3.1 Introduction
This chapter reviews consonantism in Slavic. The main focus is on the common consonantal segments, their patterns, and the role they have played in phonological theory. Section 3.2 reviews the inventories of the 12 contemporary Slavic languages. As we will see, Slavic languages have relatively large consonant inventories, which is related to three factors: (i) most Slavic languages contrast plain vs. palatalized consonants, (ii) many Slavic languages have two sets of posterior coronal fricatives, and (iii) obstruents contrast voicing.
The remaining sections look at alternations affecting consonants. Section 3.3 examines palatalization, which is particularly extensive in Slavic. In particular, I review velar palatalization (and related processes of velar fronting and iotation) as well as secondary palatalization. Section 3.4 examines two common laryngeal alternations: final devoicing and voicing assimilation. A typical Slavic language has final devoicing and regressive voicing assimilation within obstruent clusters, but there are languages without final devoicing and languages that allow clusters of voiced and voiceless obstruents. Section 3.5 examines other local processes affecting consonants, such as place assimilation, dissimilation, and decomposition.
Finally, Section 3.6 discusses patterns in which consonants affect one another at a distance. Consonant harmony is a pattern in which sibilants must agree in some property within a word; it is found in two Slavic languages. Consonant co-occurrence restrictions (described as the effect of the Obligatory Contour Principle, OCP) constitute the opposite: two instances of a segment (or a class of segments) cannot co-occur across a vowel or within a word.
3.2 Consonant Inventories
Most contemporary Slavic languages have larger than average consonant inventories. This can be attributed to the inherited contrasts of the Proto-Slavic inventory as well as additional subsequent contrasts.
The Late Proto-Slavic consonant inventory distinguishes five places of articulation, voicing of obstruents and affricates in addition to stops and fricatives, and a range of sonorants (Table 3.1). Notable is the abundance of palatalized consonants and the asymmetries in fricatives: there is no voiceless labiodental fricative or voiced velar fricative. This inventory has 27 consonantal phonemes, which is considered to be moderately large among the languages of the world (Reference Maddieson, Dryer and HaspelmathMaddieson 2013).
Table 3.1 Late Proto-Slavic consonant inventory
| Labial | Dental | Alveol. | Palatal | Velar | |
|---|---|---|---|---|---|
| Stop | pb | td | tjdj | kg | |
| Fricative | v | sz | ʃʒ | sjzj | x |
| Affricate | ʦ | ʣʧ | |||
| Nasal | m | n | nj | ||
| Rhotic | r | rj | |||
| Lateral | l | lj | |||
| Glide | j |
On the whole, the contemporary Slavic languages have inherited the Proto-Slavic inventory contrasts, with changes mostly affecting the palatalized consonants. In fact, many contemporary languages have secondary palatalization of consonants which can affect almost all consonants (such as in Russian and Bulgarian). Table 3.2 summarizes the phoneme inventories across the 12 contemporary standard Slavic languages. The segments are color-coded by frequency.Footnote 1 This table is meant to provide a general overview of what a typical Slavic consonant inventory looks like, despite the potential drawbacks of this approach, which minimizes the discrepancies among the existing descriptions and ignores dialectal variation. All languages distinguish three pairs of stops /p, b; t, d; k, ɡ/, four fricatives /f, s, z, x/, the affricate /ʦ/, two nasals /m, n/, and the palatal glide /j/. These phonemes are mostly inherited from Proto-Slavic (although subject to sound change, see the following sections). Only /f/ was introduced predominantly through borrowing, though in some languages it also derives from *xv or other groups, for example Macedonian [fati] ‘grab’ < *xvatiti.
Table 3.2 Phoneme inventories of contemporary standard Slavic languages, shaded by frequency
Note. The approximants [w] and [wj] are labiovelar.
Most Slavic languages (except Czech, Macedonian, Bosnian-Croatian-Serbian or BCS, and Slovenian) distinguish secondary palatalization. Phonetically, secondary palatalization is realized with a raised or fronted tongue body (Reference KochetovKochetov 2002). Along with postalveolars, palatalized consonants arose in Proto-Slavic as a result of various phonological processes. The contemporary Slavic languages differ in terms of how many of the original contrasts they have maintained (Reference KochetovKochetov 2006, Reference Kavitskaya, Goldstein, Whalen and BestKavitskaya 2009, Reference Kavitskaya, Iskarous, Noiray, Proctor, Reich, Babyonyshev and KavitskayaKavitskaya et al. 2009, Reference Iskarous and KavitskayaIskarous & Kavitskaya 2018). In fact, not all palatalized consonants are equally attested across Slavic: palatalized labial nasal and oral stops are found in more languages than palatalized velars and coronals. Cross-linguistically, palatalized rhotics are marked (Kavitskaya et al. 2009; Iskarous & Kavitskaya 2010; Hall & Hamann 2010; Howson 2018). Palatalized postalveolars are rare. The processes affecting secondary palatalization will be reviewed further in Section 3.3.
Posterior sibilants differ phonetically and phonologically across the Slavic languages (Reference ŻygisŻygis 2003, Reference HamannHamann 2004). While the majority of Slavic languages have postalveolar affricates and fricatives, some Slavic languages have retroflex fricatives and affricates (e.g. Lower Sorbian), others have both (Russian), while a third group displays significant dialectal variation (Polish, BCS). Lower Sorbian, Polish, and BCS also have an additional set of alveopalatals.
One striking difference among the Slavic languages is the status of /v-w-ʋ/, which can vary phonetically from a voiced fricative to an approximant. This phoneme behaves like a sonorant in many Slavic languages (e.g. BCS, Belarusian, Ukrainian); in other Slavic languages it can behave as a fricative or sonorant depending on context (e.g. Czech, Slovak), or as an obstruent entirely (Polish). The voiced labiovelar sometimes assimilates in voicing, but other times it does not affect voicing assimilation. In many Slavic languages, the voiced labiodental has multiple allophones. For instance, in Ukrainian it is realized as [ʋ] in onsets, but as [w] in codas. The voiced labiodental is discussed further in Section 3.4.
In terms of obstruent voicing, most place and manner combinations have a voiced–voiceless pair, with some notable gaps: phonemic voiced velar fricatives are only found in Belarusian, and while all languages have a phonemic /ʦ/, only five have its voiced counterpart /ʣ/. In other languages, the missing voiced obstruents are allophones, appearing before voiced obstruents.
Among the rhotics, most Slavic languages are described as having the alveolar trill. Three languages have other rhotics: Slovenian has a flap (Reference Šuštaršič, Komar and PetekŠuštaršič et al. 1995), Upper Sorbian has a uvular trill (Reference HowsonHowson 2018), and Czech has an additional trill-fricative [r̞] (Reference HowsonHowson 2017).
Overall, many Slavic languages show substantial positional, dialectal, and interspeaker variation in rhotics, which is sometimes mirrored in the descriptions of the standard varieties.
Not shown in the inventories above are the geminates, or long consonants. Most Slavic languages allow geminates across a morpheme boundary, where they can be variantly pronounced as singletons in casual speech. Perhaps the most extensive inventory of geminates is found in Russian. The Russian geminates arise due to morpheme concatenation (e.g. [vːos] ‘import’, [rɐsːkas] ‘story’) or morpheme-internally in loanwords ([bɐrokːə] ‘baroque’, [sumːə] ‘sum’). Reference Dmitrieva and KubozonoDmitrieva (2017) demonstrates that the Russian geminates have systematically longer duration than singletons. Russian geminates display some of the cross-linguistically common properties, such as the rarity of sonorant geminates or geminates appearing next to another consonant. At the same time, Dmitrieva identified several language-specific idiosyncrasies, such as the relative robustness of word-initial and voiced-obstruent geminates. Geminates are also very common in Belarusian and Ukrainian because of the historical change Cj > Cjː, for example Ukrainian [ʒɪtjːa] ‘life’ < *ʒitje.
This concludes the overview of consonant inventories in Slavic languages. In the following section, we move on to the alternations affecting consonants.
3.3 Palatalization Processes
One of the most striking properties of all contemporary Slavic languages is the various palatalization processes. Palatalization is defined as the pronunciation of consonants in the palatal or postalveolar, and more broadly in the coronal region (Reference BatemanBateman 2007, Reference Kochetov, van Oostendorp, Ewen, Hume and RiceKochetov 2011). This description encompasses both secondary articulation (e.g. k → kj) and change in place of articulation (e.g. k → ʧ).
To illustrate the phonological and morphological properties of palatalization, let us examine palatalization in BCS. The language involves several different palatalizations (Reference de Brayde Bray 1980, Reference Browne, Comrie and CorbettBrowne 1993, Reference MorénMorén 2006), which BCS inherited from Proto-Slavic. The first type of palatalization (1a) applies to stem-final velars which become postalveolars before the vocative suffix. All three velars are affected (1a-i), and the alternation changes the manner of two obstruents, but not of [x]. Voicing is maintained. The triggers of palatalization are specific suffixes which most commonly contain front vowels, as the vocative. However, not all suffixes that begin with a front vowel trigger palatalization (1a-ii) and not all triggering suffixes have front vowels (1a-iii). Thus, it is best to describe these triggers as morphological. Finally, velar palatalization is not fully productive and does not apply to all stem-final velars when followed by the triggering suffixes. In fact, sometimes palatalization applies variably (1a-iv).
The second type of palatalization affects two alveolars which become postalveolar (1b). In BCS, this alternation is also found with the vocative suffix (1b-i) and the nominative plural -ɔʋi/-ɛʋi. With the latter we can see that the velars are not affected. There are also a number of exceptions. Coronal palatalization is less productive in BCS and affects only a handful of suffixes.
(1)
BCS palatalization patterns (Reference Browne, Comrie and CorbettBrowne 1993, Reference MorénMorén 2006; prosody omitted) Velar palatalization: {k, x, ɡ} → {ʧ, ʃ, ʒ} a. i. uʧɛnik ‘pupil’ uʧɛniʧɛ ‘pupil.voc’ bɔɡ ‘god’ bɔʒɛ ‘god.voc’ sirɔmax ‘poor man’ sirɔmaʧɛ ‘poor man.voc’ ii. ʋɔjnik ‘soldier’ ʋɔjnikɛ ‘soldier.acc.pl’ iii. pɛku ‘bake.3p.pl’ pɛʧɛm ‘bake.1p.sg’ kruɡ ‘circle’ kruʒiti ‘circle.inf’ nɔɡa ‘leg, foot’ nɔʒurda ‘big ugly foot’ iv. ruka ‘hand, arm’ ruʧiʦarukiʦa ‘small hand’ Coronal palatalization: {ʦ, z} → {ʧ, ʒ} b. i. striʦ ‘uncle’ striʧɛ ‘uncle.voc’ knɛz ‘prince’ knɛʒɛ ‘prince.voc’ ii. striʦ ‘uncle’ striʧɛʋi ‘uncle.nom.pl’ ʋuk ‘wolf’ ʋukɔʋi ‘wolf.nom.pl’ ʋɔz ‘cart’ ʋɔzɔʋi ‘cart.nom.pl’ Velar fronting: {k, x, ɡ} → {ʦ, s, z} – i c. i. aɡnɔstik ‘agnostic’ aɡnɔstiʦi ‘agnostic.nom.pl’ bubrɛɡ ‘kidney’ bubrɛzi ‘kidney.nom.pl’ ɔrax ‘walnut’ ɔrasi ‘walnut.nom.pl’ ii. junak ‘hero’ junaʦima ‘hero.dat/LOC/INSTR.PL’ pɛku ‘roast.3p.pl’ pɛʦijax ‘roast.imperf’ ruka ‘hand, arm’ ruʦi ‘hand, arm.dat/loc’ iii. maʧka ‘cat’ maʧki ‘cat.dat/loc’ milka ‘(name)’ milki ‘(name).dat/loc’
The third type of palatalization also affects velars, but this time they turn into anterior coronals (1c-i). All suffixes that trigger velar fronting start with an [i], as shown in (1c-ii). Velar fronting is not without exceptions: it can be blocked phonotactically (e.g. *[maʧʦi] ‘cat.DAT/LOC’) and there are also lexically conditioned exceptions such as personal names (1c-iii). The final palatalization alternation in BCS is iotation (2). Historically, iotation applied before [j], but in contemporary Slavic languages [j] is generally no longer present. The difference between the palatalizations in (1) and iotation is in the number of segments affected: iotation applies to most segments. As in velar palatalization, velars become postalveolars (2a), while anterior coronals become posterior coronals (2b). Notice the difference between stops, which become alveopalatal (t → ʨ), and fricatives/affricates, which become postalveolar (s → ʃ).Finally, labials do not palatalize: instead the palatal lateral [ʎ] is inserted (2c).
(2)
BCS iotation (Reference Browne, Comrie and CorbettBrowne 1993, Reference MorénMorén 2006) a. {k, x, ɡ} → {ʧ, ʃ, ʒ} skakati ‘to jump’ skaʧɛ ‘jump.3p.sg’ tix ‘quiet’ tiʃi ‘quieter’ b. Coronals: {t, d, s, z, ʦ, n, l} → {ʨ, ʥ, ʃ, ʒ, ʧ, ɲ, ʎ} ʋratiti ‘to return’ ʋraʨɛn ‘returned’ nɔsiti ‘to carry’ nɔʃax ‘carry.imperf’ baʦiti ‘to throw’ baʧɛn ‘thrown’ xʋaliti ‘to praise’ xʋaʎɛn ‘praised’ c. Labials: ∅ → ʎ {p, b, f, m, ʋ} – ɡlup ‘stupid’ ɡlupʎi ‘more stupid’ ʎubiti ‘to kiss, love’ ʎubʎɛn ‘kissed, loved’ zaʃrafiti ‘to tighten’ zaʃrafʎɛn ‘tightened’ kraʋa ‘cow’ kraʋʎi ‘bovine’
The crucial difference between the palatalization patterns in (1) and iotation in (2) is that the set of triggering suffixes is different. Among all the palatalizations in BCS, only for velar fronting do all triggering suffixes share a clear common property: they all begin with [i]. As regards iotation, most suffixes are front, but not all (e.g. [nɔʃax] ‘carry.imperf’).
BCS inherited all four palatalizations from Proto-Slavic. This is why we also find palatalization in most other Slavic languages. Many of the patterns observed in BCS are similar to the ones found in the other languages, but there are also some key differences; I review these parallels in the remainder of this section. First, each palatalization type is morphologically conditioned, and the set of triggering suffixes may all share a phonological property (e.g. front vowels). Palatalization may also interact with morphology in another way: the suffix may display allomorphy, including the choice of inflectional paradigm, that is conditioned by the stem-final consonant.
Second, the outcome of palatalization is conditioned by the segmental inventory of the language. Notable in BCS is the relationship between the postalveolar and alveopalatal affricates. A more complex case is Polish, which has more sibilants and five distinct palatalization processes (see Reference Rubach, van Oostendorp, Ewen, Hume and RiceRubach 2011 for an accessible overview).
Third, most types of palatalization apply at morpheme boundaries. This type of alternation is termed a Derived Environment Effect (Reference Kiparsky, Hargus and KaisseKiparsky 1993, Reference Burzio, van Oostendorp, Ewen, Hume and RiceBurzio 2011). In the history of generative phonology, morphologically derived environments were treated together with phonologically derived environments, where an alternation applies only to a segment that is already derived by a different rule. A famous example of this type of is found in Polish, where /ɡ/ palatalizes to [ʤ] and then spirantizes to [ʒ] in specific environments: /bɔɡ-ɛ/ → [bɔʒɛ] ‘god-voc’. Underlying /ʤ/, however, does not undergo spirantization and surfaces faithfully. This type of pattern presents a substantial theoretical problem in rule- and constraint-based approaches.Footnote 2 In some cases, however, palatalization is phonologically triggered, such as secondary palatalization in Polish and Russian. In Polish, secondary palatalization is triggered by [i, j] not only at morpheme boundaries but also within morphemes (e.g. [mjit] ‘myth’) and across word boundaries ([dɔm] ‘home’ versus [dɔmj] Iwana ‘Ivan’s home’; Reference Rubach, van Oostendorp, Ewen, Hume and RiceRubach 2011).
Fourth, palatalization may be lexically restricted, which means that it does not apply to certain lexical items which meet the morphological and phonological requirements. Loanwords are a particularly likely source of exceptionality both in lacking expected palatalization (as in BCS) and showing a different set of facts. In Polish, for instance, secondary palatalization applies to postalveolars [ʒ, ʃ] but only in loanwords (e.g. [suʃji] ‘sushi’, [ʒjigɔlak] ‘gigolo’; see Reference GussmannGussmann 2007: §3.12). This sort of lexical exceptionality sometimes leads authors to conclude that a particular type of palatalization is no longer productive. In fact, Reference KapatsinskiKapatsinski (2010) shows how velar palatalization in Russian has lost productivity before a subset of suffixes. In Slovenian, velar palatalization is highly variable, but Reference Jurgec and SchertzJurgec & Schertz (2020) show that native speakers favor palatalization in nonce stems when followed by palatalizing suffixes.
Fifth, palatalization is subject to phonological restrictions, as is common in other types of sound patterns. In BCS, palatalization is blocked by certain clusters. When palatalization is variable, BCS adheres to Reference GuionGuion’s (1998) cross-linguistic generalization: /k/ palatalizes more often than the other two velars (Reference Browne, Comrie and CorbettBrowne 1993). I return to this point in Section 3.6.
3.4 Voicing Alternations
All Slavic languages contrast two sets of obstruents: voiced and voiceless. Phonetically, voiced stops are produced with voicing during the closure. Voiceless obstruents are typically unaspirated, although [kh] is aspirated in Upper Sorbian (as an allophone of /x/; Reference Ševc ŠusterŠevc Šuster 1984: 26–27)Footnote 3 and Slovenian (Reference Srebot-RejecSrebot-Rejec 1990); aspiration is not a factor in voicing alternations. The contrast between the two types of obstruents is not maintained in all contexts, and the neutralizing environments differ from language to language.
In word-final position, the voicing contrast is preserved in BCS and Ukrainian. All other Slavic languages display final devoicing, a process in which word-final obstruents are realized as voiceless. Standard Slovenian has final devoicing, but many dialects do not. Šmartno Slovenian presents an intermediate situation (3). In Šmartno, the root-final contrasts are maintained before sonorant-initial suffixes, as in the instrumental singular. Obstruents devoice word-finally, as in the nominative. The key generalization applies in the high vowel cases, such as the genitive. Roots with mobile stress paradigms, such as ‘dust’, shift stress to the suffix, but roots with fixed paradigms do not and as a result the word-final high unstressed vowel deletes. The newly final obstruent fails to devoice (/ˈbrieɣ-i/ → [ˈbrieɣ], not *[ˈbriex]). The interaction of devoicing and vowel deletion is opaque rather than transparent, because devoicing does not apply to all word-final obstruents on the surface.
(3)
Šmartno Slovenian opaque voicing (Reference JurgecJurgec 2019; vowel length omitted) Mobile Fixed /prax/ /snieɣ/ /ˈsmiex/ /ˈbrieɣ/ Instr /-m/ ˈpraxm̩ ˈsnieɣm̩ ˈsmiexm̩ ˈbrieɣm̩ Nom /-∅/ ˈprax ˈsniex ˈsmiex ˈbriex (final devoicing) Gen /-ˈi/ praˈxi sneˈɣi ˈsmiex ˈbrieɣ (no final devoicing) ‘dust’ ‘snow’ ‘laughter’ ‘coast’
While the traditional phonological analyses of final devoicing posit full neutralization between final voiced and voiceless obstruents, phonetic work has shown that the neutralization is incomplete in many languages (Reference ChenChen 1970, Reference Dinnsen and Charles-LuceDinnsen & Charles-Luce 1984, Reference Port and O’DellPort & O’Dell 1985, Reference Warner, Jongman, Sereno and KempsWarner et al. 2004). This means that while the contrast between voiced and voiceless obstruents is weaker than in presonorant positions, it is nevertheless still statistically significant in perception and production experiments. In this discussion, Russian has played an important role. Reference ShragerShrager (2012) and Reference KulikovKulikov (2012) demonstrate that voicing leaves traces in phonologically devoiced stops. Reference KharlamovKharlamov (2014) shows that this phonetic effect is dependent on the task: minimal pairs and orthographic presentation of stimuli are more likely to result in neutralization being incomplete. Reference Bishop, Oh, Zhou, Calhoun, Escudero, Tabain and WarrenBishop et al. (2019) extend the findings to Bulgarian, where neutralization is also incomplete.
The second alternation that is common across Slavic languages is voicing assimilation. In most Slavic languages, obstruent clusters must agree in voicing, with the rightmost obstruent determining the voicing of the entire cluster. This can be seen in the case of Russian prefixes (4). The presonorant position reveals the underlying voicing of the prefix-final obstruent, which can be either voiceless (4a) or voiced (4b). The distinction is neutralized before obstruents: all obstruents surface as voiceless before voiceless obstruents and voiced before voiced obstruents. This applies across morpheme boundaries, within morphemes, and across word boundaries – but not across pauses where final devoicing applies instead. Note that voicing assimilation is not inherently linked to final devoicing. For example, although BCS lacks final devoicing, it exhibits the same voicing assimilation patterns found in Russian.
(4)
Russian voicing assimilation (Reference Padgett, Borowsky, Kawahara, Shinya and SugaharaPadgett 2012) Before a sonorant Before a voiceless obstruent Before a voiced obstruent ˈs-jexətj ‘to ride down’ s-prɐˈsjitj ‘to ask’ ˈz-djelətj ‘to do’ ɐt-ˈjexətj ‘to ride off’ ɐt-stuˈpjitj ‘to step back’ ɐd-ˈbrosjitj ‘to throw aside’ pəd-njiˈstji ‘to bring to’ pət-pjiˈsatj ‘to sign’ pəd-ˈzeʧj ‘to burn’ iz-lɐˈɡatj ‘to state’ is-kljuˈʧjatj ‘to exclude’ iz-ˈɡnatj ‘to drive out’
Voicing assimilation in Russian is symmetrical: obstruents assimilate regardless of their underlying voicing. In contrast, Ukrainian shows asymmetrical assimilation, a fact that has played a key role in our understanding of obstruent voicing typologies (Reference ButskaButska 1998, Reference LombardiLombardi 1999, Reference Wetzels and MascaróWetzels & Mascaró 2001). As shown in (5a), Ukrainian allows word-final voiced obstruents and thus does not exhibit final devoicing. In obstruent clusters, voiceless obstruents become voiced (5b), but voiced obstruents do not devoice (5c): /liʒko/ → [liʒko], not *[liʃko].
(5)
a. No final devoicing Ukrainian voicing assimilation (Reference Padgett, Borowsky, Kawahara, Shinya and SugaharaButska 1998) ʋas ‘you.acc.pl’ ʋaz ‘vase.gen.pl’ plit ‘fence’ plid ‘fruit’ b. Regressive voicing assimilation prosɪtɪ ‘to request’ prozjba ‘a request’ borotɪ ‘to struggle’ borodjba ‘struggle’ c. No assimilation to voicelessness liʒok ‘bed.gen.pl’ liʒko ‘bed’ ridɪtɪ ‘to rarefy’ ridko ‘rarely’
Voicing of obstruents may also interact with sonorant voicing in several ways. First, in Polish (Reference Rubach and BooijRubach & Booij 1990, Reference RubachRubach 1996) and Russian (Reference JakobsonJakobson 1978, Reference Hayes, Aronoff, Oehrle, Kelley and Wilker Stephens.Hayes 1984, Reference KiparskyKiparsky 1985, Reference Petrova and SzentgyörgyiPetrova & Szentgyörgyi 2004, Reference RubachRubach 2008b), it has been reported that at word boundaries, voicing assimilates even across sonorants (e.g. [ɐd mɡlɨ] ‘from the haze’). Recent experimental studies have, however, revealed that sonorants are not transparent and that there is a contrast in the presonorant position in these cases (Reference StrycharczukStrycharczuk 2012, Reference KulikovKulikov 2013).
Second, obstruents typically devoice word-finally if the following word starts in a sonorant. Slovak is the exception (along with Lower Sorbian and south-western Polish): word-final obstruents are voiced if the following word starts in a sonorant (6a), even though morpheme-internally obstruent voicing is contrastive in pre-sonorant position (6b).
(6)
Slovak presonorant voicing at the end of the word (Reference BlahoBlaho 2008) a. Pre-sonorant voicing across word boundaries vɔjaka ‘soldier.gen.sg’ vɔjaɡ idɛ ‘the soldier goes’ lɛsɛ ‘forest.loc.sg’ lɛz jɛ ‘the forest is’ b. Contrastive voicing before sonorant word-internally mɔkra: ‘wet’ puzdrɔ ‘case’ tlak ‘pressure’ dlaɲ ‘palm’
Third, specific segments can behave exceptionally with respect to voicing assimilation. Shared among many Slavic languages is the v-sound, which behaves like a sonorant in some positions, but like an obstruent in others. This may have to do with its phonetic properties. For instance, Reference PadgettPadgett (2002) suggests that Russian has a ‘narrow approximant’ [ʋ̞] which explains its ambivalent behavior. In Slovenian, the labiovelar sonorant is realized simply as rounding on some consonants when not adjacent to a vowel: [wzeti] ‘take’ (Reference Srebot-RejecSrebot-Rejec 1981).
To illustrate the unusual behavior of some consonants, consider Czech (7). The basic facts in Czech are identical to the ones in Russian (4): voicing assimilation applies regressively. The prefix /s-/ ‘with’ surfaces faithfully before a sonorant, but voices before a voiced obstruent (7a). The prefix /z-/ ‘from’ again surfaces faithfully before a sonorant, but devoices before a voiceless obstruent (7b).
(7)
Czech regressive voicing assimilation (Reference HallHall 2003, Reference Hall, Arnaudova, Browne, Rivero and Stojanović2004) Before a sonorant Before a voiceless obstruent Before a voiced obstruent a. slesem ‘with a forest’ spolem ‘with a field’ zdomem ‘with a house’ b. zlesa ‘from a forest’ spole ‘from a field’ zdomu ‘from a house’
Turning to [v], we see that it behaves like an obstruent when followed by a consonant: it is underlyingly voiced but becomes devoiced before voiceless obstruents (8a-i). When [v] is rightmost in a cluster, it devoices after voiceless obstruents in some dialects, but in any case does not trigger voicing regressively (8a-ii); [v] is voiced when following a voiced obstruent (8a-iii). Another exceptional segment in Czech is the trill-fricative [r̞], which shows a different pattern: it devoices after voiceless obstruents (8b). Across word boundaries, the trill-fricative can trigger voicing, although there is some interspeaker variation (Reference PalkováPalková 1997).
(8)
Czech anomalous segments (Reference HallHall 2003, Reference Hall, Arnaudova, Browne, Rivero and Stojanović2004) a. i. vlese ‘in a forest’ fpole ‘in a field’ vdomɲe ‘in a house’ ii. tvor̞̥it ‘in a field’ tfor̞̥it iii. dvor̞̥it ‘to court’ b. i. pr̞̥i ‘near’ ii. br̞̥ex ‘shore’
3.5 Other Local Interactions
In the languages of the world, the vast majority of alternations affecting consonants are local, meaning that a consonant is affected by an immediately adjacent segment. This is the case for palatalization and voicing assimilation discussed so far. In this section, I review other local alternations affecting consonants. I limit these discussions to those that are not clearly affected by the sonority restrictions on syllable structure. For instance, obstruent-sonorant complex onsets are much more common that the reverse, and some Slavic languages do not allow complex onsets consisting of a sonorant followed by an obstruent (Sonority Sequencing Principle; see Reference Clements, Kingston and BeckmanClements 1990 for a comprehensive history and overview). Syllable restrictions on consonant combinations are reviewed in Chapter 4 in this volume.
I begin with the place of articulation of nasals in consonant clusters, which is restricted in many Slavic languages. Consider the example from Polish in (9), where I omit that palatal nasal /ɲ/ for simplicity. We can see that an underlying /m/ is possible before stops regardless of their place of articulation. This is not the case for /n/, which assimilates to the same place of articulation as the following stop, even across word boundaries. For instance, [pan] ‘mister’ alternates with [pam] when followed by a word starting in a labial stop. Before fricatives, the situation is more complex, with nasal glides surfacing in some positions. Words with historical nasal vowels have largely the same distribution as /n/.
(9)
Polish nasal place assimilation (Reference Czaykowska HigginsCzaykowska Higgins 1989, Reference Czaykowska Higgins1992, Reference PadgettPadgett 1994) Before stops m n Ṽ Labial bɔmba ‘bomb’ pam buk ‘Lord God’ zɔmp ‘tooth’ Coronal kɔmtur ‘commander’ blɔnd ‘blond’ z̢ɔndu ‘government’ Dorsal klamka ‘doorknob’ baŋk ‘bank’ vɛŋɡjɛl ‘coal’ Before fricatives Labial trɪuw̃f ‘triumph’ kɔɱflikt ∼ kɔw̃flikt ‘conflict’ (no data) Coronal xamski ‘boorish’ ʃansa ∼ ʃaw̃sa ‘chance’ mɔw̃ʂ ‘husband’
Nasals are also involved in another alternation in Polish and Upper Sorbian. In the latter, shown in (10), the palatal nasal [ɲ] in the prevocalic (or syllable onset) position alternates with the sequence [jn] in the position before a consonant or end of the word (in the coda). Note that the sequence [jn] can also appear in the prevocalic position [kɔmbajnɨ], while the [ɲ] can never occur before a consonant or the end of word. Assuming an underlying /ɲ/, the motivation for decomposition into [jn] is the retention of palatality and nasality in the coda where [ɲ] is illicit (Reference RubachRubach 2008a). Polish displays a similar pattern that is driven by segmental environment, not syllable structure: decomposition applies before stops and fricatives.Footnote 4
(10)
Upper Sorbian nasal decomposition (Reference RubachRubach 2008a) kamjɛɲa ‘stone.gen.sg’ kamjɛjn ‘stone.nom.sg’ kɔɲa ‘horse.gen.sg’ kojn ‘horse.nom.sg’ waɲa ‘bathtub.nom.sg’ wajnʧka ‘small bathtub.nom.sg’ kɔmbajnɨ ‘threshing machine.nom.pl’ kɔmbajn ‘threshing machine.nom.sg’
Next, several Slavic languages show dissimilation, a pattern in which one segment alternates when next to a similar segment. Consider Macedonian (11). Coronals (and [ɡ]) turn into labiodentals before the suffix -ʧɛ. The key generalization is that a non-labial becomes labial when next to a non-labial. This process is not fully general and applies only to this specific environment. In this sense dissimilation in Macedonian is similar to palatalization in BCS, applying in specific morphological environments.
(11)
Macedonian diminutive -ʧɛ dissimilation (Reference LuntLunt 1952:17, Reference Friedman, Comrie and CorbettFriedman 1993: 259−260) vɔz ‘cart’ vɔfʧɛ ‘small cart’ vɔʃka ‘louse’ vɔfʧɛ ‘small louse’ mɔst ‘bridge’ mɔfʧɛ ‘small bridge’ ʃamija ‘scarf’ ʃamifʧɛ ‘small scarf’
This concludes the short survey of the other local consonant patterns in Slavic languages. Further local consonant alternations are described in Chapter 5 of this volume.
3.6 Long-Distance Interactions
The alternations reviewed so far are local. For instance, the segments involved in voicing alternations are adjacent. At least two Slavic languages also involve long-distance interactions among consonants, which will be reviewed next.
The first long-distance interaction is consonant harmony (Reference HanssonHansson 2001, Reference Hansson2010, Reference Rose and WalkerRose & Walker 2004), which has been reported for Russian and Slovenian. Consonant harmony is a process in which consonants within a word must agree in some feature, such as voicing, minor place, or nasality. The most common type of consonant harmony is sibilant harmony. Most typically, a posterior sibilant (e.g. [ʒ, ʂ]) cannot co-occur with an anterior sibilant (e.g. [ʦ, z]). These restrictions may result in alternations but this is not necessarily so. Some are simply static restrictions on the shape of morphemes (or Morpheme Structure Constraints). Moreover, many cases of consonant harmony are strong tendencies rather than exceptionless generalizations (Reference Kochetov, van Oostendorp, Ewen, Hume and RiceArsenault & Kochetov 2011, Reference Ozburn and KochetovOzburn & Kochetov 2018).
In Russian, many words historically contained a sequence of an anterior and a posterior sibilant, but these were subsequently harmonized (e.g. [ˈs̢ɛrs̢ɨnj] ‘hornet’ from Old Russian sьɾʃenь). Reference Kochetov, Radisic, Reich, Babyonyshev and KavitskayaKochetov & Radisic (2009) investigate whether there is a phonetic basis for such a pattern. They asked Russian speakers to repeat words containing different sibilants. In total, 18 percent were mispronounced. Most of the mispronunciations were assimilatory, such as the target [sap ʃap] being pronounced as [ʃap ʃap]. Anterior sibilants more likely assimilate to posterior than vice versa, and plain sibilants more likely assimilate to palatalized than vice versa. While these experiments do not show that Russian has consonant harmony, they demonstrate a potential mechanism for the genesis of consonant harmony.
More systematic sibilant harmony is found in some varieties of Slovenian, particularly the western dialects. Even though Standard Slovenian does not display sibilant harmony, it is nevertheless encountered occasionally, as in (12). Note that the underlined sibilants alternate even though they neighbor the same segments. What causes this alternation? In the first column, we see that alveolar sibilants are not followed by another sibilant. When a postalveolar appears towards the end of the word as a result of velar palatalization – as in BCS (1) – the alveolars earlier in the word harmonize and become postalveolars (underlined). In short, an alveolar sibilant becomes postalveolar when followed by a postalveolar within the same word, even though it could be several segments away. This process is directional, as the reverse order is not restricted, as in [ʒelezniʦa].
(12)
Sibilant harmony in Slovenian slux ‘hearing’ ʃliʃim ‘hear.1p.sg’ ɾazlika ‘difference’ ɾaʒliʧiʦa ‘variant’ ʒelezniʦa ‘railroad’ ʒeleʒniʧki ‘railroad.adj’
The individual dialects differ, and the pattern is in decline (e.g. Reference SteenwijkSteenwijk 1992 for Resian). The key issue under discussion in the literature has been whether sibilant harmony can apply across consonants, and if so, whether a subset of consonants block it. Reference JurgecJurgec (2011) describes a variety in which only coronal stops block sibilant harmony, while Reference BonBon (2017) and Reference MisicMisic (2018) identified varieties with other kinds of blockers. Blocking is exceedingly rare in consonant harmony, and so these patterns have played a key role in the literature (see Reference Hansson and AronoffHansson 2020 for a comprehensive review).
A consonant co-occurrence restriction is another type of long-distance interaction that we find in Slavic. In Russian, for instance, roots cannot consist of two homorganic consonants (Reference PadgettPadgett 1992). In particular, roots with two labials, two dorsals, or two types of a subset of coronals are reported to be impossible or at least vastly underrepresented (13a), but roots consisting of combinations of these groups co-occur freely (13b).
(13)
Russian roots (Reference PadgettPadgett 1992) a. Impossible or exceedingly rare *map (two labials) *kaɡ (two dorsals) *lor (two coronal sonorants) *sjoz (two coronal fricatives) *dat (two coronal stops) b. Commonly attested ɡrjeb ‘dig’ brat ‘brother’ koz ‘goat’ tolk ‘explain’ poln ‘full’ sad ‘sit’ bod ‘awake’
Similar OCP effects are also found elsewhere in Russian. Reference Linzen, Kasyanenko and GouskovaLinzen et al. (2013) look at the variation in Russian prepositions, which can be pronounced with or without a vowel [sə mnoʐəstvəm] ∼ [s mnoʐəstvəm] ‘with a large amount’. Adjacent identical segments tend to be avoided, so the additional vowel is more common if the preposition contains the same consonant as the following word. There is a second, weaker generalization: non-adjacent identical consonants are also avoided. So, the extra vowel is more common if the second consonant of the noun is the same as the preposition (14).
(14)
Non-local OCP in Russian prepositions (Reference Linzen, Kasyanenko and GouskovaLinzen et al. 2013) More common variant Less common variant və dvɐreʦ ‘into the palace’ kə dvɐrʦu ‘to the palace’ və dvɐre ‘into the yards’ sə dvɐrom ‘from the yard’
Thus, the OCP effect is a driver of the vowel alternation in prepositions, resembling the overall tendency observed in the roots.
A similar restriction is found in Slovenian, where multiple non-adjacent posterior sibilants are dispreferred within a word – recall that the opposite is found in western dialects discussed above. The co-occurrence restriction can be seen when looking at palatalization. Much like in BCS (Section 3.3), palatalization in Slovenian is found with specific suffixes which turn velars into postalveolars (15). This process is variable, and the percentages below present the number of palatalized tokens in the corpus.
(15)
Slovenian palatalization (Reference JurgecJurgec 2016) Stem Non-pal. Palatalized %pal Tokens baɾok ‘baroque’ baɾok-ən baɾoʧ-ən 99.8 10,466 ‘adj’ stɾaŋk-a ‘party’ stɾaŋk-iʦa stɾanʧ-iʦa 88.3 206 ‘dim’ ɡɾax ‘pea’ ɡɾax-ək ɡɾaʃ-ək 55.7 341 ‘dim’ kɾok ‘circle’ kɾoɡ-əʦ kɾoʒ-əʦ 9.0 4,804 ‘dim’
When the stem contains a postalveolar, however, the palatalization rates are much lower. Often, palatalization is completely blocked (16).
(16)
Palatalization blocked by a distant postalveolar (Reference JurgecJurgec 2016) Stem Non-pal. Palatalized %pal Tokens ʒaɡ-a ‘saw’ ʒaɡ-ən ʒaʒ-ən 0.0 15 ‘adj’ ʃʧɪɾk-a ‘daughter’ ʃʧɪɾk-iʦa ʃʧɪɾʧ-iʦa 0.0 5,335 ‘dim’ ʃpɛx ‘fat’ ʃpɛx-ək ʃpɛʃ-ək 0.0 18 ‘dim’ ʧuk ‘owl’ ʧuk-əʦ ʧuʧ-əʦ 0.0 405 ‘dim’
Reference Jurgec and SchertzJurgec & Schertz (2020) demonstrate that the speakers extend these generalizations to derived and non-derived nonce words: the acceptability of nonce words with two postalveolars is lower than other combinations of sounds, including two velars. The authors also found a separate identity avoidance effect, mirroring the Russian preposition facts.
Only in recent decades have long-distance interactions been a major focus of phonological research. It is likely that this research will uncover similar patterns in other Slavic languages.
3.7 Conclusions
This chapter provides an overview of the most common sound patterns involving consonants in Slavic. We have seen that the inventories of contemporary Slavic languages are quite similar, with the chief differences being in the number and realization of coronals and in the presence of palatalized consonants. Slavic languages display a great range of palatalizations, which involve either primary or secondary place, and which can be conditioned by various phonological, morphological, and lexical factors. Voicing alternations have received attention in the phonological literature and display key typological differences. Finally, this chapter reports on several long-distance interactions between consonants, including consonant harmony and the OCP.
