^Ka ka'sika 'I sat; I slept (aorist)' < AG KaBiZm katHzo: 'I sit'. The texts show 'kjufu to be the unmarked form in ST: 9 instances in Diet, Samp and Har, with none of 'kasiu. Scutt (1913-14, 18) records the form Tiou^Ta tjufta for Leonidio, and Pernot (1934, 351) also reports ['t>ufu] for Costakis. This would normally indicate

an etymon of xunxw tupto: (which in AG means 'to strike', and which in fact has the Tk reflex 'tifu 'to push'), but it seems that, since [t>] and [c] were in free variation at the time, the process [t>] > [c] could be reversed.

NT form only attested in passive past participle in C. Houpis usespra'yjazu (Mal §25; Thus 36), also attested for NT by Scutt (1913-14, 27); Pernot (1934, 351) also noted that his Kastanitsiot consultant used pra'yjazu instead of 'kjufu. 'kjufu does in fact turn up in NT (Costakis 1951: 141; Mal §8 kuou^i^o 'kjupsimo 'lying down'), but 'sleep' is overwhelmingly rendered as pra'yiazu (9 instances in the corpus, as against 2 of 'kjufu.) NT pra'yiazu, PT pla'yjazo 'to lie down; to sleep' < MG pla'yjazo 'to lie down' < AG nXayiog plagios 'sideways'. ka'sio ~ ka'siko is the only stem used in PT for 'sleep' (13 instances); a rather neat instance of ambiguity is given in Samp §12v: Maapyax^a xaatfyxovrs 'xavi axo xZaKi, o yepo ^e xa ypaia, xaai xo Ka^yi xaaiora 'xavi, xo naXiKap' mx aryati'na ka'siygonde tapi sto 'djaki, o 'yero me ta 'yrea, tee to kav'yi kasi'ota tani, to pali'kar 'one evening the old man and the old woman sat by the fireplace, and the lad slept, the young man'. No forms cognate.

61. die

anoBvfjaKia apothne:isko:

neBaivrn pe'Oeno

nevaKou pe'nakhu

nevaKou pe'nakhu

neBaivrn pe'Oeno

MT pe'nakhu< *ped'naskon (/0n/ > /n/ characteristic of Tk avoidance of clusters; see Pernot 1934, 257) < *pethnasko:n < AG ano0v^aKM apotne:isko: 'die'. While there is a temptation to see in the /a/ of *ped'naskon a Doricism, by equivalence with AG /s:/, Pernot considers it more probable this is an analogical development from the past stem *pedan- < apethan-. MG, PT pe'deno < EMG ane0aivw ape'deno < AG ane0avov apethanon, aorist of apothnee:isko:. The /e/ in pe'nakhu seems likewise to point to influence from the aorist stem, and quite possibly from MG. NTpe'nakhu: Thus 74. NT, ST form cognate; PT form borrowed.

62. kill

anoKxeivrn apokteino:

CTKoxrnvrn sko'tono

ctkoxouvou sko'tunu

ctkoxouvou sko'tunu

CTKoxrnvrn sko'tono

MG sko'tono < AG ctkotm skoto: 'to darken'. A more archaic verb exists in 0Uou 'Oiu 'to slaughter' < AG 0Uœ thûo: 'to offer a burnt sacrifice; to kill', but this is restricted to killing animals, and sko'tunu is used to denote killing either animals or humans. In the ST corpus, there are 7 instances of sko'tunu (5 with human victims) against 7 of 'Oiu (6 with animal victims); for NT, 30 instances of sko'tunu (5 with animal victims) as against 2 of 'Oiu (1 with a human victim).

Tk /sk/> /kh/ has not taken place; the regular process AG /o:/ > Tk /u/ has taken place; but this occurs in a suspect context. Pernot (1934, 272) considers the -'unu present modern, displacing earlier -'ukhu. Several verbs admit both possibilities; the absence of an -'ukhu variant, like ka'tfunu, "dénote un emprunt particulièrement récent." Indeed, Pernot (1934, 258) found -'ukhu was still productive in rendering newly borrowed MG -'ono verbs: e. g. learned MG dike'ono > ditee'ukhu 'to justify'. Since MG does not distinguish between /o/ and /o:/, -'ukhu is not in fact inconsistent with sko'tunu being recently borrowed from MG; /o:/> /u/ need not be an archaism at all. The -ukhu suffix seems to have arisen by analogy with the aorist suffix 'uka (corresponding to MG -osa), and the -'ikhu class of verbs. As Pernot (1934, 273) finds, -'unu is newer than -'ukhu, but both are present for this verb (sko'tunu, sko'tukhu), so this verb has been in the language for a reasonable amount of time. Indeed, -'unu has been so recently imported in the language that it has yet to spread to the subjunctive progressive of sko'tunu, which is only sko'tukhu. In C, however, the sko'tukh stem is only listed for the passive participle skotu'khumene. NT sko'tunu: Thus 64. Kisilier (2017, 129) records ko'tukhu for K. All forms borrowed.

63. swim

vew neo:, nAew pleo:

KoAu^nm kolim'bo

npeou 'preu, KouAu^ni^ou kufa'mbizu

(a)npeou (a)'preu, KouAiou^ni^ou kutiu'mbizu

niAeuyw pi'levyo, KouA^nm khul'mbo

The primary meaning ofpleo: is 'to float'. MG kolim'bo < AG KoAu^pdw kolumbao: 'to dive'. MT 'preu, PT pi'levyo 'float, swim' < AG pleo:; the PT form has been morphologically remodelled to a different conjugation. Costakis (1951, 51) gives the ST form a'pleu as a counterexample to /pl/ > /pr/; this is not mentioned in C. ST kufau'mbizu < EMG KoAu^piZW kolim'vizo< AG KoAu^p« kolumbo:; Deffner gives the form KoAiou^nou kofaum'bu. The fact that ST retains the AG /u/ suggests that ST did not necessarily borrow the form from MG; however, Modern Laconian Greek also pronounced the noun for 'swim' as ko'lumbi (Newton 1972, 23, cited from Koukoules 1908), making a loan much likelier. C (I:xii) explicitly gives Tk KouAiou^ni ku'faumbi < kufau'mbizu as older than the MG loan ^ndvio 'bapo < Italian bagno, and dnAe^a 'aplema < pleo: as older than ku'faumbi.

There are no instances of 'swim' in the available ST or NT texts, so one cannot establish which is the unmarked variant from textual frequency. All other things being equal, the deviant form 'preu must be assumed to be the original Tk form; as a result, no forms are counted as cognate.

64.fly

neTo^ai petomai

neTw pe'to

ave|iouK ou ane'mukhu

neTou pe'tu

neTw pe'to

According to Deffner, ane'mukhu < *dve^ow anemoo:, cognate to ave^og anemos 'wind'; cf. AG dve^iZp^ai anemizdomai 'to be driven by the wind'; dve^ou^ai anemoumai (the passive counterpart of * anemoo:) 'to be filled with wind; to be inflated, swollen'. There are two instances of ST pe'tu outside C: Har p. 145 arxipie na raipetalude ena yiurepetunte 'she started seeing butterflies all around flying', and Ikonomu's 1870 poem The Sparrow (cited in Lekos 1984 [1920], 94): nsrovvra n^otip' e^dTaepe; pe'tunda phur e'zatsere 'how did you go flying?' No NT form is given by C; Houpis uses ane'mukhu exclusively (Mal §3; §17), although the definition in C of ane'mukhu gives examples only of MG pe'to in its other meaning, 'throw' (also present for ane'mukhu in Houpis), and ane'mukhu is used only in the sense 'throw' in ST texts (22 instances in Dict, Samp and Har.) This means that both NT and MG have the same polysemy for pe'to and ane'mukhu respectively: both 'fly' and 'throw'.

Now, it seems this polysemy is accidental for MG: while clearly pe'to < AG neTo^ai petomai 'fly', Andriotis (1990 [1983]) quotes Hatzidakis' (1989 [1905], 413) derivation from AG neTdvvu^i petannumi 'open, spread out, unfold'; specifically for the meaning 'throw', he adds Menardos' derivation from dnmdaa« apotasso: 'range apart; depart from', presumably via the aorist dnsTa^ov apetakson (cf. the aorist of pe'to, neTa^a petaksa.)

One suspects, therefore, that a similar case to 'adropo ~ 'atfopo 'person; man' has taken place: either pe'tu was borrowed into ST to disambiguate ane'mukhu, or ane'mukhu was extended to the meaning 'fly' from the meaning 'throw' in NT, as a calque of MG pe'to. Although in the case of 'atfopo NT was more conservative than ST in retaining it in the meaning 'man' when MG 'adropo was introduced, in general NT has been more subject to MG influence than ST, and it is improbable that an accidental polysemy in MG would be echoed by the same polysemy in an unrelated Tsakonian word. There is a possibility that the polysemy is not accidental, with 'throw' taken as the causative version of 'fly'; even so, however, it is difficult to see why the same causitivisation would occur in NT if not under MG influence. Thus, NT ane'mukhu 'fly' is a calque of MG pe'to, while ST and PT here have cognates.

65. walk

nepinaxew peripatéo:, naxérn patéo:, (Paívia baíno:, (PaSí^a badízdo:

nepnaxw perpa'to

xpé/m 'trexo, xcrd/ou 'tjaxu S, nepnaxoú perpa'tu

xpé/ou 'trexu, xad/ou 'tjaxu

xpé/m 'trexo

A tjachoumene ('walking', but given as verb). Of the AG terms, the primary meaning ofpateo: is 'to tread, to step on' (peripateo: adds the preposition peri 'around'), while badizdo: also means 'to go slowly, to pace'; both baino: and badizdo: have been extended in meaning to 'to go', a meaning baino: tends to acquire in compounds (e. g. eKpaivw ekbaino: 'to get out' >MG Pyaivw 'vyeno.) baino: refers to motion only in its present stem: its perfect aspect denotes standing. The text counts strongly suggest badizdo: was the basic form for 'walk' in AG: in Plato, there are 17 instances each of the verbs badizdo: and baino: against 7 of peripateo:, while in Aristophanes there are no less than 89 instances of badizdo: against 2 of baino: and 4 of peripateo:.

MGperpa'to < AG nepinaxewperipateo:; while badizdo: > va'dizo also survives in MG, there is no doubt that perpa'to is the basic term, with va'dizo meaning 'to pace, to walk slowly'. 'trexo is SMG for 'to run'. 'tfaxu 'to walk; to run' is the regular reflex of D xpaxw trakho: 'to go', cognate to AG xpe^M trekho: (> 'trexo) 'to run'. No instances of 'tfaxu 'to walk' have turned up in the Tk texts I have. No cognate of perpa'to is listed in C, but the verb is used by Houpis (Mal §3, §17).

66. come

ep/o^ai érkhomai

ep/o^ai erxome

napZ íou par'jiu

na(p,Z )íou pa'(r/j)iu

ep/o^eve er'xomene, '|ia 'molo ma

According to Deffner,pa'riu < napiwvpario:n, past participle of nap^iparie.mi 'let fall; pass by, pass over; allow to pass, admit'. The past tense of MTpa'riu is (e)'kana < D iKav« hika:no: 'to come' (Pernot 1934, 286), and the MT (Scutt 1913-14, 28) future is formed from PT 'molo < D ^oAeTv molein 'to go (aorist)'. NT, ST forms non-cognate.

67. sit

KdBn^ai

káthe:mai

KdBo^ai 'kaOome

(raí %eve ka'jimene)

Kacíou ka'siu,

Kaxaaívou

ka'tsenu

Kaa^Kia ka'siko, Kaaírn ka'sio

Q qi-si-mni 'seated' (Kaa%eve ka'simene). NT ka'fimene only attested in past participle form; its palatalisation is characteristic of NT (§3.2.4.5). Tk ka'si(k)u 'I sit; I sleep (present)' < Kaa^Ka ka'sika 'I sat; I slept (aorist)' < AG Ka0iZM kathizdo: 'I seat' (with /0/>/s/: §3.2.1.1) < Kaxa kata 'down' + i^w hizdo: 'to seat'. Tk ka'tsenu has the aorist eka'tsaka and the imperative 'katsa (Pernot 1934, 268); these are comparable with the MG aorist 'ekatsa, imperative 'katse, and it seems the present stem has been remodelled after the aorist. The MG aorist 'ekatsa < e'katsa < *'ekadsa< e'kadisa (Hatzidakis 1990 [1907], 156); Tk does not have a -sa aorist (as shown by the fact that the Tk aorist eka'tsaka adds ka on to the MG stem), so one would reasonably consider ka'tsenu an importation. MG 'kadome < AG Ka0^ai kaths:mai 'to sit' < Kaxa kata 'down' + ^ai M:mai 'to sit'.

Alongside ka'si(k)u, Tsakonian also has the form ka'thi(n)u 'to place'; while C derives this from kathizdo:, Deffner and Pernot (1934, 279) consider it likelier that the aspirated stop reflects a cluster formed by apocope from Kaxa kata 'down' + e0^Ka ethe:ka 'I placed'. Since the aorist of ka'siu is ka'sika rather than *kasi'aka — even though the ika aorist is characteristic of transitives rather than intransitives — and ka'siu is defective, Pernot considers ka'siu to have been influenced by its transitive equivalent ka'thinu.

Of the two ST forms, it seems ka'tsenu means 'sit down' (eKaxaaKa Kovxa xav iKapa va aoviaxou eka'tsaka konda than i'khara nafopis'tu 'I sat (aorist) near the fire to warm myself': C)

and ka'siu 'be seated' (ski Kaa%eve х^цои тсс' oki viou 'eki ka'simene 'xamu te 'okipu 'he sat (progressive) on the ground and didn't speak': C). (The two notions are distinguished only by aspect in MG.) The other meanings of ka'siu are 'dwell, stay, settle into some place, be occupied with, be idle, rest, sediment, stay awake, set (of sun or stars), go to bed, sleep', while ka'tsenu also means 'settle into its place, set (of sun or stars), go to bed, get stunted, wait'. The dichotomy is largely borne out in the corpus: in the ST corpus, 9 of 14 instances of ka'tsenu are inceptive, while the remaining 5 are to do with the meaning 'stay', which also occurs with the MG aorist 'ekatsa. For ka'siu, 15 instances are progressive and 3 inceptive — the spread to inceptive implies ka'siu is the unmarked term. In NT, 12 instances of ka'tsenu are inceptive, with one doubtful and one instance progressive and meaning 'stay'; there are 7 instances of ka'siu, all progressive.

ka'tsenu does not seem to be present in PT: ka'sio is used in both the inceptive (19 instances in the corpus) and progressive meanings (36 instances), while there are only two instances of a kats stem, which may well have been borrowed from MG. All forms cognate.

68. lie

кеТцш keîmai,

катакАгуш

kataklino:

ÇanAœvœ ksa'plono

TCTanAoûvou tsa'plunu, yeipou 'yiru, npayiàÇou pra'yiazu

тстапроикou tsa'prukhu, yeipou 'yiru

ÇanAœvœ ksa'plono, yeipœ 'yiro

MG ksa'plono < AG e^anA« eksaplo: 'to spread out'. According to C, none of the terms for 'sleep' which etymologically may have once meant 'lie down' — pra'yiazu < 'sideways', ka'siu < 'sit', 'kjufu < 'stoop' — have that meaning in current Tk. Both the NT and ST forms tsa'plunu, tsa'prukhu, are regular reflexes of the EMG form. (Kisilier 2017, 112 also reports the more MG-like forms ksa'prukhu in Tyros, ksaplukhu in K.) The alternation of older -'ukhu and newer -'unu as verb endings in Tsakonian has already been commented on for 'kill'; the ST form is quite regular, while the NT form is characteristically much closer to MG. The alternative word 'yiru has as its primary meaning 'lean', as does its MG cognate 'yerno; it occurs once in the NT corpus as against pra'yiazu occurring twice (the latter possibly under MG influence); from the example ("lay down on the bed and went to sleep"), 'yiru probably corresponds to MG 'yerno, which is inceptive rather than stative. NT 'yiru: Scutt (1913-14, 27). All forms cognate.

69. stand

1'сттг|Ц1 histe:mi

сттеко|Ш1 'stekome

сттекои 'steku

сттекои 'steku

сттекш 'steko

In MG, CTT8KM ste'ko is a less frequent variant of 'stekome; the form is ultimately a backformation from the AG aorist SCTT^Ka heste.ka, and occurs already in the Septuagint (1st century bc) and the New Testament. The form looks like a recent importation; a verb does exist with /st/ > /th/, as expected for Tk, but this is the inceptive 'thenu 'to stand up, to awake', attested by Q (ezanw 'stand up!'; interpreted by Pernot (1934, 503) as 'etha 'thanu, or possibly e'thatse 'tanu 'he stood up'.) This form points to a proto-form *'stenon, and the variant 'thendu reported by Pernot (1934, 380), to *'stezon; this indicates the morphology of this verb has been extensively remodelled. All forms borrowed.

70. give

SiSœ^i dido:mi

Sivœ 'dino

Siou 'diu

Siou 'diu

Sivœ 'dino, vwia 'dino

Apparently, MT 'diu < AG 5i5w dido:, with elision of intervocalic /8/ (§3.2.3.4)—something Pernot (1934, 276) attributes to dissimilation. MG 'dino is an analogical back-formation from the aorist form of the verb. All forms cognate; PT borrowed.

71. say

Mya> légo:, ^ÓCTKrn phásko:, phs:mí

Ma> 'leo

^a^oú la'lu

aou a u

Mía 'leo, M^w la'lo, |íouctí(k)ix> mu'si(k)o

The AG forms are a complex interlocking of four verbs — the three cited, and *emi eipo:, surviving only in the aorist elnov eipon. These verbs fill in each others' conjugations; the imper-fective ofphasko: is matched with the aorist ofphe:mi, and the aorist eipon substantially replaces the aorist of lego:. lego: is dominant in the present tense: in the first person singular, Plato uses lego: 496 times against phe:mi 166 times (Aristophanes: lego: 44 times, phe:mi 20 times, phasko: once), and in the third person singular, Xeyei legei 260 times against phe:i 44 times (Aristophanes: legei 47 times, phe:i once). In the aorist, eipon has an appreciable presence (third person singular: 29 instances of elne(v) eipe(n) in Aristophanes, 26 in Thucydides, 118 in Plato), butphe:mi is dominant (16 instances of e^n ¿ph£: in Aristophanes, 56 in Thucydides, no less than 2103 in Plato), while lego: is hardly in the running (3 instances of sXe^e elekse in Aristophanes, 8 in Thucydides, 1 in Plato).

ST a'u < NT la'lu < AG XaXea laleo: 'utter'; la'lo is also the unmarked term for 'say' in Cypriot. PT mu'si(k)o < AG ^u0ea muteo: 'say' and / or D ^ouai55ei mousiddei 'speak (3sg)'; Costakis (1951, 191) gives this as an example of a PT archaism. The glosses of mu'si(k)o in C point more towards 'speak' than 'say' (intr). The aorist of la'lu, ST e'peka PT peka, is the same stem as the aorist of MG 'leo, 'ipa, and is the reflex ofAG eipon. Alongside it, the aorist of laleo: survives: ST ali'te PT la'lika. For aorist stems, where the conjugation of la'lo has been contaminated by 'leo in PT, the 'peka stem outnumbers the (v)a'ika (< la'lo) stem in the PT corpus by 168 to 54, with four instances of mu'si(k)o. For present stems, where the verbs la'lo and 'leo are distinct, 'leo outnumbers la'lo 74 to 1. Although there is clearly still usage of la'lo in the present (e. g. C's example fiarja xio 0eo a^apea' va'io 'tio de'o sxo'res 'I tell you God have mercy'), 'leo has clearly become the dominant form in PT. NT, ST non-cognate; PT borrowed.

72. sun

^iog he:lios

All forms cognate.

73. moon

aeA.^vr| sels:ne:

^iog 'iljos ['Líos]

^e^dpi fey'gari

'iáe

^e^dpi fe'gari

'iáe

9e^d(p,íj )i fe'ga(r/j)i

^Xie We

^e^dpi fe'gari, 9e^a 'fega

MG fey'gari < AG ^e^apiov pheygarion, dim. of ^e^og pheygos 'light'. PT 'fega not mentioned in C, but given in Makris (1952). All forms cognate.

74. star

ÚCTX^p aste.r

dcxpo 'astro

dcxpe 'astre S, dxcCi 'atji S, ácÍKi

(a)axépi (a)'steri

No etymological derivation given by C, although the etymon AG aaxpov astron 'star' (> MG 'astro) is obvious. Given the attested phonetic correspondences of Tk, the probable pathway is AG aaxpa astra 'stars' > 'astri 'star' (back-formation; recorded by Kisilier 2017, 126 for Tiros and Melana) > *'astfi > 'atfi (which Kisilier 2017, 125 also records in Tiros). *'astji > 'aftfi (recorded in Kisilier 2017, 126 for Tiros and Melana) > 'afki ['asci]. The change /tji/ > [ci] is odd, but has a parallel in Costakis' pronunciation of the reflex of *a'stritis 'asp (vipera ammodytes)', noted by Pernot (1934, 327): both [a'jtjita] "avec un premier /J/ plus sifflant que le second", and [aJ'tJhita], where [tJ] was in free variation with [c]. It seems that, as a dissimilation, /JtJV was resolved to underlying /ft/ through the lenition of the second /J/; /Jt/ then immediately

went on its way to palatalising into [fc]. *'astri > 'afi (cf. MG aTpipw 'strivo > ST aou^ou 'fufu, NT axaou^ou 'Jtfufu 'twist'.)

No PT form of 'astre is given in C, although PT has the dim. aaTpouAi as'truli, and the derived form aaTpo^evT^ia astrofe'dzia 'starry night'; PT a'steri < AG dim. daTepiov asterion < astron, also present in MG. C claims a'steri is an MG loan; although this may be the case for its ST cognate aaTeZi a'steji, there is no obvious cause for doing so for PT. All forms cognate.

75. water | uSwp hudo:r | vepo ne'ro

UO io

UO 10

vepe ne re

Q iy-w 'cold water', narww 'water' — narww is ne'ro, and there is no reason to suppose this to have been part of Tk at the time. Vi (travel journal: Famerie 2007, 237) uw 'io. MT 'io < AG u5wp hudo:r 'water'. Although intervocalic /8/ drops out frequently in Tk (see 'foot'), Pernot (1934, 174) considers a direct transition unlikely, since AG /o:/ > Tk /u/; he instead postulates the process 'io < * 'ivo < back-formation from plural Tk 'ivata < EMG *'ydata < AG u5aTa hudata, plural of hudo:r. The existence of *'ivo is confirmed by the dim. PaTaouAi va'tsufai. MG form already present (as ne:ron) in 2nd century bc. Hatzidakis' thesis that the initial vowel of 'io fails to be /u/ because the word was borrowed through the Church is unlikely. While the ancient word has survived in Peloponnesian Greek, interestingly, it has survived with the /u/ vowel: ou5aTa 'udata 'urine' (Andriotis 1974 §6166). NT, ST forms non-cognate; PT form borrowed.

76. rain

úexóg huetós

Ppo/n vro'xi Ppoxn vro'xi

Ppé/o 'vrexo

Ppo/d vro'xa

The ST form has changed gender from feminine to masculine; C believes it is patterned after masculine 'ifae 'sun'. Deffner reports the form vro'xi for Leonidio; the word was unknown to Costakis (1934, 334). The PT form vro'xa seems to be the native Tk reflex of IG Ppo^n brokhe: (§3.2.1.3); with the exception of the words for 'mother' and 'daughter', feminines ending in /i/ in Tsakonian are all recent importations into the language from MG (Pernot 1934, 169; Costakis 1951, 72 says outright that vro'xi is a loan.) The ancient Greek word has survived in */ye'tia/ ^ [ju'cia] 'wet weather'. ST, PT forms cognate; NT form cognate and borrowed.

77. stone | ACQog lithos | neTpa petra | neTae petfe | neTae petfe | neTpa petra

MG 'petra 'stone' < AG neTpapetra 'rock'. The MT form displays the normal MT process /tr/> /tJV (§3.2.3.1.a), although its gender has switched to masculine from feminine; C adduces an instance of a masculine form neTpoi petroi in Strabo. As already mentioned, /tr/ > /tJV does not occur in PT; we have no way of telling whether PT 'petra is an archaism or a loan from MG, but presume the former. Kisilier (2017, 127) records the variants 'petfi (Tiros), 'petse, 'petsune (Melana), petfune (Prastos, Tiros). All forms cognate.

78. sand

a^^og ammos

All forms cognate.

79. earth

yñ ge:, X0®v

khth6:n

a^^og amos

Yn Yh

'xoma

a|i|io amo

Yn Yt,

'xoma, xoú^a 'xuma S

a|i|io amo

(i)Yn (i)Yi, Yng

Yis, x(<á/oú)|ia 'x(o/u)ma

a|i|io amo

(i)Yng (i)'Yis, avryng apiris,

Xm^a 'xoma

MG 'xoma < AG kho:ma 'earth thrown up, mound'. Two forms are involved here: yi 'ground' and 'xoma 'dirt' (although it has also acquired the meaning 'ground', absent in MG: M KaAe xw^a ka'le 'xoma 'good ground', T e^e naAe^ouvTe ^e Ta ^w^aTa 'emepaley'gunde me ta 'xomata 'we busy ourselves with the land' — cf. H ^e Ta y^g noAe^m me ta yispole'mo 'I busy myself with the land'.) The form yis seems to have taken over from yi in vernacular MG, with yi reinstated by Puristic Greek.

From an inspection of the examples given by C, the dominant form in ST is i'yi; however both Kisilier (2017) and Kassian (2018) take the main form to be 'xoma. Pernot (1934, 172) considers i'yi a loan from MG ("un emprunt aux dialectes voisins"), and it is true that i'yi does not fit the Tk native declension patterns; it is known that Doric had ya ga: for IG y^ ge:. NT y^ DH; this seems a straightforward loan from MG. As for 'xoma, the form 'xuma is clearly the native form (§3.2.2.3), with 'xoma a loan from MG. All forms borrowed from MG.

80. cloud

vé^oç néphos, ve^é^n nephéle:

CTÛvve^o 'sinefo

CTÛ(v/y)ve9o 'si(y)nefo

CTÛ(v/y)ve9o 'si(y)nefo

CTÛvve9(o/e) 'sinef(ofe)

NT 'siynefo in Houpis (e. g. Mal §23). In other dialects which have retained AG /u/ (IG /y/) as /u/ (Maniot, Old Athenian), this word occurs as suyne'fia, which is consistent with the word's etymology (< AG ativ sun 'with' + vé^oç néphos 'cloud'). Note that all words inheriting the AG sun- prefix in ST have /i/ rather than the archaic /u/ (Pernot 1934, 110). All forms cognate.

81. smoke

Kanvoç kapnos

Kanvoç kap'nos

Kan(i)vé kap(i)'ne

Kan(i)vé kap(i)'ne

Kan(i)vé kap(i)'ne

Vi Kanve kap'ne. The Tk epenthesis is characteristic of its treatment of such clusters; cf. AG 5eTnvov deipnon 'dinner' > ST 5ein(i)ve 'dip(i)ne 'dinner after a funeral'. All forms cognate.

82. fire

nùppûr

^rnxid fo'tja

Kapa 'khara

(i)Kdpa (i)'khara ^rnxiafo'tia

MG fo'tja < 9«xia photia < AG 9«g, 9«xog pho:s, pho:tos 'light'. MT 'khara < AG ea^apa eskhara: 'hearth; coal-pan' (C; Pernot 1934, 170), which survives as MG a^apa 'sxara 'grill'. Costakis (1951, 61) compares it to Hesychius eKxapeav ekkhareo:n 'of the kitchens'; indeed, he quotes Hatzidakis on this point, who goes so far as to say "this, too, is a sure piece of evidence for the descent of Tsakonian from the Laconian dialect." It may well be that ekkhareo:n reflects Proto-Tsakonian: kx could be an attempt to render aspiration at a time x was already fricated (this is in fact how modern Tsakonians write /kh/), and the semantics is consistent with the meaning of eskhara:. The PT form is not necessarily a loan: it retains archaic stress compared to MG, and it has a cognate in MT 9Kia fkia 'light'; treating it as a loan, however, is the simplest explanation for the form. NT 'khara: Thus 66.

AG pur survives as ST 'piji, kir, kiji 'fire; intense heat': the proverbial expression eZaKai Kup KuZiou oXoi e'zakai kir ki'jiu 'oli 'they all went to the Fire of the Lord (=Hell)' strongly suggests that this word was reintroduced into Tsakonian through the Church; at any rate, it is clearly not the unmarked expression for 'fire'. NT, ST forms non-cognate; PT form borrowed.

83. ash

anoSoç spodos, xé^pa téphra:

CTxa/xn 'staxti

anoïÀa spo'ila

anoïa spo'ia

CTxa/xn 'staxti, ai a/xn 'sthaxti

The textual evidence weakly suggests tephra: was used more widely than spodos: there are three instances of tephra: in Aristophanes and one in Plato, but none in either of spodos.

MG 'staxti < AG axaKx^ stakte: 'dripping'. ST spo'ia < NT spo'ila (§3.2.4.1) < *ano5iXa spo'dila < AG ano5og spodos (cf. AG ano5ia spodia 'heap of ash'). NT spo'ila: Thus 36, corroborated in Kisilier (2017, 127); the form as given is as would be phonologically expected, and the form spo'ila is also given in Oikonomou (1870). Oikonomou also givespo'ila andpho'ia, with the latter the expected reflex of spodia. There is no obvious channel for spo'ia to have been borrowed into Tk from MG; it is likely that the regularpho'ia was in use in the 19th century alongside the unexpected archaism spo'ia, and that spo'ia eventually displacedpho'ia because its sp- made it look like MG — not because it actually matched anything in MG. ST aKovi 'skopi 'ashes; powder

(as medicine)', cognate with MG okov^ 'skoni 'dust', is a borrowing, and clearly a rarer form. NT, ST forms non-cognate; PT form borrowed.

84. burn

i kaío:

'keo

Saíaou 'desu

SaÍCTOu 'desu

ksa'li(g/k/x)o

MT 'desu< AG Saiw daio: 'to burn'; Hesychius attests the use of this verb in Laconian. A problem here is the emergence of /s/; this has led Deffner to postulate *'dedon, with /0/ subsequently going to /s/ (§3.2.1.1). There are only three verbs in Tsakonian ending in -su (*a'ledon > a'lesu 'to grind', *'ynedon > 'nesu 'to spin', 'desu), and Pernot (1934, 134) considers it likeliest that in all three cases, the present stem /s/ originates by analogical extension from the aorist stem (t.lesa, ene:sa, edaisa); ironically, the Tk aorists of all three have now dropped their /s/ (a'leka, e'neka, e'daka), as the aorist suffix ka has spread in the dialect at the expense of sa (Pernot 1934, 250).

C gives no etymology for PT ksa'ligo; although the first syllable suggests the MG prefix ^e- + a- kse a- 'un-; exhaustively' (cf. H ^a^a^w ksa'lazo 'change clothes'< kse + a^a^w a'lazo; but also H ^aAuvw ksa'lino 'get untied' < kse + Auvw 'lino). AG Aryvug lignu:s 'smoke' is suggestive — then again, so is Turkish alaz 'flame'. At any rate, this is clearly not a loan from MG. Liosis (pers. comm.), in his contribution to the Cognacy in Basic Lexicon project at the Max Planck Institute, Jena, has proposed AG e^avaAiaKw eksanalisko: 'to consume', with -iygo the PT counterpart to the productive verb suffix NT ST -indu, MG -izo; the cognate kat-ana'lisko had already acquired the sense 'to burn, to destroy' in Mediaeval Greek. He admits e^aAei^w eksaleipho: as another possible etymon. No forms cognate.

85. path

ô86ç hodós

Spô^oç 'dromos

aTpaTa 'strata

CTxpdxa 'strata, CTÚp^a 'sirma, nopeia po'ria, Spó|io 'dromo

aTpaTa 'strata, Spo|ie 'drome

Vi nopeia po'ria, noting the survival of the archaic word po'ria < AG poreia 'journey, march', where other Greek dialects used aTpaTa < Italian strada. The term strata is extant in MG, but is not the most common current expression. A semantic swap seems to have taken place: C glosses Tk 'dromo as MG po'ria 'journey, march', and Tk po'ria as MG 'dromos 'way, path'. Furthermore, AG 5po^og dromos primarily means 'race', and only secondarily 'running track'. 'sirma < AG oup^a surma 'something trailed along'. Of the sundry synonyms given,po'ria is clearly predominant in ST: there are 12 instances in Diet, Samp and Har, and no other word is present. ST form non-cognate.

86. mountain

opog oros

piouvó vu'no

CTíva 'fina, CTíve 'Jine

áíva fina

piouvó vu'no, piouví vu'pi

MG vu'no < AG Pouvog bounds 'hill'. Kisilier (2017, 126) records the cognate vu'ne in Pras-tos, and 'sina in Melana. MT 'fina < D 0ig, 0ivog thi:s, thi:nos 'bank, sandy hill; strand, highlands' (C, after Deville and Hatzidakis). Scutt (1912-13, 151), after Deffner, speculates it might be a reflex ofAG pig, pivog ri:s, ri:nos 'nose'. Although /0/ tends to go to /s/ in Tk, not /JV (§3.2.1.1), Pernot (1934, 374) is inclined to accept 0ig as the etymon, having heard from a Katsanitsiot the form tho 'sina 'on the mountain'. Kisilier (2017, 111) proposes MG ori'na 'mountainous.neut. pl' as the etymon; the aphaeresis of ori'na > ri'na would explain the initial /JV by §3.2.1.2, but the accent shift is not explained, and I do not consider this development likely. NT, ST forms non-cognate; PT form borrowed.

87. red

épu9pôç eruthrós

kôkkivoç 'kokinos

KoXCTive kotsi'ne

KOXCTivS

kotsi'ne

KoXCTive kotsi'ne

AG is notorious for the difficulties in matching its colour vocabulary to that of modern languages. There were four AG terms for red: eruthros seems to have been unmarked, while ^oivi^ phoiniks corresponded to dark red, and nop^upeog porphureos to light red (although ancient enumerations of the colours of the rainbow (Maxwell-Stuart 1981, 163-169) tended to eschew erutros in favour ofporphureos and, in particular,phoiniks). In Old Greek, eruthros is less used: Homer uses eruthros 11 times, porphureos 36 times and phoiniks 25 times. By Classical times, eruthros was dominant: Plato uses it 5 times, againstporphureos 3 times andphoiniks none. (Aristophanes uses only eruthros, 3 times).

The fourth AG term is the term that survived into the modern language: MG 'kokinos < AG KOKKivog kokkinos 'scarlet' < K0KK0g kokkos 'scarletberry'. The regular reflex of this in Tk would be *'koteine; the process /ki/> /tsi/ does exist in Tk (§3.2.3.3.a), but seems to be late, and does not explain the accent shift. Pernot (1934, 72) believes this indicates the presence of adjectival formant -*ti'nos> -tsi'ne, a metanalysis of AG / MG adjectival formant -ivog inos ~ i'nos, also present in kji'sitsine 'barley (adj)' (MG Kpi0apeviog krida'renios), tji'xitsine 'made of hair' (MG xpixeviog tri'xenios). Kisilier (2017, 127) records the variant kotfi'ne in Tiros and Kastanitsa. All forms cognate.

. green

X^wpog khh:ros

npauivog 'prasinos

npauive prasine

npaaive 'prasine

npauive 'prasine

The denotation of khlo:ros is light green and yellow; it is the colour of young grass, honey, and sand. Dictionaries tend to ascribe pale green, bluish-green, and gray to yXauKog glaukos; in his monograph, however, Maxwell-Stuart (1981) argues for light blue as the primary denotation. Dark green seems to have been covered by KuavoCg kuanous, which was primarily dark blue, and by the surviving colour term, which appears to have been secondary in AG: MG 'prasinos < AG npaaivogprasinos 'leek-coloured'. Kisilier (2017, 126) records prafine in K. All forms cognate.

89. yellow

X^wpog khlo:ros, §av96g ksanthos, rn/po? o:khros

KiTpivog 'kitrinos

tct i(T)pZ ive 'tei(t)rjine

tct i^ive 'teijine

tct iTpive 'teitrine

khb:ros has already been seen; ksantos is defined as 'golden-yellow, pale yellow; red-yellow', and its major denotation (which has survived into MG) is 'blond'. Yellow was probably also covered by o:khros 'pale'.

MG 'kitrinos < AG KiTpivog kitrinos 'citrus-coloured' < Latin citrus. The normal reflex in ST would have been 'tßitfine or 'teitjine, depending on the order in which rules §3.2.3.1.a and §3.2.4.3 are applied. Presumably, as the NT form hints, the second /t/ dropped out, either in manner dissimilation from the first affricate, or because the resulting /t3/ affricate would be problematic (M3/ would be unacceptable in Tsakonian unless prenasalised.) Kisilier (2017, 126) records only the more MG-like 'tsitrine. All forms borrowed, given that the MG word itself is a loanword.

90. white I MuKÖg leukos | aunpog 'aspros | Mk6 le'ko | Mk6 le'ko | aunpe 'aspre

Vi AeuK6 lefko, recording it as a noteworthy archaism contrasted with MG aanpog. (It has been revived since in Puristic, mostly with figurative meanings calquing French blanc in the sense 'blank'.) MT le'ko < AG AeuKog leukos 'white'; MG 'aspros < Latin asper. To explain the fact that the reflex is le'ko rather than the expected lefko, Pernot (1934, 20) formulates a rule that fortitioned u drops out in Tsakonian, corroborated by *euriskon 4 e'rexu 'to find'. Kisilier (2017, 125) records the K form as 'astre, indicating a more recent MG loan. NT, ST forms non-cognate; PT form borrowed.

91. black

|iéAaç mêlas

l^aupoç mavros

Koußave ku'vane

|iaupe mavre, Koußave ku'vane

|iaupe mavre

MG 'mavros< AG ^atipog mauros 'dark'. MT ku'vane< AG KuavoCg kuanous 'dark blue'; with /v/ interpolated (§3.2.3.5). Hesychius attests the meaning 'black' for this word (Kouava kuana) in Laconian. NT ku'vane only listed for S, but used abundantly by Houpis (e. g. Mal §23). A perusal of MT texts and of the lexical productivity of the two words leave no doubt that ku'vane remains the unmarked MT form; there are 5 instances in Samp, Dict and Har of ku'vane, and none of 'mavre. NT, ST forms non-cognate; PT form borrowed.

92. night

vUÇ nuks

vu/xa 'nixta

vioUx a 'ßutha vioUx a 'ßutha voUx a 'nutha

Tk 'puta < EMG vuKTa 'nykta< AG vu£, vuKTog nuks, nuktos 'night'; AG /kt/> Tk /kh/ by §3.2.3.1.e, and D /u/> /ju/ by §3.2.2.1; PT /ju/> /u/ by §3.2.4.6. NT 'putha: Thus 28. All forms cognate.

93. hot

9ep|i6ç thermàs

ÇeCTXoç zes'tos

9ep|i6 Oer'mo, CTo^iCTxé fomi'ste

CTO|i6 fo'mo, CToviCTxé fopi'ste, Ae^ouxé lexu'te

Çeaxé zes'te

Q swmw 'dinner', A schomo 'cooked food' (a secondary meaning fo'mo has acquired in Tk). Oer'mo, fo'mo are glossed by C as 'very hot, boiling', and the texts suggest foni'ste is the unmarked term. MG zes'tos < AG ZeaTog zdestos 'boiling'. ST fo'mo < *Oremo < NT Oer'mo < AG 0ep^og thermos 'hot' (See Pernot 1934, 19 for the metathesis; initial /0r/ > /J7 by §3.2.3.1.a.) ST fopi'ste (past participle) < aovixou fo'pixu 'to heat' < NT ao^ou fo'mixu (§3.2.4.2) < *Ore'mizu < AG 0ep^iZ« t^ermizdo: 'heat'. NT fomi'ste: Thus 36. MG 0ep^og Oer'mos was reintroduced into the language through Puristic Greek; there is no reason to think that NT Oer'mo is a loan, since initial /0/ is frequently retained in Tk in words which cannot have been borrowed from MG.

Liosis (pers. comm.), in his contribution to the Cognacy in Basic Lexicon project at the Max Planck Institute, Jena, adds lexu'te as the ST form for 'hot', as opposed to 'warm'; this is a verbal adjective corresponding to the verb AexouKou le'xuku 'to pour boiling liquid on something', which Andriotis (1974, 353) derives from AG Aexm lekho: 'new mother'. (Deffner 1923 instead glosses the verb as 'to warm someone by putting them to bed', which explains the etymology, but it is not clear whether this is a gloss or etymologising; Deffner's exclamation "that explains everything" makes it suspect.) NT, ST forms non-cognate (though they use the same concept as MG: 'boiling'); PT form borrowed.

94. cold

^uxp6ç psu:khros

KpUoç 'krios

^ou/pé psu'xre

KÇavé kja'ne, ^ou/pé psu'xre

Kpuavé kria'ne,

Kp(a/e)vé

kr(œ/e)'ne

ST kja'ne < PT kria'ne < EMG 'krios + a'nos 'adjectival ending'. MT psu'xre < AG ^uxpog psu:khros 'cold'; this term survives in MG aspsi'xros, but is a reimportation from Puristic Greek in meanings other than 'chilly', and is not the unmarked term for 'cold'. There are no instances of the adjective in the main ST texts, although there is one instance each in Samp of the nouns KZa5e 'kjade (§11) and ^uxpa 'psixra (§23 = Dict §8ii) (which looks like an MG loan, notwithstanding the fact that 'psixra is attested throughout Tsakonia). With regard to the nouns, Pernot (1934, 382) notes that Costakis used both 'psixra and 'psuxra — but the first only jokingly ("en plaisantant"); while his Kastanitsiot consultant used 'krada. This suggests a split between NT and ST; but there is not enough other evidence to support it. PT /«/ by §3.2.4.7.

psu'xre is slightly more productive than kja'ne: it gives rise to MT ^oû^pa ~ Tk ^ti^pa 'psuxra ~ 'psixra 'cold (n.)', Tk ^ou/papsu'xra 'coldly', and ST ^ou/piaivoupsuxri'enu ~ ^ouxcraivou psux'fenu 'to make cold; to grow cold; to disappoint'. By contrast, kja'ne generates ST KÇaSa 'kjada ~ ST, PT KpuaSa kri'ada ~ PT KpaSa 'krada ~ PT KpaSa 'krœôa 'cold (n.)', ST KpuaviÇou kria'pizu ~ KÇaviÇou kja'pizu 'to start getting cold' and PT Kpuaivœ kri'eno ~ Kpaivœ 'kreno 'to grow cold, to feel cold'. Semantically, ST kja'pizu is clearly marked, as distinct from ST psux'fenu and PT kri'eno. So the evidence weakly suggests psu'xre is the more basic term in ST. ST, NT term non-cognate; PT term borrowed.

95. full

|nÀ.^pnçplé:re:s |ye^dxoç ye'matos |yio^dxe yio'mate |yio^dxe yio'mate |yio^dxe yio'mate

MG ye'matos < AG ye^M gemo: 'to be full'. yio'matos Qo'matos] is a widespread variant of ye'matos [je'matos] in MG; /e/> /o/ backed because of the following labial (Andriotis 1990 [1983]), subsequent to the palatalisation of /y/. Kisilier (2017, 128) gives je'mate (Prastos), jo'mate (Tiros, Vaskina), 'jomakiu (Vaskina); he addsfulari'ste (Melana) < MGfu'laro 'to pad', and xo'dre (Prastos) xon'dre (K) < AG khondros 'coarse' MG xon'dros 'thick, fat'. All forms cognate.

96. new

Kaivoç kainos, véoç néos

Kaivoupyioç ke'nuryios

Ta aivoupyie tee'nuryje S

Ta i(p)voûpTa'n tei(r)'nutei

Ta aivoupyie

tee'nuryje

Of the AG terms, kainos also means 'fresh', while néos also means 'young'. To eliminate the 'young' reading, neuter (=inanimate) singular instances of the two adjective were counted in Plato: there are 36 instances of véov néon against 9 of Kaivov kainon. (For the masculine singular, there are 47 instances of néos against only 1 of kainos.)

Pernot (1934, 377) reports that Costakis used both tei'nurtei and tei'rnutei; so the metathesis was synchronically current. ST tei(r)'nutei leads back to *ki'rnuki (metathesis) < *ki'nurki, so that the /ry/ cluster must have undergone fortition (Pernot 1934, 377). Deffner gives the form xÇivoûpxÇe tei'nurtee, whose /e/ ending arises by influence from the ending of MG ke'nuryios. An NT form for 'new' is only given for S by C; I have found no instances of 'new' in Houpis' writings. MG makes a semantic distinction between ke'nuryios (< AG Kaivoç kainos 'new' + epyov érgon 'work'), and véoç 'neos (< AG véoç néos 'new; young'): the former relates to things (cf. English brand-new), and the latter to abstractions, as well as meaning 'young'. The distinction is not made in Tk, where véo 'neo means exclusively 'young' (notwithstanding that Kisilier 2017, 128 gives 'neo as his gloss of 'new'). All forms cognate.

97. good

àya96ç agathos KaMç ka'los

KaM ka'le

KaM ka'le

ka'los < AG KaÀ6ç kalos 'beautiful, noble'. All forms cognate.

98. round

aTpo^û^oç stroygulos

aTpo^u^6ç stroygi'los

aTpovTÇuM strondzi'le

aTpovTÇuM strondzi'le

KaM ka'le

aTpovT^uM strondzi'le

Phonologically, this form is suspect because it does not reduce the /tr/ cluster to /tf/, as is normal in MT; the expected reflexes are NT *ftfondzi'le and ST *fondzi'le. strondzi'le has reduced MG /gi/ [ji] to /dzi/, as is regular in Tk; so the form has undergone some assimilation — but this only highlights the fact that */y/ here becomes /i/, which once more indicates that this is an MG loan. So all forms have been borrowed from MG, even though this is known to be an old form (Oikonomu 1888, 40): e'xouve xcr eva Kanovi arpovrZvXs 'aav xo nenovi 'exune te 'ena ka'popi strondzi'le san to pe'poni 'and they have a capon round as a melon'.) Kisilier (2017, 127) records strodji'le (Prastos, Tiros, Melana, Vaskina), strodzi'le (Tiros, Melana) and stroygi'le (Vaskina); he also records xo'dre < AG khondros 'coarse', MG xon'dros 'thick' for K.

99. dry |§r|pog kse:ros | §epog kse'ros |Taepe tse're |Taepe tse're |Taepe tse're

Tk tse're < AG ^npog kse:ros 'dry'; AG /ks/> Tk /ts/ (§3.2.3.1.d). The shortening of /e:/ in both MG and Tk is suspect. It is true, however, as Costakis (1951, 35) notes, that there are sporadic instances where /e:/ > Tk /e/ but MG /i/: AG nA^powple.roo: > MG nA^pmvwpli'rono Tk nAepouKou ple'rukhu 'to pay', AG T^yavov te.ganon > MG T^ydvi ti'yani Tk Teyave 'teyane 'frying pan', AG xa^^Aog khame:los > MG xa^Aog xami'los Tk xa^eAe xame'le 'low'.

MG differentiates between 'dry' as in 'dried up' and 'dry' as in 'not wet'; the latter is ste'ynos < AG steganos 'air-tight', and Kisilier (2017, 129) takes that as the basic form, giving its Tk equivalent as ste'yne. (He also addspsa'xne for Prastos < MGpsa'xnos 'fleshy'.) All forms cognate.

100. name | ovo^a onoma | ovo^a 'onoma | ovou^a 'onuma | ovou^a 'onuma | ovo^a 'onoma

Vi ovou^a 'onuma. NT 'onuma: Thus 74. /o/> /u/ under the influence of following labial. Kisilier (2017, 127) records 'onuma only in Vaskina; in Prastos, Tiros, Melana and Kastanitsa he records 'onoma. All forms cognate.

6. Other Lexicostatistical accounts

This paper was first drafted in 1996. Since then, three lexicostatistical accounts of Tsakonian have appeared:

• Blazek (2010) is a list of Swadesh-100 word lists for Greek; it includes Attic Greek, Modern Demotic, and Southern Tsakonian. The primary source it uses is Deffner (1923), mainly because Blazek feels Deffner reflects Tsakonian more accurately before the mass influence of Modern Greek.

• Kassian (2018) is a list of Swadesh-110 word lists for Greek; it includes Ancient Greek (Ionic: Herodotus and Attic: Plato), Modern Demotic, Southern Tsakonian, Pharasa Greek, and Aravan Cappadocian Greek. The primary source it uses is Deffner (1923), with secondary reference to Costakis (1951, 1986); Kassian objects to Costakis for his prescriptive tendencies. Kassian also uses Swadesh word lists recorded in situ in Prastos, Tyros, Melana and Kastanitsa in Vyatkina (2015).

• Kisilier (2017) concludes with a Swadesh-110 word list for Tsakonian, based on his own fieldwork.

The reliance of both Blazek and Kassian on Deffner over Costakis is in my opinion a mistake; Kassian's engagement with contemporary field work is welcome, but given language attrition, checking against corpora is also necessary. (It is true, as Liosis (2007) points out, that Houpis' grammar is that of a terminal speaker, and the reliability of his corpus is compromised; unfortunately his is the only comprehensive corpus available for NT).

Both Blazek and Kassian are affiliated with Sergei A. Starostin's programme of rehabilitating glottochronology ("recalibrated glottochronology": Starostin 1999). As part of that approach, it is critical for them to isolate loanwords from cognates, since Starostin's metholodology discounts loanwords as compromising the stability of the core vocabulary set. In that regard, it has to be said, glottochronologists are fortunate that the radical changes in Tsakonian phonotactics allow most loanwords to be identified readily.

Kassian's (2018, 11) conclusion is that there are too many loanwords in Tsakonian for it to be of much use: "Because of its mixed nature, the Tsakonian wordlist can hardly be used for lexicostatistic phylogeny of Ancient Greek dialects, not to mention for calibration of glottochro-nological formulae". On the other hand, Blazek uses the 63.4 % non-borrowed cognates between Attic and Tsakonian to argue for a divergence date of 1725 BC, close to the estimated date

of 1900 BC for the divergence of Northwest and Southeast Greek. He discards the estimate of 300 AD for a Tsakonian / Modern Greek divergence, based on 78.2 % non-borrowed cognates, as ahistorical, contaminated by the strong influence from Modern Greek (that even eliminating clear loanwords does not address).10

It is worthwhile to compare the findings of Blazek, Kassian and this paper with regard to which forms are cognate (C), which are borrowings (B), and which are not cognate (N) between Southern Tsakonian and Modern Greek (I am ignoring the very recent MG loans Kassian notes from the 2010s wordlists in Vyatkina 2015, and which Kisilier 2017 also notes). Of the non-cognates, N* indicates the Tsakonian forms which are cognate with the main Attic form, N# Tsakonian forms which are cognate with other Attic forms, and Nt Tsakonian forms with direct Doric antecedents.

Table 1. Lexicostatistical Comparison of Blazek, Kassian and Nicholas

Blazek Kassian Nicholas

1 I C C C

2 Thou C C C

3 We C (unaware of the problem raised by the difference between Doric-looking oblique and Koine-looking nominative) C (does not comment on the problem raised by the difference between Doric-looking oblique and Koine-looking nominative) B

4 This N (unaware that ende- pronouns are cognate with MG aftos, whether following Pernot's etymology, Deffner's, or Tzitzilis') C (allows Pernot's derivation, but considers etymology unclear) C

5 That C (takes on face value that e'tine is derived from Doric ts:nos, unaware of the resulting phonological difficulty) N (allows Pernot's phonological concerns) N#

6 Who C (unaware of the phonological difficulty of 'pier) B B

7 What C C C

8 Not C (unaware that MG den is marginal in Tsakonian, or that MG 'oxi 'no; non-' is not a proper rendering of 'not' N N*

10 Amusingly, even Starostin's recalibrated glottochronology takes no account of standard deviation (which to his credit Swadesh 1955 had). Starostin (1999, 10) derives a modified rate of decay of X = 0.05 on the Swadesh-100 list once loanwords are accounted for; he bases this on seven languages, which give a sample standard deviation of s = 0.01278, for use in the glottochronological formula for two languages (where c is the proportion of cognate words, and the value is in millennia):

À = 0.05, c = 0.782 in Starostin's revised glottochronological formula gives 1667 years, so (from a date of 1986) 319 AD. But a 95 % confidence interval is associated with three standard deviations (3a) either side of the mean value; using the actual mean À = 0.04625 and s (as an estimate of a) = 0.01278 gives us a 95 % confidence range for À between 0.00791 and 0.08459, namely a mean value of 1734 years, but a range between 1282 and 4193 years (so nominally 252 AD, but 95 % confidence for anywhere from 704 AD to 2207 BC.) Starostin's (1999, 11) claim that "the value for À is stable and varies only slightly between 0.04 and 0.06", is, regrettably, risible: the difference between 0.04 and 0.06 is one of SO % (and English in the sample set has a value of À = 0.08); and the relation of À to time elapsed is 1/V A, which shoots up as À approaches 0. (Hence the small difference between 252 AD and 704 AD, but the massive difference between 252 AD and 2207 BC.)

Blazek Kassian Nicholas

9 All C C C

10 Many N N N#

11 One C C C

12 Two C C C

13 Big N N N#

14 Long C B C

15 Small C C C

16 Woman C C C

17 Man N N N#

18 Person B C (argues that 'atfopo rather than the MG loan 'aOropo means 'person' based on Vyatkina 2015 data from Prastos and Melana; Vyatkina 2015 cites 'aOropo only for Tyros) B

19 Fish C B C

20 Bird B B B

21 Dog C (assumes ky5:n had genuinely survived in MG: the survival is Puristic, and rare at that) N N*

22 Louse C B (not so named in list of loanwords at Kassian 2018:11, but clearly considered a loanword in detailed discussion, Kassian 2018:99) C

23 Tree C C C

24 Seed N (different derivational suffix of MG 'sporos vs ST phrama) C C

25 Leaf C C C

26 Root C C C

27 Bark C B C

28 Skin C (assumes to'mari is the basic term, unaware ofpe'tsi or 'teepa) N (accepts Deffner's Hellenic etymology of 'teepa) B

29 Flesh C C C

30 Blood C C C

31 Bone C B C

32 Grease C (assumes the basic form is non-borrowed 'paxos, but that is the state of someone being fat; Kassian 2018 concurs) B B

Blazek Kassian Nicholas

33 Egg C (unaware that avu'yo is unlikely to be an old form in ST) B B

34 Horn C (assumes 'teerate is the basic term in ST, and did not even list 'tfoxane; we have seen the decision is difficult) B (assumes 'teerate is the basic term in ST, did not even list 'tfoxane) N#

35 Tail C C C

36 Feather C C C

37 Hair C (assumes 'trixa is the basic form in MG) N N*

38 Head C C C

39 Ear C C (because "the starting points in Tsakonian and Demotic are different", considers ST form influenced by rather than directly borrowed from MG) C

40 Eye C C C

41 Nose N N N#

42 Mouth C C C

43 Tooth C C C

44 Tongue C C C

45 Claw C (did not suspect the phonology was borrowed) B B

46 Foot C C C

47 Knee C C C

48 Hand C C C

49 Belly N N Nt

50 Neck C B C

51 Breasts C C (takes 'stiOi as the basic term; surprisingly for Kassian, does not note the phonological difficulty, which means that not only 'stiOos but 'stiOi must be recent loans) C

52 Heart C C C

53 Liver C B C

54 Drink C C C

55 Eat C B C

56 Bite N (assumes not cognate, since they look so different) C (accepts Pernot's derivation from dakno:) C

Blazek Kassian Nicholas

57 See N N N*

58 Hear N N N

59 Know C (unaware of the phonological difficulty of 'kseru) B B

60 Sleep N N N#

61 Die C C C

62 Kill C (unaware of the phonological difficulty of sko'tunu) B B

63 Swim C (assumes kuKum'bu is the basic term, unaware of preu) N N#

64 Fly C C C

65 Walk N N (considers 'eygu 'go' < AG erkhomai 'come' the basic form, glossing 'go' rather than 'walk'. Does not acknowledge the form 'tfaxu 'walk') N#

66 Come N N N#

67 Sit C C C

68 Lie N (posits the basic term in ST is 'vanu 'to set') B C

69 Stand C (unaware of the phonological difficulty of 'steku) B B

70 Give C C C

71 Say C (assumes MG 'lalo is the basic term) N (Kassian 2018:11 lists only the aorist suppletive stempe- as borrowed from MG) N#

72 Sun C C C

73 Moon C B C

74 Star C C C

75 Water N N N*

76 Rain C B C

77 Stone C C C

78 Sand C C C

79 Earth C (takes 'xoma rather than i'yi as the basic form) B (takes 'xoma rather than i'yi as the basic form) B

80 Cloud C B C

81 Smoke C C C

82 Fire N N N#

Blazek Kassian Nicholas

83 Ash N N N#

84 Burn N N N#

85 Path N N N#

86 Mountain N N Nt

87 Red C B C

88 Green C B C

89 Yellow B B B

90 White C (assumes lefkos has genuinely survived in MG; recalls that Villoison expressed surprise at its survival in Tsakonian) N N*

91 Black N N N#

92 Night C C C

93 Hot C (assumes Oer'mos is the basic term in MG) N (has difficulty finding expressions for 'warm', which he takes as the base concept, as opposed to 'hot') N*

94 Cold C (assumespsix'ros had survived in MG: it is secondary term) N N*

95 Full C B C

96 New C B C

97 Good C B C

98 Round C (unaware of the phonological difficulty of 'strondzile) C (exceptionally, unaware of the phonological difficulty of 'strondzile) B

99 Dry C C C

100 Name C C C

Non-Borrowed 97 72 86

Cognate Non- Borrowed 76 (78.4 %) 45 (62.5 %) 59 (68.6 %)

Blazek is not a Hellenist, let alone a Tsakonologist, and it shows: he is often naive about assuming Puristic cognates of ST were genuine survivals; he does not try to establish which among multiple alternatives is the basic term and allows any match among synonyms; and he is unfamiliar with Tsakonian phonology and its diagnostic value. In the case of 'seed', he has decided to discount the cognate pair for having different derivational suffixes; but the same objection could be raised for 'eye' or 'tree', and has not been.

Although Kassian has not used the range of sources this paper has, he is meticulous about trying to work out which terms are basic, and his conclusions are mostly the same as this paper's. Kassian is if anything even more strict than I have been about considering terms to be MG loans,

and on inspection it is clear his criterion has been to consider a term a loan from MG if it is a post-Koine innovation in MG, even when those innovations are still cognate, as with 'bark' (AG phloios; ST frua < MG fluda < back formation of fludion < dim. ofphloios), or 'long' (AG makros, MG changed declension ma'kris < *ma'krys, ST matching declension ma'kju). That is a methodological difference, and it is consistent with regarding Tsakonian as a Doric variant with heavy MG colouring; Pernot's contrary approach, of regarding Tsakonian as a Modern Greek variant with a Doric core,11 has less difficulty in admitting phonologically native forms like tsaprukhu 'lie' or 'Jkoki 'liver' as Tsakonian cognates of ksaplono and si'koti, rather than as early loans. Kassian also makes some different interpretations of basic terms ('warm' vs 'hot', 'go' vs 'walk').

The massive disparity in cognate proportions between the three approaches point to the methodological challenges around lexicostatistics — all the more pronounced in the different results I obtain from Kassian, although we are in basic agreement about both the data and most etymologies.

References

Andriotis, N. P. (1956) I krasi tou symplegmatos ou+e se o sti mesaioniki kai nea ellinikl [The merger of u + e into o in Mediaeval and Modern Greek]. In: Mélanges offerts à Octave etMelpo Merlier. Vol. 1. Athens: Institut Français d'Athènes Publ., pp. 1-11. (In Modern Greek) Andriotis, N. P. (1974) Lexikon der Archaismen in Neugriechischen Dialekten. (Schriften der Balkankomission Linguistische Abteilung XXII). Vienna: Österreichische Akademie der Wissenschaften Publ., 705 p. (In German)

Andriotis, N. P. (1990 [1983]) Etymologiko lexiko tïs koinïs neoellïnikïs [Etymological dictionary of standard Modern Greek]. 3rd ed. Thessalonica: Aristotle University, Institute of Modern Greek Studies Publ., xxiii, 436 p. (In Modern Greek)

d'Ansse de Villoison, J.-B.-G., Famerie, É. (ed.) (2006) De l'Hellade à la Grèce. Voyage en Grèce et au Levant (1784-1786). Hildesheim; Zurich; New York: Georg Olms Verlag, 293 p. (Altertumswissenschaftliche Texte und Studien. Vol. 40). (In French) Bergland, H., Vogt, H. (1962) On the validity of glottochronology. Current Anthopology, 3 (2): 115-153. (In English)

Blazek, V. (2010) Glottochronological analysis of the Greek Lexicon: Modern, Tsakonian, Old and Mycenaean

Greek. Graeco-Latina Brunensia, 15 (1): 17-35. (In English) Caratzas, S. C. (1976) Les Tzacones. Berlin: De Gruyter Publ., xxi, 451 p. (In French)

Costakis, A. P. (1951) Syntomï grammatikï tïs tsakonikïs dialektou [A brief grammar of the Tsakonian dialect]. Athens: Institut Français d'Athènes Publ., 224 p. (Collection de l'Institut Français d'Athènes. Vol. 35) (In Modern Greek)

Costakis, A. P. (1975-76) Ta poimenika tis Tsakönias [Animal husbandry in Tsakonian]. Laografia, 30: 45-160. (In Modern Greek)

Costakis, A. P. (1980) Deigmata tsakonikïs dialektou [Samples of Tsakonian dialect]. Athens: Tsakonian Archives, 80 p. (In Modern Greek)

Costakis, A. P. (1986) Lexiko tïs tsakonikïs dialektou [Dictionary of the Tsakonian dialect]: In 3 vols. Vol. 3. Athens:

Academy of Athens Publ. (In Modern Greek) Deffner, M. (1923) Lexikon tïs tsakonikïs dialektou (Dictionary of the Tsakonian dialect). Athens: Estia Publ., xxi, 411 p. (Lexikografikon archeion tis mesis kai neas ellinikis [Lexicographical archive of Middle and Modern Greek]. Appendix. Vol. VI). (In Modern Greek) Famerie, É. (2007) Villoison et la redécouverte du dialecte tsakonien. Anabases, 6: 235-248. (In French) Haralambopoulos, A. L. (1980) Fonologikïanalysï tïs tsakonikïs dialektou [Phonological analysis of the Tsakonian dialect]. Thessaloniki: Aristotle University Publ., 195 p. (Aristoteleio Panepistimeio Thessalonikis, Epistïmonikï Epetirida tis Filosofilkis Scholis [Aristotle University of Thessaloniki. Scholarly papers of the Faculty of Philosophy]. Appendix. No. 30). (In Modern Greek) Hatzidakis, G. (1975 [1892]) Einleitung in die Neugriechische Grammatik. Athens: Academy of Athens Publ. (In German)

11 Kisilier (2017) takes a similar approach, arguing for example that in its loanwords, Tsakonian is indistinguishable from Modern Greek.

Hatzidakis, G. (1989 [1905]) Mesaiönika kai nea elllnika [Mediaeval and Modern Greek]: In 2 vols. Vol. 1.

Amsterdam: Adolf M. Hakkert Publ. (In Modern Greek) Hatzidakis, G. (1990 [1907]) Mesaiönika kai nea elllnika [Mediaeval and Modern Greek]: In 2 vols. Vol. 2.

Amsterdam: Adolf M. Hakkert Publ. (In Modern Greek) Houpis, D. G. (1990) Etrou Enatthe [That's how it happened]. Athens, 144 p. (In Tsakonian) Houpis, D. G. (1992) Chörata sta tsakönochöria [Jokes in the Tsakonian villages]. Athens, 159 p. (In Modern Greek)

Houpis, D. G. (1993) Malevisia kai tsakönika [Malevisia and Tsakonian]. Athens, 61 p. (In Modern Greek) Joseph, B. D. (1994) On weak subjects and pro-Drop in Greek. In: I. Philippaki-Warburton, K. Nicolaidis, M. Sifianou (eds.). Themes in Greek Linguistics: Papers from the First International Conference on Greek Linguistics. Amsterdam: John Benjamins Publ., pp. 21-32. (In English) Karaliotis, N., Contossopoulos, N. (ed.) (1969) Apo tl zöl tou Nikola Karaliötl [From Nicholas Karaliotis' life]. [Unpublished manuscript]. Athens, Centre for the Compilation of the Historical Dictionary of Modern Greek, Academy of Athens, 71 p. (In Modern Greek) Kassian, A. (2018) Annotated Swadesh wordlists for the Greek group (Indo-European family). [Online]. Available

at: http://starling.rinet.ru/new100/grk.pdf (accessed 14.03.2019). (In English) Katsanis, N. (1989) Koutsovlachika kai tsakönika [Arumanian and Tsakonian]. Elllnikl Dialektologia — Greek

Dialectology, 1: 41-54. (In Modern Greek) Kisilier, M. (2017) Leksicheskie osobennosti tsakonskogo dialekta novogrecheskogo yazyka: predvaritel'nye nablyudeniya i perspektivy issledovaniya [Lexical peculiarities of the Tsakonian dialect of Modern Greek: Preliminary observations and perspectives of study]. Voprosy Jazykoznanija — Topics in the Study ofLanguage, 1: 105-136. (In Russian)

Koukoules, F. (1908) Oinountiaka ê melete peri tes historas tön ethön kai ethimön kai tou glössikou idiömatos tou dêmou Oinountos tes eparchias Lakedaimonos [Studies on the history, traditions and the dialect of the Oenus municipality of Lacedaemonprovince]. Chania: Michaël I. Salivëros Printing house, 309 p. (In Modern Greek) Koukoules, F. (1924) Ekthesis peri tou kata to etos 1919 telesthentos diagönismou tis en Athiinais Glössikis Etaireias [Presentation of the competition conducted by the Linguistic Society of Athens in 1919]. Athina, 36: 254-281. (In Modern Greek) Lauritzen, F. (2009) Michael the Grammarian's irony about Hypsilon. A step towards reconstructing Byzantine

pronunciation. Byzantinoslavica, 67: 161-168. (In English) Lekos, M. A. (1984 [1920]) Peri tsakönön kai tls tsakönikls dialektou [On the Tsakonians and the Tsakonian

dialect]. Athens: Notis Karavias Publ., 96 p. (In Modern Greek) Liosis, N. (2007) Glössikes epafes stl notioanatoliklPeloponnlso [Language contact in the Southeastern Peloponnese].

PhD dissertation (Linguistics). Thessalonica, Aristotle University. (In Modern Greek) Makris, I. (1952) To glössikon idiöma tön chorion Vatka kai Chavoutsi, tis eparchias Kyzikou [The dialect of the villages Vatka and Havoutsi in the Cyzicus District]. Mikrasiatika Chronika, 5: 197-215. (In Modern Greek)

Manolessou, I., Pantelidis, N. (2018) New data on the history of Tsakonian. 8th Conference on Modern Greek

Dialects and Linguistic Theory. [Presentation]. (In English) Maxwell-Stuart, P. G. (1981) Studies in Greek Colour Terminology: In 2 vols. Vol. 2. Charopos. Leiden: Brill Publ., 254 p. (In English)

Mirambel, A. (1960) Consonnes aspirées en Tsakonien. Bulletin de la Société Linguistique de Paris, 55: 40-73. (In French)

Newton, B. (1972) The generative interpretation of dialect: a study of Modern Greek phonology. Cambridge:

Cambridge University Press, xii, 236 p. (Cambridge studies in linguistics. Vol. 8). (In English) Nicholas, N. (1996) The diachrony of Modern Greek complementiserpos: A non-monotonic language change.

University of Melbourne Working Papers in Linguistics, 16: 195-222. (In English) Oikonomou, Th. (1870) Grammatikl tls tsakönikls dialektou [Grammar of Tsakonian dialect]. Athens: Psylliakos

Publ., 97 p. (In Modern Greek) Oikonomou, K. (1888) O Polydöros: poilma voukolikon, kai dekapentepoilmata tls tsakönikls dialektou [Polydorus, a bucolic poem; and fifteen poems in the Tsakonian dialect]. Peiraeus: Sfaira Publ., 56 p. (In Modern Greek) Panayiotou, A. (1993) I lakonikI kata ti romaïkl periodi: Dialektik! epiviosi I archaiopinls anaviosi? [Laconian in the Roman period: dialect survival or antiquarian revival?] Studies in Greek Linguistics, 14: 497-509. (In Modern Greek)

Palmer, L. R. (1980) The Greek language. London: Faber and Faber Publ., xii, 355 p. (In English) Pernot, H. (1934) Introduction à l'étude du dialecte tsakonien. Paris: Les Belles Lettres Publ., 550 p. (Collection de l'Institut Néo-Hellénique de l'Université de Paris. T. 2). (In French)

Starostin, S. (1999) Comparative-historical linguistics and lexicostatistics. In: V. Shevoroshkin, P. Sidwell (eds.). Historical Linguistics & Lexicostatistics. Melbourne: Association for the History of Language, pp. 3-50. (In English)

Swadesh, M. (1955) Towards greater accuracy in lexicostatistic dating. International Journal of American

Linguistics, 21 (2): 121-137. (In English) Tzitzilis, Ch. (In prep.) I tsaköniki dialektos [Tsakonian dialect]. In: Ch. Tzitzilis (ed.). Oi neoettinikes dialektoi

[Modern Greek dialects]. Thessaloniki: Institute for Modern Greek Studies. (In Modern Greek) Vyatkina, A. O. (2015) Relevantnye i nerelevantnyepriznaki (kprobleme izogloss v novogrecheskoj dialektologii) [Relevant and irrelevant indicators for the issue of isoglosses in Modern Greek dialectology]. Bachelor's thesis. [Unpublished Manuscript]. Saint Petersburg, Saint Petersburg State University. (In Russian)

Author:

Nick Nicholas, e-mail: opoudjis@optusnet.com.au

For citation: Nicholas, N. (2019) A critical lexicostatistical examination of Ancient and Modern Greek and Tsakonian. Journal of Applied Linguistics and Lexicography, 1 (1): 18-68. DOI: 10.33910/2687-0215-2019-1-1-18-68 Received 21 March 2019; reviewed 22 May 2019; accepted 26 June 2019.

Copyright: © The Author (2019). Published by Herzen State Pedagogical University of Russia. Open access under CC BY-NC License 4.0.