1 Introduction

In Modern Hungarian, two kinds of vowel harmony are operative: backness harmony and labial harmony (Siptár & Törkenczy 2007, 157–170). Backness harmony has been present in the language ever since its earliest reconstructible ancestral stage (Proto-Uralic, cf. Suomi 1983; Aikio 2022). It is triggered by, and affects, all vowels – modulo the possibility of neutral behavior for front non-labial vowels (Rebrus & Törkenczy 2021; Rebrus, Szigetvári & Törkenczy 2019). As a result, it causes widespread alternations in suffixes. Labial harmony, by contrast, seems to have appeared in the Old Hungarian period (10th–early 16th centuries), and is thus much more recent than backness harmony. It is also much narrower in scope: in Modern Standard Hungarian it only affects the short front non-high vowels [ɛ] and [ø], and causes alternations in only a few suffixes.

There is evidence, however, that in Late Old Hungarian, in particular in the 15th and early 16th centuries, some suffixes including non-high long vowels were able to undergo not only backness harmony but also labial harmony. This paper discusses three of the most widely attested suffixes in question, the ablative, the delative and the elative nominal case suffixes. Their Modern Standard Hungarian forms are [toːl]/[tøːl] (-tól/-től) ‘from’, [roːl]/[røːl] (-ról/-ről) ‘from top of’ and [boːl]/[bøːl] (-ból/-ből) ‘from inside’, respectively. All the three suffixes were originally grammaticalized from case-marked nouns; their participation in vowel harmony followed on their integration into the morphological structure of host nouns. They underwent backness harmony in all varieties of Old Hungarian, but in several varieties they also underwent labial harmony. However, their ability to undergo labial harmony was subsequently lost; in Middle and Modern Hungarian long vowels can only harmonize in backness. We propose an explanation for why they stopped harmonizing in labiality, an explanation based partly on the phonological parameters of variation extant in Late Old Hungarian, partly on homophony avoidance in the changing paradigmatic space of the case system.

The paper is organized in the following way. In 2 we give a brief overview of the emergence of labial harmony. In 3 we present the grammaticalization process that gave rise to several nominal case suffixes in (Pre-)Old Hungarian. This is relevant because the three suffixes in question were themselves grammaticalized from nouns. In 4, the central part of the paper, we present the three suffixes in detail, including their origins, diachronic and dialectal variants and harmonic behavior. In 5 we propose a hypothesis on why they stopped harmonizing in labiality (but not in backness). Section 6 concludes.

2 The emergence of labial harmony in Hungarian

This section presents the processes whereby labial harmony emerged and then crystallized into the pattern found in Late Old Hungarian and afterwards, up to the Modern Hungarian period. These processes are explained in much more detail elsewhere (Cser, Oszkó & Várnai 2023 and 2025, see also Cser, Oszkó & Várnai 2024); this presentation briefly summarizes the findings reported and the analyses given in those papers. The reader is also referred to the literature cited there for further details and explanations.

The Early Old Hungarian vowel system is usually reconstructed as shown in Table 1 (cf. Kiss & Pusztai 2003, 336, 344). EOHu had inherited a small inventory of labial vowels; in fact, only short [o] and [u] are securely reconstructed for Pre-Old Hungarian. It is possible that at least some varieties of the language had inherited [y], perhaps going back as far as Proto-Finno-Ugric or Proto-Uralic, but even this is tentatively suggested by etymologists only for a handful of lexical items. At any rate, it seems clear that labiality was not a robust contrastive feature before, or even in, the Early Old Hungarian period. At this stage, however, several overlapping changes greatly increased the number of labial vowels.

First, nearly all final vowels became high and labial (1), harmonizing in backness with the preceding vowel. By the end of the 13th century, these final high vowels were apocopated. In suffixed forms the original (non-high) vowels were often preserved but the innovative final vowel could also be analogically transferred from the unsuffixed form (see 2, where [ɛ] is the original stem-final vowel, [y] and [ø] are innovative).

Further changes that increased the number of labial vowels include sporadic labialization next to labial consonants (3), more regular labialization before [l] + a coronal consonant (4), and, perhaps most importantly, the contraction of vowels with following glides or non-sibilant fricatives (5).

It appears quite clearly that both analogically transferred vowels (2) and the final long vowels resulting from contraction (5) were able to initiate a regressive spreading of labiality; a large number of lexical items display variation that is most plausibly explained by such a process (6).

Regressive labialization was crucially dependent on affixed forms, which either retained the original stem-final vowel or showed the analogically levelled labial vowel. Thus, a pattern in the variation developed in which a non-labial vowel before a suffix could only follow a non-labial vowel in the stem; e.g. the plural of a word like [ɟɛrmek] ‘child’ could be any of the three forms in (7), but it could not be **[ɟɛrmøkek].

It was most probably this pattern that was then reinterpreted as progressive labial harmony. This affected suffixes grammaticalized from a word containing a labial vowel, such as allative *[xuz] > EOHu [huz] / [hoz] ultimately from Proto-Ugric *kućV ‘side’ (Benkő 1993‒1994 s.v. hozzá). The allative suffix first developed the front variant [høz] after front-vowelled stems, but then, on the analogy of forms such as [ɟɛrmekek] ‘children’ (cf. 7) it also developed a non-labial variant [hez]. In some varieties of Late Old Hungarian, the front suffixes either showed only the labial variant or showed free variation between the labial and the non-labial variants; these Old Hungarian dialects had not yet attained the stage of consistent progressive labial harmony. Others had, however, and displayed consistent progressive labial harmony that aligned with inherited progressive backness harmony.

3 The emergence of suffixes via grammaticalization

As was mentioned above, the allative suffix originates in a noun meaning ‘side’; it appears as a suffix already in the earliest text of some length (Funeral Sermon, end of 12th century). Of several other suffixes that derive from full words we briefly exemplify the process here with the help of two that are based on the noun [bɛl] ‘inside’ (> MoHu bél ‘intestines’). One is the inessive suffix [bɒn] / [bɛn], the other the illative suffix [bɒ] / [bɛ]; for both the variants indicated here crystallized in the Late Old Hungarian period and have been part of the language in this form ever since. Both are based on previous suffixed forms of [bɛl], viz. *[bɛlɛn] for the inessive and [bɛleː] or the illative.

There are three traditional criteria to judge how far grammaticalization has progressed. One is the shortening of the form. On this count the inessive was the more progressive of the two since it is not at all attested in the longer form by the time Hungarian comes to be written; its earliest documented form is [bɛn] (written ben). By contrast, the illative is attested in earlier texts as unshortened [bɛleː] (written bele); the shorter form is generalized only in the Late Old Hungarian period. The person-marked forms, however, preserved the longer variant and have stayed that way to the present (i.e. [bɛleːm] ‘into me’, [bɛleːd] ‘into you’, etc., as opposed to the shortened person-marked inessives [bɛnnɛm] ‘in me’, [bɛnnɛd] ‘in you’, etc.).

The second criterion is vowel harmony. When a word becomes a suffix, it starts displaying backness harmony. The Funeral Sermon (see above) only shows non-harmonizing [bɛn] and [bɛleː] with any stem. A slightly later text (Gyulafehérvár Lines, early 14th c.) shows variation between a front and a back variant after back stems, but scribal error cannot be excluded since there are only three relevant data in this short text. After the early 14th century, however, both suffixes show consistent backness harmony.

The third criterion is rightly considered to be the weakest, since it concerns spelling in an era long before standardization. We will not dwell on it at any length but note that there is considerable variation in the Old Hungarian period as regards whether the words that had become or were becoming suffixes are written together with the host word. This variation, however, settles into a consistent practice of not graphically separating the suffix from the stem by the beginning of Middle Hungarian.¹

4 The ablative, the delative and the elative suffixes

The ablative suffix derives diachronically from the ancestral ablative² case-marked form of a noun meaning ‘trunk, stem’. This form is reconstructed for Pre-Old Hungarian as *tiɣy-l or *tyɣy-l. The noun itself also survives as [tøː] (MoHu tő) ‘stem’. The intervocalic *ɣ was lost so early that it is not attested in writing at all, and the two vowels were contracted into a long front labial vowel. As the form was gradually grammaticalized into a suffix, its vowel underwent backness harmony (i.e. it was velarized after back stems), and in some language varieties it also underwent labial harmony (i.e. it was delabialized after non-labial stems). The main variants that thereby emerged by LOHu are as given in (8). The variation in vowel height appears to be dialectal, as it still is in Modern Hungarian.

The delative suffix derives diachronically from the ancestral ablative case-marked form of a noun meaning ‘surface’. This form is reconstructed for Pre-Old Hungarian as *roɣo-l (possibly with different back vowels). Other forms deriving from the same stem include the sublative suffix [rɒ] / [rɛ], the person-marked sublative forms ([raːm] ‘onto me’, [raːd] ‘onto you’, etc.) as well as the person-marked superessive forms ([rɒjtɒm] ‘on me’, [rɒjtɒd] ‘on you’, etc.). Just as in the case of the noun grammaticalized into the new ablative suffix, the intervocalic *ɣ was lost so early in this word too that it is not attested in writing at all, and the two vowels were contracted into a long back labial vowel. As the form was gradually grammaticalized into a suffix, its vowel underwent backness harmony (this time it meant fronting after front stems). In some language varieties the front allomorph also underwent labial harmony (i.e. it was delabialized after non-labial stems). The main variants that thereby emerged by LOHu are as given in (9). The variation in vowel height is dialectal, as it still is in Modern Hungarian. The high-vowelled variants are very rare, unlike those of the ablative suffix.

The elative suffix derives diachronically from the ancestral ablative case-marked form of the noun [bɛl] ‘inside’. Other forms deriving from this word, including the noun bél ‘inside, intestines’, the inessive and the illative suffixes, were shown in the previous section. The original ablative is attested in the form [bɛleɥl] in Early Old Hungarian. By phonological change this developed regularly into [bɛløːl]. In at least some varieties this form underwent vowel harmony, as a result of grammaticalization, before shortening into a monosyllable; hence the attested back suffix variant [bɒloːl]. By the end of the Old Hungarian period, however, the suffix largely settled into the monosyllabic variants, which in some language varieties also showed labial harmony. The person-marked forms, however, continue to show the original longer form to this day ([bɛløːlɛm] ‘from inside me’, [bɛløːlɛd] ‘from inside you’, etc.). The set of variants for the elative suffix is given in (10).

The broad outlines of this variation as well as some data are given cursory treatment in Korompay (1992, 375–6). For the present paper we have searched the entire Old Hungarian Corpus (http://omagyarkorpusz.nytud.hu/en-intro.html, cf. Simon 2014; Simon & Sass 2012) and have gathered all occurrences of the front variants of the suffixes (since only the front variants can show labial harmony). Here we disregard those source texts that do not have more than 5 occurrences of the non-labial variant of the three suffixes combined. This leaves us with 13 codices altogether. In these 13 codices there are 430 instances of a non-labial front suffix variant. In Table 2 the number of suffix variants per codex is given; the three suffixes are shown in separate columns, and the overall length of the codices is also given for comparison in number of words.

As can be seen in Table 2, the non-labial variants represent a minority pattern everywhere except for a single codex (Gyöngyös Codex) – the shortest one, as it happens. If not sporadic, the presence of suffix variants written with a letter that indicates a non-labial vowel (most typically <e>) can usually be understood as evidence that delabialization has taken place. Caution, however, is advised; copying errors, dialect shift and other factors can come into play. In the Kulcsár Codex, for instance, about one third of all front suffix variants are non-labial, which is a comparatively high proportion. In some cases, however, we have good reason not to take a given form at face value. For example, the same elative form of the word meaning ‘trap’ occurs written as twrbel on 24r but as twrbewl on 33v, the former suggesting [tyːrbeːl], the latter suggesting [tyːrbøːl]. Not only is it surprising to find a non-labial suffix variant after a labial stem vowel; on closer inspection it can also be seen that the form twrbel is found at the end of its line (see Figures 1 and 2). Its position may well have prompted the scribe to shorten the form by omitting the w, a rather wide letter.³

View Full Size

The elative of twr ([tyːr]) ‘trap’ in Kulcsár Codex 24r;⁴ note the vertical margin just before the letter l

Citation: Acta Linguistica Academica 72, 1; 10.1556/2062.2025.00915

• Download Figure
• Download figure as PowerPoint slide

View Full Size

The elative of twr ‘trap’ in Kulcsár Codex 33v⁵

Citation: Acta Linguistica Academica 72, 1; 10.1556/2062.2025.00915

• Download Figure
• Download figure as PowerPoint slide

In⁴ spite of such variation, not all of which is phonological, it seems plausible to assume that the data point to labial harmony operating to a certain extent on these long-vowelled suffixes. To wit, in the Kulcsár Codex, there are 72 suffix forms written with non-labial vowels; of these only two are found after labial stem vowels, but of these two one is almost certainly a writing error (the other is kwbel [kyːbeːl](?) ‘out of stone’ at 19v). In the remaining 70 cases the non-labial suffix variant follows a non-labial stem vowel.

Similar ratios are found in other codices too. In the Jordánszky Codex, for instance – the work of a single copyist – the front suffix variants are overwhelmingly labial (930 vs. 29), but of the 29 non-labial variants only one occurs after a labial vowel.⁵⁶ In the Debrecen Codex, the work of seven distinct hands, the suffix variants are again overwhelmingly labial (861 vs. 44), but of the 44 non-labial variants only one occurs after a labial vowel.⁷ The environments of all the 430 non-labial suffix variants are schematically summarized in Figure 3.

View Full Size

The environments of the non-labial suffix variants. Legend: I = non-labial vowel in preceding syllable, L = labial vowel in preceding syllable, V = velar vowel in preceding syllable

Citation: Acta Linguistica Academica 72, 1; 10.1556/2062.2025.00915

• Download Figure
• Download figure as PowerPoint slide

What this shows is that the distribution of the labial vs. non-labial variants was not random: there was at least a weak tendency for these suffixes to undergo labial harmony. Apparently it never became as widespread and systematic as the labial harmony seen in short-vowelled suffixes, but it nevertheless was a tendency. Why then did it disappear after a brief period? This is the question to which we now turn.

5 The end of labial harmony for the long-vowelled suffixes

As indicated in the introduction above, labial harmony is subject to three featural constraints in Modern Hungarian: (i) it does not affect high vowels (ii) it does not affect long vowels and (iii) it does not affect back vowels. Of these, (i) and (iii) have always been operative ever since labial harmony has existed,⁸ but (ii) did not obtain of certain Old Hungarian language varieties – this is what we showed in the previous section. But if constraint (ii) was apparently weaker than the other two, and could be contravened, why did these long-vowelled suffixes stop harmonizing ultimately?

We believe that there are two reasons for the loss of labial harmony in these suffixes (apart from the general instability of front labial vowels in several varieties of Old Hungarian). In the case of the ablative suffix the high-vowelled variants [tuːl]/[tyːl] were much more frequent than the high-vowelled variants of the other two suffixes. Etymologically, this can be explained with reference to the original high vowels that the predecessor form (*tiɣy-l or *tyɣy-l) contained. While the suffix that resulted from the grammaticalization of this form did develop mid-vowelled variants, the high-vowelled variants remained prominent and this may well have prevented the non-labialized forms from appearing. As is indicated by Korompay (1992, 376), those codices that consistently show a high-vowelled ablative suffix typically do not show a non-labial variant for the same suffix even if they show non-labial variants for the elative and the delative suffix. This is confirmed by our own counts presented in Table 2 above: the Keszthely Codex and the Kulcsár Codex, which entirely lack the [teːl] variant, show written twl, clearly indicating [tyːl] and [tuːl] (mid labial vowels are written ew and o in these codices).

In the case of the other two suffixes an unrelated sound change may have influenced the development of the front variants. In Old Hungarian as well as at later stages in the history of the language coda [l] was prone to deletion (and concomitantly to hypercorrect insertion). This affected the ablative, the delative and the elative suffixes too, since they all ended in [l]. We do indeed see spelling variants in the Old Hungarian codices (cf. also Korompay 1992, 369), but also in Middle Hungarian texts, cf. (11).

While some of these may well be copying errors, or printing errors in Middle Hungarian, the phenomenon is too widely attested throughout history and even in Modern Hungarian not to be taken as linguistic reality. If we now consider the non-labial front variants of the elative and the delative suffixes, when affected by coda l-deletion, these would have become [beː] and [reː], respectively. These forms, however, would have been perilously close to their semantic opposites: the elative to the illative suffix [bɛ], and the delative to the sublative suffix [rɛ] (on the emergence of these suffixes see sections 3 and 4 above). We assume that this phonological closeness to their semantic opposites was a factor in the disappearance of the non-labial variants of these two suffixes.⁹

6 Conclusion

The recently grammaticalized Late Old Hungarian ablative, elative and delative suffixes showed labial harmony for a short period. How labial harmony appeared in general is a topic treated elsewhere; in this paper the focus was on how these long-vowelled suffixes lost their capacity to undergo it. After looking at their harmonic behaviour in much more detail than seen in the earlier literature we propose two explanations. One concerns the ablative suffix, the only one of the three in which the vowel was predominantly high, though with mid variants. High vowels never underwent labial harmony, and this appears to have adversely impacted the nascent non-labial variants of this suffix. For the other two suffixes the danger of homophony with their semantic opposites was realistic when the vowel was delabialized and coda [l] was lost, and this may well have exerted a pressure to lose the non-labial suffix variant. What this shows is that the exclusion of high vowels from labial harmony is diachronically a more stringent constraint in Hungarian than the exclusion of long vowels – the latter is more of a historical accident.

7 Conflict of interest

András Cser is a member of the Editorial Board of the journal and was not involved in the review process in any capacity.