Interjections as pragmatic particles in Modern Greek: Using diverse corpora in identifying pragmatic functions

Full Text

1. Introduction

Interjections across the world’s languages have typically been treated as elements of emotional expression, whereas their pragmatic roles and functions have received comparatively limited attention. In Greek in particular, as the following section will discuss, the three basic interjections–α /a/ ‘ah, huh’, ε /e/ ‘eh’, and ω/ο /o/ ‘oh’–have not been extensively studied, despite their high frequency in everyday conversation (see evidence in 4.1). This paper seeks to address this gap by examining the frequency, functions and distribution in text types of these items in a range of corpora consisting of extensive, authentic Greek data. As will be shown, corpus evidence is crucial for revealing the full scope of their usage patterns, especially the relative frequency–and therefore the pragmatic significance–of their various functions. The study’s broader aim is to investigate how interjections evolve pragmatic functions in discourse that extend well beyond the mere expression of emotion.

2. Interjections and pragmatics

Interjections have been largely neglected in the linguistics literature, at least until the 1990s when the seminal Ameka (1992) is published,¹ among else, due to their intrinsic relation to emotion, which has been overlooked by the predominantly referentialist view of language in 20^th century linguistics (Wilce 2009: 39). It is indicative that Sapir, for instance, believes that “[i]deation reigns supreme in language […] volition and emotion come in as distinctly secondary factors” (1921: 38–39) or that “the emotional aspect of our psychic life is but meagerly expressed in the build of language” (1921: 217). It was only in the last few decades that the full extent of linguistic devices related to emotion has become apparent (see e.g. Dewaele 2010, Foolen 2012, Wilce 2009: 39ff.) to the extent that Taboada, in full reversal of Sapir’s pronouncement, finds that “the linguistic expression of emotions and opinions is one of the most fundamental human traits” (2016: 326).²

At the same time, the simplistic connection of interjections with emotion seems to have been taken as an endpoint in the related discussion, somehow associating them with pre- or non-linguistic material which is not amenable to further investigation, rather than as the foundation on which further uses and functions have been developed in the world’s languages. For instance, Triandaphyllidis’ standard grammar of Modern Greek defines interjections as monosyllabic, non-declinable words that denote feelings³ and specifies that a relates to admiration and query, e to scorn, address, irony, regret and disgust and ο (written either as ο or ω) to query, pain, sorrow, and address (1949 [1976]: 203–204). Similarly, the three major Modern Greek dictionaries define the meaning of a, e and o as related to “intense emotion” (Charalambakis 2014), “several emotions” (Triandaphyllidis Dictionary 1998) or “several intense emotions” (Babiniotis 1998), ranging from joy to terror, although they also recognize uses such as vocatives, phrases of address or self-standing answers to questions.

There are two exceptions to this undifferentiated treatment of the interjections a, e and o, which both rely on the study of authentic data. The early Tzartzanos’ grammar (1946 [1963]: 143–145) treats only e as an “interjectional particle” and, on the basis of mostly literary data, carefully distinguishes several uses, including a vocative use, pragmatic implications of contrast, conclusion or disagreement, an invariant tag use in commands and questions, that can also have ironic or disapproving overtones, and a self-standing use of surprise or disapproval. The second exception comes from three recent studies of a from a Conversational Analysis perspective, published in the same volume (Christodoulidou 2020, Karachaliou 2020, Pavlidou 2020), which all point to its use as a pragmatic marker of change-of-state, following the analysis of oh in English by Heritage (1984). Although these studies open up a new perspective for the treatment of such elements in Greek, due to the methodological framework followed, it is not easy to place the particular pragmatic function identified in the items’ overall patterning and thus evaluate its contribution to their meaning. It is also not clear whether the roughly sixty examples discussed in all three studies were selected precisely for their exemplification of the meaning focused upon or are representative (and to what extent) of their broader use.

It is important to note that a, e and o are phonologically and morphologically minimal and produce syllable-like utterances consisting of one vowel sound. In fact, the corresponding vowel phonemes take up low (central) and medial (front and back) position in the five-vowel phonological system of Greek⁴ and as such are formed by some of the most basic elements to be found in natural languages. They are thus potential candidates for manifesting universal properties through their prototypical uses in the sense of Dingemanse, Torreira & Enfield (2013) and, as a result, of further significance beyond the grammar of Greek. Their phonological and morphological simplicity is one among the features of the category of inserts in Biber et al.’s (1999) terms, pragmatic particles in Beeching’s (2002) terms or interactives in Heine’s (2023) terms.⁵ Other features include (see Beeching 2002: 53, Biber et al. 1999: 1083 ff, Heine 2023: 12, 31, Heine et al. 2024):

a) morphological invariability; they are non-declinable in the case of Greek,

b) syntactic non-attachability: they are not part of a larger grammatical structure, although they may be prosodically attached to a clausal unit or may be self-standing,

c) placement in initial or final positions, and only rarely in the middle of the clause (pace Beeching 2002, cf. Georgakopoulou & Goutsos 1998 for Greek),

d) lack of denotative meaning: their meaning is dependent on their pragmatic function,

e) lack of homonyms in other word classes,

f) inability to negate or question them,

g) multifunctionality,

h) typical occurrence in spoken rather than written discourse,

i) association with informality, which may occur with stylistic stigmatization.

The items under study in this paper seem to conform to these criteria and thus are candidates for belonging to these larger categories. This issue will be revisited at the final section of the paper, but for now it suffices to point out that interjections like a, e and o potentially have characteristics that are shared by other elements in a wider category of items with multiple pragmatic functions and a broader discourse role.

As a final note, it would be interesting to compare Modern Greek interjections with their Ancient Greek counterparts, to which they are etymologically related and which have been extensively studied (e.g. Nordgren 2015). The multifunctionality of items like a, e and o makes it possible to suggest possible paths of language change, while a corpus investigation may be also exploited for sociolinguistic explanations e.g. of the kind suggested by Denis & Tagliamonte (2016).

3. Data and methodology

The data drawn upon for this study come from a variety of corpora. Our starting point is the approximately 300,000-word conversational sub-corpus of the Corpus of Greek Texts (CGT). CGT is a 30-million-word general reference corpus, including a broad range of spoken and written genres from 1990–2010 (Goutsos 2010). Its conversational sub-corpus includes 87 texts of informal, everyday conversation between two or more intimates (e.g. friends or family), in what is considered to be the prototypical genre of spoken discourse.

Conversational data is compared to two other genres of contemporary data which involve speech representation, that is literary data and film dialogue. Literary data come from the 2.6-million-word literary sub-corpus of CGT, comprising a variety of sub-genres including novels, short stories, poetry, theatrical plays, song lyrics etc. Film data come from the Corpus of Greek Film Dialogue (CGFD), which was created with the purpose of studying Greek cinematic discourse (Goutsos 2025) and includes transcribed dialogues from 105 films in Greek from nine decades with an approximate size of 900,000 words. For purposes of comparison 30 films were selected from the 1990s and 2000s, the same period as that covered by CGT. It is significant that there were only a few instances of a, e and o in genres other than conversation, literature or film dialogue in CGT and these were excluded from further investigation.⁶

Our investigation also draws evidence from data of an earlier period of Greek drawn from the Diachronic Corpus of Greek of the 20th century (CGT20). CGT20 is a 4-million-word general reference corpus including a variety of genres from the first nine decades of the 20^th century, that is 1900–1989 (Goutsos et al. 2017). CGT and CGT20 belong to a family of Greek corpora, covering the whole of the 20^th and the first decade of the 21^st century. It was again found that only a handful of instances (less than 4%) of a, e and o occur in genres other than literature or film speech and were thus excluded from further investigation. (Comparable conversational data are not included in the CGT20). Table 1 summarizes the data used in this study.

 Table 1. Corpora used in the study 

Concordances were extracted for a, e and o in these four corpora and were cleaned in order to identify instances of interjectional uses. This means excluding other occurrences of the letters α, ε, ο and ω e.g. as part of other words, in listing (a, b, c …) etc. Other cases that were excluded are occurrences of the vowels as part of other interjections e.g. αχ, εμ, ωχ, α πα πα etc., instances in songs that are used for metrical purposes (e.g. α, μια νύχτα πριν να σβήσουν τ’ άστρα ‘a:: a night before the stars go out’ LIT-1950-0175) and the homophonous injunctive α /a/ or άι /ái/ in examples like α να χαθείς από ‘δώ πέρα ‘Get lost!/Get out of here!’ (FILM-1980-0002), which is etymologically non-related to the interjection a (see Babiniotis 1998, Triandaphyllidis Dictionary 1998).

The next step has been to assign a function to the items under investigation, on the basis of its context in a concordance line of 5 words to the left and 5 to the right of the node word (cf. Columbus 2010) and then try and classify individual examples to larger categories.

4. Findings

This section presents the findings of the study for the frequency and functions of a, e and o in the analysed corpora.

4.1. Frequency

Table 2 displays the frequencies of the three items under investigation in conversation, literary texts, film dialogues and the diachronic data of CGT20, presenting their raw (N) and normalized frequency per 10,000 words.

Table 2. Raw and normalized frequency of a, e and o in the data

The figures in Table 2 suggest that a, e and o are especially frequent in conversation, although o is clearly much less frequent than the other two items, as there are roughly 100 e and 40 a in every 10,000 words of spontaneous conversation in Greek but only one ο. Instances are much fewer in literary data: a roughly occurs slightly more than two times in every 10,000 words of literary data both in contemporary and diachronic data (20 times less than in conversation data), o occurs once or less (similar to conversation), whereas e occurs four-five times in literary and diachronic data (also 20 times less than in conversation). Certainly, speech representation takes up a small portion of literary genres, as it mainly appears in dialogic parts of fiction and drama and much less in other literary genres. Film dialogue, which aspires to represent actual conversation more faithfully, has normalized frequencies for a, e and o that are closer to conversation, namely four times less for a, less than half for e and the same for o.⁷ Even so there is a pronounced difference from non-scripted conversation.

It is interesting to compare the relative frequency of all three items in the respective corpora, as presented in Figure 1, which breaks down figures for CGT20 into literary and film data.

Figure 1. Relative frequency of a, e and o in the data

As can be gleaned from Figure 1, o occurs proportionally more in literary data, especially from the beginning of the 20^th century, whereas it is almost extinct in contemporary conversation.⁸ By contrast, a has relatively similar frequency across data (roughly one third of all items) and e takes the lion’s share (roughly two thirds of all items) with a slight exaggeration in contemporary film data and a slight underrepresentation in diachronic literary data (one fourth of all items).

Overall, it seems that e and a are especially frequent in conversation and thus potentially significant for spoken interaction. It is indicative that e is among the 15 most frequent items in the data, following grammatical words like και ‘and’, το ‘the’, να ‘to’, ναι ‘yes’, δεν ‘not’, είναι ‘is’ etc., and a is among the 50 most frequent items in conversation, whereas o is much less frequent. This finding is consistent with the overall picture found in English (see e.g. Beeching 2016: 34, Biber et al. 1999: 1053ff, 1096–1097), although in reverse, as the English oh is much more frequent than ah, something which must be related to the range of functions covered by the items in question. As Heritage (2018: 157) has observed, in Early Modern English oh’s “frequency of use accelerated, partly at the expense of a(h), which served similar functions in Middle English”.

At the same time, the representation of conversation in literature and film differs to a large extent from actual conversation both in terms of the overall frequency of these devices and in terms of the proportional distribution of the items involved; literature tends to overrepresent o, especially in earlier data, while contemporary film dialogue tends to overrepresent e. As such, literature and film seem to involve a specific view of what actually happens in conversation.

4.2. Functions

4.2.1. a

The investigation of a in the four corpora of the study has indicated that there are nine main uses associated with it, namely:

a) exclamation: a is used as a typical interjection e.g. of joy as in (1) or terror as in (2):

(1)   […] ο Στέφανος τις γελούσε δεύτερη φορά. — Α, α! ξεφώνιζαν λαχταριστά οι δυο μικρές […] (LIT-1910-0005)

       Stefanos was playing with them for a second time. — Ah, ah! the two small ones would cry delightedly

(2)   <Σόλων> ΆΤΙΜΟ ΣΌΙ (.) ΘΑ ΣΑΣ ΣΦΑΞΩ <Λία> Α:: Α:: ΠΩ ΠΩ ΠΩ ΠΩ:: (FILM-1960-0002)

       <Solon> DISGRACEFUL FAMILY (.) I WILL KILL YOU <Lia> ah:: ah:: oh Gosh

b) surprise and aside: this use involves the perceptual element of sudden encounter or recollection of something on part of the speaker. The meaning of surprise is only found in literary data and recent film dialogue, as in (3) and (4):

(3)   […] καλημέρισε την παρέα του. </p> <p> « Α, αυτή κι αν είναι έκπληξη!» κατάφερε να πει εκείνος (LIT-5001)

       […] said good morning to his friends. “Ah, that was a surprise!” he managed to say

(4)   <Γυναίκα 3> πάμε στην θάλασσα (.) τώρα <Γυναίκα 2> α εδώ είστε (FILM-2000-0015)

       <Woman 3> let’s go to the see (.) now <Woman 2> ah that’s where you are

Conversation, instead, only has the meaning of aside, especially in set phrases that introduce a new topic, further information (5) or a story (6):

(5)   <Δ> [η κυρία Γ είναι σούπερ (.) <Σ> καλά (.) α δε σου 'πα ξεκίνησε και μαθήματα κομπιούτερ (CONV-0004)

       <D> [Mrs G is super (.) <S> fine (.) ah I didn’t tell you she started computer classes

(6)   <Α> της στέλνω μήνυμα όντως μου απαντάει (.) α εντωμεταξύ δεν σας είπα (.) μου λέει πώς το βλέπεις (CONV-0029)

       <A> I send her a message and she does reply to me (.) ah meanwhile I didn’t tell you (.) she says how do you see this

The same device is found in literary and film data:

(7)   επιταγή είναι δεν τη βλέπεις; επιταγή είναι <Δημήτρης> Α:: και δε μου λες <Ρένα> ορίστε <Δημήτρης> ποιος την έφερε; (FILM-1980-0002)

       It’s a cheque can’t you see it? It’s a cheque <Dimitris> ah:: tell me something <Rena> sure <Dimitris> Who brought it?

(8)   (ΠΑΕΙ ΣΤΗΝ ΠΟΡΤΑ) <ΤΖΕΝΗ> Α... ξέχασα να σου πω... τηλεφώνησε η φίλη σου (LIT-5023)

       (GOING TO THE DOOR) <Jenny> ah… I forgot to tell you… your friend called

Obviously, this is a handy device for changing the topic or introducing a new element in discourse.

c) address: in this use a may accompany vocatives with proper names, as in (9), or — more rarely — religious invocations such as α παναΐα μου ‘a my Virgin Mary’ (CONV-5008), α Χριστέ μου ‘a my Christ’ (FILM-1940-0002):

(9) <Τ> Λουίτζι; <Λ> α Τζιοβάνι εδώ είσαι (.) να περάσω; (CONV-0032)

       <G> Luigi? <L> ah Giovanni here you are (.) may I come in?

This use also includes cases in which the speaker draws the interlocutor’s attention e.g. by objecting to something they say or do, as in (10):

(10) <Αντώνης> άσε εκεί δα να σου πατήσω ένα φιλάκι <Ρούλα> Α:: ΑΦΗΣΕ ΜΕ ΚΑΛΕ (.) ΔΕ ΘΕΛΩ (FILM-1950-0004)

       <Antonis> let me give you a kiss right there <Roula> AH:: LET ME BE, YOU (.) I DON’T WANT TO

d) evaluation: this involves reference to a third person or object (in the nominative or accusative) to express admiration or reproach:

(11) […] και να κάνει και τις κινήσεις να δείχνει <Π> α ρε φοβερός <Ε> να λέει […] (CONV-0001)

       […] and he would do the motions he’d show <P> ah re⁹ incredible <E> he’d say

(12) <B> και τι σας είπε; <Δ> ότι τα πήρατε εσείς <B> α το κάθαρμα (.) ώστε εξακολουθεί να επιμένει […] (FILM-1960-0002)

       <B> and what did he tell you? <D> that you took them <B> ah the scoundrel (.) so he keeps on insisting

e) intensification: a is commonly used as an intensifier as in (13), mainly along with evaluatives (α μπράβο/ωραία), with response signals of agreement/assent (α βέβαια/μάλιστα/ναι/γεια σου) or denial/disagreement (α δε μπορώ/όχι/μπα) or with social formulae (α ευχαριστώ/συγγνώμη) (see Heine 2023) as in (14):

(13) […] λεμονανθοί, τριαντάφυλλα, η θάλασσα. Α, πώς μοσκοβολούσε η θάλασσα! (LIT-0003)

       […] lemon flowers, roses, the sea. Ah, how fragrant was the sea!

(14) Οι σοφοί αρνήθηκαν. Α, μπα! δεν είνε δυνατόν. Μα ο Ευμορφόπουλος επίμενε. (LIT-1900-0001)

       The wise men refused. Ah, nah! It’s not possible. But Evmorfopoulos insisted.

f) implicature: this use of a is related to the change-of-state marking that has already been studied in Greek (see Section 2) and involves an implicature that the person speaking has now realized the truth of their interlocutor’s general proposition as in (15) or that a specific point has been clarified as in (16). It also involves specific implicatures such as drawing a conclusion from an element of the surrounding context as in (17).

(15) <Δ> είναι για το πώς θα σφουγγαρίσει ((προς τον Π)) <Π> α <Χ> ναι επειδή δεν ξέρω πώς να σφουγγαρίζω (CONV-0050)

       <D> it is about how he will mop the floor ((to P)) <P> ah <X> yes because I don’t know how to mop

(16) <M> το κουνουπίδι ογδόντα λεπτά <Χ> το ψωμί παιδί μου <Μ> α:: το ψωμί; α εγώ κουνουπίδι άκουσα (CONV-0045)

       <M> eighty cents for cauliflower <X> for bread <M> ah:: for bread? ah I heard cauliflower

(17) ο παπάς ιδόντας τα κατσικάκια και τες γίδες, είπε: — Α!.. βλέπω απόψε, Μαριανθούλα, έχετε και φιλινιάδες! (LIT-1900-0004)

       The priest, seeing the goats and their babies, said: — Ah!... I can see, dear Marianthi, that you have some girl friends with you tonight!

As suggested from the examples above, a may be used in self-standing utterances with this meaning, especially in cases of general implicature.

g) use in set phrases and deictic mentions: a can also be used in set phrases like α καλά as in (18) or by speakers to deictically refer to a previous use in discourse as in (19).

(18) είναι:: στο- το Ελαϊκόν είναι στο Γουδί <Χ> α:: καλά πολύ μακριά (CONV-0011)

       It’s in the- in Elaikon it’s in Goudi <X> ah:: OK very far

(19) δεν είναι ούτε μηχανικός ούτε αρχιτέκτων <Ναύαρχος> α <Θωμάς> πολλά α μου λες ναύαρχε <Ναύαρχος> τα σταματώ και προχωρώ (FILM-1960-0001)

       He’s not an engineer nor an architect <Admiral> ah <Thomas> you say many ah, admiral <Admiral> I quit them and go on

In (19) the first a is related to the drawing of an implicature that the admiral does not further specify; Thomas comments on this by his use of a. Both these examples suggest an increased degree of conventionalization, in the sense that speakers rely on well-established meanings of the item used.

h) filled pause: a, usually prolonged, may be used in Greek as a filler in pauses or a hesitation marker (see Heine 2023: 134). Most examples come from conversation (20–21), although some instances are found in other genres, too, as in (22).

(20) το:: μέτρο πρέπει να 'ναι πάνω απ' όλα α:: δεν αφουγκραζόμαστε την ψυχική μας διάθεση ούτε τίποτα (CONV-0062)

       the right measure must be above all ah:: we don’t listen to our mood or stuff

(21) με μια Μαρία (.) δε σ' το 'πα; <ΠΑ> α ε μου το 'χεις πει βασικά αλλά δεν […] (CONV-0029)

       with one Maria (.) didn’t I tell you? <PA> ah eh you have told me basically but I didn’t […]

(22) <ΣΠΥΡΟΣ> καλημέρα σας δεσποινίς <ΛΙΛΑ> α:: ε:: καλημέρα κυρ-Σπύρο <ΣΠΥΡΟΣ> σας χάσαμε δεσποινίς (FILM-1940-0003)

       <Spyros> good morning miss <Lila> ah:: eh:: good morning Mr Spyros <Spyros> long time no see miss

As suggested by (21) and (22), a can precede e as a filler, whereas instances of the reverse have not been found in the data.¹⁰

i) invariant tag in questions: a special use of a, found only in Greek data from Cyprus, both authentic (23) and scripted (24) conversation, involves a word added, usually at the end of an utterance, to elicit the listener’s response or agreement (Columbus 2010), presumably with a special intonational contour:

(23) <Ε> τελικά εν κρυάδα την νύκτα α; <Π> την νύκτα εν κρυάδα αλλά τούτες τες μέρες […] (CONV-5006)

       <E> so it’s quite cold at night ah? <P> at night it’s cold but these days […]

(24) <ΚΛΕΙΤΟΣ> Ποιος σου έμαθε να λέεις όλα τούτα τα ψέματα; Α; <ΤΕΥΚΡΟΣ> Ξέρεις πως δεν είναι ψέματα... (LIT-5040)

       <Kleitos> Who taught you to say all these lies? Ah? <Teukros> You know they’re not lies…

This function is taken up mostly by e in data from mainland Greece (see 4.2.2).

Table 3 presents the frequency of all uses of a in the four corpora.

Table 3. Functions of a in the data

The evidence in Table 3 suggests that a is mainly used as a marker of implicature, an intensifier or a filled pause in conversation and this is generally mirrored in other genres, as is also shown in more detail in Figure 2.

As can be seen in Figure 2, literature and film present a wider range of functions, with relatively more emphasis to evaluation and address. By contrast, filled pause is more significant in conversation but underrepresented in literary and filmic data. In all, it seems that scripted conversation takes advantage of the functions manifest in non-scripted conversation with the exception of filled pauses, which are less frequent in it.

Figure 2. Distribution of functions of a in the four corpora

3.2.2. e

Although e is at least twice as frequent as a in the data studied, there are less functions found for it in the four corpora, namely:

a) exclamation: this use is not particularly frequent and only occurs in literary data, as in (25) and (26):

(25) Ορχήστρα (χασαποσέρβικο) <ΚΑΡΑΓΚΙΟΖΗΣ> Ε, ώπα, ώπα, ώπα, έξω φτώχεια. (LIT-0002)

(Orchestra plays a tune) <Karagiozis> eh, opa, opa, opa, down with poverty

(26) […] τότε μία φωνή βραχνή και νυσταλέα, αλλ' απότομος. — Ε! βάρδ' απ' τα περιβόλια! Ανοιχτά! . . . Ανοιχτά! (LIT-1910-0003)

[…] then a voice hoarse and sleepy, but brisk. — Eh! straight from the gardens! In the open!... In the open!

b) address: in this use, e is found together with a proper or a common name in the vocative as in (27) or with an interlocutor that is clearly retrievable from context as in (28), in order to draw their attention.

(27) ((κοιτάζουν για λίγο τηλεόραση)) <Γ> ε παιδιά λέει για τα καλαμάρια που ξανασερβίρουνε (CONV-0008)

       ((watch TV for some time)) <G> eh guys it says about squid that gets served again

(28) […] κλείνοντας την πόρτα πίσω του. «Ε, περίμενε!...» Η Τάρκιν την άνοιξε, αλλ' αυτός είχ' εξαφανιστεί (LIT-5084)

[…] shutting the door behind him. “Eh, wait!...” Tarquin opened it, but he was gone.

This makes e a typical attention signal like hey in English (see Heine 2023: 116) that can also be extended to cases like (29) in which it issues a warning:

(29) […] συμβουλές σ’ αυτό το- σ’ αυτό το διαμάντι <Φίλος> ε ε μαζέψου <Γιάννης> ναι σε είδαμε κι εσένα (FILM-2000-0002)

       […] advice to this- this real gem <Friend> eh eh cut it off <Giannis> yes we know about you

c) evaluation: as with a (see 4.2.1), e can occur with evaluative phrases of admiration or disapproval referring to a person, as in (30), or an object (31), or expressing a general assessment of a situation as in (32).

(30) <Π> έλα ρε μωράκι ((γέλια)) <ΠΑ> ε ρε το μουρλό <Π> ναι και:: (CONV-0029)

       <P> c’mon re baby ((laughter)) <PA> eh re the crazy guy <P> yes and::

(31) <Γρηγόρης> δεν έχω φτιάξει; κοίτα ((κάνει φιγούρες)) Ε ΡΕ ΚΟΡΜΟΣΤΑΣΙΑ (.) ποια θα την πάρει (FILM-2000-0002)

       <Gregory> am I not better? Look ((striking poses)) EH RE WHAT A BODY (.) which girl is going to take it?

(32) […] για τις άσπρες μπλούζες τι τους κάναμε, να πέσεις κάτω. Ε, ρε, πλάκα. (LIT-5016)

       […] what fun we had with them for the white blouses-to fall down with laughter. E re fun

It is interesting that re (see note 4) co-occurs with e in most of these examples.

d) intensification: e is commonly used as an intensifier in cases of hedge as in (33). More frequently, it accompanies evaluatives (ε χαλαρά/ωραία), response signals of agreement/assent/concession (ε ακριβώς/βέβαια/(ε)ντάξει/ναι/σίγουρα) or denial/disagreement (ε μα/όχι/σιγά) or with social formulae (ε με συγχωρείτε/σόρι/συγγνώμη) (see Heine 2023) as in (34) and (35):

(33) […] και τούτο εκ θεού είναι. Ε, τι λες, τυχερή δεν ήμουνα, παιδί μου; (LIT-5015)

       […] and that comes from God. Eh, what do you think, wasn’t I lucky, my child?

(34) <Ι> ναι και την καρδιά σου θα ρωτήσεις <ΑΙ> ε ακριβώς <Ι> τι λες; (CONV-0001)

       <I> yes and you’ll ask your heart <E> eh exactly <I> what do you think?

(35) <Τσίτσης> άνοιξε θέλω να του μιλήσω <Βασιλική> ε δεν μπορώ τώρα να ανοίξω <Τσίτσης> είναι ανάγκη (FILM-1990-0008)

       <Tsitsis> open up I want to talk to him <Vassiliki> eh I can’t open right now <Tsitsis> it’s urgent

In most of these examples, e introduces an element of hedging or concession and that may be one aspect of its difference from a, which may also account for the different collocates of the two items in this use.

d) implicature: e invites several implicatures that refer either to the general proposition of an interlocutor’s statement (36) or a specific aspect of this (37). The meanings of these implicatures relate to cause, consequence (38), conclusion (39), correction (40), narrative continuation (37) or dismissal of a proposition (“so what”, e.g. 41). In many cases the comment following e is presented as dependent on accepting the interlocutor’s proposition as true (“if one accepts this, then this follows”, e.g. 38). Some of the collocates like άμα ‘if’, αφού ‘since’, οπότε ‘therefore’, και μετά ‘and then’, και τι έγινε ‘what happened’, λοιπόν ‘so, well’ (only found in earlier data and not in contemporary conversation, cf. Tzartzanos 1946 [1963]: 143) clarify the particular implication intended.

(36) να φύγουνε από την ΠΑΣΠ τότε αφού το καταλάβανε <Ρ> ε αφού τους διέγραψε (CONV-0004)

       they should get out of [this organization] if they realized this <R> eh since they were removed [as members]

(37) τελικά δώδεκα η ώρα γυρίσαμε από το θείο μου <Ι> ε και τι έγινε δώδεκα η ώρα βγαίνει ο κόσμος (CONV-0009)

       finally we come back at twelve o’clock from my uncle’s <I> eh so what, twelve o’clock is when people go out

(38) <Β> δε θα κάτσω θα φύγω <Ε> ε θα φτιάξουμε κάτι γρήγορο να φάμε <Β> θα φύγω (CONV-0043)

       <V> I’m not staying I’m going <E> eh we’ll make something quick to eat <V> I’m going

(39) γίνεται τόσο από μπαχαρικά όσο απ' τα ίδια τα υλικά ε οπότε λογικά τότε φτιάχνανε κάπως έτσι (.) δηλαδή μια πίτα (CONV-0002)

       it’s made both from spices and the ingredients themselves eh hence logically then they made it somewhat like this

(40) <Αλέξανδρος> τραγουδίστρια; <Ηλέκτρα> ε τραγουδίστρια τώρα δεν την έχω ακούσει ποτέ να τραγουδάει (FILM-1990-0012)

       <Alexandros> a singer? <Elektra> eh singer now I haven’t heard her sing

(41) από μετριοφροσύνη σκίζεις πάντως τι να σου πω; <ΑΓ> ε τι να κάνω; αυτογνωσία (CONV-0015)

       you’re super humble what can I say? <AG> eh what I can do? pure self-knowledge

In all cases above there is an overtone of obviousness: the implication that is drawn in the utterance followed by e is considered to be obvious or self-evident by the speaker.

e) use in set phrases and deictic mentions: like a, e can also be used in set phrases like ε καλά, reminiscent of the English ‘well’, as in (42) or by speakers to deictically refer to a previous use in discourse as in (43), in which the speaker uses e to comment on a previous use of an apology.

(42) <Μ> χτες χτες τρεις παρά τέταρτο <Λ> ε καλά άσε τώρα τις λεπτομέρειες (FILM-2000-0012)

       <M> yesterday yesterday at a quarter to three <L> eh fine (‘oh well’) leave out the details now

(43) <Ε> δεν ήταν άγριος και λέει η Σταυρούλα συγγνώμη ε τι συγγνώμη λέει τι συγγνώμη; (CONV-0047)

       <E> he was not rough and Stavroula says sorry eh what sorry she says what sorry

f) filled pause: as mentioned above, e (usually lengthened) is the main item occurring as a filler in pauses or hesitation marker in Greek (Heine 2023: 134). This is the most common use in conversation (44) and one of the more frequent uses in the other corpora (45), something which suggests that this is a well-established, conventional function of the item in question.

(44) <Δ> το πλαστικό <Μ> δεν ε:: <Κ> αποσυντίθενται [εύκολα <Μ> [αποσυντίθενται (CONV-0010)

       <D> plastic <M> is not eh:: <K> decomposed [easily <M> [decomposed

(45) […] πώς και δεν εμπιστεύτηκε σ' αυτόν την... ε... ε... την... την... αδυναμία να την πω; (LIT-0003)

       […] how didn’t he confide to him this… eh… eh… this… this… weakness so to say?

g) invariant tag in questions: e is the invariant tag per excellence in Greek, as pointed out in 3.2.1. It is a non-declinable item added, usually at the end of the utterance, to elicit the listener’s response, agreement or confirmation (Columbus 2010). This use is the most frequent one in film and CGT20, the second most frequent in the literary corpus and the third more frequent in conversation, something which suggests that it is very well-established both in non-scripted (46) and scripted (47) conversation.

(46) <Δ> πού πονάει; <Σ> ε:: ψηλά στον ώμο <Δ> στον ώμο ε; (CONV-0006)

       <D> where does she hurt? <S> eh:: up in the shoulder <D> in the shoulder e?

(47) «Λίγο πολύ μαντεύεις τι σου επιφυλάσσεται». «Έτσι λες, ε; Έτσι...» «Έτσι, παιδί μου» (LIT-0003)

       “More or less, you guess what is in store for you.” “So you think eh? So…” “So, my child”

There is obviously much scope for analysis of examples like these above in order to clarify further uses of the invariant tag, as is done e.g. in Denis & Tagliamonte (2016), Stubbe & Holmes (1995), Westphal (2024), among many else, for eh in English.

Table 4 presents the frequency of all uses of e in the data and Figure 3 their relative distribution in each corpus.

Table 4. Functions of e in the data

Figure 3. Distribution of functions of e in the four corpora

As was also found in 4.2.1 for a, conversation seems to give emphasis to the use of e in filled pauses and as an invariant tag in questions. Pragmatic uses like implicature and intensification (which always carries a pragmatic overtone) are equally prominent. By contrast, literary data seems to exploit the implicature uses of e, followed by that of filled pause and invariant tag and its use for address, while film gives emphasis on similar uses but overrepresents the use of invariant tag at the expense of the filled pause. Overall, however, uses of e seem to be conventionally well-established and this is borne out in the fact that their distribution does not differ that much in non-scripted and scripted conversation.

3.2.3. o

Most of the uses of o occur in the literary and film corpora rather than conversation. In particular:

a) exclamation: o is used as an interjection mainly in literary texts, either in cases of admonition like in (48) or in exclamations with a non-definable purpose as in (49):

(48) <Λία> Ω:: Ε ΤΩΡΑ (.) Ω Ω Ω Ω:: <Σόλων> σιγά σιγά (.) σιγά πουλάκι μου σιγά (FILM-1960-0002)

       <Lia> o:: e now (.) oh oh oh oh:: <Solon> take it easy (.) easy my bird easy

(49) […] και τίμησες τη φτωχική μου τάβλα». «Ω! Από τη Νεάπολη έρχομαι. (LIT-0003)

       […] and you honoured my poor table”. “Oh! I come from Neapoli

b) surprise: this seems to be a prototypical use of o in Greek, found in conversation (50) and exploited for dramatic effect in literary and film representations of speech (51):

(50) <Ε> θα συμφωνήσω δε θα 'θελα αλλά με προκαλείς <Π> ω:: παιδιά (CONV-0001)

       <E> I will agree I wouldn’t want so but you provoke me <P> oh:: guys

(51) ((ΑΛΛΑΓΗ ΣΚΗΝΗΣ)) <ΚΛΕΙΤΟΣ> Ω!!! Να και η αγαπητή μου γυναικούλα... (LIT-5039)

       ((Change of scene)) <Kleitos> Oh!!! That’s my lovely little wife…

c) address: this is the most common use of o in literary and film texts. It is clearly one of the earliest, prototypical uses of o, reaching back to the non-obligatory component of the vocative, preceding names and other forms of address in Classical Greek (Heritage 2018: 157, Nordgren 2015: 95).

(52) […] να μη σκεφθής, να μη συλλογισθής τίποτε. Ούτε τη μάνα, σου ακόμη. Ω παιδί μου! Κώστα μου! Τι συμφορά! (LIT-1900-0002)

[…] do not think, do not consider anything. Neither even your mother. O my child! My Kostas! What a disaster!

Apart from the vocative found in the examples above, a special case of address is found with the accusative as in the following:

(53) <Ναύαρχος> καλημέρα σας κύριε Βασιλείου <Θωμάς> ω:: τον αγαπητό κύριο ναύαρχο (.) στας διαταγάς σας (FILM-1960-0001)

       <Admiral> good morning Mr. Vasileiou <Thomas> oh:: dear Mr. Admiral (.) at your command.

Another use is found with taboo items, either swearwords (54) or religious words (55), both in the vocative, and this is one of the most frequent uses in earlier texts:

(54) «Τόχεις να κάτσης πάλε στα χαρτιά;» «Ω διάβολε, λογαριασμό θα σου δώσω!» (LIT-1910-0018)

       “Do you intend to sit down to play cards again?” “Oh devil (‘Damn you’), I won’t answer to you!”

(55) ΠΕΤΡΑΚΗΣ: Μακάρι να ήθελαν να με συλλάβουν! Κα ΠΕΤΡΑΚΗ: Ω θεέ μου... ΠΕΤΡΑΚΗΣ: Εκτελέσεις (LIT-1970-0001)

       Petrakis: I wish they meant to arrest me. Mrs Petraki: Oh my God… Petrakis: Executions

In modern film dialogue and in the only instances in conversation o occurs with swear or taboo words:

(56) <Σ> άρχισε πάλι <Χ> ω ρε γαμώτο απόψε (FILM-2000-0014)

<S> He started again <X> oh re damn tonight

d) evaluation: as opposed to address, this use, again not found in conversation, involves reference to a person or object in the accusative (57) or the nominative (58), which is the object of admiration.

(57) μια κυρία ψάχνοντας για σένα <Λακάκης> για μένα; <Φάνης> ναι <Λακάκης> ω την καημένη (FILM-1970-0003)

A lady looking for you <Lakakis> for me? <Phanis> yes <Lakakis> oh the poor one

(58) Ιδού ο περίβολος των νεκρών! Ω! ο Παράδεισος, απ' αυτόν τον κόσμον ήδη, ήνοιγε τας πύλας του (LIT-1910-0003)

Behold the garden of the dead! Oh! Paradise already from this world opened its gates

e) intensification: this is another use found in conversation, also occurring mainly in literary texts, involving co-occurrence with evaluatives like (ω ωραία), response signals of agreement (ω ναι/(ε)ντάξει/βέβαια/ασφαλώς, mainly in literature) or disagreement (ω μπα/μα/όχι, also mainly in literature) or with social formulae (ω ευχαριστώ, in literature). Conversation examples like (59) markedly differ from literary and film examples like (60) and (61) in the degree of formality or obsolescence of expressions:

(59) <Α> [ώπα <Ε> [δώσε <Π> [ω βοή (.) α:: (CONV-0001)

<A> [opa <E> [cheers <P> [oh racket (.) ah::

(60) το σπίτι (.) η κουζίνα (.) το κρασί (.) είναι στη διάθεσή σου <Τασία> ω σας ευχαριστώ πάρα πολύ (.) μα ξέρετε; είμαι πολύ κουρασμένη (FILM-1930-0001)

the house (.) the kitchen (.) the wine (.) are at your disposal <Tasia> oh thank you very much (.) but you know I’m very tired

(61) […] θα μας δώσουν την κληρονομιά μας.... — Ω! δυστυχία μας! έβγαλε δυνατή φωνή η κυρά Πανώρια. (LIT-1900-0001)

[…] they will give us our heritage… — Oh! disaster! (‘Woe to us’) cried loudly Mrs. Panoria

Examples like (61) in which o is found in the set phrases ω συμφορά/δυστυχία are characteristic of earlier literary texts.

f) implicature: this use is only found in earlier literary and film texts and involves the drawing of a conclusion as in (62) and (63):

(62) κι ύστερα έσκυψε και μου φίλησε τα χέρια <ΖΙΖΗ> ω:: (.) τώρα λοιπόν πρέπει να παντρευτείτε <ΛΙΛΑ> και βέβαια (FILM-1940-0003)

He then leaned down and kissed my hands <Zizi> oh:: (.) now then you have to het married <Lila> of course

(63) ΣΟΛΩΝ: Δεν το βλέπεις πού είναι; ΛΟΥΚΑΣ: Το γραφείο; ΣΟΛΩΝ: Ω! Είσαι και τυφλός. (LIT-1970-0009)

Solon: You can’t see where it is? Lukas: The desk? Solon: Oh! You’re deaf too.

A similar use in contemporary conversation would rather involve a or e.

g) use in set phrases and deictic mentions: like a and e, o can be used in the set phrase ω του θαύματος ‘o what a miracle’ in literary texts and as in (64) in order to comment on a previous use:

(64) ΣΟΛΩΝ: (Βγάζει ένα μήλο.) Ω! ΛΟΥΚΑΣ: Ω ; Τι ω; ΣΟΛΩΝ: Πάψε, το παίζουμε... (LIT-1970-0009)

Solon: (takes out an apple) O! Lukas: Oh? What oh? Solon: Shut up, we’re playing…

Table 5, which presents the frequency of all uses of o in the data and Figure 4, which shows their relative distribution in each corpus, confirm the impression from individual examples that contemporary conversation only has few uses of o, whereas film and — mainly — literary data, especially from an earlier period have both more uses and more emphasis on uses like address, evaluation or implicature, which are absent from conversation. To this extent, the findings about o deviate from those about a and e.

Table 5. Functions of o in the data 

Figure 4. Distribution of functions of o in the four corpora

4. Discussion and conclusions

Our investigation of a, e and o in four corpora of Greek, including texts from authentic, non-scripted conversation, contemporary and earlier literature and film dialogues, has unearthed a wealth of material for these items both in terms of their frequency and their functions. First, it has been found that a and e are extremely common in conversation, whereas o is quite marginal, and this frequency is not equally reflected in scripted conversation, as in literature o and in film e tend to be overrepresented. Overall, the frequency of these items in scripted conversation is much less than that in non-scripted, suggesting that speech representation in literature and film involves careful word selections exploited for specific effects.

Items like a, e and o that are commonly characterized as interjections have an extensive range of functions that can be thought of as belonging to a continuum from less to more conventionalized use. Thus, uses of a, e and o for exclamation, in vocatives of address or attention signals are much closer to the prototypical function of interjections as elements of emotive (rather than emotional) communication. As Caffi & Jenney suggest, emotive communication is “inherently strategic, persuasive, interactional and other-directed by its very nature” (1994: 329), something which underlies their purposefulness and explains their use for evaluation. Norrick aptly points out that “primary interjections function in the participation and information frameworks of discourse, rather than simply signaling emotional involvement” (2008: 461).

Further along the line, these items involve cognitive communication, indicating a sudden change in the cognitive state of the speaker (Heine 2023: 185) such as the “sudden discovery of something” (Nordgren 2015: 95), found in uses of surprise or aside. Further implicatures involving change-of-state in the interlocutor’s knowledge (in the case of a) or drawing a conclusion that is presented as obvious (in the case of e) are developed on this basis and are exploited for various effects.

Finally, uses of a, e and o for intensification are further conventionalized, as they only add an element of hedging or boosting, supporting agreement or disagreement etc., without cognitive implications. Interpersonal uses such as that of hesitation markers (filled pauses) or invariant tags in questions that invite the interlocutor’s involvement, assent etc. also serve textual purposes of demarcating specific parts of the interaction. The use of the analysed items in fixed phrases or as deictic mentions of previous uses is placed further along the continuum of conventionalization.

Comparing Modern to Classical Greek, as well as contemporary with early 20^th century Greek, with respect to this continuum of conventionalization may be revealing of general tendencies. For instance, Modern a, e and o seem to present a much wider range of functions than their Ancient Greek counterparts ἆ, ἔ ἔ, αἴ and ὤ. Nordgren (2015) points to the blend of cognitive and emotive content in Ancient Greek interjections, but Modern Greek seems to employ an even broader development of cognitive, interpersonal and textual functions. Without doubt, further investigation relating these two remote periods to each other through Medieval Greek is necessary for arriving at meaningful conclusions. Furthermore, the comparison of the beginning of 20^th century to late 20^th and early 21^st century data suggests a progressive loss of functions for o, the predominance of e and a general tendency for more conventionalized functions.

In order to account for the derivation of pragmatic implicatures from primary emotive uses it is imperative to distinguish purely interjectional from other functions. In general, the label “interjection”, either considered to refer to an individual part of speech or not, seems to cover a wide range of actual uses that make the items characterized thus more akin to pragmatic particles (Beeching 2002), inserts (Biber et al. 1999) or interactives (Heine 2023, cf. Heine et al. 2024), that is elements with a rich contribution to interactive discourse, both in non-scripted and scripted conversation in our case. Clearly, much further work is needed to precisely identify the function of each item in context and tease out their multifunctionality, something which is hard to do with extensive corpus data. Notwithstanding this, the contribution of corpora in the exploration of the pragmatic functions of interjections is indispensable (cf. Norrick 2008: 461), not only because of their advantages in pattern finding, systematicity, generalisation, reproducibility and transparency, pointed out by Landert et al. (2023: 7–8), but also because of their immense help in identifying what is central and what is peripheral in language and thus evaluating the significance of our findings.

_{¹ The editors have rightly pointed out that this mainly holds true for Anglo-Saxon literature. Bobińska (2015), Buridant (2003) and Stange (2016: 5ff.) offer useful entry points to a rich, even if erratic, line of research on interjections.}

_{² Certainly, this delay is related to the fact that the alternative lines of thinking introduced by Bakhtin/Voloshinov, Benveniste or Wittgenstein have not been sufficiently integrated into mainstream linguistic research.}

_{³ Greek does not distinguish between emotion, sentiment, feeling and affect in the same way as English; the word συναίσθημα ‘sinésθima/ is the most frequent catch-all term.}

_{⁴ The other two vowel sounds, the high /i/ and /u/ can also be used as interjections of extreme pain or disgust and disapproval, respectively, but are much less frequent and seem to be marginal in terms of their uses and functions.}

_{⁵ Beeching (2016: 3) includes a long list of terms used for these elements, ranging from discourse markers to hedges and boosters. Along with the — perhaps, unavoidable — proliferation of labels, there have only been few attempts to clear the terminological confusion (see e.g. Degand, Cornillie & Pietrandrea 2013).}

_{⁶ As is the case in CGT20, too, these mainly occur in magazines e.g. in interviews and -to a less extent- in private letters. Stange (2016: 6), who finds similar results in the BNC, draws the conclusion that “in writing, too, interjections are actually a feature of spoken language”.}

_{⁷ It must be noted here that film data come from modern films, which tend to give a more realistic picture of conversation (cf. Goutsos 2025: 192 ff.).}

_{⁸ Although we need a much more detailed diachronic analysis, covering the “intermediate link” of Medieval Greek (see 4 below), we can speculate here that the prototypical association of o with exclamation in high-flown literary texts (see 3.2.3) has rendered it a high register item to be avoided in everyday (low) conversation. (English seems to have followed the opposite direction). One must also not preclude the effects of language contact (see Boček 2015).}

_{⁹ re is a non-denotational item that functions as a marker of intimacy.}

_{¹⁰ In the absence of detailed annotation for prosodic features we cannot comment on the difference between prolonged and non-prolonged vowels. This is also true about writing conventions for these items, which, as known, are notoriously inconsistent.}

About the authors

Dionysis Goutsos

Author for correspondence.
Email: dgoutsos@phil.uoa.gr
ORCID iD: 0000-0002-2204-9740

PhD, is Professor of Text linguistics at the Department of Linguistics at the School of Philosophy

References

Supplementary files

Supplementary Files

Action

1. JATS XML

Download

2. Figure 1. Relative frequency of a, e and o in the data

Download (86KB)

Indexing metadata

3. Figure 2. Distribution of functions of a in the four corpora

Download (90KB)

Indexing metadata

4. Figure 3. Distribution of functions of e in the four corpora

Download (92KB)

Indexing metadata

5. Figure 4. Distribution of functions of o in the four corpora

Download (74KB)

Indexing metadata