Doctoral Thesis Progress Report
		Home > Thesis
Language: ENG ELL EPO JBO TLH LAT

Nick Nicholas; Department of Linguistics & Applied Linguistics, University of Melbourne, Parkville 3052, Victoria, AUSTRALIA.

Thesis topic: The Story of pu: the grammaticalisation and sociolinguistics of a Modern Greek complementiser.

1. The functions of (o)pu.
2. The diachrony of (o)pu.
3. The Balkan context for (o)pu.
4. The diatopy (dialectical distribution) of (o)pu.
5. (O)pu and diglossia.
6. (O)pu and the koineisation of Modern Greek.
7. Pu in Contemporary Modern Greek
8. Progress
9. Results so far
10. References

• hína-> *iná->na: Relative adverb 'whither' (pre-Classical)-> 'in order to' (Classical)-> subjunctive marker (well under way in New Testament Greek) (Burguière 1960; Joseph 1983);
• hópou(-> ópu)-> opú-> pu: Relative adverb 'where' (pre-Classical)' relativiser (by 500 AD) (Bakker 1974)-> particle introducing (usually factive) complement and adjunct clauses;
• oudèn-> dhen: 'nothing'-> realis negator 'not' (by ix AD) (Jannaris 1897; Zampelios 1986 [1852]);
• thélo hína-> thelo na-> thena-> tha: 'I want to'-> future tense particle (ix-xv AD) (Jannaris 1897; Kriaras 1969-1993).

Of these grammaticalised particles, na and pu in particular have spread in function to such an extent in Modern Greek that, almost every time a subordinate clause needs to be attached to a matrix, either na or pu will turn up. Indeed, frequently they will both turn up, either in competition or in complementary distribution.

Of the aforementioned grammaticalisations, na and tha are well documented diachronically, and are widely cited as examples of grammaticalisation.[2] The grammaticalisation of pu from relative adverb to relativiser is also fairly well known (Bakker 1974). However, there does not seem to be any systematic treatment of the diachrony of pu, covering the period after it became a relativiser (ca. 500 AD according to Bakker, although ambiguous instances date back to 95 AD.) Christidis' work in the '80s on the synchronic distribution of pu (Christidis 1981; Christidis 1982; Christidis 1983; Christidis 1986) is a welcome start. However, although he appeals to grammaticalisation theory, Christidis considers neither the diachronic career nor the diatopy of the particle. Instead, he exploits the property of grammaticalisation known as persistence (Hopper and Traugott 1993 :3), whereby the etymology of a grammaticalised word constrains its subsequent grammatical functions. This means that Christidis examined only the endpoints of the grammaticalisation of (o)pu, but not the interval in between.

In this study, I intend to document the fortunes of the particle pu since 500 AD--concentrating on its non-relativiser functions, which I consider subsequent to its grammaticalisation as a relativiser. The pecularities of Greek linguistic history mean that any account of the history of pu will not be limited to a straightforward account of language-internal change. Three language-external factors have been at work on pu: language contact in the Balkans; the koineisation of Modern Greek dialects; and Modern Greek diglossia. Therefore, my thesis will have the following outline:

1. The functions of (o)pu.

1. A relative locative adverb, indefinite or definite. This is the original meaning of (o)pu, and has remained phonologically unreduced as ópu.
2. A relativiser; pu is indeclinable, and comparable to English relativiser that or Hebrew relativiser asher. In some variants , it is a generalising relative, meaning 'the kind of... that'; in Contemporary Standard Modern Greek (CSMG), the irrealis na also needs to be present. In early Modern Greek (but not in CSMG) it could also serve as a headless relative.
3. Pu can also act as a 'pseudo-relativiser', modifying adverbs of place or time.
4. As a complementiser, pu is factive, and usually in complementary distribution with na. Note that pu is marked for factivity; the unmarked realis complementiser is oti or pos. The class of predicates pu can appear with coincides with the class of predicates identified in the linguistic literature on presupposition. Such predicates are also signalled by the Serbo-Croat complementiser shto, subjunctive complements in Spanish, and no-complements in Japanese (Christidis 1981). Pu can appear in the following contexts:
   1. After predicates of emotion (True factives (Karttunen 1971)). In this context, only pu and na are acceptable in CSMG; the unmarked realis complementisers pos and oti are not, although pos does occur in such contexts in literature written earlier this century.
   2. After predicates of perception and knowledge (Semi-factives (Karttunen 1971); Assertive factives (Hopper 1975)). Pu is in contrastive distribution with oti/pos; in free variation with na in affirmative statements; and in constrastive distribution with na in negative statements.
   3. After (non-factive) predicates of saying. This usage appears to be unacceptable in CSMG, but is frequent in Greek dialects.
5. As a connective introducing sentential adjuncts; its usage appears to be restricted to colloquial registers. The adjuncts can be causal,[3] resultative, contrastive, or temporal.
6. Opú can be used as a textual connective with the meanings 'and', 'so', 'but', or 'finally'; this usage seems limited to the Ionian islands.
7. Pu can be preceded by a closed class of prepositions or subordinators. It can also appear in other collocations, mainly with adverbs.
8. Finally, pu can appear in combination with na to introduce irrealis adjuncts and complements; these include potential result clauses, concessive clauses, optative clauses, and non-referential relative clauses.

2. The diachrony of (o)pu.

The major focus of my diachronic research is to ascertain how far (o)pu had grammaticalised by xii AD, and how far it grammaticalised afterwards. For those areas where (o)pu was not yet used in xii AD, we should expect to have textual evidence of the subsequent reanalysis and spread of (o)pu; I will document and analyse this evidence, from the perspective of current grammaticalisation theory. For those areas where (o)pu was already in use in xii AD, an internal reconstruction of the reanalysis processes involved is still possible, and I will attempt this.

My methodology will involve selecting vernacular texts at regular intervals from the period under investigation. My purely diachronic account finishes at 1850 (the Greek State was established in 1833), from which point no diachronic account of vernacular Greek can proceed in isolation from an account of its sociolinguistics. (See below.)

Where possible, the texts selected should be prose narrative.[4] At the moment, I am using 13 texts or groups of text, beginning with the semi-vernacular poems by Michael Glykas (1158/9) and Ptokhoprodromos (1170s), and ending with Makriyannis' Memoirs (1829-ca. 1850). I hope to supplement this data with material gathered in the preparation of Emmanuel Kriaras' Dictionary of Mediaeval Greek Vernacular Writing (Kriaras 1969-1993), archived in Salonica.

In the early part of this period, most texts are somehow 'corrected', displaying interference from the learned idiom. Indeed, the earliest texts are regarded as outright macaronic (Browning 1978). Therefore, any statistics on the relative ratio of occurences of opu against other particles would be meaningless. (As Browning (1983) warns, this is a trap that no less a linguist than Psichari fell into.) So my investigations in the earlier texts will be qualitative, rather than quantitative: what are the trends in the usage of opu, and what occurences of opu point to reanalysis.

Later texts (starting xv AD) yield a somewhat clearer picture of the vernacular. So beginning with Makhairas' Chronicle (after 1432) and diplomatic correspondence of late xv AD from Miklosich & Müller's (1961 [1871]) collection, I will also be quantitatively investigating the spread in functionality (if any) of opu.

3. The Balkan context for (o)pu.

In Shtokavian Serbo-Croat,[5] particularly in Serbia, the parallels are even more striking (van der Auwera and Kuchanda 1985). Etymologically, shto means 'what?', not 'where'; however, its functions include that of relativiser (headless and not); factive complementiser; adjunctiviser in combination with prepositions, as in poshto 'after' and izuzev shto 'except that'; and (rarely) temporal adjunctiviser ('when', 'as soon as'). In effect, its functions almost coincide with those of pu; and it is no surprise that shto as a factive complementiser is more frequent in Serbian than in Croatian (Mønnesland 1972). (Serbian participates in the Sprachbund to a greater extent than Croatian (Comrie 1981 :198).)

So clearly, it would be unwise to investigate the history of pu in isolation from the history of deto and shto.[6] A full account of the history of pu must incorporate a thorough diachronic and dialectological survey not only of pu, but also of deto and shto, in order to trace the areal 'currents' of connectives, and how and when they diffused in the Balkans. Such a full account is probably not feasible in the time-frame of my research, and would require much original research: for instance, as van der Auwera & Kuchanda (1985:953) found, "unfortunately, little or no detailed work has been done on the history of shto-C [complementiser-shto]." So while I hope to make a start in investigating the areal component of the story of (o)pu, drawing on whatever research in the area is already available, I do not intend to perform primary Balkanological research for this thesis.

4. The diatopy (dialectical distribution) of (o)pu.

To investigate the usage of (o)pu in Greek dialects, I intend to consult the extensive archives of the Historical Dictionary Section of the Academy of Athens. The main type of text collected by Greek dialectologists are fables; the narrativity of such texts make them comparable to the diachronic texts investigated. Most Greek dialects are now moribund, so fieldwork would be difficult--although it may prove necessary if there are significant gaps in the data at the Academy's archives.

Australia may well prove a better terrain for Greek dialectological research than Greece itself; there is less institutional pressure here for Greeks to conform to Standard Greek norms. The influence of the old dialects can also still be discerned in regional variants of CSMG; I may attempt to use these current variants to support my dialectological research.

At the end of my research, I hope to have a set of isoglosses for the various functions of (o)pu, which may cast some light on the geographical[7] (and possibly cross-linguistic) spread of these functions.

5. (O)pu and diglossia.

The issue of diglossia has become intensely politicised in Greece, and most major modern Greek linguists have become embroiled in it in some way. Diglossia is typically portrayed as an 'Us-And-Them' battle between reactionaries and progressives, which the progressives have ultimately won. In fact, Greek diglossia has been a rather more complex phenomenon, and certain facts need to be borne in mind:

• While the learned idiom has consciously distanced itself from the vernacular in Greek for the past two millenia, modern diglossia is a much different phenomenon to Hellenistic and Byzantine Atticism. As Browning (1978) points out, the Atticists never set out to make the uneducated majority of Greeks speak ancient Greek, and Atticism did not impinge on vernacular Greek; puristic Greek (katharevousa) certainly did.
• Until fairly recently, the linguistic situation in Greece was more akin to triglossia than diglossia: puristic Greek at one extreme; the various local dialects at the other; and standard Modern Greek, a koine of Modern Greek dialects, in between. This makes the linguistic behaviour of Greeks earlier this century rather complex to describe, particularly where an individual's treatment of (o)pu in their dialect might be at odds with the situation in Standard Modern Greek.
• To complicate matters even further: Modern Greek has been in a diglossic continuum, rather than being subdivided into two or three discrete variants; and this continuum persists today, even after the institutional abolition of puristic Greek. Thus, the intermediate variants of Greek identified by Mirambel in the '30s, mikti ('mixed'--the language of the press) and kathomilumeni ('spoken'--the language of the bourgoisie), are merely points on the diglossic continuum.
• Puristic Greek itself was hardly a monolithic whole; indeed, its inconsistent varieties were one of the features demoticists like Psichari attacked most consistently. For example, while the strictest katharevousianos might use the Classical relativiser hòs, others might prefer the Middle Greek hóstis, while more moderate users of purists (like Coraïs himself) might use the old Modern Greek calque from French/Italian o opios (which thereby reentered the vernacular), or (o)pu itself.
• The last point is obvious, but it needs to be asserted: puristic Greek has had a significant effect on CSMG, particularly on its more formal registers. For this reason, I follow other Greek linguists in avoiding the term 'demotic' to refer to standard Modern Greek.[8] One of the main contentions I make in my thesis is that, under the influence of the puristic (most variants of which proscribed (o)pu), formal registers of CSMG tend to avoid pu, particularly in its 'non-core' functions (bare adjunctiviser). I believe this avoidance has in fact spread throughout CSMG.

6. (O)pu and the koineisation of Modern Greek.

Several issues in the recent development of pu have a direct bearing on the history of the koineisation of Greek. For example, the phonological reduction opú-> pu seems to have taken place suspiciously quickly. While pu is attested as early as xiii AD (in a poem by Mevlânâ Jelâl ed-din Rumi), it is extremely rare throughout the millenium; in Makriyannis' Memoirs, written between 1829 and 1851, the proportion of opú to pu is still 1610:15--pu making up just 1% of all occurences of the lexeme (Kyriazidis, Kazazis and Brehier 1992). In modern literary Demotic however, from Psichari in the mid 1880s onwards, opú is unknown. Such a rapid linguistic change points towards an external cause--dialect contact--rather than to any language-internal process.

Another issue tied in with koineisation is the apparent confusion about choice of complementisers late last century and early this century. As seen, in several modern Greek dialects (including Istanbul Greek), (o)pu has spread from being a complementiser marked for factivity to becoming the generic complementiser. Furthermore, there are certain contexts (namely, after predicates of emotion) where the unmarked realis complementisers, oti and pos, are unacceptable in CSMG. However, some early Demotic literature uses pos precisely where CSMG would disallow it; and reportedly this usage persists in some peoples' idiolects. This could either be the peculiarity of a particular dialect, or originate as a hypercorrection by speakers whose own dialect uses (o)pu more extensively than CSMG. In my research, I aim to address this question.

Research on Modern Greek koineisation is inextricably bound with research on the interplay between puristic and vernacular Greek; any linguistic data from the 1830s onward will show evidence of both. Such research will also need to consider how the latest Greek koine is related to the two earlier Modern Greek koines that have been posited in the literature. (The language of both late Byzantine vernacular literature and modern folk song display a relative lack of typical dialectal features (Browning 1978; Triandafyllidis 1981 [1938]).) Therefore, I will need to draw on research performed for other sections of the thesis: analysis of Byzantine texts; folk song texts; samples of Greek dialect (which, however, tend to be rather thin on the ground for dialects related to CSMG, as dialectologists did not find anything interesting about them); and vernacular texts from the past two centuries. This line of research should prove a substantial original contribution to the history of a phenomenon little investigated by Greek linguists.

7. Pu in Contemporary Modern Greek

The value of this research for Greek linguistics is clear. For general linguistics, the history of pu provides not only a case study in grammaticalisation (diachronically, diatopically, and within the synchronic system of CSMG), but, more importantly, a study in how grammaticalisation can come into conflict with the extralinguistic forces of koineisation and diglossia; and how grammaticalisation spreads within a Sprachbund.

8. Progress

• Familiarised myself with contemporary thinking on grammaticalisation theory;
• Familiarised myself with the literature on presupposition and factivity in linguistics;
• Familiarised myself with the linguistic history of Middle and Modern Greek;
• Decided on a taxonomy for the various functions of pu in CSMG, intended as a more consistent scheme than either the taxonomy given in Tzartzanos (1991 [1946, 1963]) or by Mackridge (1985), and tested this taxonomy by analysing Modern Greek texts;
• Analysed all the occurences of pu, pos and oti in two prose works of particular linguistic importance with respect to Modern Greek: General Makriyannis' Memoirs (1829-1850; 143,000 words), and Costas Takhtsis' The Third Wedding (1963; 118,000 words);
• Done some preliminary work on the distribution of pu in Modern Greek dialects, using those resources available to me locally (this includes Italiot Greek, Cappadocian Greek, Tsakonian, Istanbul Greek, Roumeliot, Cretan, Cypriot, and Heptanesian);
• Investigated the available resources on Greek during the Dark Ages (650-1100), including the Byzantine chronographers and the works of Constantine Porphyrogenitus (as summarised in Bakker (1974)), the Proto-Bulgarian inscriptions, and the documents in Trinchera (1865);
• Done a preliminary investigation (pending my accessing critical editions of the works involved) of the distribution of pu in Early Modern Greek (1100-1400);
• Had a (very superficial) look at similar grammaticalisations in the Balkans (Bulgarian, Serbian).

9. Results so far

• There is a real cleft amongst Modern Greek dialects between those dialects where pu has spread in function to become an unmarked, non-factive complementiser, and those which (like CSMG) treat pu as marked for factivity. The former group certainly includes Istanbul Greek and Thracian, and possibly also includes Roumeliot, judging from Makriyannis' Memoirs. The variability amongst Greek dialects seems to indicate that this is a rather recent development--probably dating after the fifteenth century.
• The adjunctiviser, collocation and complementiser functions of pu are widespread in Greek dialects. There are fully present, for instance, in Italiot Greek, which has been cut off from the rest of the Greek-speaking world since the eleventh century. The collocation and adjunctiviser functions of pu are also present in Cappadocian Greek dialects, cut off from the rest of Greek at about the same time (although their contact with Turkish has been so extensive that the relativiser and complementiser functions of pu have been replaced by Turkish-like nominalisation strategies.) Furthermore, while I have not found any instances of pu-complements in twelfth-century vernacular literature, pseudo-relative and resultative pu are present. This leads me to conclude that the bulk of the grammaticalisation I am investigating had already taken place by the eleventh century, before records of the vernacular become available.
• On the other hand, a study of Makriyannis' Memoirs, in particular, shows that syntactic and semantic ambiguity between the various functions of pu is still rife in Modern Greek. Therefore, it should not prove difficult to trace the pathways for reanalysis of the particle--all the more so if I draw on the history of comparable grammaticalisations in other languages, and Balkan languages in particular.
• The most striking difference between the Makriyannis text and the Takhtsis text is the drastic reduction (almost elevenfold) in instances of pu introducing causal, temporal and contrastive adjuncts. What this seems to indicate is that pu was formerly used to loosely link any factive clause to a matrix clause, and that this tendency has become constrained as a result of the twin pressures of influence from Puristic Greek and greater literacy. I would expect, incidentally, that written CSMG would have a lesser proportion of adjunct-pu than spoken CSMG; and I intend to test this hypothesis.
• It seems probable that Old Serbian, not subject to the same prescriptive constraints as Greek, has better clues as to the development of this grammaticalisation, since the distribution of shto in Modern Serbian has striking parallels to the distribution of pu in CSMG. However, since no specific study on the development of shto in Old Serbian has been made, I will probably not pursue this angle fully.

References