The Sonority Sequencing Principle and the Structure of Slovak Consonant Clusters

Generally, the structure of the syllable in languages is governed by the so-called Sonority Sequencing Principle (also known as the Sonority Sequencing Generalization), which states that the centre of the syllable is created by the most sonorous sound and the sonority of the other segments in the syllable decreases towards its margins (see, for example: Blevins, 1995; Goldsmith, 2011; Zec, 2007). This idea is usually connected with the work of Jespersen (Jespersen, 1904). In Jespersen’s theory, each speech sound is characterized by a certain degree of sonority. Sonority – the relative loudness of a sound – is given by the functioning of vocal cords and the opening of supra-glottal cavities during the articulation of the sound (see, for example, Ladefoged, 1992), that is, it depends on the articulatory-acoustic and perceptual characteristics of sounds.1 All speech sounds are ranked on the sonority scale with vowels as the most sonorous sounds at the top of this hierarchy and obstruents at its bottom. The sonority hierarchy “vowels > liquids > glides > nasals > fricatives > stops” is considered language-universal and can also be found in the Universals Archive as Universal Number 626 (https://typo.uni-konstanz.de/rara).


Introduction
Generally, the structure of the syllable in languages is governed by the so-called Sonority Sequencing Principle (also known as the Sonority Sequencing Generalization), which states that the centre of the syllable is created by the most sonorous sound and the sonority of the other segments in the syllable decreases towards its margins (see, for example : Blevins, 1995;Goldsmith, 2011;Zec, 2007). This idea is usually connected with the work of Jespersen (Jespersen, 1904). In Jespersen's theory, each speech sound is characterized by a certain degree of sonority. Sonority -the relative loudness of a sound -is given by the functioning of vocal cords and the opening of supra-glottal cavities during the articulation of the sound (see, for example, Ladefoged, 1992), that is, it depends on the articulatory-acoustic and perceptual characteristics of sounds.1 All speech sounds are ranked on the sonority scale with vowels as the most sonorous sounds at the top of this hierarchy and obstruents at its bottom. The sonority hierarchy "vowels > liquids > glides > nasals > fricatives > stops" is considered language-universal and can also be found in the Universals Archive as Universal Number 626 (https://typo.uni-konstanz.de/rara).

The sonority hierarchy in the Slovak language
The literature on sonority offers a number of sonority scales (for details, see Parker, 2002). They all agree on vowels as the most sonorous segments and obstruents as the sounds with the lowest degree of sonority. Parker, in his account of sonority, mentions that he found more than 100 sonority hierarchies, some of them promoting their universality and others being suitable only for particular languages (Parker, 2002, p. 62). One of the most quoted sonority scales is that proposed by Jespersen (Jespersen, 1904). Considering the aim of my research, a sonority-based analysis of initial and final consonant clusters, attention will be paid to the ranking of consonants. Following Jespersen's theory, consonant Renáta Gregová The Sonority Sequencing Principle and the Structure… Page 3 of 20 sounds are arranged into the following sonority hierarchy: "r-sounds > laterals > nasals > voiced fricatives > voiced stops > voiceless fricatives > voiceless stops" (Jespersen, 1904, p. 192). When comparing this hierarchy to the sonority scale offered by Universal 626 (see above), it is clear that Jespersen's scale provides a more detailed sonority specification of voiced and voiceless fricatives and stops. Languages with affricates offer even more refined sonority hierarchies, as for example the following one for Polish: glides > liquids > nasals > fricatives > affricates > stops (Pawelec, 2012, p. 184). In the Slovak language, there are nine degrees of the sonority of consonants depending on the place and manner of articulation of these segments, as specified in Table 1 2 For a detailed phonetic and phonemic discussion about the sonority differences among obstruents in languages and the different sonority of voiced and voiceless sounds in languages, see Parker, 2002. 3 The transcription symbols for some Slovak consonant phonemes are different from the symbols used by the IPA. In particular, the Slovak symbols /[š, ž, č, ǯ, ť, ď, ň, ľ]/ stand for the IPA /[ʃ, ʒ, ʧ, ʤ, t j , d j , n j , l j ]/, respectively. See also note 5. 4 Usually, the label "sonorant" is used to indicate all consonantal segments /r, l, m, n, j, ľ, ň/ to distinguish them from obstruents (i.e. fricatives, affricates and plosives) in Slovak. Nonetheless, there are differences in the degree of sonority among sonorants, and thus the name "sonorant" is used here only to indicate the sounds with an identical degree of sonority that is lower than the sonority of the other sonorants, liquids /r, l/ and nasals /m, n/. A similar simplification is applied to nasals. In Slovak, there are three nasal phonemes /m, n, ň/; however, the sonority of /ň/ is lower than the sonority of /m, n/, which is why /ň/ is covered by the term sonorant only.
5 The symbol /ʒ/ stands for the alveolar voiced affricate as pronounced, for example, in the word priadza 'yarn', and the symbol /ǯ/ indicates the post-alveolar voiced affricate occurring, for example, at the beginning of the word džavot 'jabber'.

Renáta Gregová
The Sonority Sequencing Principle and the Structure… Page 4 of 20

The Sonority Sequencing Principle and Slovak initial clusters
According to Kuryłowicz (Kuryłowicz, 1948), the beginning of the word is at the same time the beginning of the first syllable of the word and the end of the word is at the same time the end of the last syllable of the word. In Slovak, the maximum number of consonants occurring word/ syllable-initially is four. No word in Slovak can begin with more than four consonants.6 Most of the initial consonant clusters are binary (CC). These clusters represent almost 93.9% of all consonant clusters occurring in the initial position of a word (syllable). Less frequent are three-consonant clusters (CCC), representing 5.8%, and the lowest frequency of occurrence can be observed in four-consonant clusters (CCCC), accounting for only 0.3% (Sabol & Ivančová, 2014).

The sonority-based analysis of initial two-consonant clusters
Slovak initial CC clusters fall into 34 different types7 (Table 2), and seven of them violate the SSP as specified for Slovak.8 6 Claims that some Slovak words start with a five-consonant cluster (e.g. zmrzlina 'icecream', žblnkot 'murmur' or žbrnda 'pigswill') do not respect the difference between the graphic and the sound level of a language and the important fact that the syllable is a sound unit. The word zmrzlina 'ice-cream' is three-syllabic: zmr.zli.na with '-r-', '-i-' and '-a-' as syllabic nuclei. The words žblnkot 'murmur' and žbrnda 'pigswill' are both di-syllabic: žbln.kot and žbrn.da with the syllabic nuclei '-l-', '-o-' and '-r-', '-a-', respectively. In line with the idea that a consonant cluster is a sequence of consonants that appear together in the syllable without a vowel between them (see, for example, Jones, 1976) and that the centre of the syllable can be a vowel but also a consonant with a high degree of sonority, in all three words there are only two-consonant initial clusters (see also Gregová, 2016).
7 An exhaustive list of (almost) all tokens of initial, medial and final clusters obtained by the comparison of several sources, as well as by an in-depth analysis of the Slovak national corpus, can be found in Gregová, 2012Gregová, , 2016 In the Slovak language, the combinations of two obstruents or two nasals are allowed word-initially (Zec, 2007, p. 189); for example, ktorý 'which' or mnoho 'many'. Sonority plateaus are acceptable in this language and the Minimal Sonority Distance for Slovak is 0.

Renáta Gregová
The Sonority Sequencing Principle and the Structure… Page 5 of 20 The clusters that do not follow the SSP will now be analysed in more detail. Table 2, initial two-consonant clusters that violate the Sonority Sequencing Principle, that is, their Minimal Sonority Distance is 0 (see note 8), fall into five categories, depending on their initial segment:

The analysis of the initial two-consonant clusters violating the SSP As it follows from
(1) starting with a voiceless fricative, for example: sp, st, sv, χv (2) starting with a voiced fricative, for example: vd, vď (3) starting with a nasal, for example: mň, mľ (4) starting with a liquid, for example: lk, lž (5) starting with a voiceless affricate, for example: cť The visual difference between the so-called well-formed and ill-formed syllables with well-formed and ill-formed initial clusters from the viewpoint of the SSP can be seen in Figures 1-3. The sonority degrees of consonants are captured in Table 1. Degree 10 belongs to vowels, and degree 9 indicates the sonority of the semi-vowels [î] and [û] that occur at the beginning of Slovak diphthongs (see Figure 6).
Figure 1. Sonority graph of the monosyllabic word znak 'sign' with the initial cluster 'zn'

Renáta Gregová
The Sonority Sequencing Principle and the Structure… Figure 2. Sonority graph of the monosyllabic word tkáč 'weaver' with the initial cluster 'tk' Figure 3. Sonority graph of the word spolu 'together' with the initial cluster 'sp' Figure 1 illustrates CC cluster 'zn' with rising sonority, and Figure 2 -cluster 'tk' with the plateau between 't' and 'k', representing the Minimal Sonority Distance of 0 that is acceptable in Slovak (see note 8). Figure 3 visualizes the cluster 'sp' that violates the principle of rising sonority in the syllable onset (see Pawelec, 2012, pp. 185-186 for similar findings in Polish).
The clusters in (2)-(5) are very rare in Slovak. The sample clusters in (2) can only be found in the words vdova 'widow', vďaka 'thanks' and their derivatives. The exceptional clusters in (3) occur in some onomatopoetic

Renáta Gregová
The Sonority Sequencing Principle and the Structure… Page 8 of 20 words and their derivatives (mňaukať 'mew') or in the so-called higher pronunciation style of some words (Kráľ, 2005) as a result of a neutralization process (for details, see Sabol, 1989). The clusters starting with a liquid (4) manifest themselves in some poetic and literary expressions, that is, in stylistically marked words borrowed from Czech (Sabol, 1975, p. 204), for example, lkať 'to lament' or lživo 'falsely'. The cluster 'cť' occurs only in words derived from ctený 'honoured' or ctiť 'to honour, to revere', which are considered archaic and not stylistically neutral, either. However, CC combinations with a voiceless fricative at their beginning, especially those starting with 's' ('sp', 'st', 'sv'), are the most frequent initial CC clusters in Slovak (Sabol, 1975, p. 53).

The sonority-based analysis of initial three-consonant clusters
There are 28 types of initial CCC clusters in standard Slovak (see note 7), and 16 of them do not follow the SSP (Table 3).  Table 3 shows that the initial CCC clusters violating the SSP are the following types:

The analysis of the initial three-consonant clusters violating the SSP
(1) starting with a nasal, for example: mdl, msť (2) starting with a voiced fricative, for example: vzd, zbl, zdv, zgň, zdn, hmľ (3) starting with a voiceless fricative, for example: fst, scv, svr, str, fpľ, štv, stm (4) starting with a sonorant, for example: ľst It is clear that the initial segments disturbing the principle of sonority in the initial three-consonant combinations are the sounds 'm', 'v/f',14 'z/s',15 'h' and 'ľ'. The sonority shapes of the sample CCC clusters violating the SSP are illustrated in Figure 4 and Figure 5. Figure 4. Sonority graph of the word mdlý /mdlí/ 'insipid' with the initial cluster 'mdl' 14 The occurrence of either 'v' or 'f' depends on the character of the following segment due to the regressive assimilation of voicing (neutralization of the phonological opposition voiced -non-voiced) in the Slovak language. See also note 15.
15 The occurrence of either 'z' or 's' depends on the character of the following segment due to the regressive assimilation of voicing (neutralization of the phonological opposition voiced -non-voiced) in Slovak. See also note 14.

Renáta Gregová
The Sonority Sequencing Principle and the Structure… Page 11 of 20 Clusters starting with 'h' and 'ľ' are very rare. The cluster 'hml' in hmlisto /hmľisto/ 'hazily' originated as an anagram from Old Church Slavic *mьgla (see Králik, 2015). The clusters with an initial 'ľ' can be found in borrowings from Czech (ľstivo /ľsťivo/ 'tricky'). The clusters with a nasal sound at their beginning ('mdl', 'msť') can also be found only in words with a Czech origin (see Table 3 for examples). The other sequences of consonants of the types (2) and (3), that is, those starting with a voiced/voiceless fricative, are among the most common types of initial consonant clusters in Slovak (see Gregová, 2016;Sabol, 1975).

The sonority-based analysis of initial four-consonant clusters
When considering the initial consonant clusters, the combinations with four segments (CCCC) are the least numerous. There are five types of initial four-consonant clusters in Slovak (Table 4). Four types of initial CCCC clusters violate the SSP, as indicated in Table 4 and visualized in Figure 6.

Renáta Gregová
The Sonority Sequencing Principle and the Structure… Page 12 of 20

The analysis of the initial four-consonant clusters violating
the SSP Taking into account the SSP, and as it follows from Table 4, the exceptional initial CCCC clusters are of two types: (1) starting with a plosive, for example: pstr (2) starting with a fricative, for example: vzdm, fspr, fskv Four-consonant clusters starting with a plosive, for example, the cluster in (1), are very limited. The clusters in (2) can be seen as initial three-consonant clusters starting with a fricative (Table 3) preceded by the initial fricative 'v/f'.18 Thus, the sonority is violated by the initial sequence "fricative -fricative" as a whole, as can be seen in the sample analysis of the cluster 'fspr': f -sonority degree 4, s -sonority degree 4, p -sonority degree 0, r -sonority degree 8 the sequence of sonority in the cluster 'fspr' = 4 -4 -0 -8 (see Table 1 and Figure 6 for sonority degrees) Figure 6. Sonority graph of the word vzpriamiť /fspriamiť/ 'to raise sth.' with the initial cluster 'fspr'

Renáta Gregová
The Sonority Sequencing Principle and the Structure… Page 13 of 20

The Sonority Sequencing Principle and the Slovak final clusters
The phonotactic possibilities at the end of Slovak words (syllables) are not as rich as in the initial position (see 2.1). There are only two-and three-consonant clusters in the word/syllable-final position in Slovak (see, for example, Gregová, 2016;Sabol, 1975). Final two-consonant clusters prevail over threeconsonant clusters in the final position by 91.4% to 8.6%, respectively.

The sonority-based analysis of final two-consonant clusters
In the Slovak language, the final two-consonant clusters fall into 19 types (Table  5, see also note 7), and only two types of those clusters violate the SSP. A graphical presentation of a sample cluster that does not follow the SSP is given in Figure 7. (1) ending with a fricative, for example: ps (2) ending with a nasal, for example: jn The clusters ending in a nasal are very exceptional and can be found only in words of foreign origin, for example, 'jn' in the word kombajn 'harvester' (English origin) or in the word dizajn 'design' (French origin, cf. e.g. Králik, 2015). Final two-consonant clusters ending in a fricative are not frequent, either. They manifest themselves in some onomatopoetic expressions, for example, čľups 'plop' and in foreign words, for example, kolaps 'collapse' (taken from Latin) or čips 'chips' (English origin).

The sonority-based analysis of final three-consonant clusters
There are three types of final three-consonant clusters in Slovak (Table 6) and they are all fully in agreement with the SSP, which states that the sonority lowers towards the edges of the syllable (see above).

Discussion and implications for further research
Considering the overall number of word/syllable-initial consonant clusters in Slovak (67 types), 40% of them are not in line with the principle of sonority. Although the occurrence of many of those clusters is limited to borrowings from other languages, initial two-and three-consonant clusters starting with fricatives, among them especially those starting with 's', are the most common types of consonant clusters in standard Slovak (see 2.1.1.1 and 2.1.2.1). The final clusters are less numerous and the number of those violating the SSP is lower, too (9%). The present analysis thus proves the claim about the distribution and character of tokens and types of consonant clusters per syllable position: in the Slovak language, the word/syllable-final combinations of consonants are not only less frequent (compared to the initial and medial positions) but also the least varied (cf. Gregová, 2016;Sabol, 1975).
The findings on the initial clusters raise two basic questions: (1) Is the Slovak language somehow exceptional pursuant to the principle of sonority? and (2) If the syllables whose structure violates the Sonority Sequencing Principle are considered ill-formed, why are they so prevalent in the language?

Conclusion
In accordance with the Sonority Sequencing Principle (SSP), also known as the Sonority Sequencing Generalization, the structure of the syllable in languages depends on the sonority of the individual segments: the most sonorous sound creates the peak of the syllable (the syllabic nucleus, centre) and the sonority of segments in the syllable onsets and codas increases towards the syllable peak. A detailed analysis of initial and final consonant clusters in Slovak delimited on the basis of Kuryłowicz's functional approach (Kuryłowicz, 1948) has shown that almost 40% of initial clusters in Slovak violate the SSP. Although some of those sonorously ill-formed clusters occur only in words that are on the periphery of the Slovak vocabulary, many clusters, mainly those starting with a voiced/voiceless fricative, are very common in the central vocabulary. The frequent initial ill-formed clusters violating the SSP usually have the consonant /s/ as their initial segment. Similar findings on the violation of the SSP in the word-initial position can be observed in other (not only Slavic) languages (see section 1). The data about the violation of sonority in languages indicate both that the sonority of segments is not the only criterion a language uses for the creation of words, and that the process of the creation of consonant sequences can also be governed by the combinability of segments on the basis of their distinctive features (Sabol, 1975) and/or by the perceptual advantage of consonants' combinations in languages (Engstrand & Ericsdotter, 1999). However, the verification of these two theories is a task for future research.