The emergence of stand-alone insubordinate conditional clauses in Hungarian

1 Introduction

The study of the emergence of insubordinate clauses is fraught with multiple and severe challenges for the researcher: prior discussions of the history of these clauses are few and far between (cf. D’Hertefelt 2018: 220), while quite a few rival theories have been proposed with respect to their emergence. Historical corpora do not necessarily represent the appropriate genres and registers (or at least not in sufficient quantities) in which such sentence types could typically occur. To make matters worse, their various subtypes include a high number of transitional instances, and their syntactic/pragmatic independence or their actual functions are also debatable in many cases. Nevertheless, online diachronic databases constitute extremely valuable source materials for the exploration of these clauses: they allow us to review long periods of time and large quantities of data, even if insubordinate clauses tend to occur in them in small numbers compared to subordinate clauses involving the same conjunctions, and even if their exploration invariably requires manual annotation.

The present paper studies the emergence of stand-alone insubordinate conditional clauses in four Hungarian diachronic databases (ÓMK, TMK, KED, MTSz, see Section 4) whose combined materials cover the period between Old Hungarian and Contemporary Hungarian (1200–2010), that is, more than eight hundred years. Our main aim is to answer the research question of what role interjections might have played in the emerging independence of these conditional clauses.

Our interest in this topic was aroused by two prior explorations in this general area. One of them yielded the observation that, with respect to insubordinate complement clauses introduced by the conjunction hogy ‘that’, the earliest attested data also involved interjections (Dér 2022). The other prior exploration was a questionnaire study (Dér 2023a) supporting the claim that insubordinate optative clauses expressing desire and involving conditional conjunctions (ha ‘if’, hogyha ‘on the condition that’) are more acceptable for native speakers if they also include interjections (ó ‘ah’, jaj ‘oh my’) than if they lack such elements.

This paper is structured as follows: Section 2 introduces various types of insubordinate clauses and briefly summarises some theories concerning their emergence. Section 3 describes interjections occurring alongside insubordinate clauses, followed by Section 4 in which we survey the databases used and the methods of analysis employed. Section 5 presents the results of this study organised in terms of the periods of the history of Hungarian, and enumerates the possible ways that such insubordinate clauses emerged. Finally, Section 6 presents our conclusions.

2 The types and emergence of insubordinate clauses

Complex sentences including conditional clauses typically consist of a subordinate clause known as a protasis, preceding the matrix clause and giving a condition, and an apodosis, the matrix clause following the subordinate clause, where “the truth of the proposition in the matrix clause is a consequence of the fulfilment of the condition in the conditional clause” (Quirk et al. 1985: 1088, cited by Lastres-López 2020: 24), for instance:

Conditional constructions come in several diverse kinds, and the typical way in which they are semantic expanded is when ideational conditionals with a descriptive referential meaning turn into interpersonal or textual conditionals (Halliday and Matthiessen 2014, summarised by Lastres-López 2021: 46–50). The last stage of that process is when the subordinate clause of the conditional construction becomes insubordinate (Lastres-López 2021: 149).

By insubordinate clauses we mean, citing a classic definition by Evans (2007: 367), cases where clauses retain their formal features as subordinate clauses but are conventionalised as matrix clauses in language use. Two dominant subtypes are usually differentiated: elaborative clauses that are only syntactically independent and stand-alone clauses that are independent both syntactically and pragmatically. The latter type is also described by attributes such as “performative”, “autonomous”, and “self-contained”, given that they have a specific meaning in isolation, too (see Debaisieux et al. 2019; D’Hertefelt 2018; Kaltenböck 2019). Some researchers also claim that they are prosodically delimited as well (e.g. in Spanish, they are characterised by a specific intonation and boundary-marking stress: Elvira-García et al. 2017). Consider an English (2), a Spanish (3) and a Hungarian example (4) for stand-alone versions:

As far as Indo-European languages are concerned, book-length studies have been published on insubordinate clauses in Germanic and Romance languages (mainly Spanish and French, see D’Hertefelt 2018; Lastres-López 2021). These studies describe two major productive groups of stand-alone varieties: complement clauses that are introduced by a complementizer meaning ‘that’ and insubordinate conditional clauses that involve conjunctions meaning ‘if’ (D’Hertefelt 2018).

Conditional clauses that are syntactically free but pragmatically non-independent are referred to by D’Hertefelt (2018) as “post-modifying conditional constructions” giving an extra condition. She uses the term “elaborative” for complement clauses of this kind and takes them to be functionally equivalent with the former type. Post-modifying conditionals elaborate on an earlier utterance, making it more accurate. In the following example, the clause expresses a further condition, attaching it to Speaker A’s prior utterance. The exchange takes place in a conversation between a dog trainer (A) and the dog’s owner (B):

It is important to note that D’Hertefelt (2018: 170) takes the latter type to have emerged by dependency shift, rather than by insubordination (that is, by ellipsis of the matrix clause, see below for details).^[1]

Several ideas have been raised with respect to the emergence of insubordinate clauses (cf. Cristofaro 2016: 395–399; Lastres-López 2021: 38–40); these will be introduced briefly in the order of their having been proposed.^[2] It is important to see that these theories did not all come about to describe the emergence of the same construction: some approaches address only stand-alone clauses, while other approaches also address (other) insubordinate clauses, and moreover, the same construction can emerge in a language in more ways than one.

The earliest idea was Evans’ ellipsis- or redeployment-hypothesis (Evans 2007; Evans and Watanabe 2016; cf. Heine et al. 2016; Schröder 2016): this stated that stand-alone insubordinate clauses had evolved from complex sentences by the omission of the matrix clause. This ellipsis-based construction was later conventionalised, and the clause was constructionalised as a matrix sentence (Evans 2007: 375; cf. Brinton 2014: 286). Thus, for instance, in the case of insubordinate clauses expressing modality, predicates expressing thinking, apperception, emotions, or appreciation were ellipted, while in the case of insubordinate clauses expressing indirectness or interpersonal control, predicates describing a wish, permission, command, etc. were omitted. In insubordinate clauses, linking elements perform discourse linking (cf. “redeployment of linkages from intra-clausal to general discourse links”, Evans 2007: 370). D’Hertefelt (2018: 220) claims that synchronic studies support the ellipsis hypothesis in stand-alone cases, where the process can be gradual, not merely abrupt or instantaneous. However, the ellipsis theory raises a few issues according to Van linden and Van de Velde (2014: 241): does it mean that the matrix clause is invariably recoverable? And is it uniquely recoverable in all cases, or can we think of several ways of recovering it?

Mithun (2008, 2019) describes the emergence of insubordinate clauses as extension (“extension of dependency”), during which the use of sentence-level markers extends beyond the sentence level, to the discourse situation, and such sentences describe contextual (background) information. In her view, there was no matrix clause that would have been omitted from beside the subordinate clause. D’Hertefelt (2018: 168) thinks that Mithun’s idea correctly explains the emergence of elaborative constructions that “are used to describe the different subevents of one more complex situation”.

Brinton (2014) uses the idea of constructionalisation taken from construction grammar in the case of metalinguistic constructions like if you like/choose/want etc. – that can be taken to be elaborative in view of the foregoing – for explaining their emergence, meaning the ways “in which ambient linguistic patterns may give rise to the emergence of new constructions of different degrees of schematicity and substantivity” (Hoffmann and Trousdale 2011: 12).

Croft’s (2000) concept of hypoanalysis was revived by Van linden and Van de Velde (2014: 228, 240) as a notion complementing Evans’ ellipsis hypothesis in the explanation of Dutch (semi)autonomous subordinate clauses^[3] with the illocutionary force of exclamations. Hypoanalysis is “[a] form-function reanalysis such that a contextual semantic/functional property is reinterpreted as an inherent property of a syntactic unit” (Van linden and Van de Velde 2014: 241). In the present case, subordinate constructions with the conjunction dat acquired an interpersonal meaning in a non-subordinate context, expressing strong emotional involvement on the part of the speaker.

Cristofaro (2016) takes several paths of emergence to be possible for the same conditional form (e.g. for an insubordinate clause expressing request), though she does not, in some cases, reject the ellipsis hypothesis either. Another way of emergence that differs from the foregoing is clausal disengagement, in which the relationship between the matrix clause and the subordinate clause weakens, and the latter can begin to occur on its own as time goes by (such are some uses of German weil or English because – these are syntactically independent elaborative cases, and their occurrence is characteristic of conversations).

Another explanation we should mention here is the cooptation account of insubordinate clauses (Heine et al. 2016), according to which the operation of cooptation reallocates linguistic units taken from sentence grammar (SG) to thetic grammar (TG)^[4] – this is what also happens to subordinate clauses in the present case. The linking unit they include now serves discourse linking, too, and such (in)subordinate clauses are grounded in the discourse situation. Kaltenböck (2019: 178) explains the productivity of stand-alone insubordinate conditional clauses (as opposed to non-conditional ones) by saying that a general semantic relationship emerges in them between two states or events, and hence it becomes possible that they get detached from preceding discourse.

D’Hertefelt (2018: 147–179) explains the insubordination of elaborative and post-modifying conditional clauses by a dependency shift that resembles Mithun’s extension of dependency: “This is a mechanism that serves to analyse cases in which subordinating conjunctions shift away from under the scope of their main clause and develop towards more ‘coordinate-like’ uses” (D’Hertefelt 2018: 160–161). She does not hold the idea of ellipsis probable in these cases since the utterances they modify work as a kind of anchor for them (and can be interpreted as their matrix clauses at the propositional level). Since there is no omitted main clause, there is also no main clause to reconstruct, as the conditional clause remains linked to its former matrix clause in a kind of dependency relationship.

Debaisieux et al. (2019: 372) explain the emergence of French exclamative clauses like si tu savais ‘if you knew’ in a way that differs from all the foregoing ones: “the construction has involved stance framing, with no apodosis, from the beginning”. That is, they assume that this construction was an exclamative main clause to begin with, an idea that is supported by the fact that the transfiguration of a declarative subordinate clause into an exclamative insubordinate clause is unlikely (assuming an omitted main clause one has to take it to be declarative, too): “It would not be easy to explain how the integration of a declarative clause in the subordinate clause results in a shift of illocutionary force from assertive to exclamative, including a change in prosodic contour” (Debaisieux et al. 2019: 371).

3 Interjections in insubordinate clauses

In a comprehensive paper reviewing the history of linguistic descriptions of interjections, Ameka (1992) defines interjection as a word-level unit, a part of speech that expresses speakers’ mental state and their action/attitude/reaction with respect to the speech situation (and is hence strongly context-dependent). Primary interjections are short, stressed, typically monosyllabic words or non-words^[5] (Ouch!, Wow!, Gee!, Oho!, Oops!) that can constitute an utterance in themselves, and typically do not form constructions with other words (Ameka 1992: 105–106; cf. Dingemanse 2023; Traugott 2014). Secondary interjections have a semantic value of their own, conventionally reflect mental states or attitudes (e.g. Help! Fire! Careful! Damn! Shame!), and typically emerge from content words by conversion (Deák 1936: 11). Interjections can be classified into expressive (e.g. wow!), conative (e.g. sh!, eh?) and phatic ones (e.g. uh-huh, yeah). Interjections are syntactically (relatively) independent, and as utterances they can be classified as exclamations.

In Hungarian part-of-speech classifications, interjections are usually categorized under sentence-words (e.g. Kugler 2000b: 295). Syntactically, they count as independent unarticulated sentences (Balázs 1995: 71–75), just like forms of address and disjunct adverbials. An important theoretical issue arises as to what the relationship may be between interjections and other cooccurring parts, like articulated but insubordinate clauses or other unarticulated sentences like further interjections or forms of address. All these are relevant categories for our investigation since, in the case of insubordinate clauses occurring alongside interjections, certain items can be wedged in between the interjection and the subordinating conjunction without modifying the syntactic/semantic relationship of the two: these are most often forms of address (in Late Old Hungarian, 62.3 % of the sentence types attested in the core material under investigation were forms of address, Balázs 1995: 80) or other interjections.

Balázs (1991) mentions with respect to the double interjection ó, jaj! that interjections rarely occurred on their own in written texts in the Old Hungarian period, and when they occurred between articulated sentences, “they could not be preserved in their original status, either, but both their functional and structural properties underwent modifications as they interacted with articulate linguistic constructions” (Balázs 1991: 58). This point is important to bear in mind since most of them co-occur with articulate sentences (Balázs 1991: 59), as is the case with insubordinate clauses of the kind studied here.

The relationship between interjections and articulate sentences has, in historical lingusitics, been interpreted in various ways: it is either listed under one of the types of coordinate or subordinate relationships, or taken to be similar to those (Deme 1971: 102, cited by Balázs 1991: 59–60). Balázs takes up a third view: she considers that relationship to have existed historically before the emergence of subordinate and coordinate kinds of relationships and calls it equivalence (Balázs 1991: 60). Equivalence means that the emotional, volitional or affective content that the interjection expresses in its unarticulated form is once more represented by the following sentence, but in an articulated and detailed form (Balázs 1991: 60). If this is the case, the function of the interjection will determine what modality the articulate sentence that follows may have. Gallasy (1992: 849) makes a similar statement: whenever ó and jaj are used without being included in a sentence structure, their “content will rhyme with the exclamative or optative modal content of the subsequent sentence or sentences”. We have seen this before, where interjections were referred to as exclamations (Ameka 1992), and this point will prove to be relevant with respect to my own investigation, too.

It is important that interjections may lose some of their independence when they are used alongside articulate sentences and may become either members of some major word class or grammatical-pragmatic tools (e.g. discourse markers like English oh, Schiffrin 1987; or like the Hungarian vocative marker ó, Balázs 1991). An example of the former type of change is the nominalisation (conversion) of the Hungarian interjection jaj in the meanings ‘cry, grief, bitter/grievous fate’ or ‘pain, suffering’ (Balázs 1991: 61) that can express a threat, a complaint, or fear as the nominal part of a predicate and can even take arguments. The construction jaj nekem ‘poor me!/it is painful for me that…’ (similarly in other persons and numbers) was in use in quite a few variants in the Old Hungarian period and is still used today, hence I have excluded such cases from my own investigation due to the syntactic role that jaj has in them.

Some 19th-century Hungarian grammars address interjection-initial conditional clauses, covering some insubordinate examples of wishes, albeit without characterising them further. Thus, in Ferenc Verseghy’s early 19th-century grammar the author writes about interjections: “After vajha and óh ha the verb stands in the wishing mood, as in vajha otthon volnék ‘I wish I were at home’, Óh ha én még egyszer ifiú lehetnék ‘Oh if I could be young once more’” (Verseghy 1818: 492). In the insubordinate example, the co-occurrence of ó ha is noteworthy, and perhaps not unintended. János Fogarasi, in a later grammar, classifies ó ha with optative interjections in his list of the means for expressing wishes: “We normally employ the optative mood if we want to express a possibility. That is, […] 2. After many optative interjections like beh, bár, bárcsak, ó ha, vajha, vajmi etc.” (Fogarasi 1843: 251–252).^[6] Three years later, the grammar published by the Magyar Tudós Társaság also lists optative particles among interjections and gives an example with Ó HA: “óh, ha, bárcsak; vajmi, vajha are usually found with a verb in the wishing mood^[7] e.g. óh ha tud-nám! ‘oh, if I knew’, bárcsak megnyer-néd! ‘I wish you won it’, vajmi szép len-ne ‘how nice it would be’, vajha kijő-ne! ‘I wish she came out’” (Magyar Tudós Társaság 1846: 100).

None of the aforementioned three grammars employ the term particle, which had been used in 17th-century grammars for non-inflectable parts of speech (adverbs, prepositions, conjunctions and interjections) (Vladár 2012: 11).^[8] It is possible that they did not mention ha among the conjunctions because it was classified among interjections when it occurred along with Ó, since they must have felt the two items belonged inseparably together, as they only mention them in tandem.

A seven-volume dictionary representing the state of Hungarian in the beginning and middle of the 20th century, A magyar nyelv értelmező szótára (‘Definitive dictionary of Hungarian’) defines ha (in the entry ha ¹) as an adverb, in the optative meaning ‘I wish’ (with the qualifying label “rare”) (Az MTA Nyelvtudományi Intézete 1960: 4). It also discusses ha as a conjunction, in both of the main functions I have assumed it had in insubordinate clauses, i.e. in wishes and in exclamations (Az MTA Nyelvtudományi Intézete 1960: 6–7):

rare <in main clauses expressing volition, wish, without a main clause, often after the interjections ó, hej> Ha nekem egyszer a lottón ötös találatom lenne. ‘If I once won the jackpot at lottery’, Ha ezt el tudnám felejteni! ‘If I could forget this!’ […] <In similar sentences to express that the speaker is afraid the conjecture might come true or does not want it to come true; sometimes after the interjections ó, jaj> Ha meg találna halni! ‘If she happened to die!’, Egek! ha el találja magát árulni! ‘Heavens! If he happens to break cover!’ (Ferenc Kazinczy).

As can be seen, ha does not appear in the dictionary as an interjection; its use with ó is thought to be receding. The form ha in present-day Hungarian optative sentences is taken to be practically obsolete by Kas (2005). On the other hand, a 299-participant questionnaire study (with convenience sampling, Dér 2023a) proved that over two thirds of the participants accepted the same optative sentence involving an optative particle, interjection + optative particle, or interjection + conjunction if it contained ha alone (204 subjects for written language, 225 subjects for spoken language). Versions with an interjection were accepted by even more participants (ó ha: written: 254 subjects, spoken: 258 subjects, respectively jaj ha: written: 234 subjects, spoken: 246 subjects). A corpus study complementing that examination (on the MNSz2 corpus of 1.5 billion running words)^[9] also showed that almost half a thousand hits can be attested from present-day Hungarian language use for insubordinate conditional clauses beginning with interjection + ha/hogyha.

In what follows, I will briefly review where and how interjections were treated in the international literature on insubordinate clauses, primarily (but not only) in monographs and edited volumes. Traugott (2017: 301) mentions cases involving the interjections o, oh among exclamative if- and what-monoclauses, and points out that in conditional requests it is the interjection that refers to their exclamative nature (requests involve wishes, a reason why there may be an overlap between if-requests and if-exclamations). In her monograph, D’Hertefelt (2018: 55) mentions secondary interjections like Danish tænk and Swedish tänk, both of which have lexicalised from an imperative verb form (‘think!’) and can precede insubordinate clauses beginning with a ‘that’-complementiser. Several primary interjections (e.g. English oh, Swedish ååå) are frequent, if not compulsory, elements of insubordinate wishes (D’Hertefelt 2018: 83). Beijering and Norde (2019: 81) mention social formulas (e.g. forms of thanking) and curses among interjections occurring in semi-insubordinate constructions.

To date, the topic has been treated most comprehensively by Sansiñena (2019: 202), who studied Spanish insubordinate clauses having an optative, evaluative, or discourse linking function and introduced by que ‘that’. In particular, she studied vocatives, primary interjections (ah, ay, eh, hey, ja, je, ji, jo, oh, uf, uh, uy), and expressions derived from certain lexicalised forms occurring as turn construction units (TCUs) in such clauses and referred to collectively as ‘prefaces’. She claims that polyvalent interjections may occur in a wide variety of contexts (Sansiñena 2019: 214–215) and can be of three general types: those expressing emotions, those mirroring the speaker’s mental state (without expressing emotions or will), and those trying to catch the addressee’s attention (Sansiñena 2019: 215).

4 Methods and material

The present research focuses on the emergence of stand-alone insubordinate conditional clauses in Hungarian. In our study, we employ a functional-pragmatic approach (cf. Lastres-López 2021: 19), meaning that we analyse the conditional insubordinate clauses extracted from the corpora by assigning them to discourse functions. These functions are defined based on earlier categories found in the literature on insubordinate conditionals (primarily Kaltenböck 2016: 349; Lastres-López 2021: 131–142), adapted to the data from the Hungarian language. In the initial phase, we distinguish between directives, wishes and exclamatives. We combine this perspective with a corpus-based approach, as we deem it essential to incorporate quantitative analyses based on a larger dataset when examining insubordinate subordinate clauses diachronically. Additionally, a third dimension is relevant to the current research, namely the discourse variation analysis (Levey et al. 2013; Pichler 2013), since at the outset of the analysis, we consider the insubordinate conditionals with and without interjections (e.g., ó ha vs. ha, hogyha) as variants of each other. Discourse variation analysis combines the methods of grammaticalization theory, discourse analysis, and corpus linguistics to describe discourse-pragmatic variants, as their characteristics (variety of forms, and correspondingly variation in origin, context sensitivity) necessitate a complex approach. Through this method, we compare variants of linguistic units with the same or very similar discourse functions but with varying degrees of formal difference.

The available online databases cover the whole attested era of the history of Hungarian^[10] beginning with the appearance of coherent texts: over 800 years from the very first extant coherent text (Funeral Sermon and Prayer, 1192/95) to Contemporary Hungarian, the period up to and including the present state of the language (until 2010).

In chronological order, these corpora are as follows: ÓMK (1192–1526; registers: letters, religious genres: Bible translations, parables, sermons, prayers, psalms, etc.; size: 3.2 million words); TMK (end of the 15th century–1772; registers: witness testimonies of witch trials and private letters; size: 1 million words); KED (1550–1772; registers: memoirs, dramas; size: 200,000 words); MTSz (1772–2010; registers: fiction and press language, legal, professional, personal texts; size: 30 million words).

My fundamental aim was to gather data to the fullest possible extent, and I was able to achieve that with respect to insubordinate conditional clauses by searching for their subordinating conjunctions (ha ‘if’, hogyha ‘provided that’, hogy ha ‘that if’), selected interjections (e.g. oh, jaj), and their co-occurrences (either directly next to one another or indirectly, in the vicinity of each other). I gleaned them via a manual search, given that automatic search of subordinating conditional conjunctions would deliver all conditional clauses, not just insubordinate ones. Using interjections and conjunctions as search terms, I received extremely large sets of hits. In cases when I got lists of under a thousand hits, I perused the full lists, whereas in cases when the full list was above one thousand hits, I queried a random one-thousand-hit sample. That is, I collected insubordinate items in several ways from each database, since I wanted to study insubordinate conditional clauses lacking an interjection, too, and compare them to those involving an interjection, thus I collected occurrences of the latter type, too, with the same methods (full lists under one thousand hits and one-thousand-hit samples of larger lists). In the case of each database, I will give details on the exact method of data collection and the proportions obtained. I studied a total of approximately 12,000 listed hits in the four diachronic corpora.

Regarding the method for categorizing independent clauses: since we collected stand-alone cases, we focused on identifying which subordinate clauses starting with (interjection +) conditional conjunction appear without a main clause. However, it was not only necessary for the main clause to be physically absent, but the clause could not be elliptical either, meaning that the main clause could not be reconstructed from the surrounding context as part of the given subordinate clause.

It is important to see that in the older periods, in the absence of a unified Hungarian language and spelling, lowercase versus uppercase initials, as well as the punctuation employed and spelling items as one word versus multiple words, exhibited a large scatter, a fact that has repercussions for the topic of this investigation, especially with respect to how close-knit the relationship of an interjection and a conjunction might have been.

5 Results

5.3 Insubordinate conditional clauses in Modern and Contemporary Hungarian (MTSz)

Modern Hungarian was given top priority in this research, given that after (more) sporadic data from Old and Middle Hungarian it was this period in which stand-alone conditional clauses started to occur in Hungarian in a larger number (Dér 2024).

The MTSz database^[28] contains data from between 1772 and 2010. In it, I searched for constructions involving the functionally equivalent variants of interjections o∼ó∼oh∼óh, as well as ah, jaj, aj∼áj, ej, hej, haj, hú and hű (according to TESz, all of these are attested from before 1772)^[29] and the conditional conjunctions ha/hogyha ∼ hogy ha. In all cases, I manually analysed full hit lists from the corpus of 30 million running words.

Obviously, I also wanted to search for insubordinate clauses from among the clauses lacking an interjection and beginning with the conjunctions Ha/ha, Hogyha/hogyha or Hogy ha/hogy ha. However, since the total number of hits for sentences beginning with these items was extremely large (Ha/ha: 131,371 items, Hogyha/hogyha: 2,216 items, Hogy ha/hogy ha: 3,534 items), given that they included all conditional compound sentences, and since it is typical of conditional subordinate clauses that the relevant main clauses follow rather than precede them in most cases, I analysed thousand-strong random samples for each, looking for insubordinate conditional clauses. The selection of the relevant interjections was based on Balázs (1995), Gallasy (1992) and Deák (1936). Gallasy mentions ó (oh), jaj, no as the most typical and most frequently occurring interjections from the core material of Late Old Hungarian (1301–1526) she analysed, therefore these were in focus in the present investigation, too.

Note that the version hogyha was often not written as a single word in this period,^[30] even if it was used as a conditional conjunction ‘if’ and not as a conjunction ha ‘when’ of a clause embedded in a relative clause beginning with the complementiser hogy ‘that’ (e.g. Szólj, hogy ha vége lesz, vigyelek haza! ‘Tell me that, when it ends, I should drive you home!’ ∼ Szólj, hogy vigyelek haza, ha vége lesz ! ‘Tell me to drive you home when it ends’) where the relevant spelling is necessarily in two words. However, the search engine of MTSz does not make it possible to search for three adjacent words (e.g. jaj hogy ha ‘oh that if’), so I had to try to identify insubordinate interjection + hogy ha sequences in the full hit lists of interjection + hogy. Partly due to the unregulated practice of spelling, and partly taking the lesson of my earlier research projects into consideration (e.g. in some genres, like in poems, idiosyncratic orthographical solutions may occur, and conjunctions/discourse markers may precede interjections, e.g. de ó ‘but oh’), I also permitted both lowercase and uppercase initials in the case of the interjections under study.

Since I also wanted to know if there are insubordinate variants in which several interjections occur adjacently (e.g. ó jaj), or a name, form of address, etc. occurs after the interjection but before the conjunction, and I furthermore wanted to take the peculiarities of punctuation (the optionality of a comma after an interjection) into account, I searched for four types of hits for each interjection + conjunction pair: first, cases when they are directly adjacent and there is no punctuation mark between them (single-word distance), and then cases where the distance between the interjection and the conjunction is two, three or four words. Thus, versions where the interjection is followed by a comma or an exclamation mark (two-word distance), followed by a term of address (three-word distance), and possibly another punctuation mark and only then comes the conjunction (four-word distance) could also be taken into consideration, for instance:

(27)

Hű,	barát-om,	ha	akkor	ott	jön	a
wow	friend-poss.1sg	if	then	there	come	the
vezér	és	lát-ja,	hogy	az
chief.executive	and	see-3sg	that	the
Előkelő	Bank	egy	reményteljes	ifjú	tisztviselő-je	ott
Exclusive	Bank	a	promising	young	clerk-poss.3sg	there
tántorog	evvel	a	kis	liget-i	nő-vel!
stagger	with.this	the	little	park-adj	woman-com
‘Wow, my friend, if the chief executive then comes along and sees that a promising young clerk of the Exclusive Bank is staggering there with this little woman from the park!’
(MTSz, Ernő Szép: Lilac acacia, 1920)

It is important to note that since the hits for several words of distances include those of smaller distances (thus, three-word ones include two-word ones, etc.), the data are cumulative in Table 1.

Table 1:

Ratio of the most frequent insubordinate conditional clauses involving interjections in MTSz (N ≥ 20 for occurrences with 4-word distance).

	1-Word distance (hits)	2-Word distance (hits)	3-Word distance (hits)	4-Word distance (hits)
Oh/oh + ha	15	56	63	68
Ó/ó + ha	17	65	65	66
Jaj/jaj + ha	1	36	36	36
Óh/óh + ha	10	24	27	27
Hej/hej + ha	4	19	19	27
Ah/ah + ha	3	17	19	23

A total of 258 insubordinate conditional clauses starting with interjections were found (one per 116,000 words).^[31] If we split the whole of the 19th and 20th centuries covered by the corpus into thirds, and study the frequency of such clauses in that way (Figure 1), we can see that insubordinate conditional clauses with interjections start occurring in substantial numbers in the data from the second third of the 19th century, a tendency that persists for over a hundred years.

Figure 1:

Number of interjection-initial insubordinate clauses between 1800 and 2000 across the whole material of the MTSz.

Versions of Ó (the most frequently occurring interjection) account for almost two-thirds of the hits (Figure 2), while the rest of the interjections trail far behind: over one-sixth of the clauses begin with jaj, and six other interjections together amount to one-quarter of the hits (ah, hej, no, haj, hű). Here are examples of the three most frequently occurring cases:

Figure 2:

Relative frequency of interjection-initial insubordinate clauses in MTSz. Absolute frequency of occurrence: Ó: 160; JAJ: 40; AH: 22; HEJ: 22, NO: 8, EJ: 3, HAJ: 2, HŰ: 1.

(28)

Ó	ha	panaſz-om-at	ki-önt-het-né-m!
oh	if	complaint-poss.1sg-acc	vpfx-pour-pot-cond-1sg
ha	kétſég-be	es-het-né-k!
if	despair-ill	fall-pot-cond-1sg
ha	bocsánat-ot	nyer-het-né-k	tö́lök!
if	mercy-acc	gain-pot-cond-1sg	they.abl

–	ha	csak	eggy	orá-cská-t	is	tö́lt-hét-ne-k
	if	only	one	hour-dim-acc	prt	spend-pot-cond-1sg
bóldogító	remény-ben!	ha	le-buk-hat-né-k
happy	hope-iness	if	vpfx-bow-pot-cond-1sg
láb-ai-k-hoz!	ha	ki-sír-hat-ná-m	elö́ttök
foot-poss.pl-3pl-all	if	vpfx-cry-pot-cond-1sg	before.them
maga-m-at!
refl-poss.1sg-acc
‘O if I could pour forth my complaint! If I could give in to despair! If I could gain their mercy! If I could spend a mere hour in happy hope! If I could bow at their feet! If I could have a good cry in front of them!’
(MTSz, Kazinczy: Stella, 1794)

(29)

Jaj,	ha	tud-ná-k,	mint	vár-om	ők-et
oh	if	know-cond-3pl	how	wait-1sg	they-acc
már.
prt
‘Oh, if they knew how I am waiting for them!’
(MTSz, Petőfi: John the Valiant, 1844)

(30)

Ah,	kedves-em,	ha	ismér-né-d	ez-t	a
ah	dear-poss.1sg	if	know-cond-2sg	this-acc	the
Kliméné-t!
Clymene-acc
Nem! Barátom! Sokkal szebb ő, hogysem ravaszságokra szoruljon, igen is bizonyos az ő diadalma.
‘Ah, my dear, if you knew this Clymene! No! My friend! She is far too beautiful to fall back on craftiness, her victory is indeed certain.’
(MTSz, Verseghy: Moral Letters, 1786)

The extremely large difference observed between the two conditional conjunctions (Ha/ha: 237 hits, 91.86 %; Hogyha/Hogy ha/hogyha/hogy ha: 21 hits, 8.14 %) can be explained by the fact that hogy ha is used in its earlier role even today: in the case of conditional clauses embedded in an object (or bound adverbial) subordinate clause introduced by hogy, the combination hogy ha occurs quite frequently. However, although these cases will never include insubordinate ones, such adjacent occurrences were the factor that made the change hogy|ha > hogyha and the emergence of the latter as a clear-cut conditional conjunction possible from the 14th century onwards (Haader 1995: 542, 556; Rácz 1995: 708–709). Here are two examples of hogyha written as a single word versus as two separate words as conditional conjunction of insubordinate clauses:

(31)

Melly sokſzor, ammint forrón ápolygattalak, megrözzent ſzívem, ’s azt mondotta:
Óh!	hogyha	ö́-t	el-hagy-n-om	köll-en-e!
oh	that.if	she-acc	vpfx-leave-inf-1sg	must-cond-3sg
‘How often, when I feverishly attended you, did my heart wince and say: Oh, if I had to leave her!’
(MTSz, Verseghy: What is Poetry? 1793)

(32)

Felejted-e, hogy vár jó anyád, Ki úgy szeret a mint önmagát,
Óh	hogy	ha	baj	ér-t	Szolgálat-od-ért!
oh	that	if	trouble	get-pst	service-poss.2sg-causfin
‘Are you forgetting that your good mother is waiting for you, who loves you as she loves herself, Oh, if you got into trouble for your services!’
(MTSz, János Garay: Reunion, 1836)

Stand-alone insubordinate clauses tend to have one of two discourse functions: they can be wishes (146 hits, 56.58 %) – there are no examples of threats, complaints, offers in the period, and there is a single example of a request (0.38 %) (33) – or else they can express exclamations (111 hits, 43.02 %), see (31) and (32) above.

(33)

– Merci! – mondja bűbájosan a hölgy. – Ó, mily finom volt.
Ó,	ha	süt-ne	még	egy	pirinyó-t…	ur-am…
oh	if	bake-cond	prt	a	small-acc	sir-1sg
‘ “Merci,” the lady said enchantingly. “Oh, how delicate it was. Oh, if you baked another small one… sir…” ’
(MNSz2, Andor Endre Gelléri: The laundry, 1931)

There does not seem to be any connection between the choice of conjunction (ha vs. hogyha) and those two functions, the latter item occurs in optative constructions, too:

(34)

Óh,	hogyha	vol-na	nék-em	is	Könny-em!
oh	that.if	be-cond	me.dat-poss.1sg	prt	tear-1sg
de	férfikönny	ki-szárad.
but	men.tear	vpfx-dry
‘Oh, if I had tears, too! But men’s tears dry up soon.’
(MTSz, Madách: The untrue one, 1840)

In insubordinate clauses involving interjections, the verb mood was conditional in over four-fifths of the cases (212 hits, 82.17%), followed by indicative (45 hits, 17.44%), and a single case of imperative (0.39%). In clauses involving wishes, apart from two indicative verbs, only conditional forms were attested and almost half of them also had a potential suffix -hat/-het (see [27]). On the other hand, conditional mood is also typical of the exclamative type (in two-thirds of the cases). It is not surprising that these two categories are hard to tell apart, as was pointed out several times. Indicative forms exclusively occurred in exclamations.

Next, I investigated the occurrence and functions of insubordinate conditional clauses lacking interjections. In the case of Ha, ha, hogyha ∼ hogy ha, I studied thousand-strong random samples, in the case of Hogyha, I examined a total of 481, and in the case of Hogy ha, a total sample of 227 hits.^[32] For the six spelling variants, I found 35 cases of stand-alone conditional clauses, i.e. 13.56 % (roughly one-seventh) of the cases involving interjections. The ratio of wishes and directives versus exclamative clauses was roughly equal (19 hits, 54.28 % vs. 16 hits, 45.72 %); the former set again contains wishes in the largest number (14 hits, 74 % of all deontic cases) (35), but there are also threats (3 hits) (36) and requests (2 hits), too:

(35)

Hogy’ mosolygott, örvendett; sokkal boldogabb volt, mintha az övé lett volna a ruha. Mi sok ily szép ruhája volt Esztinek, mégis mi haragosan szokott öltözködni!
Ha	neki	csak	egyetlen	ily	ruhá-ja
if	she.dat	prt	only	adj.dem	dress-poss.3sg
vol-na!
be-cond
‘How she smiled, how she beamed; she was far happier than if the dress was hers. How many nice dresses Eszti had, yet how angry she was when she was getting dressed! If she had just one such dress!’
(MTSz, Pál Gyulai: Sketches and drawings 1, 1867)

(36)

Zilia: Rám is jól vigyázz! Végezz hamar! (Erszényét odahajítja neki.) Munkád jutalma száz Tallér.
De	hogyha	meg-reszket	kez-ed…
but	that.if	vpfx-quiver	hand-poss.2sg
‘Zilia: Guard me well, too! Finish quickly! (She throws her purse to him.) The prize will be a hundred thalers. But if your hand should quiver…’
(MTSz, Jenő Heltai: The speechless soldier, 1936)

Note that over two-thirds of the cases (27 hits, 77 %) come from the 20th century, whereas there are only three such instances from the 19th century, and the five insubordinate clauses from the 18th century all come from a single text by Ferenc Kazinczy (28). That is, this type emerged later than the type involving interjections.

With respect to the two conjunctions HA, HOGYHA, the ratio is slightly different (HA: 28 hits, 80 %, HOGYHA: 7 hits, 20 %) from the cases with interjections (where the share of HOGYHA is twice as large as that of HA), but there is another difference as well: the share of wishes within interjectionless cases is almost 20 % smaller than in the group involving interjections (see Figure 3). Both differences can be explained by the difference by nearly an order of magnitude in the size of the two groups (clauses with and without interjections).

Figure 3:

The ratio of insubordinate clauses involving versus not involving interjections in the samples from MTSz (IJ = interjection). Exclamation with interjection (IJ): 111, without IJ: 16; directives with IJ: 1, without IJ: 5; wishes with IJ: 146, without IJ: 14 occurrences.

6 Conclusions

On the basis of the present investigation of online corpora, covering the whole history of Hungarian since the first appearance of extant monuments, we can conclude that stand-alone insubordinate conditional clauses beginning with an interjection are attested ever since the end of the Old Hungarian period, and that variants lacking an interjection cannot be found in the Old Hungarian samples analysed but only from the period of Middle Hungarian (instances beginning with hát ‘well’) (see Table 2). The small amount of data found in Middle Hungarian texts can be accounted for by the fact that only a handful of the genres first occurring in that period are represented in the available online diachronic databases (TMK, KED).^[38]

Beginning with Modern Hungarian, a significant quantitative difference in the data can be observed in favour of variants including interjections (but that difference can already be observed in Old Hungarian, too). Hence, we can suspect that both main types of insubordinate clauses (wishes and exclamatives) first emerged and spread from instances with interjections, since they (especially Ó) emphasise the essence of such sentences, i.e. their strong emotional/affective load (cf. “equivalence”, Balázs 1991: 60). Such heightened emotional load is also typical of stand-alone insubordinate clauses beginning with the complementiser hogy ‘that’ in Hungarian (Dér 2023b), where it is also wishes and exclamations that dominate and which also emerge, primarily with interjections, from Modern Hungarian onwards (Dér 2022). Hence, uses involving interjections are not merely occasional variants but seem to have a key role in the emergence of Hungarian stand-alone insubordinate clauses, helping to identify and emphasise their emotional content.

Language users may have felt that Ó, the most frequently occurring interjection, was strongly correlated with the conjunction ha, and may have assumed the sequence of the two to play a similar or identical role as other particles of wishing (bár, bárcsak). The case of vajha also supports this, where the interjection vaj fused with ha, but Ó HA does not appear to have reached that stage (no variants written as a single word can be attested, whereas vajha is written as one word in most cases in the Old Hungarian period).

Given that I have not found matrix clauses of general meaning and reflecting emotional consequences in the data, I think that the operation of hypoanalysis is the most probable cause of the emergence of stand-alone conditional clauses. That is, it was not the case that main clauses disappeared from beside clauses becoming insubordinate; rather, such clauses emerged on their own in certain uses. However, it is important to point out that all the explanations offered are partly or totally compatible, although they view the same phenomenon from diverse aspects, hence I think all of them are useful ideas in describing insubordinate clauses.

The main limitation of the present study is that the examined databases represent different genres and varying amounts of data, which necessitates accounting for the genre effect. In other words, the observed differences in quantity between the insubordinate clauses with or without interjection may also be explained by corpus-specific characteristics. This is further supported by the limited data in the Middle Hungarian period databases (KED, TMK). However, the only counterpoint to this interpretation is that the versions with interjections appeared more frequently in larger corpora in terms of text volume (ÓMK, MTSz). Nonetheless, as with the scarcity of data, we must always consider the specific characteristics of the corpus when interpreting the results. Our claim, namely that conditional detached subordinate clauses might have emerged with interjections and that their creation likely involved hypoanalysis, can, of course, only be applied to the available data.