Printer Friendly

A change of mood: the development of the Greek mood system *.

Abstract

This paper traces the steps in the changes of the Greek mood system, from the morphological one of Classical Greek, where mood was fused with tense and agreement within INFL, to the more syntactic one of Modern Greek, where mood occupies its own projection of MOODP. We show how the catastrophic loss of the morphological marking of the mood distinctions in the verb ending during the transition from Classical Greek to Hellenistic Koine was followed by (i) the emergence of a separate projection of a functional category MOOD inside the Comp-layer hosting the subjunctive/ indicative mood features followed by (ii) the grammaticalization of the conjunction hina to the subjunctive mood particle na and its transference from the C head to the separate MOOD head located between the C and the INFL heads and (iii) the subsequent relocation of the imperative from the INFL head to the MOOD head. We show that our analysis is consistent with the theories claiming that syntactic change is associated with formal features and the fixing of parameters.

1. Introduction

The mood (1) system of Greek provides an interesting area for theoretical research for two reasons: The richness of mood oppositions offers valuable data for synchronic theoretical claims (2) and the well-documented significant changes from Classical (ClGk) to Modern Greek (MGk) offer excellent testing ground for theories of language change. More specifically, the morphological mood system of ClGk (mood oppositions marked on the verbal ending) has been reanalyzed to the more syntactic one of MGk (mood oppositions are mainly marked by means of modal particles). In this study, we trace the steps of this development and consider the forces underlying these changes. The analysis is relevant to those theories of language change which take the view that morphosyntactic change is closely associated with the properties of the functional categories, their formal features and the parameters of the language under examination (Roberts 1992; Roberts and Roussou 1999; Clark and Roberts 1993; Lightfoot 1991, 1998; Hale 1998). (3)

2. Theoretical assumptions

Generative approaches attribute syntactic change to the resetting of parameters. According to the theoretical assumptions of the minimalist program (Chomsky 1995), parameters are associated with the feature content of functional categories that express grammatical information relevant to lexical heads, an idea that goes back to Borer (1984). These parameters involve the distinctions weak vs. strong and interpretable vs. noninterpretable, which determine the way a lexical category is associated with the relevant functional category and which are relevant to the morphological manifestation of these features. (4) Thus, a change in the morphological manifestation of formal features is expected to affect the way these features are licensed in the syntactic component. This kind of parameter resetting accounts for changes in the surface word-order pattern of a language, which is determined by the operation of head movement (Hale 1998). The process responsible for this parameter resetting is REANALYSIS, which assigns a different structure to a syntactic construction without affecting its surface realization.

Reanalysis has also been considered to be responsible for the process of GRAMMATICALIZATION (Lightfoot 1979; Harris and Campbell 1995). According to this point of view grammaticalization is an epiphenomenon, that is, the actualization of a syntactic reanalysis (see also Joseph 2001). Recently, Roberts and Roussou (1999) have developed a formal account of the grammaticalization of lexical to functional material that relies on structural simplification. They note that the formal features of a functional category F can be satisfied by one of the following operations: (i) MERGE (particle), (ii) MOVE (inflection) and (iii) MERGE + MOVE (bound morpheme). (5) The last two options are more costly than the first one, so that when the linguistic data offers the opportunity, the speakers of the new generation would try to eliminate the movement and interpret it as merging. Thus, the lexical element that was associated with a functional category by means of overt movement becomes the element that satisfies the formal features of the functional category by means of direct merging, so that it looses its lexical status and becomes a grammatical element.

3. The development of the Greek mood system

The development of the Greek mood system is mainly characterized by the way the mood features are realized in the morphosyntactic structure. Thus, whereas in ClGk the mood features were realized in the verbal morphology, in MGk they are positioned in a separate functional category MOOD situated to the left of INFL (Rivero 1994; Rivero and Terzi 1995; Philippaki-Warburton 1998; Roussou 2000). The major development involved the loss of morphological mood and as a consequence the grammaticalization of the conjunction hina to the subjunctive particle na, which eventually became the element that realized this MOOD functional category. Another development involved a change in the licensing of the imperative; in ClGk imperative was licensed within INFL, whereas in MGk it is licensed in the MOOD functional category by means of overt V-movement.

In what follows we will offer a detailed description of this development and we will argue that it was triggered by the catastrophic loss of the morphological realization of mood distinctions and the subsequent introduction of the MOOD functional category in the clause structure.

3.1. The mood system of Classical Greek

The morphological system of ClGk differentiates four main moods: the indicative, the subjunctive, the optative, and the imperative, as shown in Table 1:
Table 1. The mood paradigm of ClGk (exemplified by the 2SG of the
verb yo[??] 'to set free')

 Mood
Tense Indicative Subjunctive Optative Imperative

Present lyeis lye[??]is lyois lye
Imperfect elyes
Future lyseis lysois
Aorist elysas lyse[??]is lysais lyson
Perfect lelykas lelyke[??]is lelykois lelyko[??]s
 [ist.sup.h]i
Pluperfect elelykeis


We propose that in the CIGk system mood, tense, and subject agreement combine into one functional category INFL, (6) because the morphological exponents of mood cannot be identified independently from those of tense and subject agreement. This is in fact predicted by the mirror principle (Baker 1985), according to which syntactic structure reveals the morphological structure and vice versa (see also Thrainsson 1996 and Bobalijk; Thrainsson 1998). Thus, we follow the essence of Giorgi and Pianesi's (1997, chapter 1) proposal, according to which fused morphology reveals fused/syncretic functional categories and simultaneous checking.

(1) [ILLUSTRATION OMITTED]

The projection of NEG is to the left of INFL so that the licensing of tense, mood, and subject-agreement features is completed independently of the presence or absence of negation. As a result, in ClGk, all mood variations can be found with or without negation: (7)
(2) 2SG, PRES of lyo[??] 'I set free/loosen'
 ou lyeis (indicative)
 me[??] lye[??]is (subjunctive)
 me[??] lyois (optative)
 me[??] lye (imperative)


It is important to point out that, although the choice of negative morpheme in main clauses is apparently associated with the mood of the verb, this is not the case. (8) As it is evident from the situation in embedded clauses, negation is affected by the modality of the clause: thus, me: is the negation for deontic and ou(k) for epistemic modality. Since subjunctive, optative, and imperative prototypically express deontic modalities, their prototypical negative morpheme is me[??]. Indicative, on the other hand, expresses mainly epistemic modalities and thus, its negative morpheme is ou(k). (9) However, this pattern can be easily broken, as in the case of future indicative, when it is used with a deontic meaning in purpose clauses; in this case the negative morpheme is me[??] and not ou(k). The fact that negation choice is related to the modality of the clause may indicate that there is an abstract operator in C projection (10) that determines the modality of the clause and selects the appropriate C and NEG heads. We will not explore this aspect of ClGk clause structure here since it does not affect the analysis proposed. (11)

3.2. The mood system of Modern Greek

The MGk mood system is a mixed one (Veloudis and Philippaki-Warburton 1983): it involves a morphological distinction between imperative vs. nonimperative verb forms. The nonimperative ones include the indicative and the subjunctive differentiated by means of the particle and the negation morpheme with which they combine. We follow here the analysis of the MGk mood system proposed by Philippaki-Warburton (1998), represented in (3). (12)

(3) [ILLUSTRATION OMITTED]

The analysis in (3) accounts for the following combinations:
(4) [CP C [MOODP MOOD [NEGP NEG [FUTP FUT [TP T

a. indicative: (oti) [??] ([??]en) ([??]a) [??]rafi
 that not will be writing

b. subjunctive: [??] na / as (min) [??]raft
 SUBJ not be writing

c. imperative [??] -affix
 [??]raf-e [t.sub.i]'
 write

 [vP V ]]]]]]

a. indicative: [t.sub.i]

b. subjunctive: [t.sub.i]

c. imperative [t.sub.i]


Some of the important features of the MGk system are the following: the mood oppositions are reduced to indicative, subjunctive, and imperative, and they are represented syntactically as features within a separate functional projection MOOD to the left of NEG with certain consequences for the architecture of the clause. The verbal ending in MGk marks only two mood oppositions, namely imperative vs. nonimperative. The imperative is associated with an affax, while the nonimperative is further specified as either indicative or subjunctive by the choice of the particle ([??] vs. na/as) and the choice of the negative morpheme (min vs. [??]en), as indicated in (4). The claim made in this analysis is that, whereas all three moods are located within a single projection MOOD, only the affixal mood, the imperative, motivates overt verb movement in order to be licensed. The indicative and the subjunctive are satisfied by the merging of the relevant particles with the MOOD head. This analysis offers a satisfactory explanation for the following two characteristics of MGk:

i. The object clitic pronouns precede the verb in the indicative and the subjunctive ([5a], [5b]), but they follow it in the imperative (5c):
(5) a. [??]en [??]a tu to [??]osis
 NEG FUT he-GEN it-ACC give-2SG
 'You will not give it to him'

 b. na min tu to [??]osis
 SUBJ NEG he-GEN it-ACC give-2SG
 'You should not give it to him'

 c. [??]ose tu to
 give-IMP,2SG he-GEN it-ACC
 'Give it to him'


This difference is explained as follows: the affixal character of the imperative demands that the imperative verb form move to MOOD to license the feature of imperative. The clitics, which are left adjoined to INFL, are left behind by the V-to-MOOD movement. We thus have an instance of EXCORPORATION in the sense of Roberts (1991). (13) No such movement is necessary for a nonimperative form since it being indicative or subjunctive is decided by the merging of the appropriate particle with the head of MOOD. Since no V-movement is involved there is no excorporation and thus the clitics remain to the left of the verb form.

ii. As mentioned above, there are no negative imperatives in MGk. This is explained as follows: as claimed, the imperative verb needs to move to the MOOD head to license the affixal imperative. However, when a negative head is present, by being located between the MOOD and the INFL, it creates a minimality effect and thus obstructs the V-to-MOOD movement. As a consequence, negative imperatives cannot be produced in MGk (see [6a]). Indicative and subjunctive verb forms do not undergo V-to-MOOD movement (they only move to INFL), and, therefore, the presence of a negative head creates no problems. Thus, the analysis in (3) predicts the data in (6a) and provides a reason why prohibitions in Greek are expressed by a negative subjunctive as in (6b).
(6) a. *mi fere to aftokinito su
 NEG bring-IMP,2SG the car your
 'Don't bring your car'

 b. na mi feris to aftokinito su
 SUBJ NEG bring-2SG the car your
 'You should not bring your car/Don't bring your car'


We will not discuss the syntactic status of particle [??][??]. We simply note that in our analysis it is a particle that occupies a maximal projection FUT that prototypically expresses the future indicative. (14)

3.3. The system of Hellenistic and Roman Koine

As mentioned above, the Greek mood system started as a pure morphological one and evolved into a more complex morphosyntactic one. The main change in configurational terms concerns the association of mood features with a functional category MOOD to the right of C and to the left of NEG, outside the INFL. There are good reasons to believe that this change took place quite early, during the period of HELLENISTIC AND ROMAN KOINE (HRK) (third century BC to fourth century AD), which is characterized by major changes at all levels and is considered to be the forerunner of the MGk system (Browning 1983; Horrocks 1997). (15)

Sweeping changes in the phonological system of Greek during the postclassical (fourth to third century BC) and the Hellenistic (third to first century BC) period had far-reaching consequences on the morphological system. There was a dramatic merging of phonological distinctions (16) affecting the verb endings that were invested with the expression of mood oppositions. This merging was accompanied by a parallel weakening in the use of the more marked moods, especially the optative. The changes are summarized as follows:

i. The optative is gradually disappearing. In the literary texts, which tend to imitate the Attic dialect, although optative is still used, its use is receding, as is evident in the works of PLUTARCH and POLYBIUS (Jannaris 1897; Horrocks 1997). In nonliterary texts the so-called OPTATIVE OF INDIRECT SPEECH disappears, while in main clauses the optative is systematically replaced by either the subjunctive or the indicative, depending on the modality, with the exception of some fossilized expressions.

ii. The negative imperative becomes increasingly rare so that by the end of this period (fourth century AD) it appears only as an irregular form and prohibitions are rendered by the negative subjunctive. In fact, the imperative and the subjunctive are coming closer to each other with the subjunctive often taking on the role of the imperative. Semantically this is supported by the fact that both moods prototypically express related types of modality, namely deontic. This proximity is responsible for a number of blend structures such as the following, which became quite regular in the nonliterary papyri of the period:
(7) (P. Bour. 25, 12-14 [fourth century AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 mninaoneve ... ina i tina evriskis pembe
 remember-IMP,2SG CONJ if someone find-2SG send-IMP,2SG
 pros eme
 to me
 'Remember ... to send me a letter, if you find someone coming
 here'


In (7), ina, which in ClGk was an inherently deontic complementizer expressing purpose, is inserted in order to grammatically and semantically reinforce the imperative form pembe.

iii. The subjunctive morphologically falls together with the indicative. This leveling started from the aorist subjunctive and the future indicative, and it gradually expanded to all tenses and aspects. Example (8) represents the 2nd person singular of the verb grapho: 'I write':

(8) [ILLUSTRATION OMITTED]

iv. Another important change is the gradual disappearance of the infinitive and its replacement in complementation by either ina + subjunctive or oti + indicative (Joseph 1983, 1990):

(9) [ILLUSTRATION OMITTED]

These changes led to a regularized system with subjunctive taking over the cases of deontic modality and indicative those of epistemic modality, and eventually these two moods became the predominant mood distinction. By the first century AD the distinction between indicative and subjunctive lacks a specific grammatical exponent, since verb forms are neutral as far as this distinction is concerned. In formal terms, this development meant the loss of their morphological representation in INFL. However, the semantic distinction between epistemic and deontic modality is still operative and it becomes grammatically relevant because of the specialization of the negative particle: clauses with epistemic force are negated by u(k), whereas clauses with deontic force are negated by mi. The next move is the mapping of the semantic distinction epistemic vs. deontic on the grammatical distinction indicative vs. subjunctive. Thus, the distinction between the indicative and subjunctive is easily recovered in some of the embedded clauses, where the conjunctions introducing them indicate how the clause was to be interpreted. So, for example, the clauses introduced with oti are interpreted as indicative while those introduced with ina are interpreted as subjunctive, in spite of the fact that there is no morphophonological distinction of the mood on the verb form.

On the basis of these observations, it may be argued that the modality distinction between epistemic and deontic acquires a grammatical status and comes to be expressed by the mood distinction between indicative vs. subjunctive, (17) so that it is hosted under a separate MOOD functional category below C and above NEG:

(10) [ILLUSTRATION OMITTED]

We propose that this MOOD functional projection is the lower [C.sub.fin] in the more articulated COMP system proposed by Rizzi (1997). Roussou (2000) suggests that this lower C head is better characterized as the locus of modality or mood features of the clause. If so, then the association of modality features with the indicative vs. subjunctive mood distinction, which took place in the HRK period, rendered this functional head as a pure MOOD projection. This hypothesis implies that MOOD in the HRK period is part of the COMP-layer and accounts for the mood specialization of certain complementizers. (19)

The analysis in (10) is supported by the following evidence:

i. It accounts for the consistent difference in the choice of the negative morpheme: uk for indicative vs. mi for subjunctive.

ii. It provides us with the basis for the development of the conjunction ina < hina to the subjunctive particle na. The presence of the MOOD head creates a position in the structure to be occupied by the grammaticalized ina (initially located in C) mood marker (located in MOOD) and motivates its final reanalysis as a mood particle.

We conclude that the evidence presented above shows the existence of a MOOD functional projection in the COMP-layer hosting the modality/ mood features relevant to the indicative-vs.-subjunctive distinction. As the system is changing, this functional category acquires content by the grammaticalization of the conjunction ina and its reanalysis as the subjunctive mood particle na.

3.4. The grammaticalization of ina

Let us now consider the other major development in the history of Greek mood system, that is, the grammaticalization of the conjunction ina < hina to the subjunctive particle na, for which there are clear signs that had already started during the HRK period.

In ClGk hina was a purpose conjunction introducing adjunct purpose clauses with the subjunctive verb form. Due to the affinity between the semantics of purpose and those of the subjunctive this conjunction was increasingly associated with the subjunctive. During the Hellenistic period (third to first century BC), the use of ina is extended first to all the purpose clauses, eventually replacing the alternative conjunction opos, which becomes restricted to result clauses. During the early Roman period (first century BC to second century AD) the use of ina is further extended to introduce not only purpose clauses but also complement clauses that replaced the infinitive and the predicative participle. In fact it becomes the element that introduces clauses after verbs of wishing, verbs expressing ability, fitness, attempt, etc. (Mandilaras 1973: 260). After the second century AD, the use of ina in constructions where its original meaning as a conjunction of purpose is not appropriate is the general pattern (Mandilaras 1973: 260-264, 321-322):
(11) a. CAUSAL
 (P. Giss. 17, 5-6 [second century AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 i[??]oniasa ... ina akuso oti eno[t.sup.h]re[??]sas
 worried-1SG CONJ heard-1SG that were-2SG inert
 'I was worried ... because I heard that you made no effort'

 b. IMPERATIVAL
 (P.Abinn. 8, 11-14 [fourth century ADD
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 ina un plirosis a[??]tus tin timin
 CONJ well pay-2SG them the price
 'Well, pay them the price ...'

 c. COMPLEMENT (instead of an infinitive)
 (P.Fay. 114 [second century AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 erotise me ermonaks ina a[??]ton la[??]i ...
 asked-3SG me Hermonaks-NOM CONJ him take-3SG
 'Hermonaks asked me for permission to take him ...'


This extension in the use of ina spreads even to main clauses, as in (12), and this clearly shows that ina is now used as the marker of the subjunctive.
(12) a. (N. T. Eph. 5: 33)
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 i de [??]yni ina [??]o[??]ite ton andra
 the PRT woman CONJ fear-PASS,3SG the man
 'The woman should be afraid of the man'

 b. (P.Gen. 55, 15-16 [fourth century AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 ina a[??]ta [t.sup.h]repsis, ta me[??]ista my
 CONJ them feed-2SG the greatest me
 [k.sup.h]arizomenos
 gift-PART,NOM,MASC,SG
 'Please, look after them, so doing me the greatest favor'


The use of ina in main clauses, as in (12), where there is no need for its presence, since Greek main clauses are not introduced by a complementizer, indicates that ina is getting grammaticalized as a subjunctive marker. This development is also evident by syntagms where ina introduces an infinitival complement with deontic force (see example [13]). It should be pointed out that such infinitival complements are being replaced in this period by a subjunctive finite complement clause introduced by ina. These blend forms then, which occur frequently in the nonliterary papyri of this period as errors, indicate the instability caused in the system by the replacement of infinitival complements and the grammaticalization of ina.
(13) a. (P.Merton 63, 18-19 [57 AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 e[??]khometha se ina kalos ekain
 wish-1PL you CONJ good have-INF
 'We wish that you be all right'

 b. (P.Hamb. 86, 8-10 [second century AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 parayenu ina ... ton mikron epikrine
 come-IMP,2SG CONJ the kid question-INF
 'Come so that we question the kid'


The whole process of the grammaticalization of ina will be completed by its morphophonological reduction to na during the middle Byzantine years (sixth to tenth century AD) (see Jannaris 1897; Trypanis 1960; Horrocks 1997).

We conclude that the grammaticalization of ina involved the following steps: (i) the loss of its specific meaning as a purpose conjunction (semantic bleaching), (20) (ii) its association with the subjunctive mood and (iii) its morphophonological reduction to na.

4. The MOOD functional category and the grammaticalization of ina

As we saw, the main change from the ClGk to the HRK system is that the indicative vs. subjunctive distinction is not morphologically marked in the verb ending anymore. But since the semantic distinction between deontic and epistemic modality became grammatically relevant and was mapped on the distinction between indicative and subjunctive, we suggested that this distinction was now expressed in a MOOD functional projection immediately to the right of C in the COMP-layer. We take this change to be an instance of reanalysis: (21)
(14) [CP C [NEGP NEG [INFLP INFL[Mood/Tense] ]]] [right arrow]
 [CP C [MOODP MOOD [NEGP NEG [INFLP INFL[[+ or -] imper/
 Tense] ]]]


The imperative on the other hand was still morphologically represented in the INFL functional head since its morphological distinction was not affected by the radical readjustment of the verb system. However, imperative expresses deontic modalities, a feature that belongs to the new MOOD functional category. This double marking (in the verb ending and in MOODP) explains the blend forms ina + V-IMP (see examples [7] and [13]), where the conjunction ina is inserted to reinforce the deontic modality of the verb form.

This reanalysis was followed by other changes and developments. We suggest that these changes constitute an actualization process in terms of Timberlake (1977) and Harris and Campbell (1995). According to their definition, actualization is the gradual mapping out of the consequences of reanalysis, that is, it includes all the adjustments that bring the surface into line with the innovative underlying structure. To be more concrete, we suggest that the grammaticalization of ina to na was part of the actualization of the reanalysis that created the MOOD functional category. The first signs of the grammaticalization of ina during the second century AD (loss of its specific purpose meaning, as in [11], the existence of blend forms, as in [13], its use in main clauses, as in [12]) indicate that this reanalysis had already taken place in the first century AD.

On the other hand, the grammaticalization of ina to na involved another reanalysis, namely its recategorization and its transfer from the C head to the MOOD head:
(15) [CP ina [MOODP [MOOD [??]] ... [INFLP V] ...]] [right arrow]
 [CP [MOODP [MOOD ina] ... [INFLP V] ... ]


We may trace this reanalysis in the late third century AD based on the following evidence:

i. By the fourth century AD, ina has lost its specific purpose meaning as a conjunction and is regularly used in embedded clauses to mark subjunctive.

ii. By the fifth century AD, its use in main clauses with deontic force is generalized.

iii. During the sixth century AD, the vernacular texts show regular complementation patterns in which the ina + verb syntagm follows complementizers like oti, opos, opu.

iv. After the fourth century AD, ina is normally combined with the complementizer os in a complex form os ina or as one word osina to introduce purpose/final clauses: (22)
(16) (Athan. ii. 824A [fourth century AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 iksiosan tus e[??]nuxus osina systaseos tyxosin y apisty ariany
 honor-3PL the eunuchs CONJ support get-3PL the infidel Ariani
 'They honored the eunuchs in order for the infidel supporters of
 Arrius to get support'


The last two facts show that ina does not occupy the C head anymore because this can be filled with another complementizer. Thus, ina has already been reanalyzed as a particle occupying the MOOD head. Thus, constructions such as (16) show that a real complementizer (os) is used in order to express the meaning of purpose, and that ina is no longer a complementizer, but a mood particle under MOOD.

In addition, the fact that os + ina is used as a unit shows that this MOOD functional category is still part of the COMP-layer and closely associated with the C functional head. This is also supported by the fact that during the HRK period an XP could intervene between ina and the Verb (see example [17]; see also [13] and [16]).
(17) (P.Fay. 114 [second century AD])
 [TEXT NOT REPRODUCIBLE IN ASCII.]
 ina embiros kopi ta melonta ekoptes[t.sup.t]e
 CONJ skilfully cut-3SG the going to be cut out
 'So that those to be cut down be cut skilfully'


During the middle Byzantine years (sixth to tenth century AD) another development takes place: ina is morphophonologically reduced to na. By the twelfth century AD all subjunctive clauses (main and embedded) are marked by the presence of the particle na. The subjunctive syntagm na + verb (23) can now also follow complementizers and be used in clauses with a fronted element in the COMP-layer (wh questions, focal operators, etc.). The morphophonological reduction of na and its subsequent phonological association with other elements such as the negation markers and clitics, all of which have become proclitic to the verb, shows that the MOOD functional category is no longer associated with the COMP-layer, but has been reanalyzed so as to belong to the extended INFL-domain of the clause in an articulated structure similar to that of MGk: (24) na at this time being proclitic to the main verb does not allow for any constituent to intervene between the two of them apart from the clitic pronouns and the negation particle. (25) Thus, the functional projection of MOOD that originated as part of the COMP-field in the HRK period ended up to belong to the extended INFL-layer.

(18) [ILLUSTRATION OMITTED]

In addition to these developments, it will be very useful to briefly discuss the parallel development of the imperative mood. Though less research has been done on the chronology of this development, we note here the following facts: in the Hellenistic and Roman period, negative imperatives could be easily found, which shows that the imperative mood was still licensed in the INFL functional node and not in the new MOOD functional category. However, at the end of the Roman period, prohibitive subjunctive was gradually taking over the negative imperative so that at the end of the period (fourth century AD), negative imperative, though existent, was an irregularity, and the most frequent pattern was the prohibitive subjunctive. This pattern continued in the Byzantine years, and by the tenth century, negative imperatives seem to have disappeared. We may therefore suggest that between the fifth and the ninth century AD, a further reanalysis took place with the result that imperative mood is now licensed in the new MOOD functional projection outside the INFL and above NEG:

(19) [MOODP [MOOD] ... [INFLP [INFL V-IMP] ]] [right arrow] [MOODP [MOOD V-IMP] ... [INFLP ]]

The development of the Greek mood system is summarized in Figure 1.

[FIGURE 1 OMITTED]

5. Discussion

In the previous sections we presented the major developments of the Greek mood system. We will now consider what these developments reveal about the system of the language and the theories of morphosyntactic change.

5.1. The distinction between indicative and subjunctive

As mentioned in section 3.1, in ClGk there was a morphological distinction among four moods, namely indicative, subjunctive, optative, and imperative. In the course of its history, Greek lost the morphological distinction of indicative vs. subjunctive and the optative disappeared. But, while the language eventually found a way of representing the distinction between indicative and subjunctive, optative was not recovered. This development can be attributed to the relative importance of mood oppositions and their association with the fundamental distinction between epistemic and deontic modalities. Bybee (1985) and Bybee et al. (1994), on the basis of an extensive language database, have argued that the grammatical distinction between indicative and subjunctive is fundamental and exists in almost every language expressed by different means (bound or free morphemes, particles, auxiliaries, etc.). So a language that loses the morphological expression of this distinction is expected to try to recover it. The development of the Greek mood system may be thus considered an example of this tendency: when the morphophonological distinction between indicative and subjunctive was lost, a syntactic way of representing it by means of the MOOD functional category and the grammaticalization of ina (26) was implemented. This development was facilitated by (i) the fact that the semantic distinction between epistemic and deontic modality was mapped on certain morphosyntactic structures, which were eventually grammaticalized in the indicative vs. subjunctive distinction, and (ii) the disappearance of the infinitive and all the nonfinite forms in complementation, which strengthened the syntactic importance of the distinction between indicative and subjunctive by offering new environments for its occurrence.

On the other hand, the optative as a distinct grammatical category became a luxury for the language, as the distinction between indicative and subjunctive was being reinforced, so that it eventually disappeared completely and its semantic functions were assigned to the already existing moods. It should be noted that the disappearance of optative was not in fact caused by the morphophonological changes during the Hellenistic period, since the merging of its characteristic vowel in the verb ending was completed much later than its disappearance. (27) So, when the optative disappeared from use, it was still a morphologically distinct category. Thus, the reason for the disappearance of the optative was mainly semantic and functional: (a) its functions were taken over by the subjunctive and the indicative, and (b) after the leveling between indicative and subjunctive and the mapping of the epistemic vs. deontic modality on certain syntactic structures, the distinct morphological marking of the optative was out of tune with the rest of the system.

5.2. Morphosyntactic features, parameters, and functional categories

As mentioned above, the development of the Greek mood system involved a number of changes regarding the morphosyntactic representation of mood features. The main change was the association of mood features with a separate MOOD functional category. Thus, in Greek the mood features acquired a syntactic status by means of an overt syntactic effect. In ClGk mood features were licensed in a fused INFL functional category. Thus, the different mood choice had no effect in the syntactic structure, since the licensing of mood features was completed inside the INFL functional category. In MGk the mood features have acquired a syntactic role by being associated with a separate MOOD functional category. Thus, the subjunctive is satisfied by merging a subjunctive particle (na or as), whereas the imperative requires overt V-movement resulting in the reverse V-cl order and the unavailability of negative imperatives. This means that mood features in MGk are strong and must be satisfied before Spell-Out giving rise to an overt manifestation. (28)

In formal terms, the development outlined above is also relevant to the syntactic status of mood features. Thus, the parameters involved are related to the morphosyntactic representation of mood features and their +strong value. Let us consider the first parameter, namely the syntactic representation of mood features. This parameter decides on whether mood features will have a syntactic representation in terms of formal features and subsequently whether they will be hosted under a separate functional category or be fused with another. The importance of this parameter for determining the clause structure at a synchronic and diachronic level is therefore obvious and has been pointed out by Roberts and Roussou (1999). The development of the Greek mood system offers a nice illustration of the effects of the resetting of this parameter.

The second parameter regards the strong vs. weak value of these features. Strong features are associated with an overt morphosyntactic manifestation either by means of overt movement of the relevant lexical category realizing these features, or by the merging of a particle with the functional head hosting these features. Weak features on the other hand require no overt operation and their licensing is performed by means of covert movement or an AGREE operation. The importance of this parameter has been pointed out by Hale (1998) and it is characteristically exemplified by the development of the English clause structure (Roberts 1992). In Greek, mood features have a strong value so that they require an overt manifestation by means of either an overt V-movement (imperative) or by merging a particle (subjunctive), having a crucial effect on clause structure. The parameters and their options are illustrated as follows:
(20)
 Parameters of features

A. Syntactic representation B. Formal features are
 strong

I. Separate II. Syncretic to II. Overt II. Merge
F functional another functional head- (particle)
head head F movement


The development of the Greek mood system and its effect on the clause structure shows that morphosyntactic change concerns the whole morphosyntactic status of grammatical features. A formal theory of features therefore is a powerful descriptive tool for explaining morphosyntactic change. Such a theory may provide a formal account of grammaticalization, since if grammaticalization is viewed as an epiphenomenon of a syntactic reanalysis accompanied by phonological and semantic erosion, it is expected to be associated with the morphosyntactic representation of formal features. Roberts and Roussou's (1999) approach to grammaticalization in terms of structural simplification is consistent with this spirit.

However, the grammaticalization of ina to na constitutes a case that cannot be accounted for by an approach that relies on structural simplification for the following reasons: (29) (i) ina was already an element merged to a functional category, that of C, which occupied a position above MOOD. (ii) ina was not related to MOOD by means of movement. The grammaticalization of ina did not involve the reanalysis of a lexical item (open class item) to a functional element (closed class), but rather a reanalysis regarding the functional category with which ina is merged. Such a case is not relevant to the computational labor of MERGE and Mow, so that an economy measure based on these two operations may not provide an explanation. We may speculate that the driving force for the reanalysis of ina from C to MOOD has to do with the fact that the structure contained an empty functional category MOOD with syntactically active features. Such functional categories require overt realization, favoring the MERGE option, and the reanalysis of ina can be seen as the result of the tendency of the acquirer to meet this requirement. (30) Thus, the grammaticalization of ina is attributed to the morphosyntactic requirements of mood formal features.

6. Conclusion

As is obvious from the presentation of the evolution of the Greek mood system, there are languages in which mood features play a crucial role in determining their clause structure. In such languages, mood features require a certain morphosyntactic representation and thus they are hosted under a relevant MOOD functional projection, which may be either syncretic to another functional category (mainly the INFL), or independent. The way these features are satisfied has a radical effect on the clause structure of the language. This effect may be obvious at both a synchronic and a diachronic level. Thus, the differences in the Greek morphosyntax of the imperative on the one hand and the indicative and the subjunctive on the other illustrate the effect of the different satisfaction of the MOOD functional category in each case. On the other hand, the development of the Greek mood system shows that in a language with morphosyntactically active mood features, the catastrophic loss of the morphological mood distinctions may also lead to a process by which these active mood features are satisfied by means of overt syntactic operations such as MERGE and MOVE. The nature and the results of this process support the view that morphosyntactic change is relevant to the nature of the specific formal features and the parameters that decide on whether (i) they are morphosyntactically active in the configuration, and (ii) the way they are licensed in the syntactic component.

Received 3 May 2002

Revised version received

5 November 2002

Notes

* This is an updated version of a paper read at the DIGS VI (Maryland, May 2000) and an extended version of a study of the mood system of Hellenistic and Roman Koine, a part of which has been presented at the 20th Annual Meeting of the Department of Linguistics of The Aristotle University of Thessaloniki (April 1999) and has been published as philippaki-Warburton and Spyropoulos (2000). We would like to thank these audiences for their useful suggestions. The paper has also been benefited by the comments of Julian-Victor Mendez Dosuna, Geoffrey Horrocks, Brian Joseph, David Lightfoot, Lindsay Whaley, and two anonymous reviewers. Any errors are our own responsibility. Correspondence address: Irene Philippaki-Warburton, School of Linguistics and Applied Language Studies, University of Reading, Whiteknights, P.O. Box 218, Reading RG6 6AA, United Kingdom. E-mail: llswarbn@reading.ac.uk.

(1.) The term mood is used in this paper in its strict sense as a morphosyntactic category rather than a purely semantic/functional one. Thus, mood refers to a morphosyntactically distinct category (which might be marked by means of an affix, or a particle, or even a modal element) and not to the possible uses/functions of this paradigm.

(2.) See for instance the work by Philippaki-Warburton (1994, 1998), Rivero and Terzi (1995), and Roussou (2000).

(3.) Our study is not a quantitive one, that is, it is not based on the statistical analysis of a specific corpus. It relies mostly on traditional descriptions of the history of Greek, such as those by Jannaris (1897), Goodwin (1959 [1879], 1965 [1889]), Schwyzer (1950), Trypanis (1960), Mandilaras (1973), Browning (1983), Moulton and Turner (1963), and more recently Horrocks (1997). We recognize the limitations of this and we hope that our study will provide the starting point for a future research of this kind. To the best of our knowledge, quantative analyses of the history of Greek are rather few (see Taylor 1990; Manolessou 2000), and therefore any research of this type will be welcome.

(4.) Although this is the main tendency, there are cases which show that the manifestation of overt features is not directly linked to the notion of strength (see Bobalijk 2000). More research is required on this issue; however, such a correlation seems to be valid for most of the cases.

(5.) See also Alexiadou and Anagnostopoulou (1998) on a similar parameter for the satisfaction of the extended projection principle.

(6.) We assume that ClGk clause structure involves overt V-movement to INFL (see also Rivero and Terzi 1995). The evidence for this assumption is mainly theoretical, since the normal diagnostics (position of adverbs and negation) cannot apply in ClGk due to the freedom of word order. Thus we assume that the elaborated verbal morphology of ClGk, which in formal terms may be translated into strong formal features that require overt movement, motivates overt V-to-INFL movement. We note here that Taylor (1994) has suggested that SOV orders may indicate that verb remains in VP in ClGk. Her hypothesis relies on an analysis of OV structures in terms of a linearization parameter [head first--head final]. However, these structures do not exclude verb movement; on the other hand, if we follow recent assumptions (Chomsky 1995, 2000) that linearization parameters are not a matter of syntax, then verb movement is the only option.

(7.) We point out that aorist imperatives cannot be negated in Greek. We believe that this is an idiosyncrasy of ClGk that has no structural explanation and may be attributed to the functions of this imperative due to its aspect properties and limitations.

(8.) The distribution of negation markers in ClGk is quite complex and we refer to the work of Goodwin (1959 [1879]) and Schwyzer (1950) for details. However, more research is required on the formal patterns that determine the choice of the negation marker in ClGk.

(9.) The mood of the embedded clause was also affected by the lexical meaning and by the tense specification of the main verb. In the latter case, the mood of the embedded clause was always an instance of optative, the so-called optative of indirect speech.

(10.) Relevant to the existence of these modality features seems to be the fact that some complementizers were incompatible with some moods. Thus, oti was incompatible with subjunctive and hina with indicative. This incompatibility may be attributed to the modality associated with these complementizers, that is, episteimc with oti vs. deontic with hina.

(11.) It has been suggested by an anonymous reviewer that mood licensing in ClGk may be analyzed in two alternative ways. It can be argued (i) that mood is syncretic to NEG rather than to INFL, or (ii) that there is a separate MOOD functional projection which is situated in the COMP-layer in a Split-COMP structure like the one proposed by Roussou (2000).

The first alternative presupposes a covert agree operation between the mood features in NEG and the V, which remains in INFL. This alternative is rejected on both theoretical and empirical grounds: the verb morphology clearly indicates that mood features are syncretic to tense and subject agreement inside the verbal inflection. It will therefore be odd to ignore this morphological indication and propose a syncretism that is not morphologically indicated. In addition, a syncretism with NEG implies a direct relation between negation and mood. However, as argued above, negation seems to be related to the modality of the clause rather than to the mood of the verb. This fact indicates that the licensing of mood is completed before and independently of negation. The verb form in ClGk with all its inflections seems to have its own independent status, and this is exactly what is captured by the proposal that all inflectional checking has been completed in the INFL-domain indepedently of the merging of other grammatical elements such as the negation.

The second alternative again disassociates mood licensing from all the other inflectional licensing in the INFL-domain, by situating a MOOD functional projection in the Split-Comp-layer. This alternative presupposes the system of COMP projections proposed by Roussou (2000), according to which the lower of the C heads is in fact a MOOD head and NEG is situated above this and before C, inside this extended Comp-layer, as shown in (i):

(i) [[sub.Cclause-typeP] [C.sub.clause-type] [[sub.CompP] COMP [[sub.NEGP] NEG [[sub.MOODP] MOOD]]]]

Verb movement then takes place all the way up to MOOD head so that mood features are licensed. Such an analysis has certain implications regarding the licensing of the imperative in both ClGk and MGk. We will not discuss this possibility in detail. We want, however, to point out that this analysis overlooks the syncretic character of mood in ClGk, which is morphologically indicated in the verbal inflection of tense and subject agreement. By assuming the mirror principle as a guidance for the order of functional categories, the morphological pattern clearly indicates that mood is licensed in the INFL-domain rather than in the COMP-layer. Therefore we will not entertain this suggestion until we have further explored its implications for both the synchronic and diachrohic aspects of the Greek mood system.

(12.) See also Philippaki-Warburton (1994) and Philippaki-Warburton and Spyropoulos (1999). For alternative analyses see Rivero and Terzi (1995), Drachman (1994), and Roussou (2000).

(13.) An anonymous reviewer questions whether excorporation is compatible with minimalist assumptions and suggests that clitics may occupy the heads of relevant clitic projections above INFL. Such an analysis does not affect the points made here, namely that the order Vimp-CL is derived syntactically by moving the imperative V head over the clitics. The literature on Greek clitic pronouns is vast. Here we have adopted Philippaki-Warburton and Spyropoulos' (1999) analysis.

(14.) The exact nature of this functional category is under examination. In Philippaki-Warburton (1998) it is considered to be a FUT (Future) functional category, due to the main function of [??]a, namely the expression of the future indicative. However, given the frequent modal interpretations of the verb syntagms with [??]a (see Tsangalidis 1999), it may be suggested that this is a second MOOD functional category. We will not discuss this alternative here. We just want to stress that [??]ads its own functional category (see Rivero 1994; Drachman 1994; Roussou 2000 for alternative analyses of [??]a)

(15.) This period of the history of Greek is a radically developing one so that it is quite hard to identify a coherent system. More specifically, during the first period (Hellenistic Greek: third to first century BC), we can only identify the damaging effects of the phonological, morphological, and syntactic changes to the old system. It is by the middle and the end of the second period (Roman years: first century BC to fourth century AD) when the system acquires some stability and coherence. Thus, we will focus on these years, and we will mainly take account of the nonliterary registers, as the most revealing of the emerging system of the language. Our data are drawn from the texts of the nonliterary papyri and from the NEW TESTAMENT, and more specifically we make extensive use of the detailed descriptions of the systems of the varieties used in these texts, namely Mandilaras (1973) for the nonliterari papyri and Moulton and Turner (1963) for the NEW TESTAMENT. Koine examples will be cited in Greek followed by a phonological transcription relevant to their date and place of origin. We used the review in Horrocks (1997) as a reference guide for the phonological system(s) of this period.

(16.) For a detailed exposition of these phonological changes that destroyed certain morphological differences see the reviews in Browning (1983) and Horrocks 0997).

(17.) It should be pointed out that despite their mapping modality and mood distinctions do not correspond in an one-to-one way. Mood distinctions are grammatical distinctions expressed in morphology and syntax, whereas modality is a semantic notion; thus epistemic vs. deontic readings may depend on other factors such as tense, aspect and person. It is the prototypical functions of these distinctions that lead to their mapping. We thank an anonymous reviewer for pointing out this fine detail.

(18.) We point out that, as in ClGk, the category of the imperative is still licensed within the projection of INFL. This hypothesis is supported by the existence of negative imperatives during this period--a fact which shows that the licensing of the imperative mood takes place in a projection to the right of negation, that is, the INFL head.

(19.) Another alternative would be to reject the existence of a separate MOOD functional head, and propose that the features relevant to the distinction of indicative vs. subjunctive were hosted under the C head itself. These features would be satisfied by either an appropriate complementizer or a conjunction (ina, opos) in subordinate clauses, or by an empty operator in main clauses. However, such an analysis faces the following problems: first, one of the predictions it makes is that all complementizers and conjunctions should be specified and highly specialized as far as the mood choice in their clause is concerned. For example, ina is marked to introduce subjunctive clauses and oti is associated with indicative. However, a problem arises with certain conjunctions (e.g. conditional, temporal; see Mandilaras 1973: 266-270) that can take either mood as indicated by the different negation marker with which they can combine. In these cases it would be counterintuitive to assume that the same conjunction satisfies both a subjunctive C in the one case and an indicative one in the other. A second problem concerns the operator capacity of C. It is well established that C can license only one operator or feature. If subjunctive were always licensed in C it would mean that no other element could be licensed in that position at the same time. However, this would imply that KOINE does not exhibit subjunctive wh clauses because the wh element could never be licensed in a C which is already filled by a subjunctive operator, contrary to the facts (for the nonliterary papyri see Mandilaras 1973: 266). Related to the above observations is the issue of the licensing of the second position elements (particles, pronoun clitics), which is still operative in the language. If we accept the analysis of second position phenomena which assumes that the licensing of these elements is performed by the fronting of a word or phrase to the CP (Rivero and Terzi 1995), then we must conclude that this fronting cannot take place if C is already filled by the empty operator licensing its subjunctive features. Based on this evidence we believe that an analysis suggesting the existence of a separate MOOD functional head hosting the features relevant to the indicative vs. subjunctive distinction provides a better explanation for the Koine structures.

An anonymous reviewer suggests that these problems may be solved in a more articulated Comp-system, by assuming that those complementizers that are not sensitive to mood distinctions occur in a higher C head. Although such a proposal may provide a solution to the problems of the MOOD-in-C hypothesis, our analysis in terms of a separate MOOD projection is still superior. This analysis can accommodate the structures in which ina has already started to behave as a mood particle rather than as a complementizer, as is the case in main clauses and in combination with pure complementizers such as oti and os (see the discussion in the text below).

(20.) The semantic bleaching of hina from a purpose meaning to the more transparent grammatical function of introducing complement clauses follows a universal path of semantic grammaticalization identified by Haspelmath (1989). Haspelmath has shown that most of the complement infinitives derive from a purposive function of this nonfinite form. He proposes the following universal path of semantic grammaticalization:
(i)

benefactive
allative [right arrow] purposive [right arrow] irrealis
causal directive

[right arrow] irrealis [right arrow] realis
 potential nonfactive

[right arrow] (realis factive)


He also points out that the grammaticalization of hina follows this path, reinforcing his proposal.

(21.) For our purposes, REANALYSIS is defined as the assigning of a different underlying structure without changing the surface manifestation. The addition of the empty MOOD head is thus an instance of a reanalysis in the functional layer of the clause structure.

(22.) Such constructions can be found in papyri as early as the second century AD:
(i) (P.M5ch. 505, 2-4, [second to third century AD])

 [TEXT NOT REPRODUCIBLE IN ASCII.]

 o ypiretis par o ton [k.sup.h]alkon [t.sup.h]ematisamen
 the official-NOM by who-DAT the bronze-ACC deposited-1st,SG
 le[??]i
 say-3rd,SG
 [??]e[??]okene to praktori osina sy meta[??]alite
 give-INF the collector-DAT CONJ you-DAT tranfer-3rd,SG
 'The official by whom we deposited the bronze, says that he has
 given it to the collector so that he may transfer it to you'


It must be pointed out that a similar construction can be found in ClGk: the infinitive, which had lost its purposive meaning, was introduced by the conjunction oste in nonfinite purpose clauses:
(ii) (Thuc. 1, 44 I)

 [TEXT NOT REPRODUCIBLE IN ASCII.]

 metegn[??]san ksymma[k.sup.h]ian m[??]
 changed mind-3rd,PL alliance-ACC NEG
 poi[??]sas[t.sup.h]ai [??]ste to[??]s awto[??]s
 make-INF CONJ the same-ACC
 e[k.sup.h][t.sup.h]ro[??]s kai [p.sup.h]ilo[??]s nomidzein
 enemies-ACC and friends-ACC consider-INF
 'They changed their minds not to make an alliance so as not
 to have the same enemies and friends'


In MGk also, subjunctive clauses may be also introduced by ja ('for, in order') when they are used in purposive function.

(iii) ir[??]e ja na ton pari
 came-3SG CONJ SUBJ him take-3SG
 'He came in order to pick him up'


In that case as well, the element ja occupies the C head and it is introduced to reinforce the purposive meaning of the clause.

(23.) See also Tsangalidis and Valetopoulos (1999) for a discussion on the development of the subjunctive during the early MGk period.

(24.) It should be pointed out that according to Roussou's (2000) analysis of MGk clause structure, MOOD functional category belongs to the COMP-layer.

(25.) Consider the following example
(i) (o janis) na (*o janis) min (*o janis) fi[??]i
 the John-NOM SUBJ the John-NOM NEG the John-NOM go-3rd,SG
 (o janis)
 the John-NOM
 'John should not go'


(26.) Similarly, the loss of the morphological marking of subjunctive in English has been argued to be closely associated with the development of the modal auxiliaries (see Roberts 1985; Kroch 1991; Los 1999).

(27.) The merging of the diphthong [oi] to [i] was much slower than the other phonological developments and its completion belongs to the Middle Byzantine period (sixth to tenth century AD). During the HRK period the ClGk [oi] was pronounced as [y].

(28.) This creates a problem with the licensing of indicative in MGk. Given the absence of an overt indicative particle the MOOD functional category would receive no realization. Two alternatives present themselves: (i) MOOD does not project in that ease and (ii) there is a null indicative particle to satisfy MOOD. Although there is no principled reason for rejecting the first option we will follow Philippaki-Warburton (1998) and opt for the second alternative. We thank all anonymous reviewer for pointing out this problem.

(29.) But, see Roberts and Roussou (2003) for an account of this fact.

(30.) This change in the grammar of the acquirer was facilitated by the fact that be/she was exposed to data from embedded clauses showing that ina introduced clauses with subjunctive force. He then reanalyzed ina as a subjunctive mood marker.

References

Alexiadou, Artemis; and Anagnostopoulou, Elena (1998). Parametrizing Agr: word-order, verb-movement and EPP-checking. Natural Language and Linguistic Theory 16, 491-539.

Baker, Mark (1985). The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16, 373-415.

Bobaljik, Jonathan D. (2000). The implications of rich agreement: why morphology does not drive syntax. GLOW Newsletter 46, 28-29.

--; and Thrainsson, Hoksuldur (1998). Two heads aren't always better than one. Syntax 1, 37-71.

Borer, Hagit (1984). Parametric Syntax. Dordrecht: Foris.

Browning, Robert (1983). Medieval and Modern Greek. 2nd ed. Cambridge: Cambridge University Press.

Bybee, Joan (1985). Morphology. Amsterdam: Benjamins.

Bybee, Joan; Perkins, Revere; and Pagliucka William (1994). The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: UCP.

Chomsky, Noam (1995). The Minimalist Program. Cambridge, MA: MIT Press.

--(2000). Minimalist inquiries: the framework. In Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, Roger Martin, David Michaels and Juan Uriagereka (eds.), 89-155. Cambridge, MA: MIT Press.

Clark, R.; and Roberts, Ian (1993). A computational model of language learnability and language change. Linguistic Inquiry 24, 299-345.

Drachman, Gaberel (1994). Verb movement and minimal clauses. In Themes in Greek Linguistics: Papers from the First International Conference on Greek Linguistics, Irene Philippaki-Warburton, Katerina Nikolaidis, and Maria Sifianou (eds.), 45-52. Amsterdam: Benjamins.

Giorgi, Alessandra; and Pianesi, Fabio (1997). Tense and Aspect: from Semantics to Morphosyntax. Oxford: Oxford University Press.

Goodwin, William W. (1959 [1879]). A Greek Grammar. London: Macmillan.

--(1965 [1889]). Syntax of the Moods and Tenses of the Greek Verb. London: Macmillan.

Hale, Mark (1998). Diachronic syntax. Syntax 1, 1-18.

Harris, Alice; and Campbell, Lyle (1995). Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press.

Haspelmath, Martin (1989). From purposive to infinitive--a universal path of grammaticalization. Folia Linguistica Historica 10, 287-310.

Horrocks, Geoffrey (1997). Greek: A History of the Language and its Speakers. London: Longman.

Jannaris, Antonius N. (1897). An Historical Greek Grammar, chiefly of the Attic Dialect. Hildesheim: Georg Olms Verlagsbuchhandlung.

Joseph, Brian D. (1983). The Synchrony and the Diachrony of the Balcan Infinitive. Cambridge: Cambridge University Press.

--(1990). Morphology and Universals in Syntactic Change: Evidence from Medieval and Modern Greek. New York: Garland.

--(2001). Is there such a thing as "grammaticalization"? Language Sciences 23, 113-161. Kroch, Anthony (1991). Reflexes of grammar in patterns of language change. Language Variation and Change 1, 199-244.

Lightfoot, David W. (1979). Principles of Diachronic Syntax. Cambridge: Cambridge University Press.

--(1991). How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press.

--(1998). The Development of Language: Acquisition, Change and Evolution. Oxford: Blackwell.

Los, B. (1999). Infinitival Complementation in Old and Middle English. The Hague: Holland Academic Graphics.

Mandilaras, Basil (1973). The Verb in the Greek non-literary Papyri. Athens: Hellenic Ministry of Culture and Science.

Manolessou, Io (2000). Greek noun phrase structure: a study in syntactic evolution. Doctoral dissertation University of Cambridge.

Moulton, James H.; and Turner, Nigel (1963). A Grammar of New Testament Greek. 111: Syntax. Edinburgh: T. & T. Clark Continuum International Publishing Group.

Philippaki-Warbuton, Irene (1994). The subjunctive mood and the syntactic status of the particle na in Modern Greek. Folia Linguistica 28, 297-328.

--(1998). Functional categories and Modern Greek syntax. Linguistic Review 15, 158-186.

--; and Spyropoulus, Vassilios (1999). On the boundaries of inflection and syntax: Greek pronominal clitics and particles. In Yearbook of Morphology 1998, Gert Booij and Jaap van Marle (eds.), 45-72. Dordrecht: Kluwer.

--; and Spyropoulos, Vassilios (2000). Diachroniki theorisi tis eglisis: To sistima ton egliseon tis Elinistikis ke Romaikis Kinis [Diachronic perspectives of mood: The mood system of Hellenistic and Roman Koine]. Studies in Greek Linguistics 20, 567-578.

Rivero, Maria-Louiza (1994). Clause structure and V-movement in the languages of the Balkans. Natural Language and Linguistic Theory 12, 63-120.

--; and Terzi Arhonto (1995). Imperatives, V-movement and logical mood. Journal of Linguistics 31,301-332.

Rizzi, Luigi (1997). The free structure of the left periphery. In Elements of Grammar: A Handbook in Generative Syntax, Lilianne Haegeman (ed.), 281-337. Dordrecht: Kluwer.

Roberts, Ian (1985). Agreement parameters and the development of English modal auxiliaries. Natural Language and Linguistic Theory 3, 21-58.

--(1991). Excorporation and minimality. Linguistic Inquiry 22, 209-218.

--(1992). Verbs and Diachronic Syntax. Dordrecht: Kluwer.

--; and Roussou, Anna (1999). A formal approach to grammaticalization. Linguistics 37, 1043-1065.

--; and Roussou, Anna (2003). Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press.

Roussou, Anna (2000). On the left periphery: modal particles and complementizers. Journal of Greek Linguistics 1, 65-94.

Schwyzer, E. (1950). Griechische Grammatik. Zweiter Band: Syntax und syntaktische Stilistik. Munich: Beck.

Taylor, Ann (1990). Critics and configurationality in Ancient Greek. Unpublished Ph.D. thesis, University of Pennsyvania.

--(1994). The change from SOV to SVO in Ancient Greek. Language Variation and Change 6, 1-37.

Thrainsson, Hoksuldur (1996). On the (non)-universality of functional categories. In Minimal Ideas: Syntactic Studies in the Minimalist Framework, Werner Abraham, Samuel D. Epstein, Hoksuldur Thrainsson, and Jan-Wouter Zwart (eds.), 253-281. Amsterdam: Benjamins.

Timberlake, Alan (1977). Reanalysis and actualization in syntactic change. In Mechanisms of Syntactic Change, Charles N. Li (ed.), 141-177. Austin: UTP.

Trypanis, C. A. (1960). Early medieval Greek'iva. Glotta 38, 312-313.

Tsangalidis, Anastasios (1999). Will and tha: a comparative study of the category future. Thessaloniki: University Studio Press.

--; and Valetopoulos, Friderikos (1999). I "ipotaktiki aoristu" stin Proimi Nea Elliniki [The "aorist subjunctive" in Early Modern Greek]. Studies in Greek Language 19, 523-534.

Veloudis, Ioannis; and Philippaki-Warburton, Irene (1983): I ipotaktiki sta Nea Elinika [The subjunctive in Modern Greek]. Studies in Greek Linguistics 3, 151-168.

IRENE PHILIPPAKI-WARBURTON

University of Reading

VASSILIOS SPYROPOULOS

University of the Aegean
COPYRIGHT 2004 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2004 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Philippaki-Warburton, Irene; Spyropoulos, Vassilios
Publication:Linguistics: an interdisciplinary journal of the language sciences
Date:Jul 1, 2004
Words:10122
Previous Article:Combinatory conditions on Halkomelem causatives (1).
Next Article:Initial and final position for adverbial clauses in English: the constructional basis of the discursive and syntactic differences (1).
Topics:


Related Articles
Mood swings: fact or myth? An inescapable rite of teen life is riding an emotional roller coaster--euphoric one minute, despondent the next. Science...
Jury is still out on use of HRT for Perimenopausal Depression. (Few Older Women Participate in Trials).
NIMH charts strategy for Mood Disorders Research: expect funding to be tied to priorities. (Objectives Defined).
Compared with sham scans: MRI scans may perk up bipolar patients.
Most women are unlikely to experience premenstrual mood change with pill use.
`Magical Games' in store ( Rogge.
Verbal aspect, the indicative mood, and narrative; soundings in the Greek of the New Testament.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters