Reframing the Phylogeny of Asia Minor Greek: The View from Pontic Greek

Citation with persistent identifier:

Ioanna Sitaridou. “Reframing the Phylogeny of Asia Minor Greek: The View from Pontic Greek.” CHS Research Bulletin 4, no.1 (2015).


1§1 Greek historical dialectology remains an understudied area of research –understandably so because of millennia-long diglossia– despite recent theoretical advances in the study of syntactic variation, comparative linguistics and phylogenies.

1§2 The aim of this paper[1] is, precisely, to fill this gap through the study of the evolution of Pontic Greek (PG) within the broader context of Asia Minor Greek (AMG). The need for explanatory grammatical analyses aside, there are critical historical and classificatory issues, namely to what extent PG:

  1. participated in the koineization process which, in its end result, yielded (Standard) Modern Greek ((S)MG)); and
  2. contributed to the dialect formation processes that resulted in the emergence of the major MG dialects.

1§3 Given the lack of sufficiently old textual evidence for PG,[2] which could provide clues as to its evolution, the archaic character of Romeyka, an endangered Greek variety still spoken in Turkey, essentially means that it can be used as a ‘window on the past’ (Sitaridou 2013).

1§4 In North-Eastern Turkey, in the area traditionally known as Pontus, there remain three Greek-speaking enclaves: Of/Çaykara, Sürmene, and Tonya (Parcharidis 1880, Deffner 1878, Mackridge 1987, Sitaridou 2013). Islamization of Greek speakers in the area is reported in the fifteenth-eighteenth centuries (Vryonis 1986, Lowry 2009). However, it is unclear, how many of them were native populations or Turks entering Pontus from the eleventh century onwards, who adopted Greek. During post-Islamization of Greek-speaking populations in Pontus (seventeenth century), PG branches out into: a variety spoken by Muslims (referred to as /romé(á)ika/ by its speakers and which I spell ‘Romeyka’ to distinguish it from other historical/contemporary varieties of AMG); and the one spoken by Christians (also called /roméika/, PG being a recently adopted label). The Exchange of Populations between Greece and Turkey (under the Treaty of Lausanne) in 1923 saw all Orthodox Christians of Asia Minor ‘relocate’ to Greece where incipient contact with ‘high’ Greek prior to the expulsion (through schooling and nineteenth century immigration to Russia) has since intensified leading to some attrition and subsequent alignment with MG; and all Muslims of Greece ‘relocate’ to Turkey. With religion as the defining criterion, Greek-speaking Muslims were allowed to stay in their Asia Minor homeland, but Greek-speaking Christians had to leave, which explains why Greek survives only in small enclaves in this area.

The classification problem of Greek dialects

1§5 Romeyka belongs genetically to the PG group, along with Cappadocian, Marioupol Greek, Pharasiot, at the core of the AMG group (Andriotis 1995:100–107, Drettas 1999:15, Horrocks 2010:398–404, Kontossopoulos 1981, Karatsareas 2011).

1§6 Affinities among AMG varieties led Dawkins (1931:399) to hypothesize that a medieval AMG Koine must have existed, whose idiosyncratic development possibly preceded, and was facilitated by the incipient Seljuk invasions of the eleventh century (Dawkins 1916:205, 213, Browning 1983:130, Horrocks 2010:382).

1§7 Some claim, however, that at least some distinctive AMG developments originate in the regional Koine Greek spoken in Asia Minor and adjacent islands during Hellenistic and Roman times (Thumb 1914:199, Kapsomenos 2003:63, Drettas 1999:15). Nevertheless, according to Horrocks (2010:113–114), there is little relation between the grammatical innovations shared by the modern dialects and the region-specific characteristics of the Hellenistic (HelGr) Koine of Asia Minor recorded by Brixhe (1987).[3]

Towards the recasting of the internal classification of Greek

The reconstruction method

2§1 Until recently reconstruction of syntax was not possible since there was no scientific study of syntax as such, but also because, first, of the difficulty of establishing syntactic comparanda: ‘Due to the nature of syntactic variation, it is impossible to establish correspondences in syntax’ (Lightfoot 2002:119-121); and, second, syntax is not as variable as the lexicon, hence similarities are less probative (Longobardi & Guardiano 2009:1684).

2§2 However, Campbell & Harris (2002), Willis (2011) and Longobardi & Guardiano (2009) i.a. have all defended the view that syntactic patterns can be compared across languages, with innovations being stripped away to reveal aspects of the protolanguage.

2§3 This is essentially the view taken here. I have been approaching the syntactic classification of Pontic Greek by means of, first, comparing the syntax of specific phenomena in Romeyka to the ones in Hellenistic, Medieval and Pontic Greek (Sitaridou 2014a) to see which one Romeyka matches best; second, assessing whether changes/innovations could have sprung out of a Hellenistic or Medieval Greek pool of grammatical cues (in the sense of Lightfoot 2010). Crucial to this modus operandi is the idea that reanalysis takes place during child language acquisition and the distinction between the abstract grammatical system and the surface output of that system. On this view, it follows that reanalysis is constrained both by pre- and post-reanalysis grammars and that it must be acquirable on the basis of the same primary linguistic data. This imposes limits on the possible hypotheses that can be entertained (see also Willis 2011) –– the same would hold even if we simply adopt Sapir’s (1921) drift, which refers to the predisposition to undergo certain changes given certain precursor traits.

How to treat homoplasy?

2§4 It is often said that the entire morphosyntax of AMG is remodeled on Turkish (see, i.a., Thomason and Kaufman 1988:215-222); it is therefore important to discuss how the above methodology can distinguish/deal with cases where it is not clear whether we are dealing with homoplasy, namely when the structure/pattern is not present in the ancestor but innovated by diffusion across contiguous areas, or genuine inheritance. Let us take a look at how this actually works.

SMG is like English in requiring the presence of an overt pronominal object with transitive verbs (see (1)):

(1) Q: Foras        to            palto        su? (SMG)
wear.2SG the.ACC coat.ACC your
‘Are you wearing your coat?’
A: To        forao.
it.ACC wear.1SG
A’: *Forao.
‘I am wearing it.’

However, in certain contexts in SMG, this object may be omitted if there is an indefinite discourse antecedent (Dimitriadis 1994, Giannakidou & Merchant 1997, Tsimpli & Papadopoulou 2006). The same holds for Romeyka, which, however, also allows null objects (see (2)) in contexts where this would not be possible in SMG:

(2) Q: Pios iðe                   ton          Ademi? (Romeyka:OK;*SMG)
who see.Past.3SG the.ACC Ademis.ACC
‘Who saw Ademis?’
A: O            Mehmetis               iðe.
the.NOM Mehmetis.NOM see.Past3SG
‘Mehmet saw him.’

Importantly there can be two diachronic explanations for null objects –see Tables 1 and 2:

(a) Classical Greek has null objects (Luraghi 2003, 2004).
(b) Hellenistic Greek has null objects (Luraghi 2003, 2004; Lavidas 2013).
(c) Medieval Greek does not have null objects.
(d) Romeyka has null objects.
(e) Romeyka descends from Hellenistic Greek (independently proved
(f) Therefore, Romeyka null objects are inherited (from HelGr)

Table 1: Romeyka null objects as an inherited feature

(a) Classical Greek has null objects (Luraghi 2003, 2004).
(b) Hellenistic Greek has null objects (Luraghi 2003, 2004; Lavidas 2013).
(c) Medieval Greek does not have null objects.
(d) Romeyka has null objects.
(e) Romeyka descends from Medieval Greek (independently proved).
(f) Turkish allows null objects (Şener & Takahashi 2010).
(g) Therefore, Romeyka null objects are back-mutation (contact-induced by Turkish).

Table 2: Romeyka null objects as back-mutation

2§5 Although it lies beyond the scope of this paper to discuss the diachronic trajectory of null objects in Romeyka, the reconstruction method discussed above (or any other microparametric method in fact) can shed light on distinguishing parameters with deep phylogenetic value from parameters more prone to contact-induced change/areal effects (see Sitaridou in prep./b, 2012).

The Hellenistic Roots of Pontic Greek

The strong thesis

3§1 Sitaridou (2014a), on the basis of the properties of the Romeyka infinitive, argued that AMG participated partially in the processes that resulted in the major MG dialect formations; in particular, for Proto-Pontic, it was claimed that the terminus ante quem is the Hellenistic times (strong thesis), not the middle of the Late Medieval period, as claimed by Horrocks (2010:382) and Holton & Manolessou (2010) for other MG dialects.[4]

The Romeyka infinitive

3§2 The Classical Greek (CG) infinitive (see (3)) was eventually lost (Joseph 1990) and gradually replaced by hina-clauses during Hellenistic times and subsequently by na-clauses in MG (see (4)).

(3) ἡ              γυνὴ πάλιν φρούδη, πρὶν    εἰπεῖν          ἐσθλὸν (CG)
he            gune palin phroude prin     eipein         esthlon
the.NOM woman.NOM again gone before say.AOR.INF good.ACC
ἢ  κακὸν      λόγον
e  kakon      logon
or bad.ACC word.ACC
‘The woman left, before saying either a good or a bad word.’
(Sophocles’ Antigone, 1245 in Sitaridou 2014a:24)


(4) I                 jineka              efije                    prin     na                pi (MG)
the.NOM woman.NOM leave.Past.3SG before PRT.SUBJ say.PNP.3SG
ute        kalo            ute kako        loγo.
neither good.ACC nor bad.ACC word.ACC
‘The woman left before saying either a good or a bad word.’

Against this background, Romeyka (see (5)) has retained the CG prin cum infinitive construction, which, albeit retained in HelGr (see (6)), only survives as a learned borrowing in Medieval Greek (MedGr) (see (7)).

(5) prin    pisini         fain, prin    spudžisini so mandrin,   tši    pao (Romeyka)
before make.INF food.ACC before clean.INF at.the barn.ACC NEG go.1SG
‘Before I cook and before I’ve cleaned the barn, I’m not going.’
(Sitaridou 2014a:135)


(6) πρὶν   αλέκτορα  φωνῆσαι, τρὶς   απαρνήσῃ με (HelGr)
prin    alektora    phonesai tris    aparnesei me
before cock.ACC crow.AOR.INF thrice deny.PNP.2SG me
‘before the cock crows, you will disown me three times.’
(Matthew 26:34 in Sitaridou 2014a:42)


(7) πρὶν    τοῦ          τὸ μοναχικὸν σχῆμα λαβεῖν, (…) (MedGr)
prin    tu          to monaxikon sxima lavin
before the.GEN the.ACC monastic.ACC schema.ACC receive.AOR.INF
‘before receiving my monastic habit (i.e., before becoming a monk), (…).’
(Pseudo-Sfrantzes, 152.7 in Sitaridou 2014a:24)

Moreover, in the complementation, the Romeyka infinitive follows the MedGr ‘shrinking’ of the infinitive to monoclausal domains (see (8)). Essentially, this means that in Romeyka the infinitive as a complement only survives as a complement to negated past tense nonveridical verbs such as “can”, “want” etc. (see (9)).

(8) Οὐκ ἠμποροῦν την εὕρειν (ΜedGr)
uk imborun   tin evrin
NEG can.3PL  her find.INF
‘They cannot find her.’
(Digenis Escorial, 124 in Sitaridou 2014a:58)


(9) utš eporesa evrin=ä (Romeyka)
NEG can.Past.1SG find.INF=it
‘I couldn’t find (them).’


Hellenistic cues, not Medieval ones

3§3 Crucially, the distribution of the Romeyka infinitive which is more complex than described here (but see Sitaridou 2014a/b for a detailed analysis) shows that innovative change also takes place, namely reanalysis of the infinitive into a negative polarity item (licensed in similar conditions to English “nothing”); that is, it came to operate under a rather constrained use tied to the degrees to which a proposition does not assume its own truth-value. Crucially, such a development could not have taken place just by virtue of the MedGr grammatical cues, which lacked an adequate number of antiveridical infinitival contexts for a frequency association to be conventionalized as a grammatical rule between antiveridicality and the infinitive. Rather, the pool of cues must have been Hellenistic.

‘Leap-frog’ contact with Medieval Greek

3§4 The fact that Romeyka maintains CG/HelGr prin cum infinitive yet, at the same time, exhibits MedGr characteristics, such as a very similar distribution of infinitival complements, as well as counterfactual conditionals containing an infinitive –a MedGr innovation in itself, indicates that contact of ‘local’ Greek in Pontus with other MedGr varieties was discontinuous, that is, the innovations spread in a ‘leap-frog’ manner (in the sense of Chambers & Trudgill 1980). The sociolinguistic evidence seems to be scarce although this is more a reflection of us not looking for it (but, for a recent such endeavor, see Andriollo 2013). Despite the scarcity of the evidence, it is well known that the monasteries established in the Black Sea region in Turkey functioned as important religious and cultural centers, often entrusted with a political role maintaining stability in geographically fragile areas of Byzantium. The monks would often come from areas outside Pontus, as is the case of Barnabas and Sophronius, the founders of the Soumela monastery in CE 386, who both came from Athens and would therefore not be speakers of the ‘local’ Greek variety. This essentially means that native speakers of Greek would be exposed to the non-Pontic varieties through the various dealings with the monastery, as well as villagers who were going to the commercial port of Trebizond for work or travel. Or, consider the influx of non-Pontic Greek speakers following the dissolution of Byzantine rule in Constantinople, due to the fourth crusade in 1204, and the move of some of the members of the Byzantine imperial family to Trebizond.

Negation in Romeyka: Hellenistic or ‘Byzantine Porridge’?[5]

Binary negator system and nonveridicality in the history of Greek

4§1 Greek belongs to the majority of the world’s languages that manifest a binary negator system which is regulated by (non)veridicality (Giannakidou 2006:589, et seq.). This distinction holds, to our knowledge, in all diachronic and dialectal forms of Greek (Horrocks 2010, Chatzopoulou & Giannakidou 2011, Chatzopoulou 2012).

The Romeyka negator system

4§2 Yet the complexity through which the relation between negation and nonveridical functions is grammaticalized is unique in Romeyka (Sitaridou 2014b, Chatzopoulou & Sitaridou 2014, Sitaridou in prep./a), as there is distinction among multiple instances of sentential negation –see Table 3:

Standard negation (in declaratives) Prohibitives Embedded in the scope of nonveridical predicate in imperfect tense only Counterfactual conditional protasis and apodosis with ixa/wishes Nonveridical conditional protasis Embedded in the scope of nonveridical predicate Apodosis of counterfactual conditionals with imperfect only Directives and other nonveridical (future and conditionals) Exhortatives
Romeyka NEG1u(tš)(i) NEG2 mi/mu NEG3 miðen NEG4 xe NEG5 tše(n) NEG6 utšas
Standard Modern Greek NEG1ðen NEG2 mi(n) NEG2 mi(n) [for directives] / NEG1 ðen [elsewhere] NEG2 mi(n)
Classical Greek NEG1οu(k) NEG2 me (rarely NEG1 οu(k))

Table 3: Comparative distribution of negative particles in three varieties of Greek

Although the continuity of the CG NEG1 ouk[6] is striking in not exhibiting Jespersen’s Cycle as in MG (see Table 4), we note two puzzling aspects of the Romeyka negators, which prima facie challenge the strong thesis:

  1. The binary negator distinction present throughout the history of Greek does not seem to hold in Romeyka;
  2. Nonveridicality is not at work given that NEG4, NEG5, NEG6 seem to originate from negators typically associated with veridical contexts;
Proto-Indo-European *ne vs *meh
Homeric Greek ou(k[h]) vs me
Classical Greek ou(k[h]) vs me
Koine Greek ou(k) vs me
Late Medieval Greek u(k) and (u)ðen vs mi and miðen
Modern Greek ðe(n) vs mi(n)
Romeyka u(tš)(i) vs mi/     miðen/xe/tše(n)/utšas
NEG1 NEG2    3       4       5         6

Table 4: Jespersen’s Cycle in Greek

‘Nothing makes sense except in the light of diachrony’ [7]

Table 5 presents a summary of the etymology I propose for the Romeyka negators:

Romeyka negators Etymology
NEG1 utš < CG NEG1 (u(kh)(i) with palatalization)
NEG2 mi < CG NEG2
NEG3 (a)miðen < an (“if”) + MedGr NEG2
NEG4 xe < MedGr (θe “want.inv”<θena)
NEG5 tšen < CG NEG1 + MedGr en< θen (“want.inv”)
NEG6 utšas < CG NEG1 + as

Table 5: The etymology of the Romeyka negators

4§3 Leaving aside NEG2, which has receded but otherwise poses no diachronic puzzle, and NEG3, which is a MedGr innovation exhibiting Jespersen’s Cycle, for the innovative negators (namely NEG4/5/6 as well one unexpected use of NEG1 as the only negator allowed for nonveridical complements in coreference –see (14) below) the claim I put forward is that they result from the synergy of the ‘upstairs’ CG NEG1 ouk and the lexical head of the matrix clause which becomes grammaticalized into a modal/auxiliary head, thus creating a restructuring context (that is, a monoclausal domain).

Monoclausality and Romeyka negators

4§4 Crucial to my explanation is the notion of restructuring, that is, the syntactic operation whereby two domains/clauses (one matrix, one embedded) are being restructured as a single one through the reanalysis of the matrix lexical verb as a functional one (modal or auxiliary). For instance, in the case of the development of the future in the history of Greek, the change consisted of the grammaticalization of the volitional verb θelo “I want” from being a lexical head in the upstairs clause into a functional element (see (10)) due to loss of V features in coreference contexts (see Markopoulos 2009, Roberts & Roussou 2003, Pappas & Joseph 2001, Joseph 1983, 1990):

(10) Lexical>auxiliary>particle

Therefore, lexical θelo splits into two patterns: (i) it goes from being merged in V, having argument structure and taking a TP/CP complement to moving to T (via v where it is now merged), as shown in (11)a-b; and (ii) continues the biclausal structure in (11)c:

(11) a. [TP to         θelis         [VPtv [TP maθi]]]
it.ACC want.2SG               learn.PNP.3SG
‘You want him/her to learn it.’
b. [TP to         θelis         [VP maθi]]
it.ACC want.2SG      learn.PNP.3SG
‘You want him/her to learn it.’
c. [TP θelis         [VPtv [CP na               to          maθis]]]
want.2SG                PRT.SUBJ it.ACC learn.PNP.2SG
‘You want to learn it.’
(Roberts & Roussou 2003:60)

4§5 Evidence that reconstruction yields monoclausality/clause-union comes from clitic climbing in (11)b. Crucially, (11)c shows that lexical θelo “I want” does not trigger clitic climbing. The prediction would then be that, during the reanalysis stage of (11)b, the infinitive cannot be negated as used to be the case in CG (12); moreover, during the same stage we predict that objects cannot be scrambled either or the adjacency conditions for a monoclausal reanalysis would not arise. The prediction seems to be borne out since in the medieval record of Greek we consistently find that negation is always before the volitional verb, not between the volitional and the infinitive (13).

(12) “τὴν Κέρκυραν ἐβούλοντο μὴ προέσθαι” (CG)
“ten Kerkuran eboulonto me proesthai”
the.ACC Corfu.ACC want.IMPF.3PL NEG2 AOR.INF
‘They wished not to give up Corfu.’
(Thucydides’ History of the Peloponnesian War, 1.44)
(13) “Ὁ θεὸς νὰ μὲ σκεπάσῃ, (MedGr)
“O θeos na me skepasi,
the.NOM God.NOM PRT.SUBJ me look.after.PNP.3SG
ὅτι ἂν μοὔδωκεν ὅλην του τὴν ἀφεντίαν
oti an mu=ðoken olin tu tin afentian
COMP if me=give.Past.3SG all.ACC his the.ACC majesty.ACC
[δὲν θέλω ἀρνηθεῖν] τὴν ἁγίαν Τριάδαν.”
[ðen θelo arniθin] tin aɣian Triaðan.”
NEG1 want.1SG refuse.INF the.ACC holy.ACC Trinity.ACC
‘God look after me, since you gave me all your mercy, I will not refuse the Holy Trinity.’
(Leontios Machairas’ Chronicle, §24)

4§6 Importantly, one of the peculiarities of the Romeyka negator distribution is precisely this: the embedded complement of a nonveridical predicate when in coreference cannot be negated in the downstairs clause, only an upstairs negation is acceptable –compare Romeyka (14)a with SMG (14)c:

(14) a. *aɣapo/ θelo xe /NEG na troo. (Romeyka)
love.1SG/ want.1SG NEG4/NEG PRT.SUBJ eat.1SG
‘I don’t like eating.’
b. utš aɣapo/ θelo na troo.
NEG1 love.1SG/ want.1SG PRT.SUBJ eat.1SG
‘I don’t like eating.’
c. θelo na min troo. (SMG)
want.1SG PRT.SUBJ NEG2 eat.1SG
‘I don’t want to eat.’

Although clitic climbing is not available as an additional diagnostic for restructuring in Romeyka due to strict enclisis, (14)a-b shows that even lexical θelo “I want” exhibits clause union in Romeyka.

Emergence of NEG4 XE[8] and NEG5 TŠEN

4§7 I would like to put forward the hypothesis that in AMG, modal-like θelo, as a result of the reanalysis in (10), broke down to at least two particles, namely (i) xa (underlyingly θa) and (ii) en/enna (see also Joseph & Pappas 2003 for a discussion of the sporadic appearance of enna in MedGr, and references therein).

4§8 Evidence for the split of θelo into various particles comes from several sources. First, data from a contemporary Romeyka variety from a different locality within Çaykara (where current data presented here also stem from) shows this xa particle (see Asan 1998). Second, Dawkins’s (1931:79) data from the Christian village of Giga in Pontus (collected in 1914, see Sitaridou 2014a:32) show the form en, which has a futurate interpretation and is compatible with na, is always preceded by negator NEG1 k’.[9] Third, the strongest evidence that θelo split into various particles in AMG comes from Pharasiot, which in fact shows several particles whose modal/futurate meaning is not entirely clear so far (Metin Bağriaçik, p.c.), but which, when negated, are always negated with NEG1 tšo which also has to precede any sequence of particles (see Anastasiadis 1976:263).

4§9 Crucially, for all these modal/futurate particles to have emerged we need to assume that the volitional θelo was morphologically reduced/impersonalized prior to any such reanalysis. Therefore, a recent claim put forward by Markopoulos (2009:198) namely, that θe<θelo was geographically restricted to Crete and Cyprus and that it arose due to ‘contact between Greek and Romance speakers, especially Old Venetian and Old French’ has to be refuted.

4§10 The claim is then that NEG5 emerged out of a contraction of upstairs NEG1 utš and en, which derives from θe(n)(a)<θelo,[10] although en, as such, does not exist anymore in Romeyka (*en na pao).

4§11 For NEG4 I will assume exactly what I have assumed for NEG5 with one additional step, namely Croft’s Cycle (1991), which is a less frequent type of negation renewal when a verbal head is reanalyzed as the negation head proper (see also Givón, 1978:89). This type of negative cycle, as well as the more common Jespersen’s Cycle, has been explained by van Gelderen (2008) as driven by economy considerations; in our context, we would need to claim that θe>xe[11] has been reanalyzed as a negator proper given that Merge is more economical than Move (see also Willis 2013 for a similar analysis of Welsh cau).

Emergence of NEG6 UTŠAS

4§12 I assume that exactly the same development took place between as and θa, namely they were both originally a lexical head (φες>imperative of ἀφίημι “let”) of the upstairs clause which was reanalyzed as a functional element (first instances are recorded in sixth/seventh centuries CE) thus yielding monoclausality with an operator in C0. As previously argued, we would then predict that the only available negation would then be the NEG of the upstairs clause. Indeed this is confirmed by the medieval data –– consider (15):

(15) ἂς   ἀναλάβοντ(αι) τὰ μέρη ἀπτον, ἠ   δ᾽ οὐχεὶ ἂς
as    analavont(e) ta meri apton, i   ð’ uxi  as
PRT undertake.3PL the.NOM parts.NOM it.ACC or  PRT NEG PRT
ἔχη ἀπτὰ ἡ ἐκλισήα
exi apta   i eklisia
have.3SG them the.NOM church.NOM
‘The members should undertake it, otherwise they should not be members of the Church (anymore).’
(1051/1052, S. Italy, Guillou 1972: 7, 62.22)

Second step in the emergence of NEG6 utšas would be the contraction, viewed here as phonological/prosodic cliticization (see Broadbent & Sifaki 2013) rather than in syntactic terms (because head adjunction is predicted to yield *asutš, whereas we get utšas), and triggered by strict adjacency between NEG1 utš and the NPI as (which could also be construed as stage II of Jespersen’s cycle). Interestingly, contraction itself offers further evidence for restructuring (see Roberts 1997).

The ‘palimpsest’ effect: disentangling Hellenistic cues from Medieval ones

4§13 Given the proposal so far we can postulate that the rule which yields the distributions in Table 3 is, in fact, the one in (16); in other words, nonveridicality is still at work but a new cycle has started.

(16) (i) In matrix clauses with modal elements (en, as, θelo) restructuring applies →  NEG1 utš (morphologized as NEG4/5[12]/6)
(ii) Elsewhere →  NEG2 mi (moprhologized as NEG3)

At this point, there are (at least) two remaining issues: first, why Romeyka developed all these negators; and, second, which of these innovations are due to HelGr cues and/or MedGr.

(i) Grammatical systems are not adaptationist (‘everything is made for the best purpose’); if they were we would expect that a SMG type of development would have been the best fit given the CG system. In fact, diachronic change seems to generate some ‘by-products’ –– consider for instance, the ‘spandrels’, that is, the necessary architectural by-products, which were claimed to exist in evolution too, according to Gould & Lewontin (1979). The reason why we get these ‘spandrels’ in Romeyka is probably to do with the substrate in conjunction with the discontinuous type of contact in the region, which prevents Romeyka from fully participating in the repartition of labor between CG NEG1 and CG NEG2, which happens during the Late Medieval Greek stage (see Chatzopoulou 2013).

(ii) To arrive at today’s Romeyka negator system, the following cues were needed:

(17) MedGr structural conditions in the input:
(i) NEG3 (see (18i) below);
(ii) Grammaticalization of the future (incipient since HelGr) and exhortative particles.
(18) HelGr structural conditions in the input:
(i) CG NEG2 in conditionals (which disappeared elsewhere by medieval times);
(ii) Enclisis, which created the necessary adjacency conditions and, therefore, facilitated monoclausality reanalysis;
(iii) Verb plus verb strategy, which led to no complementizers; should they have existed they would have forced clausal boundaries and, therefore, ban monoclausality;
(iv) Verb plus verb strategy with control verbs on the non-obligatory control interpretation as well (see Markopoulos 2009), thus delaying the spread of hina-complements.
(v) The extended survival of the infinitive as a complement and its late partial replacement by a hina-clause, which would have provided the necessary “room” for a downstairs negator, as in SMG in (14c).

Corroborating this new phylogeny further

5§1 Does generative linguistics have the heuristics to computationally measure linguistic differences among languages? For instance, can we measure the syntactic distance between Southern Italian Greek and Romeyka –– the two varieties spoken in the extremities of the Greek-speaking world –– vis-à-vis HelGr?

5§2 Recently, ‘Modularized global parameterization’, a radically new comparative method of non-Bayesian computational phylogeny, has been implemented (Longobardi 2003, Longobardi & Guardiano 2009).[13] Based on the notion of microparameters, it offers an exhaustive systematization of syntactic variation in one specific module of grammar, namely nominal syntax, as it appears to display a limited amount of interaction with other domains of syntax and a lesser degree of susceptibility to information-structural pressures, which are often supposed to be triggers of diachronic instability.

5§3 The nominal syntax of Romeyka was tested out among other Greek and Romance varieties and the results obtained in Guardiano et al. (in press) indicated compatible conclusions with the ones reached in Sitaridou (2014a) despite the different (albeit compatible) methodologies used:

  1. In spite of plausible Romance and Turkic influence in Southern Italy and Pontus, respectively, both Italiot Greek and Romeyka cluster with Greek, with Romeyka being expectedly the outlier of the Greek group.
  2. There is an (even larger) number of parameters (and resulting surface patterns) which differentiate Romeyka more strongly from the rest. At the same time, there is also a significant number of parameter values that keep it significantly more conservative (as it also emerges from its position in the Principal Component Analysis), e.g. adjectival possessives, prenominal genitive, as well as innovations also due to the influence of surface patterns the language was exposed to through its contact(s), e.g. the loss of free-reduced relatives.


5§1 ‘Syntax carries a salient historical signal and cannot be discarded when producing phylogenies’ (Guardiano et al. in press).

5§2 PG emerges out of HelGr. Romeyka can be reconstructed to be the most archaic surviving variety out of all the AMG varieties. However, PG follows some of the developments occurring in other Medieval Greek varieties; crucially, on the basis of the medieval data alone, we would not be able to explain a series of grammatical phenomena occurring in PG and especially Romeyka.

5§3 Despite the rich historical record of contact in the region (between Greeks, on the one hand and the Sassanids, the autochtonous populations, Laz, Armenian and Turkish, on the other) establishing what changes result from contact is not so obvious because of the problem of homoplasy. However, beyond sheer speculation and the problems with homoplasy, phylogenetic analyses based on parameters seem to be able to shed some light on contact; for instance, it has recently been shown that if a language in the Circumpontic region has articles, then they would be bound morphemes (Guardiano et al. in press).


Anastasiadis, V. 1976. Η σύνταξη στο φαρασιώτικο ιδίωμα της Καππαδοκίας. Ph.D. diss., University of Thessaloniki.

Andriollo, L. 2013. Constantinople et les Provinces: le rôle de l’aristocratie aux IXe-XIe siècles. Unpublished Ph.D. diss., University of Paris IV Sorbonne, University of Pisa.

Andriotis, N. P. 1995. Ιστορία της ελληνικής γλώσσας (τέσσερις μελέτες). Thessaloniki..

Asan, Ö. 1998. Ο πολιτισμός του Πόντου. Thessaloniki.

Brixhe, C. 1987. Essai sur le grec anatolien au début de notre ère. Nancy.

Broadbent, J., and E. Sifaki. 2013.To-Contract or Not To-Contract? That is the Question.” English Language and Linguistics 17:513–535.

Browning, R. 1983. Medieval and Modern Greek. Cambridge. DOI: 10.1017/CBO9780511554148.

Bryer, A. 1988. People and Settlements in Anatolia and the Caucasus 800-1900. London.

Bubenik, V. 2001. “The Decline of the Ancient Dialects.” In A History of Ancient Greek: From the Beginnings to Late Antiquity, ed. A. F. Christidis, 382–385. Cambridge.

Campbell, L., and A. C. Harris. 2002. “Syntactic Reconstruction and Demythologizing Myths and the Prehistory of Grammars.” Journal of Linguistics 38.3:599–618.

Chambers, J. K., and P. Trudgill. 1980. Dialectology. Cambridge. 

Chatzopoulou, K., and I. Sitaridou. 2014. “Jespersen’s Cycle for NEG2 and Conditional Inversion in the History of Greek: Evidence from Romeyka Conditionals.” Talk in the 16th Diachronic Generative Syntax Conference. Research Institute for Linguistics, Hungarian Academy of Sciences, Budapest, 3-5 July 2014.

Chatzopoulou, K., 2012. Negation and Nonveridicality in the History of Greek. Unpublished Ph.D. diss., University of Chicago.

———. 2013. “Re(de)fining Jespersen’s Cycle.” University of Pennsylvania Working Papers in Linguistics 19.1:31–40.

Chatzopoulou, K., and A. Giannakidou. 2011. “Negator Selection in Attic Greek is a Polarity Phenomenon.” Poster presented at the 13th Diachronic Generative Syntax Conference (DiGS XIII). University of Pennsylvania. June 2–5.

Croft, W. 1991. “The Evolution of Negation.” Journal of Linguistics 27:1–27.

Dawkins, Richard McGillivray. 1916. Modern Greek in Asia Minor: A Study of the Dialects of Silli, Cappadocia and Pharasa with grammar, texts, translations and glossary. Cambridge.

———. 1931. “Folk Tales from Sourmena and the Valley of Ophis.” Αρχείον Πόντου 3:79–122.

———. 1932. Leontios Makhairas: Recital Concerning the Sweet Land of Cyprus entitled Chronicle. Oxford.

Deffner, M. 1878. “Die infinitive in den pontischen Dialekten und die zusammengesetzten Zeiten im Neugriechischen.” Monatsberichte der Königlich Preussichen Akademie der Wissenschaften zu Berlin, 191–230.

Deville, G. 1866. Études du dialecte tzaconien. Paris.

Dimitriadis, A. 1994. “Clitics and Island-insensitive Object Drop.” Studies in the Linguistic Sciences 24.2.

Dobzhansky, T. 1973. “Nothing in Biology Makes Sense except in the Light of Evolution.” The American Biology Teacher 35.3:125-129.

Drettas, G. 1999. “Το ελληνο-ποντιακό διαλεκτικό σύνολο.” In Διαλεκτικοί θύλακοι της ελληνικής γλώσσας/Dialect Enclaves of the Greek Language, ed. A. F. Christidis, M. Arapopoulou, and G. Giannoulopoulou, 15-24. Athens.

Gelderen, E. van. 2008. “The Negative Cycle.” Linguistic Typology 12.2:195–243.

Giannakidou, A. 2006. “Only,[should that comma be there or is a word missing?] Emotive Factive Verbs, and the Dual Nature of Polarity Dependency.” Language 82:575–603.

Giannakidou, A., and J. Merchant. 1997. “On the Interpretation of Null Indefinite Objects in Greek.” Studies in Greek Linguistics 17: 290–303.

Givón, T. 1978.Negation in Language: Pragmatics, Function, Ontology.” In Syntax and semantics, vol. 9: Pragmatics, ed. P. Cole, 69–112. New York.

Gould, S. J., and R.C. Lewontin. 1979. “The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme.” Proc. Roy. Soc. London B 205:581–598.

Guardiano, C., D. Michelioudakis, A. Ceolin, M. A. Irimia,. G. Longobardi, N. Radkevich, G. Silvestri, I. Sitaridou. Forthcoming. “South by Southeast. A Syntactic Approach to Greek and Romance Microvariation.” L’Italia dialettale.

Guillou, A. 1972. “L’école dans l’Italie byzantine.” La scuola nell’Occidente latino dell’alto medioevo (Settimane di Studi del Centro Italiano di Studi sull’Alto Medioevo 19). Spoleto: 291.

Holton, D., and I. Manolessou. 2010. “Medieval and Early Modern Greek.” A Companion to the Ancient Greek Language, ed. E. J. Bakker, 539–563. Malden. DOI: 10.1002/9781444317398.ch36.

Horrocks, G. C. 2010. Greek: a History of the Language and its Speakers. 2nd ed. Malden.

Jones, H. S., and J. E. Powell. 1967-1970. Thucydidis historiae, 2 vols. Oxford.

Joseph, B., and P. A. Pappas. 2003. “On Some Recent Riews Concerning the Development of the Greek Future System.” Byzantine and Modern Greek Studies 26:247–273.

Joseph, B. 1983. The Synchrony and Diachrony of the Balkan Infinitive: A Study in Areal, General, and Historical Linguistics. Cambridge.

———. 1990. Morphology and Universals in Syntactic Change: Evidence from Medieval and Modern Greek. New York.

Kapsomenos, S. G. 2003. Ἀπὸ τὴν Ἱστορία τῆς Ἑλληνικῆς γλώσσας: Ἡ Ἑλληνικὴ γλώσσα απὸ τὰ Ἑλληνιστικὰ ὣς τὰ Νεότερα χρόνια. Ἡ Ἑλληνικὴ γλώσσα στὴν Αἴγυπτο. Thessaloniki.

Karatsareas, P. 2011. A Study of Cappadocian Greek Nominal Morphology from a Diachronic and Dialectological Perspective. Ph.D. diss., University of Cambridge.

Kontossopoulos, N. 1981. Διάλεκτοι καὶ ἰδιώματα τῆς Νέας Ἑλληνικῆς. Athens.

Lavidas, N. 2013.Null and Cognate Objects and Changes in (In)transitivity. Evidence from the History of English.” Acta Linguistica Hungarica 60.1:69–106.

Lightfoot, D. W. 2010. “Language acquisition and language change.” WIREs Cogn Sci 1:677–684. DOI: 10.1002/wcs.39.

Lightfoot, D. W. 2002. “Myths and the Prehistory of Grammars.” Journal of Linguistics 38.1: 113–136.

Longobardi, G., and C. Guardiano. 2009. “Evidence for syntax as a signal of historical relatedness.” Lingua 119.11:1679–1706.

Longobardi, G. 2003. “Methods in parametric linguistics and cognitive history.” Linguistic Variation Yearbook 3:101–138.

Lowry, W. H. 2009. The Islamization & Turcification of the City of Trabzon (Trebizond), 1461-1583. Istanbul.

Luraghi, S. 2004. “Null Objects in Latin and Greek and the Relevance of Linguistic Typology for Language Reconstruction.” Proceedings of the Fifteenth Annual UCLA Indo-European Conference, Los Angeles, November 7-8, 2003, ed. K. Jones-Bley, M. E. Huld, A. D. Volpe, and M. Robbins Dexter, 234–256.

Luraghi, S. 2003. On the Meaning of Prepositions and Cases: The Expression of Semantic Roles in Ancient Greek. Amsterdam.

Mackridge, P. 1987. “Greek-speaking Muslims of North-East Turkey: Prolegomena to Study of the Ophitic Sub-dialect of Pontic.” Byzantine and Modern Greek Studies 11:115–137. DOI: 10.1179/030701387790203037

Manolessou, I., and N. Pantelidis. 2012. “Velar Fronting in Modern Greek Dialects.” Proceedings of the 5th International Conference on Modern Greek Dialects and Linguistic Theory, ed. M. Janse, B.D. Joseph, A. Ralli, and M. Bagriacik.

Markopoulos, T. 2009. The Future in Greek: from Ancient to Medieval. Oxford.

Pappas, P. A., and B. Joseph. 2001. “The Microcosm of a Morphological Change: Variation in Thelo+infinitive futures and Ethela+infinitive Counterfactuals in Early Modern Greek.” Diachronica 18:59–92.

Parcharidis, I. 1880. “Γραμματικὴ τῆς διαλέκτου Τραπεζοῦντος: Περὶ ἐπιθέτων, ἀντωνυμιῶν και βαρυτόνων ρημάτων.” Χειρόγραφο 335. Κέντρο Ἐρεύνης τῶν Νεοελληνικῶν Διαλέκτων και Ἰδιωμάτων, Ἀκαδημία Αθηνών.

Pavlou N., and J. Merchant. 2015. A Surprising Morphological Span in Cypriot Greek. Ms., University of Chicago.

Roberts, I., and A. Roussou. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge.

Roberts, I. 1997. Diachronic Syntax. Oxford.

Sapir, E. 1921. Language: An Introduction to the Study of Speech. New York.

Schwyzer, E., and A. Debrunner. 1950. Griechische Grammatik, II: Syntax und sytaktische Stilistik. Munich.

Şener, S., and D. Takahashi. 2010. “Ellipsis of Arguments in Japanese and Turkish.” Proceedings of the 6th Workshop on Altaic Formal Linguistics. MIT Working Papers in Linguistics.

Sitaridou, I. Forthcoming. The Evolution of Negation in Romeyka. Ms. Harvard Center for Hellenic Studies and University of Cambridge.

———. Forthcoming. Null Objects in Romeyka: Contact with Turkish or Internal Conditioning? Ms. Harvard Center for Hellenic Studies and University of Cambridge.

———. 2013. “Greek-speaking Enclaves in Pontus Today: The Documentation and Revitalization of Romeyka.” In Keeping Languages Alive. Language Endangerment: Documentation, Pedagogy and Revitalization, ed. M. Jones and S. Ogilvie, 98–112. Cambridge.

———. 2012. “The Null Objects in Romeyka and the Turkish Dialects of Trebizond: A Case of Reciprocal Contact-induced Change or Internal Conditioning?” Second Workshop on ‘Romeyka and Asia Minor Greek,’ Queens’ College.

———. 2014a. “The Romeyka Infinitive: Continuity, Contact and Change in the Hellenic varieties of Pontus.” Diachronica 31.1:23–73.

———. 2014b. “Modality, Antiveridicality, and Complementation: The Romeyka Infinitive as a Negative Polarity Item.” Lingua 148:118–146.


Thomason, S. G., and T. Kaufman. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley.

Thumb, A. 1914. “On the Value of Modern Greek for the Study of Ancient Greek.” The Classical Quarterly 8(3):181–205. DOI: 10.1017/S0009838800008223.

Tsimpli, I. M., and D. Papadopoulou. 2006. “Aspect and Argument Realization: A Study on Antecedentless Null Objects in Greek.” Lingua 116:1595–1615.

Vryonis, S. 1986. The Decline of Medieval Hellenism in Asia Minor: And the Process of Islamization from the Eleventh through the Fifteenth Century. Berkeley.

Willis, D. 2011. “Reconstructing Last Week’s Weather: Syntactic Reconstruction and Brythonic Free Relatives.” Journal of Linguistics 47:407–446.

———. 2013. Modality, Negation and Negative Concord: Ongoing Change in Northern Welsh Dialects. :


[1] The Acts of Vazelon (14th century AD), the oldest/earliest extant text, displays few PG features and maintains a ‘high’ register.

[2] For the Romeyka project, see

[3] Τhe only exception is considered to be Tsakonian, which is taken to be the heir to the Lakonian dialect formerly spoken in the same localities (Deville 1866, i.a.) –the ‘strict’ Doric dialect which ‘proved to be the most stubborn in ceding their place to Hellenistic Koine’ (Bubenik 2001:482).

[4] The weak thesis suggests that the terminus post quem was the eleventh century (see also Dawkins 1931).

[5] Bryer (1988:96).

[6] Negation in Tsakonian also has an inherited negator from CG ouk; we leave the comparison with Romeyka to future work given that very little is known about the syntax of the Tsakonian negation other than that the negation system is a binary one as in SMG.

[7] ‘Nothing in biology makes sense except in the light of evolution’ (Theodosius Dobzhansky 1973).

[8] Prima facie, it looks tempting to etymologise xe from Ancient Greek oukí ouxí > xi (see PG ki). Schwyzer & Debrunner (1950) point out that Pontic  presupposes oukí (besides ouk) (II 592), and that both oukí and ouxí are strengthened forms, and that the shortened forms ouk and oux became distributed according to a following Spiritus asper and Spritus lenis. Crucially, this would be the wrong analysis for our data because Romeyka of Of (ROf) exhibits what is traditionally called tsitakism (Manolessou & Pantelidis 2012), namely the affrication of [c]>[ts]. Essentially this rules out the ouki explanation. Similarly, in Romeyka we observe fronting of [ç]>[š], which would rule out the form of xe originating in ouxi/e because we would then need to explain how it escaped palatalization when other forms with [ç] have become [š], for instance šimos>χειμὼν “winter”. At this point the question is if the ROf [ç] may be the result of contact with Turkish. Speech material was analyzed with the Praat software package and ANOVA tests were carried out to test this and the results indicate that ROf employs only one voiceless velar consonant [χ], whose phonetic properties are closer to the Greek voiceless velar fricative consonant [χ], rather than the Turkish voiceless velar glottal [h].

[9] This variety, as well as Sürmene varieties, lacks tsitakism.

[10] An alternative analysis would be to derive en from 3SG “to be”, as it has been very recently suggested by Pavlou & Merchant (2015) for Cypriot Greek. Even if this analysis turns out to be true, the overall reasoning regarding monoclausality being at work would not be affected.

[11] /θ/ > /χ/ is attested in Cypriot Greek and marginally in our Romeyka data too.

[12] Here we can argue that restructuring has been lexicalized.

[13] For the Langelin project, see: