The article was written in the context of the project Giving gifts and doing favours: Unlocking Greek support-verb constructions (University of Oxford, 2020-2024), which was funded by the Leverhulme Trust (grant n. ECF-2020-181) with the author as the PI.

1. Support-verb constructions as phraseological units

3By our definition of support-verb constructions above, to pay attention, to spill the beans, and to take aim qualify (Mel’čuk [2023: 34-35, 72-75, and 164]). By Mel’čuk’s subdivision into types of phrasemes, to pay attention is a collocation containing the semantic pivot attention and the quasi-unilexeme, i.e. a degenerate lexeme (Mel’čuk [2023: 46]), pay; to spill the beans is an idiom as it withstands syntactic transformations (see Section 3); and take aim! is a pragmateme, which is context-dependent. We thus observe an internally heterogenous group of constructions. That support-verb constructions form an internally heterogenous group of constructions, Mel’čuk’s [2023] Sens-Texte approach, Kamber’s [2008] function-verb construction approach, Gross’ [1998] Lexicon-Grammar approach in the context of the Laboratoire d’Automatique Documentaire et Linguistique, and the Natural-Language-Processing-inspired PARSEME terminology all find. In line with Mel’čuk, the present chapter adopts Gross’ [1998] definition of support-verb constructions as opposed to narrower definitions such as the PARSEME light-verb constructions.

4Support-verb constructions consist of a predicative noun and a support verb. The predicative-noun slot can be filled by any eventive noun (e.g. Meinschaefer [2016: 393]; Huyghe et al. [2017: 127]; Grimshaw [1990]; Bel, Coll & Resnik [2010: 46]) including nouns that assume an eventive meaning only due to reconceptualization, e.g. picture in to take a picture (Radimský [2011]), or due to metaphorical extension, e.g. heart in to take heart (Sheinfux et al. [2019]). The predicative noun can be derivationally related to a verb, e.g. συντάττω suntattō ‘to order’ and σύνταξις suntaxis ‘order’ by means of the derivational suffix -σι- -si- (van Emde Boas et al. [2019: 262-269]), but this does not have to be the case.

6The support verb is morphologically marked for tense, mood, and grammatical aspect, e.g. I was making a suggestion vs I make a suggestion, but in Gross’ [1998] framework can also indicate lexical aspect, e.g. to begin a speech vs to give a speech, voice, e.g. to receive a favour vs to do a favour (cf. Keenan & Dryer [2007]), and polarity, e.g. to lack an idea vs to have an idea (Fendel [2023a]) through lexical replacement (cf. Joshi [2020]). Support-verb constructions are commonly involved in transitivity operations, both transitivity reducing, e.g. to suggest vs to make a suggestion (cf. Marini [2010]), and causativisation, e.g. German jemandem Angst machen ‘to induce fear in somebody’ vs Angst haben ‘to have fear / to be afraid’ (cf. Marini [2018]).

7Support verbs differ from auxiliaries in that they contribute to the event structure (Butt [2010]) as visible e.g. in their profiling of the subject component. In Paul lacked a good idea for the project, the subject component (Paul) is a frustrative; in Paul had a great idea for the project, the subject component (Paul) is a Volitional undergoer (Næss [2007]). The change in the participant role of the subject component is regulated by the support verb. Support verbs also differ from auxiliaries in that they are not fully productive (Kamber [2008: 143]; Bonial [2021]), i.e. they cannot predictably be applied to any noun, and in that they do not appear in a phonetically reduced form (Loporcaro [2022: 228 n. 16]).

10Support-verb constructions are lexically speaking verbal multi-word expressions which are ambiguous (Savary et al. [2019]; Krstev & Savary [2017]; Butt & Lahiri [2013]), variable (Tutin [2016]; Savary et al. [2018]), and discontinuous (Sag et al. [2002]; Lehmann [2020]). All of these aspects impact on lexical affinity measures, such as the logDice (Rychlý [2008]), due to how they are computed. Section 3 looks in more detail at how these lexical characteristics of support-verb constructions are intertwined.

12As the present chapter is interested in origins and pathways of δίκην δίδωμι dikēn didōmi ‘to pay the price for one’s actions’, χάριν ἀπολαμβάνω k^harin apolambanō ‘to receive a favour’, and προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention’ in classical and postclassical Greek, we embrace the diversity within the group of support-verb constructions. Diversity, or internal heterogeneity, is visible not only synchronically (e.g. diatopic and diastratic distinctions) but also diachronically due to support-verb constructions hovering at three interfaces: the syntax-lexicon interface (as obvious from constructions such as lexical passives) (Heine [2020: 15]; Croft [2022: 423]; Boye [2023]), the syntax-semantics interface (as obvious from the variations triggering foreseeable and unforeseeable changes sometimes even pushing structures out of the group) (Gross [1998]), and the syntax-pragmatics interface as they can take Heine and Kaltenböck’s [2021] ‘quantum leap’ towards Mel’čuk’s pragmatemes like any other phraseme (see Section 6). Thus, a primarily lexical approach such as the Sens-Texte framework cannot fully capture them.

13The present chapter is interested primarily in the aspect of lexical and morpho-syntactic variability of support-verb constructions and the question when this variability creates a form-function pairing that differs semantically more than expected based on the formal change from its presumed base. Mel’čuk [2023: 71] offers morphological, morpho-syntactic, and syntactic tests to differentiate between his collocations and idioms, however without a strict order of proceeding through these tests (see Section 3).

15Due to the decision tree procedure, a combination could qualify in one corpus of Greek as an LVC (light-verb construction) and in a different one as a VID (verbal idiomatic expression) (similarly Mel’čuk [2023: 76]). Such shifts are not limited to a diachronic scenario. The three structures of interest here qualify as Not MWE (not a multi-word expression), VID (verbal idiomatic expression), and LVC.full (light-verb construction full) respectively in classical Greek (5^th / 4^th c. BC) but change categories over time apart from the lexical passive formation which is currently excluded from the PARSEME terminology.

16The issue of variability of form and function in support-verb constructions is not specific to classical and postclassical Greek nor to Greek. In fact, support-verb constructions exist across languages and are considered a near-universal category (Savary et al. [2018: 96]; Hoffmann [2023: 29-31]). They are a comparative concept (Haspelmath [2010: 665]) the instantiation of which differs between languages and varieties, as a descriptive category (Haspelmath [2010: 664]), and thus for a corpus language, between corpora (Fleischman [2000: 34]).

17The shared event structure of support-verb constructions feeds into research question 1, their lexical variability into research question 2, and their sitting at three interfaces into research question 3:

1. How do changes in analyticity and compositionality affect the status of a phraseme (as a collocation vs idiom)? [test case: δίκη & δίδωμι dikē & didōmi ‘to pay the price for one’s actions’]

2. What degree of lexical renewal do we accept as modification of a phraseme before positing a new phraseme? [test case: χάρις & ἀπολαμβάνω k^haris & apolambanō ‘to get thanks / to receive a favour’]

3. • 12 The translations provided in brackets reflect only one of the phrasemes that are identified in the (...)

   Why are pragmatemes instances of rebirth rather than continuity? [test case : προσέχω & νοῦς prosek^hō & nous ‘to pay attention’]12

18After this introductory section, the chapter falls into six sections. Section 2 introduces the data sample and methods of data collection. Section 3 details how variability, ambiguity, and discontinuity interplay in support-verb constructions and how they can be utilised to determine when a new phraseme is born. Section 4 considers research question 1, Section 5 research question 2, and Section 6 research question 3. Section 7 summarises the results and offers conclusions.

2. Data samples and data collection

20The distinction between literary and documentary texts is adopted from studies on Greek from the perspectives of Classics, Papyrology, and related disciplines. A literary text is thus a text that has been written with an artistic intent and literary texts thus include oratory, historiography, and prose along with texts in verse. Instances of carefully crafted technical writings also fall into this category. Literature is not limited to fictional writings. The genre divisions adopted are the following: (i) technical writing: commentaries, medical writing, lexicography, homilies, (ii) oratory: speeches, sermons, (iii) prose: philosophical writings, dialogues, vitae, epistles, (iv) historiography. By contrast, documentary texts are those that are written for specific purposes in everyday life, both at personal and at higher (e.g. governmental) levels. Documentary text types are e.g. letters, contracts, wills, receipts, and reports of court dealings. The present chapter adopts the text types assigned to texts in the Duke Database of Documentary Papyri.

21Both literary and documentary texts are drawn upon because (i) documentary texts are less influenced by genre-related pressures in the sense of artistic creativity, (ii) documentary texts are less influenced by intertextual links stylistically speaking, and (iii) documentary texts allow for idiolectal idiosyncracies to surface (Biber & Conrad [2009]; Mel’čuk [2023: 71]). However, both literary and documentary texts involve a range of degrees of flexibility as regards register, genre, and style displayed, consider e.g. the texts often labelled ‘official’ amongst the documentary texts, such as court proceedings (cf. Palme [2009: 361-363]). The bulk of documentary texts dates from the 3^rd c. BC onwards, whereas the earliest literary texts date from around the 8^th c. BC. The variety of sources allows us to trace the three structures of interest through the centuries in varying environments and thus to trace traditions and independent developments.

25The following data samples inform Sections 4 to 6 and are fully accessible here: DOI 10.5287/ora-zovd9gjv0 and DOI 10.5287/ora-g2op5v0em.

Table 1. Datasets

3. Support-verb-construction fields as a metric

27Mel’čuk [2023: 71] offers morphological, morpho-syntactic, and syntactic tests to differentiate between his collocations and idioms. Consequently, a phraseme could qualify as a collocation in one corpus but an idiom in another (Mel’čuk [2023: 76]). The syntactic transformations used to distinguish between a collocation and an idiom are the following:

(1) Tests (Mel’čuk [2023: 72])
1. Passivization of a verbal idiom involving its nominal lexemic components;
2. semantic inflection of an internal nominal component of the idiom (an internal component of an idiom is a lexemic component that is not the idiom’s syntactic head);
3. internal syntactic modification of the idiom – that is, adding a modifier to an internal lexemic component;
4. changing linear order of the idiom’s components.

This makes Mel’čuk [2023: 74-75] classify to spill the beans as an idiom but to pull strings and to pay attention as a collocation with a quasi-unilexeme (strings and pay). However, with regard to support-verb constructions in classical and postclassical Greek, three aspects of Mel’čuk’s, and by extension the above-mentioned PARSEME terminologies, are problematic: morphological passivisation, the attestation of canonical forms, and creativity as opposed to productivity.

30Since the issue of variability of form and function in support-verb constructions is not specific to classical and postclassical Greek nor to Greek but support-verb constructions are a comparative concept (Haspelmath [2010: 665]), the following paragraphs explore based on French data the interplay between Mel’čuk’s passivisation, internal inflection, addition of a modifier, and order of components before returning to classical and postclassical Greek. We use modern French support-verb constructions as a case study here (i) because significant work has been done (unlike for Greek), (ii) annotated large corpora exist (unlike for Greek), and (iii) French is a spoken language for which canonical forms can be reverse-engineered and modifications tested with native speaker judgements (unlike for Greek) (e.g. Bonial [2021]). Note that French word order is syntactically rather than information-structurally driven and that French morpho-syntax is left-branching, i.e. we have auxiliaries and prepositions where Greek (and French’s predecessor Latin) would have had endings.

31The segmentation of support-verb constructions at the syntactic and semantic levels differs. While the predicative noun and support verb fill the predicate slot at the semantic level (the f-structure in the terms of Lexical-Functional Grammar), they fill two slots at the syntactic level (the c-structure in the terms of Lexical-Functional Grammar) (see further Andrews [forthcoming]; Butt [2019]). The mismatch between their syntactic and semantic structures accounts for claims of their diachronic potential to univerbate, that is ‘the union of two syntagmatically adjacent word forms into one’ (Lehmann [2020: 206]). Their discontinuous structure (Nagy, Vincze & Farkas [2013: 329]) goes hand in hand with their analyticity and accounts for their synchronic variability (Cordeiro & Candito [2019] ; Pasquer et al. [2018] ; Sag et al. [2002] ; Savary et al. [2018]).

33The discontinuity of the structure means that the lexical unit is interrupted and discontinuity often comes with internal analyticity, in that each component of the support-verb construction can be modified separately, e.g. I quickly made an important decision, where the adverb quickly modifies the verb and the adjective important the noun (Pasquer et al. [2018 : 2585] ; Savary et al. [2018 : 88-90] ; Tutin [2016]). Modifications can trigger a literal rather than a support-verb-construction reading, e.g. he broke her chocolate heart when he stepped on it vs he broke the heart of his class mate (Pasquer et al. [2018 : 2585]). Thus, synchronically, we need to define precisely what can intervene between the two components of a support-verb construction without breaking up the structure, i.e. triggering a verb-object rather than a support-verb-construction reading. The number and type of permissible intervening items are termed the support-verb-construction field henceforth.

• • 28 Accessible here: https://lindat.mff.cuni.cz/services/kontext/first_form?corpname=parseme_fr_a (last (...)
  • 29 A type is a unique form-function pairing, in the case of a support-verb construction forming a lexi (...)

  The PARSEME 1.0 French28 corpus consists of 48,539 words and is tagged for verbal multi-word expressions. It contains 1,633 light-verb constructions (Savary et al. [2018: 118]), of which 1,298 take the form verb-noun (Pasquer [2017: 166]). However, most types are represented by relatively few tokens.29

• • 30 Accessible here: https://www.sketchengine.eu/frtenten-french-corpus/ (last accessed 14 March 2024).

  The Sketch Engine corpus frTenTen1730, made up of internet data, consists of 5.7 billion words and is part-of-speech tagged and lemmatised. Support-verb constructions for defined support verbs and/or predicative nouns can be accessed through Sketch Engine’s concordancing tool, yet with the caveat that concordances need manual correction for support-verb constructions due to the issues of ambiguity, variability, and discontinuity mentioned.

41The support-verb-construction field in the selected structures never exceeds a mean of two items. The types of items that appear are modifications of the predicative noun (determiners, adjectives) or modifications of the support verb (adverbs, negatives). Modifications of the predicative noun do not appear with structures 1 to 4, which have the smallest support-verb-construction fields. Modifications of the predicative noun appear from 5 downwards – yet notice aucune ‘no’ with 4 once. Operations such as pronominalisation / passivisation / co-ordination of predicative nouns appear from 6 downwards.

42In Table 2, the singular and plural forms of the selected nouns are considered separately as they can form distinct form-function pairings with the support verbs – compare 8 and 13, 9 and 14 for illustration. For prendre décision and prendre décisions, the formal change (pluralisation) does not provoke a functional change greater than what is expected due to the formal change. They are kept separate in Table 2 partly to align them with other selected predicative nouns, but also because the singular form is significantly more frequent than the plural form but seems to allow for a larger and more diverse support-verb-construction field than the plural form.

43For comparison, the support-verb-construction types from Table 2 are run through the PARSEME 1.0 corpus. The results are summarised in Table 3. There were no hits for faire peur ‘to scare’, redonner confiance ‘to restore confidence’, avoir confiance ‘to have confidence / trust’, restaurer confiance ‘to restore confidence’, avoir raison ‘to be right’, avoir raisons ‘to have reasons’, donner raison ‘to confirm’, attirer attention ‘to draw attention’, retenir attention ‘to retain attention’.

Support-verb construction	Tokens		Support-verb-construction field		Le Monde 1998
	Total	Active declarative	Type	Mean	SVC field
avoir peur ‘to be afraid’	5	5	NEG	0.2	0.45
faire confiance ‘to trust (somebody)’	6	5	PRT, ADJ	0.4	0.5
avoir envie ‘to want’	5	5	NEG (aucune), ADV	0.4	0.48
avoir espoir ‘to have hope’	1	1	ADJ	1	1
prendre décisions ‘to take decisions’	5 (1 passive, 1 pronominalised)	3	DP	1	1
prendre décision ‘to take a decision’	3 (1 passive)	2	ADV, DP	1.5	1.81

44The PARSEME 1.0 corpus is smaller than the Le Monde 1998 corpus, which explains the smaller number of relevant tokens. However, similarities between Table 2 and Table 3 are visible with regard to the size of the support-verb-construction field and the type of permissible items with specific types (cf. right-most column of Table 3 for comparison).

• In this very large corpus, it appears that even highly fixed redonner confiance ‘to restore confidence’ (5,491 tokens) appears at times with a determiner phrase (ils redonnent un peu de confiance au combat syndical ‘they restore some confidence in the union battle’, lamarseillaise.fr) or an adverb (nos succès d’antan redonneront peut-être confiance en nos capacités ‘our past success may perhaps restore confidence in our ability’, metiseurope.eu) intervening.

• • 42 The same is true also for avoir raison. The modifier may disambiguate between a mass-noun and a cou (...)

  For donner raison (30,034 tokens), the addition of modifications on the predicative noun breaks up the lexicalised support-verb construction (donner raison ‘to confirm’) and triggers a different form-function pairing (donner une raison ‘to have a reason’).42

• • 43 Procedure: We draw a lemma-based concordance for ‘peur’, query this concordance for the lemma ‘avoi (...)

  For avoir peur ‘to be afraid’ (226,425 tokens), the addition of a determiner phrase seems to have a particularising effect (‘to have a specific fear’) (la bureaucratie a une peur terrible de la population ‘the bureaucratic apparatus has a terrible fear of the population’, free.fr; J’ai cette peur qui revient ‘I have this fear which comes back’, cowblog.fr). Determiner phrases with avoir peur ‘to be afraid’ are rare in the frTenTen17 corpus.43

• • 44 Searches can be carried out in Sketch Engine as described for ‘avoir’ and ‘peur’.

  Finally, modifications of the noun in avoir envie (316,717 tokens) are rare and seem to have a particularising function (tous les laquais avaient une grande envie de dormir ‘all the lackeys had a great desire to sleep’, roi-mathias.fr).44

These impressionistic findings align with Sheinfux et al.’s [2019: 66] that even the seemingly most fixed structures show variations in corpora that are large enough.

46However, the variations may be rather in the sense of Finkbeiner’s [2008: 401] options that are theoretically available but by language users rarely drawn upon. In the present chapter, the key question is when do we accept a combination (and its variations) as a phraseme and which instances do we dismiss as creativity (Bonial [2021]; Baayen [2009]; Hoffmann [2005: chap. 8]; Matushansky & Zwarts [2021]). The frTenTen17 corpus is a corpus representing little-curated language usage on the internet, whereas the Le Monde 1998 corpus is a highly curated journalistic corpus. Thus, we expect innovative non-canonical structures in the former but less so in the latter.

47Overall, support-verb constructions with larger fields allow for more operations such as passivisation / pronominalisation / co-ordination of predicative nouns than those with smaller fields. With the selected support-verb constructions, the support-verb-construction field seems constrained to no more than two items and only modifications of the noun and/or the verb (determiners, adjectives, negatives, adverbs). All these items have limited flexibility in the syntactic frame, such that their displacement to the outside of the support-verb construction would indicate that the formerly discontinuous lexical unit is tending towards continuity (e.g. if the negative ne … pas was placed around the noun and the verb). Thus, the mean length of and the types of items permissible in the support-verb-construction field seem a good measure for the analyticity of support-verb constructions. The low frequency of large support-verb-construction fields may resemble what has been found e.g. for phrasal verbs in English, which allow for items to intervene but speakers disprefer heavy / long constituents (e.g.?I looked the formerly unknown word up vs I looked it up) (Gries [2003]) as well as for tmetic constructions in archaic Greek (pre 5^th c. BC) (Bertrand [2014]; Pompei [2014]).

49While the French data illustrates the interplay between variability, discontinuity, and ambiguity of the lexical unit, we have to bear in mind that for comparative concepts (Haspelmath [2010: 665]), here support-verb constructions, the instantiation differs between languages and varieties, as a descriptive category (Haspelmath [2010: 664]). We noted above that classical Greek support-verb constructions disprefer morphological passivisation, the first of Mel’čuk’s tests. Another of Mel’čuk’s tests (passivisation, internal inflection, addition of a modifier, and order of components) deserves commenting too, that is the order of components.

51With the notion of the support-verb-construction field in place, Sections 4 to 6 turn to the three case studies which reflect diastratic, diatopic, and diachronic variation and development in support-verb constructions and due to this variation the births of phrasemes.

4. Diastratic differentiation and the syntax-lexicon interface

52The first support-verb-construction phraseme to be considered is the oldest of the three selected for analysis, δίκην δίδωμι dikēn didōmi ‘to pay the price for one’s actions’. δίκην δίδωμι dikēn didōmi is a semantic-lexemic phraseme. It seems to act as a short-lived idiom in classical Greek legal jargon, whereas the phraseme in general exists from archaic times onwards and into the medieval period as a collocation. The present section is interested in the origins of the idiom from the perspective of how changes in analyticity and compositionality affect the status of a phraseme (as a collocation vs idiom).

54Plato (5^th / 4^th c. BC), in his dialectic dialogue Gorgias, has Socrates debate the meaning of δίκην δίδωμι díkēn didōmi. In (2), Socrates encourages his interlocutor to compare the meanings of the lexical passive δίκην δίδωμι díkēn didōmi and the morphological passive κολάζεσθαι kolazest^hai ‘to get punished’:

(2) Plato, Gorgias 476a τὸ διδόναι δίκην καὶ τὸ κολάζεσθαι δικαίως ἀδικοῦντα ἆρα τὸ αὐτὸ καλεῖς;
to didonai dikēn kai to
ART.ACC give.INF.ACT punishment.ACC and ART.ACC
kolazest^hai dikaiōs adikounta ara to
punish.INF.PAS just.ADV wrong.PTC.ACT.ACC PRT.ITR ART.ACC
auto kaleis?
same.ACC call.PR.ACT.2sg
‘Are you saying that ‘paying the price for one’s actions’ and ‘justly getting punished’ when one does wrong are the same?’

The subsequent exchange clarifies that Socrates (and by inference) Plato does not consider the lexical and morphological passives semantically equivalent. Socrates even questions whether δίκην δίδωμι díkēn didōmi is a passive in (3):

(3) Plato, Gorgias 476d Τούτων δὴ ὁμολογουμένων, τὸ δίκην διδόναι πότερον πάσχειν τί ἐστιν ἢ ποιεῖν
toutōn dē ^homologoumenōn to dikēn
these.GEN PRT agree.PTC.PAS.GEN ART.NOM punishment.ACC
didonai poteron pask^hein ti estin ē
give.INF.ACT whether suffer.INF.ACT what be.PR.3sg or
poiein
do.INF.ACT
‘With these things being agreed upon, is ‘paying the price for one’s actions’ suffering or doing (something)?’

The subsequent exchange sees Socrates and his interlocutor agree that δίκην δίδωμι díkēn didōmi is in fact a passive even if not semantically equivalent to the morphological passive of the simplex verb (κολάζεσθαι kolazest^hai). This is not surprising given that while δίκην δίδωμι díkēn didōmi is a recipient passive, i.e. profiling the subject as a volitional undergoer, κολάζεσθαι kolazest^hai is a patient passive, i.e. profiling the subject as the patient (Keenan & Dryer [2007]; Næss [2007]).

55Noticeably, the fine semantic nuancing seems to be no longer accessible in later scholia (3^rd c. BC to 3^rd c. AD) which equate δίκην δίδωμι díkēn didōmi with the morphological passive τιμωρέομαι (ὑπό) timōreomai (^hupo) ‘to get punished (by)’, see (4):

(4) Scholia in Euripidem, 1052
δίκην δέ μοι δέδωκε: τετιμώρηται ὑπ’ ἐμοῦ. Gr. ἔδωκε δίκην ὁ τιμωρούμενος, λαμβάνει δίκην ὁ τιμωρῶν.
dikēn de moi dedōke: tetimōrētai
punishment.ACC PRT I.DAT give.PRF.ACT.3sg punish.PRF.PAS.3sg
^hup’ emoũ. Gr. edōke dikēn ^ho
by I.GEN Gr. give.AOR.ACT.3sg punishment.ACC ART.NOM
timōroumenos, lambanei dikēn ^ho
punish.PTC.PAS.NOM take.PR.ACT.3sg punishment.ACC ART.NOM
timōrōn.
punish.PTC.ACT.NOM
‘He pays the price for his actions to me: he is punished by me. Gr. The one who is being punished paid the price for his actions, the one who punishes collects punishment / fines.’
(see also Scholia in Aelianum 1.2.128)

Yet, like in Plato’s metalinguistic debate, the simplex verb chosen to illustrate the meaning of the phraseme is not formally related to the predicative noun (such as δικάζομαι dikazomai from δίκη díkē by means of the derivational suffix -αζω -azō) and is fully transitive, i.e. with a patient object in the accusative case required (Lavidas [2009: 92-93]; Luraghi [2010: 69]). Noticeably, the agent is expressed in the dative case with the support-verb construction but in the form of a prepositional phrase with the simplex verb.

(5) Ich habe ihm Angst gemacht.
I.NOM AUX.1sg he.DAT fear.ACC make.PTC
‘I have scared him.’
(6) J’ ai envie de lui écrire
I have.1sg desire of him write.INF
‘I want to write to him’
(7) I made the suggestion that she join us.

The former two options would attest to the analysability of the support-verb construction (i.e. syntagm-like), the latter option to the support-verb construction being treated akin to a lexical simplex verb (i.e. word-like).

58The phenomenon of support-verb constructions being treated as words rather than syntagms is observed by George although without explicit mention of support-verb constructions. He discusses the passage in (8):

(8) Herodotus, Histories 4.35.2 τὴν δὲ Ἄργην τε καὶ τὴν Ὦπιν ἅμα αὐτῇσι τῇσι θεοῖσι ἀπικέσθαι. Λέγουσι καί σφι τιμὰς ἄλλας δεδόσθαι πρὸς σφέων·
tēn de Argēn te kai tēn Ōpin
ART.ACC PRT Arge.ACC PRT and ART.ACC Opis.ACC
^hama autēsi tēsi t^heoisi apikest^hai.
together they.DAT ART.DAT god.DAT.PL arrive.AOR.INF
Legousi kai sp^hi timas allas
say.PR.ACT.3pl also they.DAT honour.ACC.PL other.ACC.PL
dedost^hai pros sp^heōn;
give.PRF.INF.PAS by they.GEN
‘[they say that] Arge and Opis arrived together with the gods and they say that other honours have been given to them by them’

60The earliest attestations of the phraseme δίκην δίδωμι dikēn didōmi appear in archaic Greek (pre 5^th c. BC), 7 instances in total, and show either a dative indicating the recipient (5 instances) or no recipient, see (9) and (10):

(9) Hymni Homerici, To Hermes (8^th-6^th c. BC), l. 312
δὸς δὲ δίκην καὶ δέξο παρὰ Ζηνὶ Κρονίωνι.
dos de dikēn kai dexo
give.AOR.IMP.2sg PRT punishment.ACC and receive.AOR.IMP.2sg
para Zēni Kroniōni.
from Zeus.DAT son.of.Kronos.DAT
‘Give justice and receive [it] from Zeus, the son of Kronos.’

(10) Hesiod, Works and Days (8^th / 7^th c. BC), ll. 225-227
Οἳ δὲ δίκας ξείνοισι καὶ ἐνδήμοισι διδοῦσιν
ἰθείας καὶ μή τι παρεκβαίνουσι δικαίου,
τοῖσι τέθηλε πόλις, λαοὶ δ᾽ ἀνθεῦσιν ἐν αὐτῇ:
^hoi de dikas xeinoisi kai endēmoisi
REL PRT judgement.ACC.PL stranger.DAT and inhabitant.DAT
didousin
give.PR.ACT.3pl
it^heias kai mē ti parekbainousi dikaiou,
straight.ACC.PL and not anything deviate.PR.ACT.3pl just.GEN
toisi tethēle polis, laoi d᾽
they.DAT thrive.PRF.ACT.3sg city.NOM people.NOM PRT
antheusin en autē:
blossom.PR.ACT.3pl in it.DAT
‘For those who give straight judgements to strangers and inhabitants and do not deviate from what is just, the city thrives and the people blossom in it [sc. the city].’

62In classical times (5^th / 4^th c. AD), the situation changes. Table 4 shows the 536 classical attestations of the support-verb construction divided by the syntactic frame they appear in, i.e. with a dative (pointing to a compositional collocation), with an agent expression (pointing to a non-compositional idiom), or without either (and thus inconclusive). Percentages apply to rows :

63The correlations that are illustrated by Table 4 are: (i) instances with agent expressions appear prevalently in prose contexts, allow for limited modification and inflection only, and constrain support-verb-construction fields with regard to their size and the type of items appearing in the field. Conversely, (ii) instances with a dative appear in prose and verse yet with a prevalence in prose, allow for some modification and inflection, and constrain support-verb-construction fields comparatively less with regard to their size and the type of items in the field. The majority of instances appear without either an agent expression or a dative and thus without a syntactic cue as to the meaning of the support-verb construction.

64Generally speaking, the addition of modifiers, here attributive and determiner phrases with the predicative noun, and internal inflection, here of the predicative noun, is limited in support-verb constructions (both Mel’čuk’s collocations and idioms) at the point when it forces a referential meaning onto the noun, e.g. to break one’s heart vs to break one’s chocolate heart (Pasquer et al. [2018: 2583]; Savary et al. [2019]). Determiner phrases take the form of definite articles, negative determiners, and indefinites in Greek. Attributive phrases take the form of qualitative genitives or adjectives.

(11) Euripides, Orestes, ll. 612-614 (tragedy)
μολὼν γὰρ εἰς ἔκκλητον Ἀργείων ὄχλον
ἑκοῦσαν οὐχ ἑκοῦσαν ἐπισείσω πόλιν
σοὶ σῇ τ᾽ ἀδελφῇ, λεύσιμον δοῦναι δίκην.
molōn gar eis ekklēton Argeiōn ok^hlon
go.PTC.NOM PRT to select.ACC Argive.GEN.PL crowd.ACC
^hekousan ouk^{h h}ekousan episeisō polin
voluntary.ACC NEG voluntary.ACC shake.FUT.1sg city.ACC
soi sē t᾽ adelphē, leusimon dounai
you.DAT your.DAT and sister.DAT stoning.ACC give.AOR.INF
dikēn
judgement.ACC
‘For, going to the selected crowd of Argives, I will shake/urge the city, voluntary or involuntary, to give the judgement of stoning to you and your sister.’

By contrast, the result meaning ‘penalty/punishment’ does not allow for modification or any count-noun morpho-syntax but seems to be treated as a mass noun.

66At the latest from about the time of Plato’s metalinguistic debate onwards we can assume that a lexical passive meaning existed. Usually metalinguistic commentary reflects upon current practice. In fact, agent expressions indicating unambiguous passive patterns are rare. Table 5 shows all the standard and non-standard agent expressions attested with the phraseme, note that no relevant examples appear in non-literary texts:

(12) Plato, Gorgias 525b5
καὶ δίκην διδόντες ὑπὸ θεῶν τε καὶ ἀνθρώπων
kai dikēn didontes ^hupo t^heōn te kai
and punishment.ACC give.PTC.NOM by god.GEN.PL and and ant^hrōpōn
man.GEN.PL
‘and getting punished by both gods and men’

68In later Christian literature, the quote is adapted to Christian thought, i.e. ‘god’ is used in the singular (e.g. Georgius, Chronicon 110.132.22, Medieval Greek (MG)). All Roman-period instances show co-ordination of the support-verb construction with the simplex verb κολάζω kolazō (in the passive voice) (Plutarch, De superstitione 168B4; Plutarch, De tranquilitate animi 470D1). Noticeably, ἀπό apo + genitive ‘away from’ never takes over as would be expected in postclassical times (Bortone [2010: 184-185]; George [2005: 130]). All the instances with ἐκ ek + genitive ‘out of’ are ambiguous and can be interpreted as a reason/cause rather than an agent.

69Furthermore, several of the non-standard agent expressions are in fact replacements of the dative in the role of the indirect object especially when the indirect object refers to a recipient or goal in post-classical times (Blass [1990: paras. 187-202]; Mayser [1906: paras. 91-98]; Wahlgren [2014]; Stolk [2017: 188 and 209]). This means that rather than functioning as agent expressions they function as goal / recipient expressions and thus point to the collocation rather than the idiom. This ambiguity affects specifically παρά para + dative ‘beside / before’ and πρός pros + dative ‘near / with’.

70Thus, what we seem to witness is a collocation ‘to judge / to give judgement(s)’ existing from archaic and into medieval times and an idiom ‘to pay the price for one’s action’ during the classical period. The scholia are the terminus ante quem at which point in time the idiom apparently needed glossing. The idiom seems to appear very specifically in legal contexts where the lexical and the morphological passives seem to assume different underlying situations of liability. (13) from one of Lysias’ courtroom speeches illustrates this:

(13) Lysias, Speech 30.23-24 ἐὰν δὲ καταψηφισάμενοι τῶν ἐσχάτων αὐτῷ τιμήσητε, τῇ αὐτῇ ψήφῳ τούς τε ἄλλους βελτίους ποιήσετε καὶ παρὰ τούτου δίκην εἰληφότες ἔσεσθε. ἐπίστασθε δέ, ὦ ἄνδρες δικασταί, ὅτι παράδειγμα τοῖς ἄλλοις ἔσται μὴ τολμᾶν εἰς ὑμᾶς ἐξαμαρτάνειν οὐχ ὅταν τοὺς ἀδυνάτους εἰπεῖν κολάζητε, ἀλλ’ ὅταν παρὰ τῶν δυναμένων λέγειν δίκην λαμβάνητε. τίς οὖν τῶν ἐν τῇ πόλει ἐπιτηδειότερος Νικομάχου δοῦναι δίκην; τίς ἐλάττω τὴν πόλιν ἀγαθὰ πεποίηκεν ἢ πλείω ἠδίκηκεν;
ean de katapsēp^hisamenoi tōn esk^hatōn autō timēsēte, tē autē psēp^hō tous te allous beltious poiēsete kai para toutou dikēn eilēp^hotes esest^he. epistast^he de, ō andres dikastai, ^hoti paradeigma tois allois estai mē tolman eis ^humas examartanein ouk^{h h}otan tous adunatous eipein kolazēte, all’ ^hotan para tōn dunamenōn legein dikēn lambanēte. tis oun tōn en tē polei epitēdeioteros Nikomak^hou dounai dikēn? tis elattō tēn polin agat^ha pepoiēken ē pleiō ēdikēken?
‘If you have condemned him and award the utmost penalty to him, by the same vote you will make the others better people and you will have punished this man. Know, men of the jury, that he will be a paradigm to others not to dare to wrong against you – not when you punish those who are unable to speak but when you punish those who are able to speak. For who would be more suitable of those in the city than Nicomachus to pay the price for his actions? Who has done less good and has wronged more against the city (than him)?

71The contrast between the simplex verb (κολάζω kolazō) and the support-verb construction (δίκην λαμβάνω (παρά) dikēn lambanō (para)) seems to be one between simple punishment and punishment within the framework of the law. In simple punishment, the defendant is fully affected, as reflected in the syntactic structure by the reference to him/her in the form of an accusative object, and is unable (possibly due to not being given the opportunity) to defend himself/herself (τοὺς ἀδυνάτους εἰπεῖν tous adunatous eipein). In punishment within the framework of the law, the defendant is less affected, as reflected in the syntactic structure by the reference to him/her in the form of a prepositional phrase, and is able to influence the severity of the punishment through his/her defence (τῶν δυναμένων λέγειν tōn dunamenōn legein) (Fendel [2023b]).

74This requoting may be partially motivated by the backward-looking movements of Atticism (ca. 1^st. c. BC) (Benaissa [2012]; Horrocks [2014: 133-141]; Sean Adams [2013]; Wisse [2001]; Lee [2013]; Caragounis [2014]; la Roi [2022a]) and the Second Sophistic (ca. 1^st-3^rd c. AD) (Bowie [2012]; Kemezis [2014]) during which aspects of classical language that had since fallen out of use were artificially resuscitated especially in higher-register contexts (Bentein [2019: 146-148 and 155]). The only two later papyrological instances (p. Louvre 3.172, edict, Roman Greek (RG) and P. Berl. Frisk 4, petition, Early Byzantine Greek (EBG)) seem to confirm this, in that they come from a Roman-period edict and an Early Byzantine petition, both of which are higher-register text types (Palme [2009]).

75However, in the case of the idiom in question its ties to the legal domain may be relevant too and may even explain its quick demise. Language acts as a semiotic system expressing meanings beyond the referential meaning of the words used. Language, thus, like any sign, can be an icon (resemblance between form and meaning), an index (contiguity of form and meaning), and a symbol (arbitrary relationship between form and meaning) (cf. Peirce [1933]; Bentein [2019]). We are interested in the indexical function of language here. Bentein [2019: 123] defines linguistic indexes as ‘“structures” (lexemes, affixes, diminutives, syntactic constructions, emphatic stress, etc.) that have become conventionally associated with a particular situational dimension, and that invoke that situational dimension whenever they are used (cf. Ochs [1996: 411])’.

5. Diatopic renewal and the syntax-semantics interface

77The second support-verb-construction phraseme to be considered of the three selected for analysis, χάριν ἀπολαμβάνω k^harin apolambanō ‘to receive a favour’, is a lexical passive and only appears from the classical period onwards, like the idiom δίκην δίδωμι dikēn didōmi ‘to pay the price for one’s actions’ discussed in Section 4. However, it has an active counterpart in the same support-verb-construction family that appears in archaic times already, χάριν ἀποδίδωμι k^harin apodidōmi ‘to give / render / return thanks / a favour’ (first attested in the Septem Sapientes, Sententiae 215.12 (7^th-6^th c. BC), then several times in Aesop’s Fables (6^th c. BC), and in Simonides, Epigrammata 7.77.2 (6^th-5^th c. BC) in archaic times).

78In English, one does and gets a favour or favours, in Greek the opposition is between to give and to take/to receive. This is a semi-productive alternation both in English and in Greek which applies to a large range of predicative nouns, unlike the idiom discussed in Section 4 (cf. Bonial [2021]; Kamber [2008: 143]; Butt & Geuder [2001]). χάριν ἀπολαμβάνω k^harin apolambanō ‘to receive a favour’ also has a counterpart in χάριν λαμβάνω k^harin lambanō ‘to receive a favour’ which seems subject to fewer morpho-syntactic and pragmatic/contextual constraints.

79The question is what degree of lexical renewal we accept as modification of a phraseme before positing a new phraseme. The question falls into two sub-questions: Is χάριν ἀπολαμβάνω k^harin apolambanō a (semi-)productively formed lexical passive of χάριν ἀποδίδωμι k^harin apodidōmi and hence not a new phraseme but a derivation? Is χάριν ἀπολαμβάνω k^harin apolambanō distinct from χάριν λαμβάνω k^harin lambanō and should be considered a different phraseme? The two questions are addressed one after the other below.

81Lexical substitution also seems to affect lexical passives, with archaic λαγχάνω lagk^hanо̄ being ousted by τυγχάνω tugk^hanо̄ ‘to receive’ in classical times. They differ etymologically, in that the former is derived form a root ‘share / lot’ whereas the latter comes from a root ‘chance / lot’, but not significantly in meaning (Beekes [2010]). According to the Thesaurus Linguae Graecae statistics, none of the lemmata mentioned dies out at any point over the course of the history of Greek, substitution only seems to affect their use as support verbs.

82Section 3 discussed that Greek support-verb construction generally fail Mel’čuk’s [2023] passivisation test, possibly for reasons of reduced transitivity of the verb phrase (being a complex predicate) (see also Marini [2010]) but possibly also for lexical reasons (being a unit) (Mel’čuk [2023: 74-75]). Theoretically, four ways of passivisation of a support-verb construction are imaginable. The examples are taken from the classical ECF Leverhulme corpus. τιμή time ‘honour’ is used as the predicative noun as all four structures are attested for it:

(14) τιμὴ γίγvεται ‘there is honour’ (ø) >> 2 attestations
timē gignetai
honour.NOM become.PR.IND.ACT.3sg
(15) τιμὴ δίδοται ‘honour is given’ (by) >> 2 attestations
timē didotai
honour.NOM give.PR.IND.PAS.3sg
(16) τιμὴν λαμβάνω ‘I receive honour’ (from) >> also τιμῆς τυγχάνω
timēn lambanō timēs tunk^hanō
honour.ACC receive.PR.IND.ACT.1sg
(17) τιμὴν ἔχω ‘I have honour / I am honoured’ (ø)
timēn ek^hō
honour.ACC have.PR.IND.ACT.1sg

84The situation for χάρις k^haris in the classical ECF Leverhulme corpus is slightly more pointed in that the morphological passive option is not attested and the existential appears only twice (Aristotle, Rhetoric 1385a30, Demosthenes, Speech 18.36). Agent expressions with παρά para + genitive ‘from’ appear with the recipient passive. If indeed the only option to form a processual (as opposed to a statal) passive with support-verb constructions is lexical substitution of the support verb by means of a verb of receiving, could this qualify as a categorial periphrastic?

86Recipient passives can be the only option of passivisation that is permissible, e.g. German helfen ‘to help’ and rückerstatten ‘to reimburse’ do not morphologically passivise but recipient passives can be created with bekommen ‘get’ (Hilfe bekommen ‘to get help’, eine Rückerstattung bekommen ‘to get a reimbursement’). Greek morphological passives can be patient or recipient passives, similarly to English, but unlike e.g. German where we find a split between recipient passives and patient passives and only the latter can be morphological passives.

(18) to get a refund / to get refunded
The purchase will be refunded to you. AND You will be refunded for the purchase.

(19) zurückerstatten / eine Rückerstattung erhalten
Der Einkaufswert wird Ihnen (zurück)erstattet. BUT NOT #Sie werden (zurück)erstattet für den Einkaufswert.
‘The purchase will be refunded to you.’ BUT NOT ‘You will be refunded for the purchase.’
INSTEAD Sie erhalten eine Rückerstattung für den Einkaufswert.
‘You will receive a refund for the purchase.’

88Is χάριν ἀπολαμβάνω k^harin apolambanō distinct from χάριν λαμβάνω k^harin lambanō and should be considered a different phraseme (akin to to have a shower vs to take a shower (Leech 2009), or to do a favour vs to return a favour)? The two combinations differ in that ἀπολαμβάνω apolambanō is a compound verb with a prefix that seem to add intensifying or reciprocal semantics, whereas λαμβάνω lambanō is a simplex verb. Table 6 shows the numerical preference for the compound over the simplex in the classical period when both combinations are first attested. Percentages apply to columns:

89We may thus classify λαμβάνω lambanō as the canonical support verb but ἀπολαμβάνω apolambanō as a prototypical verb of realisation. Smith [2022: 73-74] distinguishes between prototypicality and canonicity as follows: ‘Prototypicality involves psychological salience, which is often the result of frequency, whereas canonicity is an analytical tool, which predicts that truly canonical forms are ‘likely to be rare or even non-existent’ … Prototypicality is an empirical psycholinguistic notion, whilst canonicity is a theoretical idealization’ (cf. also Brown et al. [2012 : 237] ; Finkbeiner [2008]).

90Table 7 catalogues by author the earliest attestations of both combinations, i.e. χάριν ἀπολαμβάνω k^harin apolambanō and χάριν λαμβάνω k^harin lambanō. The attestations are sorted diachronically from oldest to newest, approximate dates of birth and death for the authors are provided in round brackets (all BC) and the number of attestations in square brackets. The genre / text type is provided in italics with each author:

Table 7. Distribution across authors in classical times

ἀπολαμβάνω apolambanō	λαμβάνω lambanō
	Sophocles (497-406 BC) [1] Verse
	Euripides (480-406 BC) [2] Verse
	Critias (460-403 BC) [1] Verse
	Aristophanes (446-386) [1] Verse
Lysias (445-380) [4] Oratory Isocrates (436-338) [1] Oratory	Lysias (445-380) [1] Oratory
Xenophon (430-355) [3] Historiography	Xenophon (430-355) [1] Historiography
	Ephorus (400-330) [1] Oratory
	Lycurgus (390-325) [1] Oratory
	Hyperides (390-322) [1] Historiography
Demosthenes (384-322) [2] Oratory	Demosthenes (384-322) [2] Oratory
Aeschines (389-314) [2] Oratory
Aristotle (384-322) [1] Prose	Aristotle (384-322) [3] Prose
Axionicus (400-320) [3] Middle comedy
Demades (380-318) [2] Oratory

91Table 7 shows that while χάριν λαμβάνω k^harin lambanō starts out in verse literature (tragedy and comedy), χάριν ἀπολαμβάνω k^harin apolambanō seems to arise amongst Attic orators without ever ousting χάριν λαμβάνω k^harin lambanō. While χάριν ἀπολαμβάνω k^harin apolambanō seems to remain the preferred option in oratory, by the latter classical period, Aristotle in his prose writings prefers χάριν λαμβάνω k^harin lambanō which is going to numerically outdo χάριν ἀπολαμβάνω k^harin apolambanō from the Ptolemaic period onwards, as Table 6 above shows.

92For χάριν ἀπολαμβάνω k^harin apolambanō, there seems to be a gradual retreat into higher registers possibly along the lines of what we described in Section 4. Tables 8 to 10 show the development of both combinations over the course of the classical, Ptolemaic, and Roman periods with regard to the permissibility of modification and internal inflection, their appearance in verse as opposed to prose, and their co-occurrence with agent expressions pointing towards the lexical passive function still being in place. Percentages apply to rows.

	DP	ATT	Plural	Verse	Agent expression	Total
ἀπολαμβάνω apolambanō	3 17 %	5 28 %	–	3 17 %	6 (παρά para + genitive) 33 %	18
λαμβάνω lambanō	3 20 %	2 13 %	2 13 %	5 33 %	4 (παρά para + genitive) 27 %	15
Total						33

Table 9. Ptolemaic Greek

	DP	ATT	Plural	Verse	Agent expression	Total
ἀπολαμβάνω apolambanō	3 43 %	2 29 %	3 43 %	3 43 %	2 (παρά para + genitive) 29 %	7
λαμβάνω lambanō	3 18 %	6 35 %	–	7 41 %	3 (παρά para + genitive) 18 %	17
Total						24

Table 10. Roman Greek

	DP	ATT	Plural	Verse	Agent expression	Total
ἀπολαμβάνω apolambanō	2 15 %	4 31 %	1 8 %	1 8 %	5 (παρά para + genitive) 38 %	13
λαμβάνω lambanō	27 30 %	22 24 %	5 6 %	–	20 (παρά para + genitive [9], ἀπό apo + genitive [5], ἐκ ek + genitive [6]) 22 %	90
Total						103

93Tables 8 to 10 show that χάριν ἀπολαμβάνω k^harin apolambanō numerically yielded to χάριν λαμβάνω k^harin lambanō from the Ptolemaic period onwards. However, it became the preferred option in verse, i.e. in the higher registers. Agent expressions appear throughout comparatively speaking more often with χάριν ἀπολαμβάνω k^harin apolambanō than with χάριν λαμβάνω k^harin lambanō which may point to the passive value being in need of being indicated (cf. George [2005: 178]). The phraseme used regularly in Attic oratory became gradually associated with technical contexts and subsequently with higher registers (Bentein [2019]). However, indexing may also have happened through the route of the Attic dialect and Attic oratory as its exponent gaining a standing as elitist (cf. Atticism and the Second Sophistic).

94In the papyri, χάριν ἀπολαμβάνω k^harin apolambanō appears only once in a literary papyrus (p. Oxy. 13.1606 A Roll with Lysias, Contra Hippothersem, Contra Theomnestum and Four Other Unknown Orations). χάριν λαμβάνω k^harin lambanō, on the other hand, seems to have retreated into epistolary formulae (Fendel 2022: chap. 9), yet seems to have assumed a stative meaning:

(20) P. Michael 26, ll. 9-11 (unknown, AD 300, fragment of private letter) κέλευε δὲ ἐμοὶ ἐν οἷς βούλει καὶ γνώσε̣ι ὅτι καὶ χάριν λαμβάνω. 
keleue de emoi en ^hois boulei kai
order.IMP.2sg PRT I.DAT in REL.DAT want.2sg and
gnōsei ^hoti kai k^harin lambanō.
know.FUT.2sg that also gratitude.ACC take.PR.1sg
‘Tell me whatever you want (to have) and know that I am grateful (sc. to get it for you).’

(21) P. Neph. 4, ll. 23-25 (Alexandria, AD 301-400, private letter)
ἐγὼ γὰρ χάριν λαμβάνω, ὅτε καταξιοῖς παρʼ ἐμοῦ δέξασθαι ὁδήποτε. 
egō gar k^harin lambanō, ^hote kataxiois parʼ
I.NOM PRT gratitude.ACC take.PR.1sg that consider.right.PR.2sg from
emou dexast^hai ^hodēpote.
me.GEN receive.INF whatever.ACC
‘For I am grateful that you consider it right to receive anything from me.’

While in (20) and (21), χάριν ἀπολαμβάνω k^harin apolambanō appears in a fixed letter formula and the entity that the writer is grateful to is the addressee, in (22), the entity the writer addresses is God and the agent expression with ἐκ ek points to a lexical passive.

(22) SB 14 11494, ll. 1-9 (unknown, AD 401-500, prayer regarding eye illness)
⳨ λαβοῦσα χάριν ἐκ τοῦ μονογενοῦς̣ σου υἱοῦ στῆσον τὸ ῥεῦμα τοὺς πόνους τῶν ὀφθαλμῶν Φοιβάμμωνος υἱοῦ Ἀθανασίου ὁ κατοικῶν ἐπὶ βοη̣θίᾳ τ[ο]ῦ̣ ὑψί[στο]υ ἐν σκέπῃ [τοῦ Θεοῦ τοῦ οὐ]ρ̣ανοῦ αὐλισ[θήσεται -ca.?-]
labousa k^harin ek tou monogenous
receive.AOR.PTC.NOM favour.ACC from the.GEN of.one.blood.GEN
sou ^huiou stēson to r^heuma
you.GEN son.GEN stop.AOR.IMP.2sg the.ACC discharge.ACC
tous ponous tōn op^ht^halmōn P^hoibammōnos
the.ACC pains.ACC the.GEN eyes.GEN Phoibammon.GEN
^huiou At^hanasiou ^ho katoikōn epi
son.GEN Athanasius.GEN the.NOM inhabitant.NOM for
boēthia t[o]u ^hupsi[sto]u en skepē [tou
help.DAT the.GEN highest.GEN in cover.DAT the.GEN
T^heou tou ou]ranou aulis[t^hēsetai -ca. ?-]
God.GEN the.GEN heaven.GEN lie.in.the.courtyard.FUT.3sg
‘Having received a favour/blessing from your son of one and the same blood, stop the discharge, the pain of the eyes, of Phoibammon, son of Athanasius! For the help of the highest under the cover/protection of God of the heaven, the inhabitant will lie in the courtyard […]’

95One aspect that we need to consider is to what extent the structure may have changed meaning in later centuries as indicated by attributes but possibly also by the preference for the simplex over the compound verb. In other words, was a new phraseme born in later centuries? Given the attributes with the predicative noun, such as πνεύματος pneumatos ‘of the spirit’ (Origines, Selecta in Psalmos, 12.1637.13), διδασκαλικήν didaskalikēn ‘of/for teaching’ (Origines, Homiliae in Job, 17.80.52), θειοτέραν t^heioteran ‘more divine’ (Origines, Commentarii in evangelium Joannis, 6.6.35.8) in Roman times, it seems that a meaning ‘blessing’ has developed in Christian contexts (possibly as a nomen rei from the earlier nomen actionis ‘favour’ and nomen acti ‘gratitude’) (cf. Panagl [2020: 394]). This may underlie (22).

6. Diachronic reanalysis and the syntax-pragmatics interface

97The third and final support-verb-construction phraseme to be considered of the three selected for analysis, προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention to’, is the only one of the three that survives throughout the periods under investigation, yet after taking what Heine and Kaltenböck [2021: 5] describe as a ‘quantum leap’: ‘With the label ‘quantum leap’ we are referring to the operation of cooptation, whereby segments of sentence grammar, such as main clauses without their complements, are taken out of sentence grammar and transferred to the metatextual level of discourse processing […]. This operation […] does not involve parameters of grammaticalization.’ Consider English pay attention in which the form of the predicative noun (and the determiner phrase) is fixed and which relates to a specific context. Pay attention is used as a syntactically independent exclamative functioning at the discourse level.

98The support-verb construction seems related to earlier τὸν νόον / νοῦν ἔχω πρός ton noon / noun ek^hō pros (Aesop, Fables 40.1.3 τὸν νοῦν ὅλον ἔχων πρὸς τὸν οὐρανὸν ton noun ^holon ek^hōn pros ton ouranon ‘directing all attention to the sky’ (6^th c. BC), Aesop, Fables 198.1.8 ὅλον τε τὸν νοῦν πρὸς τὴν ὁπλὴν ἔχοντος ^holon te ton noun pros tēn hoplēn ek^hontos ‘directing all attention to the hoof’ (6^th c. BC)), which remains in use in classical times (e.g. Thucydides, Histories 3.22.5.6 πρὸς αὐτοὺς τὸν νοῦν ἔχοιεν pros autous ton noun ek^hoien ‘they may direct (their) attention to them’) and postclassical times (e.g. Josephus, Antiquitates Judaicae 16.221.3 καὶ τὸν νοῦν ἔσχεν πρὸς αὐτήν kai ton noun esk^hen pros autēn ‘and he directed (his) attention to her’ (1^st c. AD)) but which is outnumbered by προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention to’ in literary texts throughout periods of time and does not appear in the papyri. The present section considers προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention to’ from the perspective of why Mel’čuk’s pragmatemes are instances of rebirth rather than continuity (see (24) for his definition of pragmatemes).

99Table 11 provides an overview of the 1010 instances of the support-verb construction in question in the Thesaurus Linguae Graecae (Dataset 9). The columns word order, determiner phrase, and support-verb-construction field provide information about the permissibility of Mel’čuk’s modification, internal inflection, and word order changes. The columns on object patterns provide information about the morpho-syntactic integration of the support-verb construction. The columns ‘verse’ and ‘technical writing’ consider the text type in which the relevant attestations appear. Percentages refer to rows not columns.

Time period	Word order	DP	Field		Object			Alternative syntax	Verse	Technical writing	Total
	VN	Yes	Size	Type	DAT	Other	none
AG	0 %	50 %	0.5	VOC	50 %	50 %	–	100 %	100 %	–	2
CG	85 %	99 %	0.35	various	69 %	11 %	20 %	–	9 %	17 5 %	316
PG	81 %	98 %	0.26	ADV, DAT, PRT, 1x NOM (pronominal)	85 %	3 %	12 %	– (1× implied PN)	7 %	10 17 %	59
RG	91 %	99 %	0.41	various	73 %	9 %	18 %	(3× ; 3× PN implied)	(1×)	306 (Galen = 295) 78 %	393
EBG	83 %	96 %	0.39	various	73 %	13 %	14 %	8 % (+ 7× PN implied)	(1×)	118 49 %	240
TOTAL										451	1010

100There is only one instance in the papyri that appears neither in a literary papyrus (representing classical writing) nor shows the restoration of the predicative noun and determiner phrase in a damaged passage, which is problematic as Modern Greek προσέχω prosek^hō has the function of older προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention to’ without needing the predicative noun. This usage is foreshadowed in the New Testament (cf. Bauer [1957]; Lampe [1961]). The one relevant instance comes from the Roman petition P. Oxy. 8.1119 (l. 9). The support verb appears as a participle in the nominative case (masculine, singular) and is accompanied by a dative referring to the object of the attention.

101Heine and Kaltenböck’s [2021: 5] ‘quantum leap’ is often referred to as pragmaticalisation. Pragmaticalisation is the process by which an item / phrase moves ‘from a fully propositional (or referential) meaning to meanings that are more textual/discourse-oriented ... and/or interpersonal’ (Claridge & Arnovick [2010: 179]). This divorces the item / phrase from the sentence grammar, which is visible in the extension of scope (Claridge & Arnovick [2010: 180]), the decategorialisation ‘in the sense that the item loses morphological and syntactic properties that previously identified it as a member of a (more) major grammatical category’ (Claridge & Arnovick [2010: 181]), and the appearance of the item in extra-sentential positions (Claridge & Arnovick [2010: 181]; Koev [2022: 18]). However, the pragmaticalised and non-pragmaticalised patterns often co-exist. Debaisieux [2016] observes this e.g. for French parce que and labels the syntactic pattern ‘construction’ and the discursive usage ‘configuration’. Examples of pragmaticalised phrasemes include you know (Erman [2001]), I mean (Beeching [2011]), Latin rogo ‘I ask / please’ (Molinelli [2010]), postclassical (Egyptian) Greek ἀξιωθείς axiōt^heis ‘please’ (Fendel [2019]), and ἀμέλει amelei ‘of course’ (la [Roi 2022b]).

(23) Isocrates, Speech 2.11 Ὧν ἐνθυμούμενον χρὴ προσέχειν τὸν νοῦν, ὅπως ὅσονπερ ταῖς τιμαῖς τῶν ἄλλων προέχεις, τοσοῦτον καὶ ταῖς ἀρεταῖς αὐτῶν διοίσεις.
^hōn ent^humoumenon k^hrē prosek^hein ton
they.GEN consider.PTC.PR.GEN it.is.necessary pay.INF the
noun, ^hopōs ^hosonper tais timais tōn
attention.ACC that as.much the.DAT honour.DAT the.GEN
allōn proek^heis, tosouton kai tais
others.GEN surpass.PR.2sg so.much also the.DAT
aretais autōn dioiseis.
virtues.DAT they.GEN differ.FUT.2sg
‘It is necessary that (you) while considering this pay attention that inasmuch as you surpass the others in honour, you differ from them also in virtue.’

103The object clause in the form of ὅπως ^hopōs + future indicative (διοίσεις dioiseis) indicates that it is dependent on the support-verb construction as a whole as object clauses of this form in classical Greek only appear with verbs of making an effort (van Emde Boas et al. [2019: 526-527]). No component of the support-verb construction on its own qualifies as a verb of making an effort but the support-verb construction as a whole, as a lexical unit, does. This ties in further with Mel’čuk’s idioms.

104In Ptolemaic times, tighter constraints on the size of the support-verb-construction field and the type of permissible items in it seem to suggest a solidification of the support-verb construction as a reasonably invariable idiom. In the Roman period and very specifically Galen’s idiolect which dominates the Roman-period attestations, the support-verb construction seems to become a technical term. Galen, similarly to Hippocrates in classical times, seems to use the non-pragmaticalised support-verb construction in order to draw attention to the symptoms of a medical condition. The dative object refers to the symptoms the patient is showing. In Mel’čuk’s terms that would make it a pragmateme (similar to classical δίκην δίδωμι dikēn didōmi ‘to pay the price for one’s actions’ in legal contexts, cf. Section 4):

(24) Definition 9-1: Pragmateme (Mel’čuk [2023: 160])
‘A linguistic expression is called a pragmateme if and only if (iff) it is constrained pragmatically, that is, iff a special communication situation requires its use.’

Mel’čuk defines a ‘special communication situation’ as follows:

(25) Special communication situation (Mel’čuk [2023: 162])
‘A normal situation of linguistic communication is either an oral or written monologue (a narration, a journalistic or technical paper, a literary text), or a dialog. Then special situations of linguistic communication include:
– written warnings, prescriptions and instructions coming from an officially recognized authority, that is, texts on signs, on merchandise packing/labels, on medical prescriptions, etc.
– written messages: letters of all sorts, including circulars
– telecommunication: a telephone conversation, texting, emails
– oral signs: commands, warnings, calls for help
– ritualized social situations.’

In the above, we have preferred the more concise sociolinguistically based definition along the lines of genre, register, and style by Biber and Conrad [2009] (adopted by Bentein [2019]) and the functionally driven definition of a ‘quantum leap’ from the morphosyntactic onto the interpersonal level by Heine and Kaltenböck [2021]. In the early Byzantine period, the link between technical contexts and the idiom becomes more pronounced.

(26) Aristophanes, Clouds l. 575 (Classical Greek (CG))
ὦ σοφώτατοι θεαταί, δεῦρο τὸν νοῦν προσέχετε.

ō sop^hōtatoi t^heatai, deuro ton
oh most.sophisticated.VOC spectators.VOC here the.ACC
noun prosek^hete.
attention.ACC pay.PR.IMP.2pl
‘most sophisticated members of the audience, pay attention right here!’

(27) Phrynichus, Praeperatio sophistica Frg 133, l. 2 (Roman Greek (RG))
†ἄκουε, σίγα: Κρατῖνος [fr. 284] ‘ἄκουε, σίγα, πρόσεχε τὸν νοῦν, δεῦρ’ ὅρα’.
†akoue, siga: Kratinos [fr. 284]
listen.PR.IMP.2sg be.quiet.PR.IMP.2sg Kratinos.NOM [fr.284]
‘akoue, siga, prosek^he ton
listen.PR.IMP.2sg be.quiet.PR.IMP.2sg pay.PR.IMP.2sg the.ACC
noun, deur’ ^hora’.
attention.ACC here look.PR.IMP.2sg
‘listen, be quiet, Kratinos [fr. 284] ‘listen, be quiet, pay attention, now look!’’

(28) Eusebius, Praeperatio evangelica 11.13.4.3 (Early Byzantine Greek (EBG))
ὅμως δ’ ἐννόει καὶ σφόδρα πρόσεχε τὸν νοῦν
^homōs d’ ennoei kai sp^hodra prosek^he
nonetheless PRT consider.PR.IMP.2sg and very pay.PR.IMP.2sg
ton noun;
the.ACC attention.ACC
‘nonetheless, consider (it) and pay attention very much!’

106From a conversation-analytic point of view, προσέχε(τε) τὸν νοῦν prosek^he(te) ton noun highlights the following piece of information. The speaker/writer is likely to use the phrase when they hope that heightened attention to what is said will result in the preferred response (e.g. acceptance after a request or agreement after an assessment) (Emde Boas [2017: 11]). The ‘quantum leap’ seems to happen in spoken registers (comedy and the Platonic dialogues) rather than primarily written registers but may be imported into technical writing secondarily. Although, note that (27) seems a quote from Plato’s Kratinus and is thus not indicative, (28) is more informative in this regard.

107If we apply Debaisieux’s distinction between a construction and a configuration to the two (!) phrasemes προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention to’, we would qualify the idiom in technical (especially medical) writing as a construction and the discursive guidepost that emerges in dialogic contexts as a configuration.

7. Summary and conclusion: Birth and propagation of support-verb-construction phrasemes in classical and post-classical Greek

108The present chapter investigated the birth and rebirth(s) of three classical and postclassical Greek support-verb constructions: δίκην δίδωμι dikēn didōmi ‘to pay the price for one’s actions’, χάριν ἀπολαμβάνω k^harin apolambanō ‘to receive a favour’, and προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention’ by means of Mel’čuk’s lexicographic Sens-Texte framework (and with reference to the PARSEME Natural-Language-Processing inspired framework). Support-verb constructions cut across Mel’čuk’s classes of semantic-lexemic phrasemes, which are divided into compositional collocations and non-compositional idioms, and pragmatemes. The chapter explored the diversity, or internal heterogeneity, within the group of support-verb constructions, which is visible not only synchronically (e.g. diatopic and diastratic variation) but also diachronically due to support-verb constructions hovering at three interfaces. The chapter drew its data from the Thesaurus Linguae Graecae, which compiles literary texts from the archaic to the early modern periods and the Duke Database of Documentary Papyri, which compiles documentary texts of the postclassical and early medieval periods.

109Mel’čuk’s [2023: 71-72] morphological, morpho-syntactic, and syntactic tests were used in order to differentiate between types of semantic-lexemic phrasemes. For these, Section 3 highlighted the issue of a reverse-engineered canonical form (in order to carry out tests) for corpus languages in general, as the native speakers of a corpus language are the texts (Fleischman [2000]), along with the Greek-specific issue that support-verb constructions withstand morphological passivisation (Mel’čuk’s test 1). The section proceeded to introduce the metric of the support-verb-construction field, that is the space between the support verb and the predicative noun when they appear in an active, declarative clause. This metric was first tested on French data, a living language for which reverse engineering of a canonical form is unproblematic, and a language that allows for morphological passivisation in some support-verb constructions. Furthermore, word order in modern French is syntactically driven (unlike in (post)classical Greek) such that there are some predictable items in the support-verb-construction field (e.g. negatives, determiner phrases, adverbs) (cf. Mel’čuk’s test 4). The French data allowed us to gain an impression of how the permissibility of operations such as passivisation (Mel’čuk’s test 1), and the degree of permissibility of modification and internal inflection (Mel’čuk’s tests 2 and 3) correlate. Given that other verbal multi-word expressions in (post)classical Greek show the same sensitivity to the number and type of items intervening between their components, the metric was adopted for the three case studies in Sections 4 to 6.

110Using δίκην δίδωμι dikēn didōmi as the test case, we asked how changes in analyticity and compositionality affect the status of a phraseme (as a collocation vs idiom) (Section 4). It appeared that while δίκην δίδωμι dikēn didōmi existed as a collocation ‘to give judgement(s) / to judge’ from archaic times and into the medieval period, in classical times, the idiom δίκην δίδωμι dikēn didōmi ‘to pay the price for one’s actions / to get punished’ arose. The idiom became indexed for technical (legal) and subsequently higher registers but survived beyond the classical period primarily only in quotes. The non-analytic and non-compositional idiom acted as a word rather than a syntagm, as agent expressions show.

111Using χάριν ἀπολαμβάνω k^harin apolambanō ‘to receive a favour’ as the test case, we asked what degree of lexical renewal we accept as modification of a phraseme before positing a new phraseme (Section 5). We broke this overarching question into two sub-questions first evaluating χάριν ἀπολαμβάνω k^harin apolambanō ‘to receive a favour’ vs χάριν ἀποδίδωμι k^harin apodidōmi ‘to return a favour’ against Haspelmath’s notion of categorial periphrasis as a lexical passive within the support-verb-construction family around the predicative noun, and subsequently χάριν ἀπολαμβάνω k^harin apolambanō vs χάριν λαμβάνω k^harin lambanō from the perspective of prototypicality as opposed to canonicity of the support verb. Due to the semi-lexical nature of the support verb, Haspelmath’s notion of categorial periphrasis does not seem to apply perfectly to χάριν ἀπολαμβάνω k^harin apolambanō ‘to receive a favour’. The prototypical compound (ἀπολαμβάνω apolambanō) seems to be diatopically (cf. English to take a shower vs to have a shower (Leech [2009])) and subsequently diastratically (cf. English to do a favour vs to return a favour) indexed but retreats into higher registers after the classical period; the canonical simplex verb (λαμβάνω lambanō) predates it in early classical verse and postdates it as well as forms the base for the later phraseme ‘to receive a blessing’ without reciprocity implied.

112Using προσέχω τὸν νοῦν prosek^hō ton noun ‘to pay attention to’ as the test case, we asked why Mel’čuk’s pragmatemes are instances of rebirth rather than continuity (Section 6). προσέχω τὸν νοῦν prosek^hō ton noun exists as an idiom, in the sense of Debaisieux’s construction, especially in medical discourse from at least classical times onwards (cf. Hippocrates). προσέχω τὸν νοῦν prosek^hō ton noun, however, also exists as a pragmateme (with the support verb in the imperative and no object expressed) in the sense of Debaisieux’s configuration, from archaic times onwards especially in dialogic contexts there serving as a discursive guidepost. The pragmateme has taken a ‘quantum leap’ (Heine & Kaltenböck [2021: 5]) out of the sentence grammar (morpho-syntactic level of functional grammar) and into the discourse grammar (interpersonal level of functional grammar). Thus, there is no continuity but rupture in the transition.

113Table 12 summarises the phrasemes discussed and the sub-types identified, their period of birth, and their context(s) of application:

Phraseme	Meaning	Mel’čuk’s type	Time period(s)	Context(s)
δίκην δίδωμι dikēn didōmi + agent expression	to pay the price for one’s actions	non-compositional semantic-lexemic phraseme (idiom)	CG only	technical writing (legal texts) – higher register
δίκην δίδωμι dikēn didōmi + dative or equivalent	to give judgement / to judge	compositional semantic-lexemic phraseme (collocation)	from AG onwards	n/a
χάριν ἀπολαμβάνω kharin apolambanō	to get a favour in return	compositional semantic-lexemic phraseme (collocation)	CG primarily	Attic oratory – higher register
χάριν λαμβάνω kharin lambanō	to receive a favour	compositional semantic-lexemic phraseme (collocation)	from CG onwards (with possible rebirth in postclassical times)	n/a
προσέχω τὸν νοῦν prosekhō ton noun	to pay attention	non-compositional semantic-lexemic phraseme (idiom)	from AG onwards	technical writing – medical texts
προσέχω τὸν νοῦν prosekhō ton noun support verb in the imperative, no object	to pay attention	pragmateme	from AG onwards (marginal)	primarily dialogic contexts initially

114If implemented into the PARSEME terminology, which is Natural-Language-Processing inspired and fiercely synchronically focused, (i) δίκην δίδωμι dikēn didōmi + agent expression would be a VID (verbal idiomatic expression), (ii) δίκην δίδωμι dikēn didōmi + dative or equivalent an LVC.cause (light-verb construction causative), (iii) χάριν ἀπολαμβάνω k^harin apolambanō and χάριν λαμβάνω k^harin lambanō would be NotMWE (not a verbal multi-word expression) because lexical passives are not implemented currently, (iv) προσέχω τὸν νοῦν prosek^hō ton noun would be an LVC.full (light-verb construction full), and (v) προσέχω τὸν νοῦν prosek^hō ton noun with the support verb in the imperative and no object expressed would currently also be marked as an LVC.full (light-verb construction full).

115In authoritative dictionaries, such as Liddell-Scott-Jones, a consistent implementation of support-verb constructions would be necessary. The following shows the cataloguing (or not) of the support-verb constructions discussed:

(29) Liddell-Scott-Jones (LSJ)
s.v. νόος / νοῦς noos / nous I.1.2.b mentions in brackets προσέχω τὸν νοῦν prosek^hō ton noun as the equivalent of τὸν ν. ἔχειν πρός τινα ton n. ek^hein pros tina or τι ti ‘to have one’s mind directed to someone / something’;
s.v. δίκη dikē IV.3 lists δίκην dikēn or δίκας διδόναι dikas didonai ‘to suffer punishment’;
s.v. χάρις ΙΙ.2 (‘gratitude’) lists χάριν λαβεῖν τινος k^harin labein tinos receive thanks from one; ἀπολαβεῖν παρά τινων apolabein para tinōn;
III.1.a (‘favour’) lists χ. δοῦναι k^h[arin] dounai= χαρίζεσθαι k^harizest^hai.

116προσέχω τὸν νοῦν prosek^hō ton noun fares best with even its predecessor mentioned briefly, although no distinction is made between the idiom and the pragmateme. For δίκην δίδωμι dikēn didōmi, only the idiom is mentioned yet with some inaccuracy as to the permissibility of internal inflection, whereas the collocation is not catalogued. For χάριν ἀπολαμβάνω k^harin apolambanō, the lexical passive is mentioned but the distinction between simplex and compound verbs remains unexplained. Another issue appears in the paragraph after where the corresponding active χάριν δίδωμι k^harin didōmi is equated with the formation with derivational suffix -ίζω -izō, which as explained in Section 4 does not have the same function morpho-syntactically, semantically, or pragmatically. Thus, while authoritative dictionaries show some awareness of all three support-verb constructions (and the corresponding six phrasemes identified), the entries would require an update to reflect their actual diversity.