Towards the identification of verbal periphrasis in ancient Greek: a prototype analysis.

Bentein, Klaas

1. Introduction

Verbal periphrasis is a problematic and much discussed issue (cf. the introductory chapter by Pusch & Wesch 2003). This is not to say that there is no consensus whatsoever with regard to its general characteristics, which are summarised in recent review articles by Haspelmath (2000) and Spencer (2006). Spencer (2006:287) gives the following description of this grammatical phenomenon: 'the term "periphrasis" is most commonly used to denote a construction type in which a grammatical property or feature is expressed by a combination of words rather than a single (inflected) word form.' As an example of periphrasis in English, Spencer mentions the perfect aspect construction formed with 'have', as in 'the girls have sung'. He contrasts this with the expression of past tense, as in 'the girls sang', where the past tense form of the content verb itself is used (Spencer 2006: 287-88) (in the case of 'sang', scholars speak of a 'synthetic' or 'monolectic' verb form). Haspelmath (2000:660-61) makes an important distinction between two main types of verbal periphrasis, viz. 'suppletive' and 'categorial' periphrasis (cf. similarly Aerts 1965:3 and Evans 2001: 221), with the former filling a gap in the inflectional paradigm and the latter expressing 'some additional semantic distinction' (Haspelmath 2000:656). As an example of the latter type consider the Ancient Greek form [TEXT NOT REPRODUCIBLE IN ASCII] (Lc. 5.17) next to [TEXT NOT REPRODUCIBLE IN ASCII] (Lc. 5.3), both meaning 'he was teaching'. Furthermore, both Spencer and Haspelmath argue for a number of semantic, morphological and paradigmatic criteria to identify verbal periphrasis.

Even in these overview articles, however, various problems readily surface, most importantly with regard to the proposed criteria. As Haspelmath has to admit, none of his criteria is completely unproblematic, and it is not entirely clear whether they should be considered a necessary condition for periphrastic status (Haspelmath 2000:661). That the identification of periphrastic constructions in individual languages is by no means self-evident is well illustrated by Ancient Greek (1) constructions consisting of a finite verb and a participle, which form the topic of this paper. While Porter (1989) only accepts constructions with the verb [TEXT NOT REPRODUCIBLE IN ASCII] as periphrastic, other authors such as Dietrich (1973/1983) mention a large number of 'periphrastic' constructions with finite verbs such as [TEXT NOT REPRODUCIBLE IN ASCII]. In fact, some twenty-seven constructions occurring in Ancient Greek have been considered periphrastic by one or more authors. An overview of these is given in Table 1. As a result, there is a feeling of confusion and arbitrariness in the secondary literature, succinctly worded by Campbell (2008:32) as follows: 'verbal periphrasis in Ancient Greek is a problematic issue' (cf. similarly, Porter 1989:452, Adrados 1992:451, Evans 2001:221 and Rijksbaron 2006:127).

One important insight of recent years has been to relate the phenomenon of verbal periphrasis to the concept of grammaticalisation (3) (cf. Wischer 2008 for a recent treatment). Haspelmath (2000:661), among others, stresses that 'we need a comprehensive theory of grammaticalisation in order to understand periphrasis.' Another way to put it is to say that periphrasis involves 'auxiliaries' (cf. Markopoulos 2009:12), which Heine (1993:12) defines as 'a linguistic item covering some range of uses along the Verb-to-TAM chain.' As Haspelmath (2000:661) notes, this perspective makes it much easier to define periphrasis--'the more grammaticalised a construction is, the more it can claim to have periphrastic status'--although its precise identification in individual languages remains problematic.

What I would like to propose here is to consider verbal periphrasis as a 'prototypically organised category' (cf. a.o. Langacker 1987:16-19, Givon 1989; Taylor 1998, 2003 and Cruse 2011:57-67 for the 'prototype model'). Such a view is closely related to the grammaticalisation-perspective, but is not explicitly diachronically oriented (though readily allowing diachronic observations). The prototype model in general has been opposed to what Langacker (1987:16) calls the 'criterial-attribute model', 'which characterizes a class by means of a list of defining features.' With the latter model, each category member of a given category has an equal status, and the category has clear-cut boundaries (cf. Taylor 1998: 179). The prototype model, on the other hand, acknowledges that a given category has both central, 'prototypical' members and more peripheral ones, and that there are not always clear-cut boundaries between categories (as Cruse 2011:60 notes, category boundaries may be 'fuzzy'). In illustration, consider the category of FURNITURE: 'chair', 'sofa' and 'table' constitute prototypical members, while objects such as 'clock', 'vase' and 'telephone' are much less representative. This can be extended to linguistic concepts, such as TRANSITIVE CONSTRUCTION. Compare, for example, the prototypically transitive 'the child kicked the ball' with the more peripheral 'he swam the Channel', where 'Channel' is a path rather than a patient (cf. Taylor 2003:231-39). I strongly believe the prototype model is much more suitable for the analysis of verbal periphrasis than the criterial-attribute model (of which Porter 1989 may be considered an exponent; cf. Porter 1989:452-53 for a list of necessary criteria of periphrasis in Ancient Greek). In fact, when adopting such a view, one can consider the various criteria proposed for identification of verbal periphrasis--going from the three proposed by Haspelmath (2000) to the list of ten criteria by Bertinetto (1990)--as semantic, morphological, syntactic and paradigmatic dimensions along which prototypical periphrastic constructions are identified.

In this paper, I give an--inevitably rough--image of how this prototypically organised category looks like in Ancient Greek, by applying a number of recognised criteria of periphrasticity to a corpus of examples. More specifically, I discuss the criteria of 'tempo-aspectual relevance', 'conceptual integration', 'syntactic contiguity', 'clitic climbing', 'paradigmaticity' and 'restricted paradigmatic variability', for which I base myself on the studies of Bertinetto (1990), Haspelmath (2000), Ackermann & Stump (2004) and Langacker (2005). I do not explicitly discuss any morphological criteria here, because they do not help us to distinguish periphrastic from non-periphrastic constructions, with regard to Ancient Greek at least, and thus raise more questions than they solve. My corpus covers all the examples given by the major studies on Ancient Greek participial periphrasis (Kontos 1898; Harry 1905; Stahl 1907; Rosen 1957; Aerts 1965; Dietrich 1973/1983), amounting to a total number of about 1700. Taken together, these studies comprise a large part of Ancient Greek literature, both prose and poetry. (4) While their main focus is on Classical Greek, they also mention a limited number of archaic examples (mostly Homeric), which have been included in the present study. With regard to the constructions with [TEXT NOT REPRODUCIBLE IN ASCII] I have taken account of the study of Wheeler (1891), who does not explicitly call his examples periphrastic. (5)

2. Criteria of periphrasticity and the analysis of Ancient Greek

2.1 Semantic criteria

Many definitions of verbal periphrasis attach great importance to semantic criteria. A first criterion is that of 'tempo-aspectual relevance' (Bertinetto 1990:334), which most commonly points to the fact that a periphrastic construction should express a 'grammatical' meaning. Bybee, Perkins & Pagliuca (1994:660), for example, observe that periphrastic constructions often 'develop to express meanings that are more specific than the meanings already expressed grammatically in the language at the time.' The real difficulty is, of course, how to define a 'grammatical meaning' (cf. Haspelmath 2000:660). Here, some approaches stress the importance of the synthetic paradigm: for Ackermann & Stump (2004:128) we can speak of periphrasis when a construction 'expresses grammatical properties that are expressed elsewhere in the synthetic paradigm' (cf. Spencer 2006:292; this is called 'feature intersectivity'). Similarly, for Evans (2001:222) a periphrastic construction is either suppletive or 'more or less equivalent to an existing synthetic form.' For Ancient Greek, this criterion helps us to distinguish between periphrastic complementation and regular participial complementation, as in examples (1) and (2) below. While in (1) [TEXT NOT REPRODUCIBLE IN ASCII] more or less functions as an alternative (6) to the synthetic form [TEXT NOT REPRODUCIBLE IN ASCII] (cf. Stahl 1907:145), both of which mean 'he is speaking'; [TEXT NOT REPRODUCIBLE IN ASCII] in (2) certainly does not function in the same way.

(1) [TEXT NOT REPRODUCIBLE IN ASCII] (Pl. Hp. mai. 286b). (7)

So after that we have Nestor speaking.

(2) [TEXT NOT REPRODUCIBLE IN ASCII] (Xen. An. 5.5.24).

For we see that all you say is true.

While allowing to make this basic distinction between type of complementation, the criterion of tempo-aspectual relevance does raise some further questions. While in (1) the construction of [TEXT NOT REPRODUCIBLE IN ASCII] with the present participle may reasonably be considered a close equivalent to a synthetic form, it is not clear whether the same goes for the verbal form found in (3), [TEXT NOT REPRODUCIBLE IN ASCII], 'they continued lavishing...', which expresses imperfective ('continuative') aspect, a grammatical property which also occurs with synthetic forms. Several authors, among whom Jannaris (1897:490), Dietrich (1973/1983:243-45) and Adrados (1992: 453), hold the opinion that the constructions of the verbs [TEXT REPRODUCIBLE IN ASCII] with present participle, all three meaning 'to continue V-ing', should be considered periphrastic.

(3) [TEXT NOT REPRODUCIBLE IN ASCII] (Dem. Adv. Lept. 10). ... but they continued lavishing even their own fortunes.

A second criterion, called 'conceptual integration' (I use this term after Langacker 2005; Bertinetto 1990:332 uses 'integrazione semantica') brings more clarity. In both (1) and (3) there is a strong semantic bond between the finite verb and the participle, as they are co-temporal and co-referential (cf. Givon 2001, ch. 12 for an in-depth discussion of the semantic dimensions of event integration). We may contrast them with (2), where the finite verb and the participle clearly are not co-referential (the finite verb optuev has an unexpressed subject 'we' and the participle [TEXT NOT REPRODUCIBLE IN ASCII] an expressed subject [TEXT NOT REPRODUCIBLE IN ASCII] 'everything'). Clearly, however, there is a semantic difference between the finite verbs [TEXT NOT REPRODUCIBLE IN ASCII] in (1) and [TEXT NOT REPRODUCIBLE IN ASCII] in (3) when combined with a present participle: while the latter keeps its full lexical value ('to continue') the former does not (in casu 'to exist'). We may call this a difference in degree of 'schematicity' (or 'generality') of the finite verb: in (1) the semantic contribution of the finite verb is much more 'invisible' (cf. Langacker 2005:180) than in (3) (Bertinetto 1990:333 speaks of 'desemantizzazione dei modificatori' and Lehmann 1995[1982]: 127 of 'loss of semantic integrity').

Let us extend the discussion beyond [TEXT NOT REPRODUCIBLE IN ASCII], and have a look at the schematicity of the other finite verbs listed in Table 1. It is important to note here that I analyse schematicity in terms of lexical and not aspectual semantics. The latter was suggested by Porter (1989:452). In his opinion, only [TEXT NOT REPRODUCIBLE IN ASCII] builds periphrastic constructions, because it is the only finite verb which can be called 'aspectually vague', as it does not morphologically distinguish between perfective and imperfective aspect. For the other participial constructions, Porter uses the term 'catenative constructions', since 'the auxiliary inherently maintains its integrity as an independent contributor to the semantics of the clause' (Porter 1989: 487). (8) An example such as (4), where the construction of [TEXT NOT REPRODUCIBLE IN ASCII] with an aorist participle ([TEXT NOT REPRODUCIBLE IN ASCII]) is used (the so-called [TEXT NOT REPRODUCIBLE IN ASCII]), does not qualify as periphrastic (contra Aerts 1965, Drinka 2003), because there are other examples (Porter mentions Her. Hist. 1.75.1) where the verb exto is (quite exceptionally) used in the aorist, 'making it not only difficult to grasp the relation between the auxiliary and periphrastic but to see in what sense the auxiliary is ... "weakened" when a marked form predominates' (Porter 1989:490).

(4) [TEXT NOT REPRODUCIBLE IN ASCII] (Soph. El. 589-90).

You have cast out the earlier born, the pious offspring of a pious marriage.

Porter's view entails some serious methodological problems, as Evans (2001:222) notes: 'it lacks diachronic scope and yields an artificially narrow definition of periphrasis.' Indeed, it does not allow the grammaticalisation of other periphrastic constructions (an essential insight, as I noted above), and restricts the notion of periphrasis to the verb [TEXT NOT REPRODUCIBLE IN ASCII] based entirely on the notion of aspectual vagueness. Evans (2001:223), on the other hand, proposes the approach I follow here, namely that it is the lexical and not aspectual semantics of the verb that decide whether it can be used as an 'auxiliary' in a periphrastic construction (cf. also Bybee, Perkins & Pagliuca 1994:5).

In their landmark study, Bybee, Perkins & Pagliuca (1994:289) analyse desemanticisation in terms of 'generalisation', which they define as 'the loss of specific features of meaning with the consequent expansion of appropriate contexts of use.' In this paper, I take into account two specific indicators of generalisation: (a) compatibility of the finite verb with inanimate subjects, and (b) compatibility of the finite verb with participles of content verbs belonging to different aspectual classes. With regard to the latter, I make use of the well-known Vendlerian classification of lexical aspect or Aktionsart, which distinguishes between verbs of State, Activity, Accomplishment and Achievement (Vendler 1957). (9) One concomitant factor I will take into account is frequency. The importance of frequency or 'repetition' with regard to grammaticalisation has been argued for by Bybee (2006) among others. It should be noted, however, that not all scholarly works on which my corpus is based strove for exhaustivity, so that the numbers mentioned here can only give a general indication of frequency.

I divide the constructions under analysis into two groups: those with verbs of movement and those with verbs of state as finite verb. Constructions with the verbs [TEXT NOT REPRODUCIBLE IN ASCII] are not further discussed.

In Table 2, I present the constructions with verbs of movement. As can be seen, finite verbs of movement most frequently occur with a future participle. This concerns the verbs [TEXT NOT REPRODUCIBLE IN ASCII]. Although these verbs are used with participles whose content verbs belong to different classes of Aktionsart, they do not combine with inanimate subjects, so they cannot be considered fully 'generalised' or 'schematic'.

The constructions of [TEXT NOT REPRODUCIBLE IN ASCII] with future participle seem to be developing in Ancient Greek, as the study of Letoublon (1982) points out. This scholar indicates that Herodotus always uses the construction with a verb of saying, as in (5). She concludes that Herodotus' use of the construction should be considered 'metaphorical' rather than truly periphrastic. Metaphorical use is, of course, well-known as a mechanism of semantic change from lexical to grammatical meaning (cf. Bybee, Perkins & Pagliuca 1994:285). This is what Letoublon suggests for Plato, where she finds the participle expanded to other lexical types, as for example in (6), where the verb [TEXT NOT REPRODUCIBLE IN ASCII] 'I pay' is used.

(5) [TEXT NOT REPRODUCIBLE IN ASCII] (Her. Hist. 1.194.1).

I will now show what seems to me to be the most marvellous thing in the country, next to the city itself.

(6) [TEXT NOT REPRODUCIBLE IN ASCII] (Pl. Prt. 311e).

C'est donc en tant que sophiste que nous allons le payer? (tr. Letoublon 1982).

The construction of [TEXT NOT REPRODUCIBLE IN ASCII] with future participle is quite similar to the construction with [TEXT NOT REPRODUCIBLE IN ASCII], in that it occurs with roughly the same frequency and quite often with verbs of saying. It is, however, much more difficult, if at all possible, to find examples which are clearly periphrastic. Example (7), which Dietrich (1973/1983:274) translates with 'je vais l'expliquer ...' ('I am going to tell you ...') can just as well be rendered with 'I have come to tell you ...' In fact, the latter reading seems much more plausible, as Creon, who utters this sentence, has just entered the scene.

(7) [TEXT NOT REPRODUCIBLE IN ASCII] (Eur. Phoen. 706). I have come to tell you what is most immediate.

Constructions consisting of a verb of movement with a present participle occur much less frequently. Dietrich (1973/1983) discusses some, mostly Homeric, examples with [TEXT NOT REPRODUCIBLE IN ASCII], where the finite verb maintains a strong lexical sense (cf. Dietrich 1973/1983:237, 240). As can be seen in Table 2, these verbs combine with participles whose content verbs are restricted to one class of Aktionsart. The construction of [TEXT NOT REPRODUCIBLE IN ASCII] occurs in combination with the verbs [TEXT NOT REPRODUCIBLE IN ASCII], with the sense of 'I carry/take away', as in (8). (11) The single example with [TEXT NOT REPRODUCIBLE IN ASCII] cited by Dietrich, our example (9), is quite problematic, not only because it is not clear whether [TEXT NOT REPRODUCIBLE IN ASCII] should be considered a verb of movement, (12) as suggested by Dietrich (1973/1983:237: 'es kann sich hier lediglich um eine Variante der Periphrase mit [TEXT NOT REPRODUCIBLE IN ASCII] + Part. Pras. handeln'), (13) but also because the finite verb might well maintain its lexical value, as suggested by the translation provided here. Constructions with the finite verbs [TEXT NOT REPRODUCIBLE IN ASCII] and [TEXT NOT REPRODUCIBLE IN ASCII] with a present participle are much more clearly periphrastic. Note that [TEXT NOT REPRODUCIBLE IN ASCII] in example (10) takes an inanimate subject.

(8) [TEXT NOT REPRODUCIBLE IN ASCII] (Hom. Il. 1.391).

The heralds have just now gone and taken away the other woman from my tent.

(9) [TEXT NOT REPRODUCIBLE IN ASCII] (Hom. Il. 5.511).

For she it was who was bringing aid to the Danaans.

(10) [TEXT NOT REPRODUCIBLE IN ASCII] (Pl. Cra. 421b).

For once more that which is held back and forced to be quiet is found fault with.

My findings for constructions with verbs of state are presented in Table

3. As can be seen, many constructions occur infrequently: [TEXT NOT REPRODUCIBLE IN ASCII] with present, aorist and perfect participle; ext with present and perfect participle; [TEXT NOT REPRODUCIBLE IN ASCII] with present participle; [TEXT NOT REPRODUCIBLE IN ASCII] with perfect participle and [TEXT NOT REPRODUCIBLE IN ASCII] with present and perfect participle. Not surprisingly, in these constructions the content verbs are mostly restricted to one or two different classes of Aktionsart (with the exception of [TEXT NOT REPRODUCIBLE IN ASCII] with aorist participle and [TEXT NOT REPRODUCIBLE IN ASCII] with present participle). Moreover, there are no examples attested with an inanimate subject.

Other constructions are used much more frequently, among others [TEXT NOT REPRODUCIBLE IN ASCII] with present, perfect and aorist participle (the construction with aorist participle clearly being least frequent), and [TEXT NOT REPRODUCIBLE IN ASCII] with aorist participle (for an example of this construction, see [4]). These constructions do occur with inanimate subjects, and with content verbs belonging to four different classes of Aktionsart.

The constructions with [TEXT NOT REPRODUCIBLE IN ASCII] also occur very frequently, especially those with the present and perfect participle (in fact, the numbers presented here are based on a selection of the examples mentioned by Wheeler 1891, cf. note 5). As can be seen in Table 3, they occur with inanimate subjects and combine with participles whose content verbs belong to all four types of Aktionsart. These elements thus seem to attest to a high degree of generalisation.

There has been, however, and there still is, discussion with regard to the periphrasticity of the constructions with [TEXT NOT REPRODUCIBLE IN ASCII]. While scholars such as Rosen (1957:140) and Adrados (1992:452) argue that they are periphrastic, others such as Bjorck (1940:64) stress the notion of 'chance' or 'coincidence' inherent in the finite verb, which is often rendered in English by the phrase 'I happen to... In the latter view, an example such as (11) cannot be considered periphrastic, because the finite verb is not desemanticised.

(11) [TEXT NOT REPRODUCIBLE IN ASCII] (Xen. Hell. 1.6.27).

On the same day it chanced that the Athenians took dinner on the Arginusae islands.

In accordance with the latter view, we could analyse [TEXT NOT REPRODUCIBLE IN ASCII] as an epistemic modal auxiliary (15) (cf. Nuyts 2001, ch. 4 for a cognitive analysis of modal auxiliaries [based on Dutch, German and English]) and exclude it on this basis from our present discussion. In my opinion, this does not do justice to the complexity of the matter.

Firstly, it should be noted that a notion of 'coincidence' is very much context-dependant. (16) Often, it is not necessary, and sometimes contextually irrelevant, for example when there is a strong emphasis on the event denoted by the participle. In (12), the speaker has little reason to give his lifelong good conduct a nuance of coincidence or chance. Moreover, there are examples which show that the constructions are used in contexts very similar to those with [TEXT NOT REPRODUCIBLE IN ASCII] etui: in (13) and (14) Isocrates uses both finite verbs with [TEXT NOT REPRODUCIBLE IN ASCII], followed by a consecutive clause.

(12) [TEXT NOT REPRODUCIBLE IN ASCII] ... (Isoc. Antid. 322).

All my past life up to this day I have lived in a manner that Befits ...

(13) [TEXT NOT REPRODUCIBLE IN ASCII] ... (Isoc. Phil. 125).

We have dropped so far behind the barbarians that, while they did not hesitate ...

(14) [TEXT NOT REPRODUCIBLE IN ASCII] (Isoc. Panath. 209).

The Lacedaemonians have fallen so far behind our common culture and learning that they do not even try to instruct themselves in letters.

Further research is much needed here, especially from a diachronic point of view. Interestingly, Ljungvik (1926:45) notes that in Post-classical Greek [TEXT NOT REPRODUCIBLE IN ASCII] hat, wie es scheint, die Bedeutung der Zufalligkeit u. dgl. fast ganz eingebusst und wurde wohl fast nur als ein volleres 'sein' empfunden' (cf. also Rydbeck 1969:193), though he does not cite any examples where [TEXT NOT REPRODUCIBLE IN ASCII] is used with a participle. A preliminary analysis of the forms in my corpus with the most frequently occurring construction, that with the present participle, indeed shows signs of an evolution in degree of generalisation. While Thucydides, Lysias and especially Plato have a marked preference for the combination with stative content verbs, the percentages in Isocrates are spread more equally. Especially participles of activity verbs are used more frequently.

2.2 Syntactic criteria

Certainly the most prominent syntactic criterion is that of 'contiguity', which is designated by some as 'cohesion' (Lehmann 1995[1982]:147) or with the Italian term 'compatezza' (Bertinetto 1990:339). Scholars discussing this criterion generally stress the iconic nature of constituent structure: in general, two linguistic elements which are semantically close, are syntactically contiguous. Givon (2001:64) calls this 'the proximity principle'. In fact, in many languages auxiliaries and their complements cannot be separated (Wakker 2006:243).

What about periphrastic constructions in Ancient Greek? According to Porter (1999:45-46) 'no elements may intervene between the auxiliary verb and the participle except for those which complete or directly modify the participle.' This rule has been criticised, however. According to Evans (2001:232) it is 'entirely artificial and ignores the natural flexibility of word order.' When we look at some examples, we see that Evans is quite right: they show that various elements can 'intervene' (to use Porter's terminology), such as the subject in (15) and the comparative genitive in (16). Porter's rule is problematic, as it does not take into account the fact that word order is influenced by complex pragmatic factors (cf. Matic 2003). Devine & Stephens (2000:132) mention (17) as an example of the interaction of auxiliary and participle with so-called 'modifier hyperbaton', whereby [TEXT NOT REPRODUCIBLE IN ASCII] represents a weak focus and [TEXT NOT REPRODUCIBLE IN ASCII] a second weak focus.

(15) [TEXT NOT REPRODUCIBLE IN ASCII] (Isae. De Cir. 23).

Immediately afterwards he casually remarked that Ciron had left nothing at all.

(16) ... [TEXT NOT REPRODUCIBLE IN ASCII] (Isoc. Nicocl. 40).

... when they ought to cherish this relationship the more Faithfully inasmuch as it is more intimate and more precious than all others.

(17) [TEXT NOT REPRODUCIBLE IN ASCII] (Lys. De Vuln. 19).

Nothing of this kind have I ever done.

Despite the fact that contiguity does not seem to be absolutely necessary in Ancient Greek, we can discern clear differences between the constructions listed in Table 1 with regard to their syntactic contiguity. In Table 5, I present some figures for the most frequently occurring constructions. With regard to the constructions occurring less frequently, the criterion of syntactic contiguity has little relevance. The construction of [TEXT NOT REPRODUCIBLE IN ASCII] with present participle, for example, which has only one occurrence in my corpus, is one hundred percent contiguous, which obviously does not entail that it is fully grammaticalised.

Although there are several factors to be taken into account, such as the fact that the corpus consists of both prose and poetry, we can say that the results presented in Table 5 more or less correspond to the semantic observations made earlier on: with verbs of state, the constructions of [TEXT NOT REPRODUCIBLE IN ASCII] with aorist participle and [TEXT NOT REPRODUCIBLE IN ASCII] with perfect participle are at the top, while for verbs of movement the constructions of [TEXT NOT REPRODUCIBLE IN ASCII] with future participle are. Noticeable is also the relatively high position of the constructions of [TEXT NOT REPRODUCIBLE IN ASCII] with perfect and present participle, and the clear difference with the construction with aorist participle. Somewhat surprising is the construction at the bottom of this list: [TEXT NOT REPRODUCIBLE IN ASCII] with present participle. This position may be due to a particular use, namely that where the participle has an 'adjectival' function, as illustrated in (18), where we see the large distance between the component parts 'Hv... [TEXT NOT REPRODUCIBLE IN ASCII]. (17)

(18) [TEXT NOT REPRODUCIBLE IN ASCII] (Xen. An. 2.2.13).

This plan of campaign meant nothing else than effecting an escape, either by stealth or by speed.

A second syntactic criterion often mentioned in discussions of verbal periphrasis is so-called 'clitic climbing', whereby a clitic moves from the participial complement to which it belongs up to the finite verb, as illustrated with an example from Old Spanish (cf. Torres Cacoullos 2000:47) in (19), where we see 'lo fueron diziendo' instead of 'fueron diziendolo'. According to Myhill (1988) the phenomenon of clitic climbing is directly related to the desemanticisation of the finite verb.

(19) ... e assi lo fueron diziendo.

... and soon everyone was saying it.

What about Ancient Greek? My corpus contains about fifty examples of clitic pronouns accompanying the following constructions: [TEXT NOT REPRODUCIBLE IN ASCII] with aorist participle; [TEXT NOT REPRODUCIBLE IN ASCII] with present, perfect and aorist participle; [TEXT NOT REPRODUCIBLE IN ASCII] with present and future participle; [TEXT NOT REPRODUCIBLE IN ASCII] with future participle; [TEXT NOT REPRODUCIBLE IN ASCII] with aorist and perfect participle; [TEXT NOT REPRODUCIBLE IN ASCII] with future participle and [TEXT NOT REPRODUCIBLE IN ASCII] with present and perfect participle. The large majority of these clitics appears in so-called 'Wackernagel-position', i.e. in clause-second position, whether or not in combination with one or more discourse particles (cf. Janse 2008:172-73). Overall, there are only a few examples which qualify for clitic climbing. In these examples, the clitic comes in between the finite and the non-finite verb (in other words, the clitic pronoun is not fully preposed; Torres Cacoullos 2000:47 calls this the 'midway position'). Next to the verb [TEXT NOT REPRODUCIBLE IN ASCII] with perfect participle in (20), we notice [TEXT NOT REPRODUCIBLE IN ASCII] with present and perfect participle in (21) and (22). The appearance of the latter verb is, of course, very interesting in light of the discussion concerning its grammatical status, more particularly its schematicity (cf. supra).

(20) [TEXT NOT REPRODUCIBLE IN ASCII] (Her. Hist. 3.90.1).

For one joint tribute was laid on them by him.

(21) ... [TEXT NOT REPRODUCIBLE IN ASCII] (Pl. Grg. 453c).

... suppose I happened to ask you what Zeuxis was among painters.

(22) [TEXT NOT REPRODUCIBLE IN ASCII]... (Isoc. Ad Phil. 1).

And if I had not on a former occasion given you with most kindly intent such counsel...

Overall, however, these few examples are quite problematic. It is quite likely that their clitic placement has been influenced by pragmatic and phonological factors (cf. Janse 2008:173, who stresses the fact that clitic placement in Ancient Greek is 'not a matter of syntax, but of [discourse] phonology'). This raises a number of questions which cannot be answered in the present paper: Are we dealing with Wackernagel-clitics at the subclausal level? Could the verb [TEXT NOT REPRODUCIBLE IN ASCII] be used in a V2 (verb second) position? Has the finite verb been fronted for pragmatic reasons?. Again, the main point is that we should not ignore the word order flexibility of Ancient Greek.

2.3 Paradigmatic criteria

The first paradigmatic criterion I discuss is that of 'paradigmaticity'. This criterion, which points at the fact that a construction is integrated in the inflectional paradigm and as such is obligatory, plays an important role in most discussions of periphrasis. Matthews (1981:55), for example, the author of a textbook on syntax proposes the following definition: 'when a form in a paradigm consists of two or more words it is periphrastic', which seems to imply that periphrastic constructions can only occur within the paradigm (cf. Lehmann 1995[1982]:135).

In the case of Ancient Greek, very few constructions comply with this criterion. There is the well-known case of [TEXT NOT REPRODUCIBLE IN ASCII] with perfect participle. The reference grammars state that the construction is suppletive in the third person of the medio-passive indicative perfect and pluperfect of verbs ending in a stop, and the medio-passive subjunctive and optative perfect. The future form of the finite verb is also commonly used to circumscribe the active future perfect. We may note, moreover, that the construction has spread through the paradigm: my corpus shows that it is also used with verbs ending in a vowel, in the active voice, outside the indicative, subjunctive or optative mood, and not exclusively with the third person.

Less well-known is the fact that the construction of [TEXT NOT REPRODUCIBLE IN ASCII] with aorist participle was often used for forms which did not have an active synthetic perfect (Drinka 2003:111), and in these cases should be considered suppletive. With regard to Sophocles, for example, Aerts (1965:131-40) mentions the use of [TEXT NOT REPRODUCIBLE IN ASCII] with verbs such as [TEXT NOT REPRODUCIBLE IN ASCII] (Ant. 21-2), A [TEXT NOT REPRODUCIBLE IN ASCII] (OT 731) and [TEXT NOT REPRODUCIBLE IN ASCII] (Trach. 412), for which no attested synthetic perfect forms can be found (at the time of Sophocles at least). The forms of this construction did not spread through the paradigm as those of [TEXT NOT REPRODUCIBLE IN ASCII] with perfect participle. Its use is mainly limited to the singular forms of the present indicative, which represent 74 percent of the examples. (18) It is occasionally used in the infinitive and imperative moods. For the subjunctive, optative and participle moods there are only a few examples, and it is not quite clear whether they should be interpreted periphrastically. Various scholars point to the fact that the verb [TEXT NOT REPRODUCIBLE IN ASCII] is also used with participles of verbs which did have a synthetic perfect, such as [TEXT NOT REPRODUCIBLE IN ASCII] (cf. ex. [4]; Rijksbaron 2006:130).

As I have already mentioned, scholars also recognise 'non-paradigmatic' periphrases. Haspelmath (2000:660-61) recognises, next to 'suppletive periphrasis', so-called 'categorial periphrasis' for constructions which do not replace any synthetic forms. He gives the examples of 'je vais chanter' in French and 'estoy cantando' in Spanish. Such forms are related to the paradigm, as they are felt to be 'roughly equivalent' to synthetic forms. In the introduction, I have given the similar example of [TEXT NOT REPRODUCIBLE IN ASCII] (Lc. 5.17) next to [TEXT NOT REPRODUCIBLE IN ASCII] (Lc. 5.3) and in the discussion of the semantic criteria I compared [TEXT NOT REPRODUCIBLE IN ASCII] (Pl. Hp. Mai. 286b) to the synthetic form [TEXT NOT REPRODUCIBLE IN ASCII] (cf. Stahl 1907:145). Here we are on the borderline between the semantic and paradigmatic criteria: constructions which have tempo-aspectual relevance resemble synthetic forms.

A second paradigmatic criterion is that of restricted paradigmatic variability (Lehmann 1995[1982]:138), which means that the number of constructions expressing a similar aspecto-temporal meaning should be restricted. As Dietrich (1973/1983) shows throughout his work, this is not so much the case in Ancient Greek. In Table 6, I have grouped semantically similar constructions, mainly based on Dietrich's observations. Dietrich uses the German terms 'Winkelschau' for Group 2, 'sekundare parallel-prospektive Perspektive' for Group 4, 'retrospektive, prospektive und komitative Schau' for Group 5, and 'kontinuative Schau' for Group 6. He does not discuss our Groups 1 and 3, which may be characterised as realising perfect and perfective aspect respectively. Dietrich furthermore makes the important observation that the number of variants seriously diminishes in Post-classical Greek, suggesting further grammaticalisation of the remaining constructions (Dietrich 1973/1983:279).

3. The category of verbal periphrasis in Ancient Greek

As will be clear by now, the criteria I have applied to the corpus of Ancient Greek examples do not have the same status (cf. Bertinetto 1990:342-48 and Cruse 2011:60). Some criteria are of limited relevance, because they pertain to all of the constructions listed in Table 1 (the criteria of 'tempo-aspectual relevance' and 'restricted paradigmatic variability') and therefore do not help us to distinguish between constructions, or because they are theoretically problematic ('clitic climbing'). Other criteria, however, do help us to distinguish between constructions with regard to their periphrasticity, most importantly the criteria of 'conceptual integration', 'syntactic contiguity' and 'paradigmaticity'. Next to these three criteria, frequency has been an important factor throughout (cf. Givon 1989:40-43 for frequency and categorisation).

In an attempt to clarify the make-up of the category of verbal periphrasis in Ancient Greek, I would like to propose the division represented in Table 7. I distinguish between four groups, which form a gradation from central 'prototypical' members to more peripheral ones (cf. Langacker 1987:17). The first group consists of constructions which show a high degree of conceptual integration, which are most often syntactically contiguous (more than 70% of the cases, cf. supra) and are paradigmatically integrated. Moreover, they occur frequently, especially the construction of [TEXT NOT REPRODUCIBLE IN ASCII] with perfect participle. Constructions of the second group comply to a much lesser degree with the proposed criteria. They show signs of a lesser degree of conceptual integration (as indicated by the fact that various constructions only occur with an animate subject), they are less often syntactically contiguous than those of the first group, and they are not paradigmatically integrated. While some constructions occur frequently, especially that of [TEXT NOT REPRODUCIBLE IN ASCII] with present participle, others do not. The constructions of [TEXT NOT REPRODUCIBLE IN ASCII] with present participle and [TEXT NOT REPRODUCIBLE IN ASCII] with future participle form the transition to the third group: the former because it is quite infrequent (with only five instances in the corpus) and the latter because it is almost always ambiguous. The third group contains constructions which are conceptually integrated only to some degree (they do not take inanimate subjects and their participles are formed with content verbs limited to only one or two classes of Aktionsart) and which are not paradigmatically integrated. These constructions are characterised by the fact that they occur infrequently, which is why it is hard to discuss their syntactic contiguity. It would seem that they can be considered expressive alternatives ('exploratory expressions', as Harris & Campbell 1995:72-75 would say), only occasionally used by a select group of authors (in prose mostly Demosthenes and Plato). Similarly, constructions of the fourth group are situated at the periphery of the category of verbal periphrasis. As finite verbs, [TEXT NOT REPRODUCIBLE IN ASCII] distinguish themselves by the fact that they are least schematic, as they fully retain their lexical value, viz. 'to continue'. (20) Givon (2001, ch. 12) uses the term 'implicative modality verbs' for their English counterparts. He puts verbs such as these at the right end of a complementation scale, which explains why they are interpreted periphrastically by some.

The discussion in this paper has mainly been synchronically oriented. As I noted above, however, the prototype model is not incompatible with diachronic considerations, which I have made with regard to the generalisation of the constructions of [TEXT NOT REPRODUCIBLE IN ASCII] with future participle and [TEXT NOT REPRODUCIBLE IN ASCII] with present participle. From this perspective, one could call the constructions of the first group 'grammaticalised', those of the second group 'grammaticalising' and those of the third group 'non-grammaticalised'. Constructions of the first group are formed with prototypical auxiliaries, those of the second group with so-called 'quasi-auxiliaries' (Heine 1993) or 'semi-auxiliaries' (Wakker 2006).

As I hope to have shown, Ancient Greek had a large number of periphrastic constructions at its disposal. It is important to note, however, the difference with periphrastics in the modern languages (e.g. the English have-perfect mentioned in the introduction). Binnick (1991:32) words this crucial insight as follows: 'the classical languages had complex [periphrastic, KB] forms, but the use of complex forms was sporadic and played no essential role in the grammatical systems, so that the periphrastics of the classical languages parallel those of the modern languages little in regard either to form or to content.' The constructions of [TEXT NOT REPRODUCIBLE IN ASCII] with perfect participle and [TEXT NOT REPRODUCIBLE IN ASCII] with aorist participle, which I have characterised here as most prototypical, did play some role in the Ancient Greek verbal system, albeit a small one. As Binnick notes (1991:35): 'these constructions are neither systematic nor obligatory, but merely supplement the regular temporal devices of the language.'

I would like to close this paper by pointing out the main advantages of a prototype analysis of verbal periphrasis in Ancient Greek: (a) It takes into account the fact that each construction occupies a position along a scale and is 'always on the move'. We have seen the examples of [TEXT NOT REPRODUCIBLE IN ASCII] with future participle or [TEXT NOT REPRODUCIBLE IN ASCII] with present participle, among others, showing signs of grammaticalisation in Ancient Greek. (b) It anticipates the fact that the category may be 're-shaped' in Post-classical Greek, with some members becoming more central, and others more marginal. (c) It is in line with research on other languages, where similar observations have been made (e.g. Giacalone Ramat 2001). (d) It explains the considerable amount of confusion in earlier publications on verbal periphrasis in Ancient Greek. In summary, I believe my proposal improves upon previous analyses--which have not bothered to define this grammatical category at all, or have defined it in terms of a list of necessary (though partially artificial) criteria--by providing a flexible but theoretically well-founded approach.

Bibliography

Ackermann, F. & Stump, G.T. 2004. 'Paradigms and periphrastic expression: a study in realization-based lexicalism.' In L. Sadler & A. Spencer (edd.), Projecting Morphology, 111-158. Stanford: CSLI Publications.

Adrados, F. 1992. Nueva sintaxis del griego antiguo. Madrid: Gredos.

Aerts, W.J. 1965. Periphrastica. Diss. Amsterdam. Amsterdam: Hakkert.

Beekes, R.S. & Van Beek, L. 2010. Etymological Dictionary of Greek. Leiden: Brill.

Bentein, K. forthcoming. 'Adjectival periphrasis in Ancient Greek.'

Bertinetto, P.M. 1990. 'Perifrasi verbali italiane, criteri di identificazione e gerarchie di perifrasticita.' In G. Bernini & A. Giacalone Ramat (edd.), La temporalita nell'acquisizione di lingue seconde, 331-350. Milano: Franco Angeli.

Binnick, R. 1991. Time and the Verb: A Guide to Tense and Aspect. Oxford: Oxford University Press.

Bjorck, G. 1940. Hv [TEXT NOT REPRODUCIBLE IN ASCII]. Die periphrastischen Konstruktionen im Griechischen. Diss. Uppsala. Uppsala: Almqvist & Wiksell.

Bybee, J., Perkins, R. & Pagliuca, W. 1994. The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press.

Bybee, J. 2006. 'From usage to grammar: The mind's response to repetition.' Language 82:711-733.

Bybee, J.L. 2010. Language, Usage and Cognition. Cambridge: Cambridge University Press.

Campbell, C. 2008. Verbal Aspect and Non-indicative Verbs. New York: Lang.

Cruse, D.A. 2011 (3rd ed.). Meaning in Language: An Introduction to Semantics and Pragmatics. Oxford: Oxford University Press.

Devine, A.M. & Stephens, L.D. 2000. Discontinuous Syntax: Hyperbaton in Greek. New York & Oxford: Oxford University Press.

Dietrich, W. 1973. Der periphrastische Verbalaspekt in den romanischen Sprachen. Tubingen: Niemeyer. Translated into Spanish (1983): El aspecto verbal perifrdstico en las lenguas romanicas. Madrid: Gredos.

Drinka, B. 2003. 'The formation of periphrastic perfects and passives in Europe: an areal approach.' In B. Blake & K. Burridge (edd.), Historical Linguistics 2001, 105-128. Amsterdam & Philadelphia: Benjamins.

Evans, T. 2001. Verbal Syntax in the Greek Pentateuch: Natural Greek Usage and Hebrew Interference. Oxford: Oxford University Press.

Giacalone Ramat, A. 2001. 'Emergent auxiliaries and the theory of grammaticalization.' In C. Schaner Wolles, J. Rennison & F. Neubarth (edd.), Naturally! Linguistic Studies in Honourof Wolfgang Dressler, 121-131. Torino: Rosenberg e Sellier.

Givon, T. 1989. Mind, Code and Context: Essays in Pragmatics. Hillsdale N.J.: Erlbaum.

Givon, T. 2001. Syntax: An Introduction, Vol. 2. Amsterdam: Benjamins.

Harris, A.C. & Campbell, L. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press.

Harry, J.E. 1905. 'The perfect subjunctive, optative and imperative in Greek.' CR 19:347-354.

Haspelmath, M. 2000. 'Periphrasis.' In G. Booij, C. Lehmann & J. Mugdan (edd.), Morphology. An International Handbook on Inflection and Word Formation, Vol. 1, 654-664. Berlin: Mouton de Gruyter.

Heine, B. 1993. Auxiliaries. Cognitive Forces and Grammaticalization. New York: Oxford University Press.

Jannaris, A. 1897. An Historical Greek Grammar Chiefly of the Attic Dialect. Hildesheim: Olms.

Janse, M. 2008. 'Clitic doubling from Ancient to Asia Minor Greek.' In D. Kallulli & L. Tasmowski (edd.), Clitic Doubling in the Balkan Languages, 165-202. Philadelphia: John Benjamins.

Joseph, B. 2001. 'Is there such a thing as "grammaticalization"?' Language Sciences 23:163-186.

Kontos, K.S. 1898. [TEXT NOT REPRODUCIBLE IN ASCII] 10:269-306; 307-324.

Kuhner, R. & Gerth, B. 1976 [1898-1904] (3rd ed.). Ausfuhrliche Grammatik der griechischen Sprache. Zweiter Teil: Satzlehre. Hannover: Verlag Hahnsche Buchhandlung.

Langacker, R.W. 1987. Foundations of Cognitive Grammar. Vol. 1. Stanford: Stanford University Press.

Langacker, R.W. 2005. 'Integration, grammaticization, and constructional meaning.' In M. Fried & H.C. Boas (edd.), Grammatical Constructions: Back to the Roots, 157-190. Amsterdam: Benjamins.

Lehmann, C. 1995[1982]. Thoughts on Grammaticalization. Munchen: Lincom Europa.

Letoublon, F. 1982. 'Les verbes de mouvement en grec: De la metaphore a l'auxiliarite?' Glotta 60:178-196.

Ljungvik, H. 1926. Studien zur Sprache der apokryphen Apostelgeschichten. Uppsala: Lundequistska bokhandeln.

Markopoulos, T. 2009. The Future in Greek. From Ancient to Medieval. Oxford: Oxford University Press.

Mateos, J. 1977. El aspecto verbal en el Nuevo Testamento. Madrid: Ed. Cristiandad.

Matic, D. 2003. 'Topic, focus, and discourse structure: Ancient Greek word order.' Studies in Language 27:573-633.

Matthews, P.H. 1981. Syntax. Cambridge: Cambridge University Press.

Myhill, J. 1988. 'The grammaticalization of auxiliaries: Spanish clitic climbing.' Berkeley Linguistics Society 14:352-363.

Newmeyer, F.J. 1998. Language Form and Language Function. Cambridge, Mass.: MIT Press.

Nuyts, J. 2001. Epistemic Modality, Language, and Conceptualization. Amsterdam: Benjamins.

Porter, S.E. 1989. Verbal Aspect in the Greek of the New Testament, with Reference to Tense and Mood. Diss. Sheffield. New York: Peter Lang.

Porter, S.E. 1999 (2nd ed.). Idioms of the Greek New Testament. Sheffield: Sheffield Academic Press.

Pusch, C.D. & Wesch A. 2003. 'Verbalperiphrasen zwischen Grammatik, Lexikon und Pragmatik.' In C.D. Pusch & A. Wesch (edd.), Verbal periphrasen in den (ibero)romanischen Sprachen, 1-10. Hamburg: Buske.

Rijksbaron, A. 2006 (3rd ed.). The Syntax and Semantics of the Verb in Classical Greek. Chicago & London: University of Chicago Press.

Rydbeck, L. 1969. 'Bemerkungen zu Periphrasen mit [TEXT NOT REPRODUCIBLE IN ASCII] + Prasens Partizip bei Herodot und in der Koine.' Glotta 47:186-200.

Rosen, H.B. 1957. 'Die 'zweiten' Tempora des Griechischen: Zum Pradikatsausdruck beim griechischen Verbum.' MH 14:133-154.

Smyth, H.W. 1984 [1920]. Greek Grammar. Cambridge, Mass.: Harvard University Press.

Spencer, A. 2006. 'Periphrasis.' In K. Brown (ed.), Encyclopedia of Language and Linguistics, 287-294. 2nd ed. Amsterdam: Elsevier.

Stahl, J.M. 1907. Kritisch-historische Syntax des griechischen Verbums der klassischen Zeit. Heidelberg: Winter.

Taylor, J.R. 1998. 'Syntactic constructions as prototype categories.' In M. Tomasello (ed.), The New Psychology of Language, 177-202. Mahwah, N.J.: Erlbaum.

Taylor, J.R. 2003. Linguistic Categorization. Oxford: Oxford University Press.

Torres Cacoullos, R. 2000. Grammaticization, Synchronic Variation, and Language Contact: A Study of Spanish Progressive--ndo Constructions. Amsterdam: Benjamins.

Vendler, Z. 1957. 'Verbs and time.' Philosophical Review 66:143-160.

Wakker, G. 2006. 'Future auxiliaries or not?' In E. Crespo, J. de la Villa & A.R. Revuelta (edd.), Word Classes and Related Topics in Ancient Greek, 237-255. Louvain-la-Neuve: Peeters.

Wheeler, J.R. 1891. 'The participial constructions with [TEXT NOT REPRODUCIBLE IN ASCII].' HSPh 2:143-157.

Wischer, I. 2008. 'Grammaticalization of periphrastic constructions.' In E. Verhoeven, S. Skopeteas, Y. Min Shin, Y. Nishina & J. Helmbrecht (edd.), Studies on Grammaticalization, 241-250. Berlin: Mouton de Gruyter.

Klaas Bentein

Ghent University

Klaas.Bentein@UGent.Be

* Parts of this paper were presented at the Patras International Conference of Graduate Students in Linguistics (Patras, 4-6 June 2010). I would like to thank Mark Janse, Gerry Wakker and two anonymous referees of Acta Classica for their valuable comments. My work was funded by the Special Research Fund of Ghent University (grant nr. 01D23409).

(1) I use the term 'Ancient Greek' here to refer to the Archaic (9 BC-6 BC) and Classical periods (5 BC-4 BC).

(2) Since [TEXT NOT REPRODUCIBLE IN ASCII] figure in this table, one could wonder why [TEXT NOT REPRODUCIBLE IN ASCII] do not. Constructions with these verbs have in fact been called periphrastic (e.g. [TEXT NOT REPRODUCIBLE IN ASCII] by Mateos 1977:33), but--to the best of my knowledge--not by scholars commenting on Archaic or Classical Greek. Dietrich 1973/1983 also mentions the verb [TEXT NOT REPRODUCIBLE IN ASCII], but he gives no examples from Archaic or Classical Greek. For reasons of space, I have grouped the constructions according to finite verb, though I believe each combination of a finite verb and a participle should be considered a separate construction (e.g. [TEXT NOT REPRODUCIBLE IN ASCII] + part. pres., perf., aor. constitutes three different constructions).

(3) Note that the phenomenon of grammaticalisation has been been quite forcefully questioned in recent years. See, among others, Newmeyer 1998 and Joseph 2001. Bybee 2010:112-14 gives a good overview of and reply to recent criticisms.

(4) Prose: epistolography: Demosthenes, Isocrates and Plato (dub.); historiography: Thucydides, Theopompus and Xenophon; oratory: Aeschines, Andocides, Antiphon, Demades, Demosthenes, Hyperides, Isaeus, Isocrates, Lysias, Lycurgus; philosophical prose: Aristotle, Plato and Xenophon; technical treatises: Aristotle and Xenophon. Poetry: comedy: Anaxilas, Antiphanes, Aristophanes, Aristophon, Menander, Phrynichus and Plato Comicus; didactic poetry: Hesiod; epic poetry: Homer; hymnography: Homeric hymns; lyric poetry: Archilochus, Callinus, Pindar and Xenophanes; tragedy: Aeschylus, Euripides and Sophocles.

(5) I have concentrated on five randomly chosen prose authors: Isocrates, Lysias, Plato, Thucydides and Xenophon. For Plato, I limited myself to the examples from the Gorgias, Leges, Respublica and Sophista.

(6) One referee draws attention to the fact that [TEXT NOT REPRODUCIBLE IN ASCII] cannot simply be considered alternatives with the same meaning. Indeed, [TEXT NOT REPRODUCIBLE IN ASCII] has a descriptive, stativising, force which the synthetic [TEXT NOT REPRODUCIBLE IN ASCII] does not possess.

(7) The Greek text of the examples is that of the Teubner series. The translations are largely based on the Loeb series, unless otherwise indicated.

(8) Note that Porter 1989:449, 487 uses the term 'auxiliary' to denote the finite verb in both 'catenative' and 'periphrastic' constructions.

(9) These four classes are mostly defined in terms of the features 'dynamicity', 'durativity' and 'telicity' (States: -dynamic +durative -telic; Activities: +dynamic, +durative -telic; Accomplishments: +dynamic +durative +telic; Achievements: +dynamic -durative +telic).

(10) Here as in the other tables, I have included not only clearly periphrastic examples but also those which are ambiguous, in that the finite verb could be interpreted lexically as well. Cf. note 4 for the specific corpus.

(11) Dietrich 1973/1983:240 seems to suggest periphrastic interpretation ('fueron llevando a la joven hija de B.'). As one referee notes, however, [TEXT NOT REPRODUCIBLE IN ASCII] maintain much of their semantic force (as indicated by the English translation).

(12) In their treatment of the etymology of [TEXT NOT REPRODUCIBLE IN ASCII], Beekes & Van Beek 2010:1169 interestingly cite Sanskrit cdrati, 'to move around, wander, drive (on the meadow), graze' and Albanian siell, 'to turn around, turn, bring'.

(13) Cf. Dietrich's Spanish translation: 'pues iba socorriendo a los danayos' (1973/ 1983:237).

(14) Note that the verbs ????? and t?????? diverge from the other verbs listed in this table in that they are not stative when used lexically (in which case they have the meaning of 'to hit').

(15) Note that the same could be argued for with regard to the verb Kupeoo, though this verb occurs much less frequently (cf. Kuhner & Gerth 1976[1904]:63).

(16) Cf. Smyth 1984:467: 'Tuyxavoo often loses the idea of chance, and denotes mere coincidence (I am just now, I was just then) in time or simply I am.'

(17) When etui is combined with a participle that has an adjectival force, the term 'adjectival periphrasis' is commonly used. As I argue elsewhere (Bentein forthcoming), I do not consider 'adjectival' and 'verbal' periphrasis to be mutually exclusive terms, contra Bjorck (1940) and Aerts (1965).

(18) This seems to comply with one of the morphological criteria of periphrasticity proposed by Haspelmath (2000:661), namely 'reduced verbal behaviour'. Remarkably, however, the construction of etui with perfect participle shows exactly the opposite behaviour, as it has spread through the paradigm.

(19) There is some discussion whether this construction expresses perfective or perfect aspect. The latter is most commonly assumed (Aerts 1965:159).

(20) It is worth mentioning--and I owe this point to an anonymous referee--that the meaning of a verb such as [TEXT NOT REPRODUCIBLE IN ASCII], 'to continue' in fact already represents a semantic shift from the original lexical meaning 'to carry across'.

(21) The constructions in each group are ordered on the basis of their frequency in my corpus.

(22) As I noted in [section]2.1, [TEXT NOT REPRODUCIBLE IN ASCII] in combination with a present participle maintain a strong lexical sense. We may be dealing with a construction consisting of a finite (lexical) verb and a conjunct participle here.

Table 1: Participial constructions considered 'periphrastic' (2)

[TEXT NOT REPRODUCIBLE IN ASCII] [TEXT NOT REPRODUCIBLE IN ASCII] +
 + part. pres. part. pres., fut.
[TEXT NOT REPRODUCIBLE IN ASCII] [TEXT NOT REPRODUCIBLE IN ASCII] +
 + part. pres., perf., aor. part. pres., perf., aor.
[TEXT NOT REPRODUCIBLE IN ASCII] [TEXT NOT REPRODUCIBLE IN ASCII] +
 + part. pres. part. pres., fut.
[TEXT NOT REPRODUCIBLE IN ASCII] [TEXT NOT REPRODUCIBLE IN ASCII] +
 + part. pres. part. pres.
[TEXT NOT REPRODUCIBLE IN ASCII] [TEXT NOT REPRODUCIBLE IN ASCII] +
 + part. pres. part. pres., perf.
[TEXT NOT REPRODUCIBLE IN ASCII] [TEXT NOT REPRODUCIBLE IN ASCII] +
 + part. pres., perf., aor. part. pres., perf., aor.
[TEXT NOT REPRODUCIBLE IN ASCII] [TEXT NOT REPRODUCIBLE IN ASCII] +
 + part. pres., fut. part. pres., perf.

Table 2: 'Periphrastic' constructions with verbs of movement

Construction Instances in Different Animate
 corpus (10) types of subject
 Aktionsart only

[TEXT NOT REPRODUCIBLE 4 1 X
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 5 3 X
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 32 4 X
 IN ASCII] + part.fut.
[TEXT NOT REPRODUCIBLE 12 4
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 24 4 X
 IN ASCII] + part.fut.
[TEXT NOT REPRODUCIBLE 16 4
 IN ASCII] + partpres.
[TEXT NOT REPRODUCIBLE 28 3 X
 IN ASCII] + part.fut.
[TEXT NOT REPRODUCIBLE 1 1 X
 IN ASCII] + part.pres.

Table 3: 'Periphrastic' constructions with verbs of state (14)

Construction Instances Different Animate
 in corpus types of subject
 Aktionsart only

[TEXT NOT REPRODUCIBLE 6 1 X
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 5 3 X
 IN ASCII] + part.aor.
[TEXT NOT REPRODUCIBLE 2 2 X
 IN ASCII] + part.perf.
[TEXT NOT REPRODUCIBLE 182 4
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 30 4
 IN ASCII] + part.aor.
[TEXT NOT REPRODUCIBLE 687 4
 IN ASCII] + part.perf.
[TEXT NOT REPRODUCIBLE 1 1 X
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 87 4
 IN ASCII] + part.aor.
[TEXT NOT REPRODUCIBLE 5 2 X
 IN ASCII] + part.perf.
[TEXT NOT REPRODUCIBLE 6 3 X
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 1 1 X
 IN ASCII] + part.perf.
[TEXT NOT REPRODUCIBLE 429 4
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 39 4
 IN ASCII] + part.aor.
[TEXT NOT REPRODUCIBLE 101 4
 IN ASCII] + part.perf.
[TEXT NOT REPRODUCIBLE 1 1 X
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 4 2 X
 IN ASCII] + part.perf.

Table 4: [TEXT NOT REPRODUCIBLE IN ASCII] with present participle in
Plato, Lysias, Thucydides, Xenophon and Isocrates

Author Total Activity Accomplishment Achievement State

Plato 83 5 % 2 % 1 % 92 %
Lysias 26 15 % 0 % 0 % 85 %
Thucydides 69 13 % 3 % 9 % 75 %
Xenophon 133 18 % 5 % 3 % 74 %
Isocrates 118 21% 2 % 6 % 71 %

Table 5: Syntactic contiguity of frequently
occurring constructions

Construction Total Zero
 distance

[TEXT NOT REPRODUCIBLE 87 88 %
 IN ASCII] + part.aor.
[TEXT NOT REPRODUCIBLE 687 72 %
 IN ASCII] + part.perf.
[TEXT NOT REPRODUCIBLE 32 69 %
 IN ASCII] + part.fut.
[TEXT NOT REPRODUCIBLE 24 67 %
 IN ASCII] + part.fut.
[TEXT NOT REPRODUCIBLE 101 60 %
 IN ASCII] + part.perf.
[TEXT NOT REPRODUCIBLE 12 58 %
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 429 57 %
 IN ASCII] + part.pres.
[TEXT NOT REPRODUCIBLE 16 56 %
 IN ASCII] + partpres.
[TEXT NOT REPRODUCIBLE 30 50 %
 IN ASCII] + part.aor.
[TEXT NOT REPRODUCIBLE 28 46 %
 IN ASCII] + part.fut.
[TEXT NOT REPRODUCIBLE 39 46 %
 IN ASCII] + part.aor.
[TEXT NOT REPRODUCIBLE 182 42 %
 IN ASCII] + part.pres.

Table 6: Groupings of semantically similar constructions

Group 1 Group 2 Group 3

[TEXT NOT [TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part. ASCII] + part.
 perf. pres. aor.
[TEXT NOT [TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part. ASCII] + part.
 perf. pres. aor.
[TEXT NOT [TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part. ASCII] + part.
 aor. (?)19 pres. aor. (?)
[TEXT NOT [TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part. ASCII] + part.
 perf. pres. aor.
[TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part.
 perf. pres.
[TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part.
 perf. pres.
[TEXT NOT
 REPRODUCIBLE IN
 ASCII] + part.
 perf.

Group 1 Group 4 Group 5

[TEXT NOT [TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part. ASCII] + part.
 perf. fut. pres.
[TEXT NOT [TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part. ASCII] + part.
 perf. fut. pres.
[TEXT NOT [TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part. ASCII] + part.
 aor. (?)19 fut. pres.
[TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part.
 perf. pres.
[TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part.
 perf. pres.
[TEXT NOT
 REPRODUCIBLE IN
 ASCII] + part.
 perf.
[TEXT NOT
 REPRODUCIBLE IN
 ASCII] + part.
 perf.

Group 1 Group 6

[TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + + part.
 perf. pres.
[TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part.
 perf. pres.
[TEXT NOT [TEXT NOT
 REPRODUCIBLE IN REPRODUCIBLE IN
 ASCII] + part. ASCII] + part.
 aor. (?)19 pres.
[TEXT NOT
 REPRODUCIBLE IN
 ASCII] + part.
 perf.
[TEXT NOT
 REPRODUCIBLE IN
 ASCII] + part.
 perf.
[TEXT NOT
 REPRODUCIBLE IN
 ASCII] + part.
 perf.
[TEXT NOT
 REPRODUCIBLE IN
 ASCII] + part.
 perf.

Table 7: Groupings of 'periphrastic' constructions (21)

Group 1 Group 2

[TEXT NOT REPRODUCIBLE IN ASCII] + [TEXT NOT REPRODUCIBLE IN ASCII]
 part. perf. + part. pres.
[TEXT NOT REPRODUCIBLE IN ASCII] + [TEXT NOT REPRODUCIBLE IN ASCII]
 part. aor. + part. pres.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. perf.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. aor.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. aor.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. fut.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. fut.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. fut.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. pres.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. pres.
 [TEXT NOT REPRODUCIBLE IN ASCII]
 + part. pres.

Group 3 Group 4

[TEXT NOT REPRODUCIBLE IN ASCII] + [TEXT NOT REPRODUCIBLE IN ASCII]
 part. pres. + part. pres.
[TEXT NOT REPRODUCIBLE IN ASCII] + [TEXT NOT REPRODUCIBLE IN ASCII]
 part. pres. + part. pres.
[TEXT NOT REPRODUCIBLE IN ASCII] + [TEXT NOT REPRODUCIBLE IN ASCII]
 part. aor. + part. pres.
[TEXT NOT REPRODUCIBLE IN ASCII] +
 part. perf.
[TEXT NOT REPRODUCIBLE IN ASCII] +
 part. perf.
[TEXT NOT REPRODUCIBLE IN ASCII] +
 part. perf.
[TEXT NOT REPRODUCIBLE IN ASCII] +
 part. pres.
[TEXT NOT REPRODUCIBLE IN ASCII] +
 part. pres.
[TEXT NOT REPRODUCIBLE IN ASCII] +
 part. perf.
([TEXT NOT REPRODUCIBLE IN ASCII]
 + part. pres.) (22)
([TEXT NOT REPRODUCIBLE IN ASCII]
 + part. pres.)