Vowels in positions which do not immediately precede stressed syllables undergo extreme reduction. Typically, contrasts among non-high vowels are neutralised into schwa after hard consonants and into [i] after soft consonants.
It has been suggested in the literature that this type of reduction is driven by the pressure to reduce sonority in prosodically recessive positions (Crosswhite, 2001; de Lacy, 2006).14 Based on a functionally-oriented model of de Lacy (2006), I assume that reduction in atonic syllables results from an interplay of the sonority markedness scales with prosodic positions. Specifically, de Lacy (2006) suggests that prosodic heads and prosodic non-heads impose conflicting demands as to the sonority of the melodic content they contain: prosodic heads favour high sonority segments, such as the low vowel [a], whereas prosodic non-heads prefer low sonority high central vowels, such as [ɨ]. In OT, this gen-eralisation is expressed by the constraints which are formed by crossing the sonority scale (ɨ,ʉ > ә > … > a) with prosodic prominence scale (prosodic heads
> prosodic non-heads). The constraints in (10) below are formulated based on de Lacy (2006: 288).15 Note: the symbol ‘-∆ω’ stands for prosodic non-heads and ‘∆σ’ represents syllable heads. For instance, *∆σ{ɨ,ʉ} reads as follows:
13 Reduction to [i] in the context of palatalised segments is discussed in Chapter 4.
14 A similar notion of vowel reduction as a simplification of the comlex vocalic structures in pro-sodically weak positions is expressed by the Element Theory of Government Phonology (Kaye et al. 1985; Harris, 1990, 1994; Kaye, 1990; Charette, 1991). However, unlike the present model, the Element Theory views the relative markedness as ‘directly read-off from the representations rather than extrinsically encoded on the basis of observation’ (Cyran, 2010: 5).
15 De Lacy (2006: 288) formulates these constraints as context generalisations, which stand in a stringency relation.
‘Incur a violation for every head of a syllable that contains a segment with sonority which equals {ɨ, ʉ}’.
(10) a. *-∆ω{a} >> *-∆ω{ɛ,ɔ} >> *-∆ω{e,o} >>*-∆ω{i,u} >> *-∆ω{ә} >> *-∆ω{ɨ,ʉ}
b. *∆σ{ɨ,ʉ} >> *∆σ{ә} >> *∆σ{i,u} >> *∆σ{e,o} >> *∆σ{ε,ɔ} >> *∆σ{a}
Vowels in unstressed syllables are heads of the syllabic domain but non-heads of prosodic words. As only low sonority vowels [ɪ], [ᵻ], [ʊ] and [ә] are allowed in atonic positions in Russian, extreme reduction can be assumed to be driven by the constraints in (10a) above exerting the pressure to reduce sonority in prosodic non-heads.
The decrease in the sonority of atonic vowels takes place at the expense of violating identity constraints, such as ideNt-V[-high] and ideNt-V[+back], which are outranked by the markedness constraint *-∆ω{i,u}. According to the sonority hierarchy, there are two vowels, [ɨ] and [ә], which are less sonorous than [i,u]. Both vowels occur in pretonic positions in East Slavic dialects. The high vowel [ɨ] is found in a number of systems with dissimilative [a]-reduction, e.g. in Žizdra (Vajtovič, 1968).16 Given that there is a pressure to reduce sonority in prosodic non-heads, less sonorous [ɨ] makes a better reduced vowel than [ә].
However, [ә] is also found in pretonic positions, in fact, it is more widespread than [ɨ].17 It is often assumed in the literature that schwa is a defective vowel which is devoid of a melodic content (Anderson, 1982; Oostendorp, 1995;
Cyran, 2010; among others). In this view, schwa is represented as an empty root node, defined only for the major class feature [-cons]. Consequently, reduction of a full vowel to [ә] does not lead to the violation of ideNt constraints because there is no value of a given feature in the output which can be compared to the input. In dialects which reduce atonic vowels to [ɨ], schwa is disallowed by the constraint militating against empty root nodes, here informally referred to as *ә (‘A vocalic root dominates a vocalic place node’, see van Oostendorp, 1995:
138). As raising to [ɨ] entails a violation of ideNt[-high], the ranking of *ә vs.
ideNt[-high] determines which vowel, [ɨ] or [ә] will surface as optimal in a given dialect. This is illustrated in Tableau (11) by the hypothetical evaluation of the vowel //o//. Note that the candidates (11a) and (11b) fare equally well on
16 In dissimilative [a]-reduction, the quality of the reduced vowel in the pretonic position depends on the quality of the stressed vowel. Different patterns of dissimilative reduction are discussed in detail in Chapter 4.
17 Vajtovič (1968: 115) reports that very often a vowel of an intermediate quality between [ɨ] and [ә] is heard in Belarusian dialects.
the constraint *-∆ω{i,u}. The ranking *ideNt[-high] >> *ә chooses schwa in candidate (11a), while the reverse ranking, *ә >> ideNt[-high], gives preference to [ɨ].
(11) Extreme reduction
//o// *-∆ω{i,u} ideNt[-high] *ә
a. ә *
b. ɨ *
c. o *!
Tableau (12) below shows how the interaction of constraints generates reduc-tion to [ә] in the Standard Russian word volos [ˈvolәs] ‘hair’, nom. sg. (cf. the nom. pl. dim. form volosiki [vaˈlosjɪkjɪ]). Note that the candidates (12a) and (12b) fare equally well on the constraint *-∆ω{i,u}. The tie is resolved by the constraint ideNt[-high] because the reduction to a featureless schwa comes at no cost, while reduction to any other vowel violates faithfulness.
(12) Extreme reduction: Standard Russian
//ˈvolos// *-∆ω{i,u} ideNt[-high] *ә
⇒ a. ˈvolәs *
b. ˈvolɨs *
c. ˈvolos *!
d. ˈvolas *!
Interestingly, high vowels are resistant to reduction and only non-high vowels are neutralised to schwa in dialects with dissimilative reduction. Pretonic //o// and //a// are reduced to [ә] after hard consonants whereas underlying //ɨ//
surfaces as [ɨ]. The nonreduction of high vowels follows from the constraint set established above, as shown in Tableau (13) below. ideNt[-high] is mute with respect to an input containing a high vowel and, consequently, candidate (13b) with a non-reduced [ɨ] harmonically binds a candidate with a schwa (13a).
(13) Non-reduction of high vowels
//ɨ// *-∆ω{i,u} ideNt[-high] *ә
a. ә *
⇒ b. ɨ
c. o *!
To summarise, extreme reduction takes effect because prosodically recessive positions are required to contain low-sonority vowels. As it leads to the emer-gence of the least marked vowel [ә], Russian atonic reduction can be viewed as a TETU (The Emergence of The Unmarked) effect (McCarthy & Prince, 1994).
In the next section, we turn to the reduction pattern attested in positions imme-diately preceding tonic syllables.
4 . Tone-driven reduction