Diminutive formation in Hungarian

We survey templatic diminutive formation in Hungarian. We conclude that there is an intricate system of endings that are added to bases which are truncated if they contain more than one vowel. Bases are also subject to vowel length changes in both directions, as well as the palatalization of the last consonant. The templatic diminutive forms are not subject to vowel harmony occurring in suffixes which prevails in the regular additive morphology of the language. Nevertheless, these forms conform to the vowel patterns found in disyllabic monomorphemic or disyllabic suffixed word forms.

The paper begins with a brief introduction to the phoneme inventory of Hungarian ( §1). We then define what we mean by a "diminutive" form: we use this as a cover term for a wider semantic/pragmatic field, including hypocoristics, humorous, slangy, informal words, whose phonological shapes are similar ( §2). We then introduce what we call additive (or concatenative or agglutinative) morphology, which constitutes the bulk of morphology in Hungarian ( §3). In a rather limited se of cases we find vowel-zero alternation in additive morphology, ( §4). This process is distinct from truncation ( §5), a characteristic only of the formation of templatic diminutives. The catalogue of diminutive templates follows in the next section ( §6), with a subsection for each template vowel. We briefly discuss the role of reduplication ( §7), and the curious absence of vowel harmony in diminutive forms ( §8). Conclusions discussing the differences between additive and truncative morphological processes end the paper ( §9).

VOWELS AND CONSONANTS IN HUNGARIAN
The reference accent of Hungarian distinguishes between seven short vowels. It contains the canonical five-vowel inventory complemented by two front rounded vowels. The textbook tradition holds that each of the seven vowels has a long counterpart. A number of morphophonological alternations support this analysis and the standard spelling also reflects it. We list these vowels in (1). (1) The vowel system of Hungarian front unrounded front rounded back high i iː y yː u uː mid _ eː ø øː o oː low ɛ _ ɑ aː The transcription symbols we use in this paper suggest that there is also a quality difference between the low ([ɑ]- [aː]) and the front unrounded nonhigh short-long pairs ([ɛ]-[eː] (Hayes et al. 2010): [ɛ] exhibits less characteristics of neutrality than [eː], which in turn exhibits less characteristics of neutrality than [i] or [iː]. Another reason for assigning [ɛ] and [ɑ] to a natural class is that they both lengthen before suffixes, cf. (4f). In (2) we tabulate the consonant phonemes of Hungarian. We take [j], [ʋ], and [h] to be approximants, not fricatives (the latter two contra Sipt ar & T€ orkenczy 2000, 76ff), although all three consonants have fricative allophones. The status of [ʣ] is debatable: most of its 1 A lesser quality difference also exists between the other mid vowel pairs, [ø]- [øː] and [o]- [oː], but this is often ignored in transcriptions. Lass (1984) transcribes these vowels as [oe], [øː], [ɔ], [oː], respectively. In this paper the choice between the two analyses is irrelevant, so we apply the simpler set of symbols.
occurrences are geminates suggesting that it is a [d]þ[z] cluster. In our transcriptions we will include [ŋ], the nasal occurring before velar plosives. (2) The consonant system of Hungarian labial dental postalveolar palatal velar glottal plosives p b t d c ɟ k ɡ affricates ʦ (ʣ) Hungarian has voicing assimilation of obstruents: adjacent obstruents agree in voicing. The voicing of an obstruent cluster is determined by its last member. This is also true of clusters separated by a word boundary.

DIMINUTIVES
We are going to compare diminutive forms to their nondiminutive counterparts, which we call their base. This is important, because there are some regularities that can only be formulated by reference to the base. A base can often be associated with several diminutive forms ([mɑri], [ We use the term diminutive (DIM) for several different semantic categories, which do not seem to be distinguished morphophonologically in Hungarian. Some of these forms are hypocoristics, terms of endearment or belittling, others are simply informal, slangy, or humorous versions of their base. Kiefer & Lad anyi claim that diminutive forms have no semantic correlates, their use is governed by pragmatic factors (2000b, 171). In Hungarian none of these forms can be used exclusively as vocatives, unlike in Italian (Alber 2010). Since a semantic/pragmatic 2 In transcriptions we represent long vowels using the IPA length mark, but geminate consonants by doubling the relevant symbol. delineation of diminutive forms is rather uncertain, we are going to compare the phonological shape of potential diminutive forms to their bases when there appears to be a firm semantic connection between the two.
The base of most diminutive forms is a noun, typically a given name, but it may also be a common noun, as well as a family or a geographical name. In addition, we also find adjectives ([unɑlmɑʃ]  There are also cases where both the meaning and the phonological form of a word form is diminutive-like, but there is no other word in the current language which it could be synchronically related to, which could be identified as its base ( In yet other cases there is a semantically identifiable base, but the phonological connection is unprecedented. This occurs either because the base became obsolete ([lɑsti] 'ball.DIM', from an earlier [ɛlɑstik], a brand name (B arczy 1931-1932) , etc. These patterns are not productive and most speakers do not identify these forms as diminutives. We will ignore such pairs in this paper, that is, we do not posit diminutive templates that they fit in. Let us note here, that these diminutive endings harmonize, unlike any of the truncative diminutives endings we have identified. 4 We use the greater-than sign not to mean 'turns into', but to symbolize that the base form is semantically 'greater than' the diminutive to its right. 5 Adjectives are optionally also subject to being suffixed by [ʃ] (2017), web forums and blogs (like Papp 2018), as well as our native speaker competence. In some cases we have based our decisions on our own intuitions.
Diminutives can be formed by additive and also by truncative morphology. Additive diminutives follow the regularities shown by most other additive morphological processes of the language, briefly described in §3 (the stem is not truncated, the suffix either harmonizes with the stem vowel(s) or contains a netural vowel, etc.), therefore they are irrelevant for the purposes of this paper. Truncative diminutives, on the other hand, are radically different. For the sake of the comparison, we first introduce some features of additive morphology in Hungarian.

ADDITIVE MORPHOLOGY
The bulk of Hungarian morphology is additive, involving mostly suffixation. This means that phonological material is added to the end of the stem, which itself appears in its entirety (Kiefer & Lad anyi 2000a, 140 The stem may undergo relatively minor changes, thereby becoming a bound stem. These changes include the loss of the independent place or voice specification of the stem final consonant, as in (4a) and (4b), or its palatalization caused by a suffix-initial [j], as in (4c). A consonant, (4d), or a vowel, (4e), may also occur between the stem and the suffix. This "augment" is only marked in this display, later on we will mark it as belonging to the suffix. A stem-final low vowel lengthens before most suffixes, (4f). The last vowel of the stem may shorten in suffixation. This shortening may occur to low vowels, (4g), high vowels, (4h), and mid vowels too, (4i (5a) shows the most robust pattern, front/back harmony: the suffix occurs in its front-or back-vowelled allomorph depending on the front or back quality of the stem vowel. Rounding harmony is parasitic on front/back harmony: it is only a harmonizing short front vowel that may agree with the rounding of the stem vowel and only when it also harmonizes in frontness, (5b). It is a lexical property of suffixes if they have one (front unrounded), two (front and back), or three (front rounded, front unrounded, and back) allomorphs. The choice is also dependent the height of the vowel, the three-way alternation is only available with short mid . Front unrounded vowels are called neutral vowels. They often exhibit disharmonic behaviour. This is manifested in antiharmony, (5c), or in transparency, (5d), as well as the fact that almost all suffixes that do no alternate harmonically contain a neutral vowel. 9 The high [i] is almost always transparent, [eː] may be transparent, (5d). In many words containing [eː] and [ɛ], we find vacillation, (5e). Antiharmonic roots 10 are almost exclusively monosyllabic: in them a neutral (i.e., front) vowel, most often [i(ː)], sometimes [eː], governs back harmony, (5c). A monomorphemic, disyllabic stem is almost never antiharmonic, (5f). 11 A polymorphemic stem, however, may be antiharmonic. In (5g) the stem contains two neutral vowels, just like [ʦivil] in (5f). 9 In unsuffixed words containing both front and back vowels, the front ones are typically unrounded, i.e., neutral. 10 By root we mean the first morph in the word. The stem of a word is what remains when the last suffix is removed. That is, a root is a monomorphemic stem. But the root [hiːɡ] is antiharmonic (as the comparative form shows). The suffix immediately following the root also contains a nonharmonizing neutral vowel, thus [hiɡ-iːt] is a disyllabic stem containing two neutral vowels. Crucially, this stem is polymorphemic and it preserves the harmonic properties (back harmony) of its root, [hiːɡ]: [hiɡ-iːt] governs back harmony, because its root, [hiːɡ], governs back harmony. We will refer to this property of harmony as Harmonic Uniformity (Rebrus & Szigetv ari 2016). We return to Harmonic Uniformity in §8.

THE LOSS OF A WORD-FINAL VOWEL
In some cases a root may lose its ending even in what we categorize as additive morphology. This occurs only before a small set of derivational suffixes. It involves the loss of a single rootfinal vowel, short, as in (6a), or long, as in (6b). (All the suffixed forms include a verbalizer, we spare their gloss.) The loss of this vowel is linked to the verbalizing suffixes in (6), it is not merely a phonological motive (hiatus avoidance) that lies behind it. We do not find this loss before the homonymous essive-modal suffix ([meːrʦɛ] 'gauge', [meːrʦeː-yl]; [kucɑ] 'dog', [kucaː-ul]). This operation is different from the truncation observed in diminutive forms to be discussed in §5 on at least two counts: (i) in only affects a single root-final vowel, (ii) in does not limit the size of the resulting word form.

TRUNCATIVE MORPHOLOGY
What we call morphological truncation only occurs in diminutive forms. As we have seen, truncation is different from the loss of a word-final vowel, (6), because it is not limited to a single vowel, but may affect longer strings, including consonants and multiple vowels. Truncation applies in diminutive forms because they must satisfy a template that contains a given number of syllables (5 vowels). Templatic diminutive forms are one, two, or three syllables long, the two-syllable template accounting for the vast majority of examples. Since the point of truncation is to reduce the number of syllables in the stem to one, it will inevitably target at least one vowel: truncating only consonants does not influence the syllable count. While in (7a) only a vowel is deleted from the stem before the adjectivizer and the verbalizer suffix. In (7b) a much longer string is lost from the stem before the diminutive ending. The reason why stems are so radically truncated in diminutive forms is that these forms must fit a disyllabic template. We split stems and diminutive endings by a vertical bar in (7) and below. We will encounter diminutive forms where the position of this vertical bar is indeterminate. In this case we omit it. As shown in (7c-d), the output of vowel loss and truncation may coincide: truncation, (7d), may trim no more than a word-final vowel to create the base of a diminutive form, just like vowe loss occurring before certain verbalizers, (7c).
There are several diminutive templates. In addition to disyllabic diminutives, we also find a small number of trisyllabic and monosyllabic diminutive forms, although the analysis of the latter group is debatable. In template formulae we represent the truncated base by a dash: "-". Thus we say that [tøri] satisfies the diminutive template -i. In the following we discuss the kinds of relationships there exist between nondiminutive bases, their truncated versions, and diminutive templates.

DIMINUTIVE TEMPLATES
The ending of all disyllabic diminutive templates is fixed. This ending contains minimally a vowel. In some endings this vowel is preceded or followed by a consonant too.  [ɛ]. In fact, [ɛ] is obligatorily followed by a consonant. There are no templates with a fixed templatic consonant on both sides. 12 We are going to catalogue the endings by their vowel.
Since the size of diminutive templates is fixed and most templates contain a vowel, the base of diminutive forms has to be truncated if longer than one syllable (in the case of disyllabic templates) or two (in the case of trisyllabic ones). Typically the end of the base is truncated, but it is sometimes the beginning (see Tompa 1964 for some examples). Initial truncation is more common with bases beginning with a vowel ( occur on their own). Apparently, with names this is not a common possibility, we have found a single example, (8c).
There is a small set of monosyllabic diminutives. This is most common with phatic elements, as in (9a), and some common nouns, (9b), especially among specialists, but rather rare with given names, (9d). 13 We are uncertain if the compounds in (9c) belong here at all. We will ignore this set in the present discussion.
6.1. Templates containing [i] The most common template vowel is [i]. In the most common case this vowel is not accompanied by any consonant. This template is represented by the formula -i. The other templates containing [i] are -ʦi, -ʃi, -ʧi, -si, -ʒi, and -zi.

Acta Linguistica Academica
Monosyllabic bases are unaffected by diminutive formation: in this case the process cannot be distinguished from regular additive suffixation, (10a). Any base that is longer is subject to truncation, which leaves behind a single vowel in the base. If there is a single consonant between the first two vowels of the base, it usually remains in the diminutive form, (10b). Some apparent exceptions include [h] and [j], as shown in (11) (Vago 1980;van de Weijer 1989) are based on very little data (only given names) and make empirically false predictions. They make reference to possible syllable-or word-final clusters. One problem is that the status of some clusters is dubious ( 16 While in many cases the vowel of the base remains unchanged, we find both shortening, (12a), and lengthening, (12b), of this vowel in [i] The examples in (13) show that most consonants after the first vowel of the base are trimmed, The -ʧi template is less strict than any of the other templates containing [i]. It forms a consonant cluster with both a homorganic and nonhomorganic nasal, (15a) and (15e), with approximants, (15b-d), and even with plosives and fricatives, (15f-g), creating clusters unprecedented within a morpheme. This tolerance of pre-[ʧ] consonants in the -ʧi template must be the reason why we hardly find cases where [ʧi] follows a vowel. 23 (15h)  This is a rather varied set. In (17a) we see a verbal base that does not contain [ʒ]. 24 The adjectives they are paired with probably exemplify a -ʒi template. The pairs in (17b) satisfy both this and the general -i template. In (17c)

Templates containing [ɑ]
Unlike diminutives ending in [i], the diminutive template -ɑ imposes a restriction on the first syllable, it must be light. 26 This means that a long vowel in the base is shortened, (19b, e), and a consonant cluster is trimmed, (19c, f

This kind of palatalization is different from what we see in regular additive morphology.
There it is a [j] segment in the suffix that palatalizes the preceding dental consonant, resulting in a palatal geminate after a vowel, 28 (4c). Diminutive palatalization never results in a geminate, therefore we do not posit a -jɑ template. In want of a better formula, we will represent this as -ʲɑ.  Like diminutives in [i], diminutives in [ɑ] also have templates in which a consonant precedes the template vowel. In the -ʦɑ template we have found mostly names, other items that could qualify semantically do not have a base ([ʦiʦɑ] 'pussycat', [ʦoʦɑ] 'piggy'). Many of these names are disyllabic, (20a), hence they fit among all other disyllabic templates discussed here. However, there is also a trisyllabic set, which contains exclusively female names, 30 (20b). We could claim that these are diminutives of diminutives (like some items in (13f) (1931)(1932)90). 30 There is also a common noun in this set containing an augmented bound stem: [fɑr (o) Here too we find a trisyllabic template, exemplified by the items in (21d). In fact, all of these diminutive forms have a consonant before [-kɑ], but this consonant is also present in the base. Yet the base is truncated in a way that is untypical of the vowel loss described in §4.

Templates containing [o]
Short [o] (and to a lesser extent short [ø], but see §7) does not occur word finally in Hungarian. As a result, we find long [oː] in diminutives whose template contains this vowel word finally, (22a). In a more limited set of diminutives, the template is consonant final. This consonant is 31 [raːkʧɑ] and [siporkɑ] may also be analysed as cases of backformation: [-l]

Templates containing [u]
Like in templates containing [i], those with [u] also may end in [ʃ]. The words in (24a) could also be considered instances of clipping, but those in (24b-c) show that there does exist a diminutive template -u. The diminutives in (24a-b) do not end in [u], those in (24c)  It is a unique feature of diminutives ending in [-ɛr] that the consonant before the ending is geminated, hence we assign the formula -ːɛr to this template. We could not identify the source of the [k] in [bɑkkɛr]. We assume that this template imitates disyllabic German and English loanwords in which the consonant between the two vowels is also often geminated ( (1931)(1932)91) claims that [-ɛk] is abstracted away from words of Slavic origin (more specifically, probably Slovak).
In a smaller set of words we find a trisyllabic template, -iŋɡɛr. This is less productive than -ːɛr. The form [kɑlɛf], (26g), is unique in having the ending [-ɛf], we could not discover its provenance.

Summary of disyllabic diminutive templates
We summarize the disyllabic diminutive templates we have identified so far in (27). For the templates that are in parentheses we have only found one example. 38 The [j] in superscript represents the palatalization of the consonant before the templatic vowel. We have seen that this is not a reflex of a consonant [j], but the representation is not meant to suggest an autosegmental analysis for this phenomenon. In this chart we ignore vowel length alternations in the stem. 38 In this case we only considered diminutives in which the palatal or the [ʃ] is not present in the base. So although [ɑɲuʃ] 'mother.DIM' or [tɛʃoː] 'sibling.DIM' match the template -ʲuʃ and -ʃoː, they are here taken to belong only to -uʃ and -oː.

(27)
Disyllabic diminutive template endings -i -ɑ -o -u -ɛ -i, -iʃ -ɑ, (-ɑj) -oː, -os, -oːk -u, -uʃ -ɛs, -ɛk, -ːɛr, Based on this chart, we can make the following generalizations:  (27) as a markedness hierarchy, where templates on the left and at the top have a larger type frequency than those on the right and at the bottom. It follows that the templates with a single example are all at the marked edges of the hierarchy. Also the templates are in contiguous scales with two exceptions: no [k] before [i] and the -ʧoː template seems to be missing, although there is at least one item in the more marked -ʃoː template.

REDUPLICATIVE DIMINUTIVES
The reduplicative diminutive template is also disyllabic. Typically the first CV of the base is copied, As before, there are some diminutive forms that fit several templates. The items in (28g) are both diminutives and examples of the -oː, -i, and -u templates. While most of our examples are names, common nouns are also subject to this process. Many of them belong to child language, (28h). This is a rather universal feature of language as some of the English glosses for these words show.

TEMPLATES AND VOWEL HARMONY
The dative suffix following the nondiminutive and diminutive forms of the names in (31a-d) prove that templatic diminutives are not subject to Harmonic Uniformity. In (31a-c) the first name contains a back vowel in its last syllable, accordingly the suffix also shows up with its back-vowelled allomorph. In the diminutive forms this back vowel is replaced with a neutral vowel. As a result the names in (31a-b) will govern front harmony, like polysyllabic, monomorphemic words, as we saw in (5f) in §3, and unlike the words in (30c-e). In (31c)

CONCLUSIONS
In this paper we have discovered several aspects in which additive (agglutinative) morphology (AM), the regular method of inflection and derivation in Hungarian, is different from truncative morphology (TM), applied in the creation of diminutive forms only. One set of differences concern vowel alternations. These are listed in (32). The shortening of the last vowel and the loss of the final vowel of the stem or base occurs in both additive and truncative processes, (32a-b), but this is where the similarities end. Although the stem vowel may lengthen in both types of process, in additive morphology we only find the lengthening of a stem-final low vowel, in truncative morphology it is only a stem-internal vowel that may lengthen, and not only low [ɛ] and [ɑ], but also mid [o]. As for vowel/zero alternations, it is only a single short vowel that may be lost in additive morphology. Truncative morphology, on the other hand, may disappear not only short, but also long vowels, and not only one, but any number of them. (In addition, consonants may also be lost only in this type of process.) Furthermore, while only the last vowel of the stem, potentially followed by a single consonant, may be targeted by additive morphology, truncation can affect any vowel of the base.
It must be pointed out that the output of additive morphology and diminutive truncation may be indistinguishable when the latter targets a single stem-final vowel. The former process cannot target more than one short vowel, as in (33a), but in the diminutive form in (33b) we also see that a single short vowel is lost. However, as we have pointed out above and try to hint at by not using the hyphen as a morpheme boundary symbol, the diminutive form [eːʋji] behaves as a monomorphemic word, which accordingly must govern front harmony (cf. (5f)), and not as a polymorphemic word, like [beːn-iːt], which governs back harmony due to Harmonic Uniformity. In (33) we enclose the target of truncation in angle brackets. Additive and truncative morphology are also different in the harmonicity of the suffixes (or endings) added to stems/bases. We summarize the possibilities in (34).