Palatalization and consonant-vowel interactions

Basic facts and terminology

Basic facts: hard and soft consonants

The distinction between ‘hard’ and ‘soft’ consonants is traditional in both synchronic and diachronic phonology. Very broadly, ‘soft’ consonants are those that are produced by a process that involves front vowels or glides.

Two key processes are secondary palatalization, where a consonant becomes ‘soft’ before a front vowel (that may have disappeared later), and coalescence with a following *j. In principle, any consonant can have a palatalized counterpart, but synchronically we find that we can roughly divide the Slavic languages into two groups:

Northern (all of East Slavic, Polish, Sorbian) and (Eastern) Bulgarian: basically all consonants¹ have soft counterparts, usually realized as palatalized versions of the hard consonant.
- In some languages (Polish, Belarusian, parts of Ukrainian), some soft coronals (especially sonorants) are palatal rather than palatalized anterior
- A common secondary development is enhancement of the soft quality. A typical example is Polish and Belarusian: t’ d’ s’ z’ > [t͡ɕ d͡ʑ ɕ ʑ] and similar
South Slavic other than (Eastern) Bulgarian, Czech and Slovak: no secondary palatalization contrast, but a distinctive series of palatals: /c ɟ ɲ ʎ/ (and the Czech ř) as the ‘soft’ versions of /t d n l r/

¹ With some complications around velars

Hard	Soft	Source	Example
t d n l	c ɟ ɲ ʎ	*Cj coalescence	Svk žena ‘woman’ ≠ baňa ‘mine’
		Secondary palatalization	Cz prst ‘finger’ ≠ prsť ‘soil’
			Cz když ‘where’ ≠ divný ‘strange’
p b t d s z	pʲ bʲ tʲ dʲ sʲ zʲ	Secondary palatalization	Uk pit ‘sweat’ ≠ myt’ ‘moment’
n l	nʲ/ɲ lʲ/ʎ	Secondary palatalization	Uk den’ ‘day’ ≠ son ‘dream’
		*Cj coalescence	Uk kin’ ‘horse’

Basic facts: velar palatalizations

A special place belongs to posterior coronals č ž š, and in some languages the affricates c dz. Diachronically, they derive from palatalization processes.

Process	Change	Example	Gloss
First velar palatalization	k g x > č ž š / _i e ě ь	BCMS muka ~ mučiti	‘torment’ N ~ INF
Cj coalescence	sj zj > š ž	Po zwisać ~ wiszę	‘hang’ INF ~ PRS.1SG
	tj dj > various outcomes	Ru xod ~ xožu	‘walk’ N ~ PRS.1SG
		P chód ~ chodzę	‘walk’ N ~ PRS.1SG
		BCMS rod ~ rođen	‘kin, birth’ ~ ‘born’
		Bu rod ~ rozhden	‘kin, birth’ ~ ‘born’
Second velar palatalization	k g x > c dz š/s	BCMS ruka ~ ruci	‘hand’ NOM ~ DAT

Since these segments usually derive from ‘C + front vowel’ or *Cj sequences, they share many behaviours with soft consonants. However, with few exceptions² they are phonetically ‘hard’ (not palatalized or palatal), and usually do not have soft counterparts. They also show some ‘hard’ phonological patterning, as we shall see.

² Notably c’ <ц> in Ukrainian, [с ɟ] <ќ ѓ> in Macedonian, and ć đ in at least parts of BCMS.

Key alternations and examples

Surface palatalization: C → Cʲ, usually before a front vowel
(First) velar palatalization: k g/ɣ/ɦ x → č ž š, usually before a front vowel or j
Transitive palatalization: t d s z → T D š ž, before a historical j but often without a clear context synchronically
Labial iotation: ∅ → lʲ/ʎ after labials, before a historical j but often without a clear context synchronically.
(Second) velar palatalization: k g/ɣ/ɦ x → c dz/z s/š in a very restricted number of contexts

Warning

Transitive palatalization and labial iotation often occur in the same contexts as each other. Sometimes — but not always — you also see the first velar palatalization in those contexts, too, but 1VP can also co-occur with Surface Palatalization.

Alternation	Example	Gloss
Surface palatalization	P kosa [s] ~ kosić [ɕ]	‘scythe’ ~ ‘mow.INF’
Surface palatalization	Ru korm ~ kormit’ [mʲ]	‘feed’ N ~ INF
First velar palatalization	Ru muka ~ mučit’	‘torment’ N ~ INF
First velar palatalization	BCMS jak ~ jači	‘strong’ POS ~ CMP
Transitive palatalization	P kosa [s] ~ koszę [ʂ]	‘scythe’ ~ ‘mow.PRS.1SG’
	BCMS ljut ~ ljući [tɕ]	‘angry’ POS ~ CMP
	P chód [d] ~ chodzę [d͡z]	‘walk’ N ~ PRS.1SG
Labial iotation	Ru korm ~ korml’u	‘feed’ N ~ PRS.1SG
Labial iotation	BCMS glup ~ gluplji	‘stupid’ POS ~ CMP
Second velar palatalization	Bu vъlk ~ vъlci	‘wolf’ SG ~ PL
Second velar palatalization	P rąka ~ ręce	‘hand’ NOM ~ DAT

The problem of /y/ and velars

Many, but not all, present-day languages distiguish between [i] and a second non-back, non-round, high (or near-high) vowel, traditionally symbolized y.³

³ Not IPA [y]!

Sequence	Russian	Polish	Ukrainian
ti di	✗	marginal	✓
tʲi dʲi	✓	✓ (basically)	✓
tɨ dɨ	✓	✓	✓
ki ɡi	✗	✗	✗
kʲ ɡʲi	✓	✓	✓
kɨ ɡɨ	restricted	restricted	✓
tɛ dɛ	marginal	✓	✓
tʲɛ dʲɛ	✓	✓	restricted
kɛ ɡɛ	restricted	restricted	✓
kʲɛ ɡʲɛ	✓	✓	✗

The analysis of /i/ and /y/

Because of the largely complementary distribution of [i] and [ɨ], they were generally considered to be allophones of /i/ in structuralist phonology
This is central to Jakobson (1929), is acknowledged by Trubetzkoy (1934), and remains the case in Halle (1959)
To state the distribution, we have to assume that /C/ and /Cʲ/ are phonemically distinct.

Trubetzkoy, Nikolai S. 1934. Das morphonologische System der russischen Sprache (Travaux Du Cercle Linguistique de Prague 5.2). Prague: Jednota československých matematiků a fyziků.

Halle, Morris. 1959. The sound pattern of Russian: A linguistic and acoustical investigation. ’s Gravenhage: Mouton.

Analysis of Russian [pɨl] ‘ardour’, [pʲil] ‘drink.PST.SG.M’, [pɨlʲ] ‘dust’
Consonant	Word-final		Before /i/
Hard	/pil/ [pɨl]
Soft	/pilʲ/	[pɨlʲ]	/pʲil/	[pʲiɫ]

Consonant palatalization in generative phonology

What needs an account?

Front vowels generally follow Cʲ
C and Cʲ can contrast when there is no following front vowel
There hard → soft alternations before front vowels
After surface hard consonants, we only find [ɨ]

The generative approach

The foundational hypothesis of Slavic generative phonology

A large proportion of Slavic morphophonological patterns can be explained if soft consonants always derive from hard consonants followed by a front vowel

As we discussed yesterday, all kinds of alternations are produced by the single mechanism of phonological rule. On this basis, generative phonology can use morphophonological alternations to posit rules and ‘unspool’ them to recover abstract underlying representations.

Two kinds of /i/

Some derivations of **mǫk-* ‘torment’ and **xod-* ‘walk’ in Russian and Polish
Language	Item	Infinitive suffix	Nominative plural
Russian	xod	xodit’ [dʲi]	xody [dɨ]
Russian	muka	mučit’	muki [kʲi]
Polish	chód	chodzić [d͡ʑi]	chody [dɨ]
Polish	męka	męczyć	męki [kʲi]

Synchronic derivation of the Russian forms
Rule	/xod-itʲ/	/muk-itʲ/	/muk-ɨ/
Velar palatalization		mučitʲ
Post-velar fronting			muki
Surface palatalization	xodʲitʲ		mukʲi

What is going on?

On the face of it, the č in mučit’ cannot derive from a rule turning k into č before [i], because forms like muki show that [kʲi] sequences are allowed. However, such surface sequences occur precisely in contexts where the vowel turns up as [ɨ] after hard consonants when those consonants are not velars. Conversely, the rule that looks like 1VP is triggered by vowels that trigger Surface Palatalization of preceding consonants.

Because of the different morphophonological behaviours, the two suffixes are analysed as containing two different underlying vowels:

The infinitive has a real front vowel, which triggers assimilation of preceding consonants: Surface Palatalization for non-velars, 1VP for velars
The nominative plural has a back vowel, which unsurprisingly does not trigger softening — unless a different rule (Post-Velar Fronting) makes it front. If it does end up front, it is able to trigger palatalization like any other front vowel

Two key takeaways

This is all phonology, and it works with phonological segments, which have real (phonetically non-trivial) feature specifications and all the rest. The vowels behave this way because they really are front or back, not because of an abstract ‘morphophonemic’ structure
The fact that this analysis closely tracks diachronic developments is neither surprising nor problematic: this is the system working exactly as intended.

A fun fact: the idea of Post-Velar Fronting has a deep Jakobsonian pedigree. The idea that *ky gy xy developed to *ki gi xi and then the front vowel palatalized the preceding velars is a centrepiece of Jakobson (1929). His explanation remained foundational in structuralist historical accounts, but this history seems to have been largely forgotten in generative analyses; one exception is Padgett (2003).

Jakobson, Roman. 1929. Remarques sur l’évolution phonologique du russe comparée à celle des autres langues slaves (Travaux du Cercle linguistique de Prague 2). Prague: Jednota československých matematiků a fyziků.

Padgett, Jaye. 2003. Contrast and post-velar fronting in Russian. Natural Language & Linguistic Theory 21. 39–87.

Getting the complementary distribution

The analysis so far explains the fact that we have [Cʲi] and [Cɨ] but not *[Ci]

Underlying /Cɨ/ remains
Underlying /Ci/ → [Cʲi] by Surface Palatalization

There is still a problem with the Polish.

Predicted derivation for Polish cognates
Rule	/xɔd-itʲ/	/mɛNk-itʲ/	/xɔd-ɨ/	/mɛNk-ɨ/
First Velar Palatalization		mɛNčitʲ
Post-Velar Fronting				mɛNki
Surface Palatalization	xɔdʲitʲ	mɛNčʲitʲ		mɛNkʲi
Minor rules	xɔd͡ʑit͡ɕ	mɛnčʲit͡ɕ		mɛŋkʲi
Predicted surface form	xɔd͡ʑit͡ɕ	mɛnčʲit͡ɕ	xɔdɨ	mɛŋkʲi
Actual surface form		mɛnt͡ʂɨt͡ɕ

This is what I meant when I said that the velar palatalization products behave as soft in the phonology but end up being surface-hard. The cz dż sz ż series in Polish sound hard and also condition a following [ɨ], and we don’t have an account of that yet.

The backness switch

Better derivation for the Polish forms
Rule	/xɔd-itʲ/	/mɛNk-itʲ/	/xɔd-ɨ/	/mɛNk-ɨ/
First Velar Palatalization		mɛNčʲitʲ
Post-Velar Fronting				mɛNki
Surface Palatalization	xɔdʲitʲ			mɛNkʲi
Postalveolar hardening		mɛnčitʲ
Retraction		mɛnčɨtʲ
Minor rules	xɔd͡ʑit͡ɕ	mɛnt͡ʂɨt͡ɕ	xɔdɨ	mɛŋkʲi

The product of 1VP is soft. When Surface Palatalization stops being relevant,⁴ a Hardening rule applies and feeds retraction of [i] to [ɨ]

⁴ The eagle-eyed will notice that Surface Palatalization fails before [ɛ]. Ask me about it if you’re still wondering by the end!

We now correctly derive surface [Cɨ], whether it from underlying /Cɨ/ or from another process where the consonant ends up hard for other reasons, but our derivations are getting really rather long.

What’s the backness switch?

Note that underlying /kɨ/, with a back vowel (and a hard consonant), surface with a front vowel and a soft consonant, and underlying /ki/, with a front vowel, surfaces as [t͡ʂɨ], with a hard consonant and a back vowel. The term is due to Rubach (2000)

Rubach, Jerzy. 2000. Backness switch in Russian. Phonology 17(1). 39–64. https://www.jstor.org/stable/4420162.

Further developments

Taking the system further

The basic recipe for analysing palatalization:

Soft consonants come from following front vowels (or glides)
If a vowel has a softening effect, it must be underlyingly front
If a vowel does not have a softening effect, it must be underlyingly back
Rule ordering will keep us straight

Regressive palatalization revisited

Languages like Czech show coronal → palatal alternations before some, but not all, front vowels

Softening and non-softening suffixes in Czech
Context	Non-palatalizing	Palatalizing
Morpheme-internal	když [di] ‘where’	divný [ɟi] ‘strange’
Adjective inflection	pěkný [ni] ‘beautiful.NOM.SG.M’	pěkní [ɲi] ‘beautiful.NOM.PL’
Nominal inflection	hradem [dɛ] ‘city.INS.SG’	hradě [ɟɛ] ‘city.LOC.SG’

This can be accounted for by positing back underlying vowels with subsequent fronting

Analysis of Czech i vs y
Rule	/pjɛkn-ɨː/	/pjɛkn-iː/
Palatalization		pjɛkɲiː
Vowel fronting	pjɛkniː

Note

In general colloquial Czech, y is realized [ɛj]. Does this matter?

Tackling unexpected softness

Some Russian verbs
Infinitive	PRS.1SG	Imperfective	Gloss
lʲez-tʲ	lʲez-u	-lʲezatʲ	‘clamber’
ɡrɨz-tʲ	ɡrɨz-u	-ɡrɨzatʲ	‘gnaw’
žečʲ	žɡ-u	-žɨɡatʲ	‘burn’
ža-tʲ	žm-u	-žɨmatʲ	‘press’
ža-tʲ	žn-u	-žɨnatʲ	‘reap’
mʲa-tʲ	mn-u	-mʲinatʲ	‘knead’
ras-pʲa-tʲ	ras-pn-u	ras-pʲinatʲ	‘crucify’

What’s the deal with soft consonants before [a]? Three observations:

Infinitive stems have the shape CVC or CV
Imperfective stems always have the shape CVC-a
Present stems are either CVC or CC
- Cʲa occurs in items that CV- in the infinitive and CN- in the present

Analysis of surface [Cʲa]
Rule	mIn-tʲ	mIn-u	mina-tʲ
Surface Palatalization	mʲIntʲ	mʲInu	mʲinatʲ
Nasal vowel formation	mʲĩtʲ
Vowel deletion		mʲnu
ĩ → a	mʲatʲ
Softness assimilation		mnu

We posit that the root is CVC, with a final nasal
The vowel is plausibly front
- It triggers Surface Palatalization
- It alternates with a real front vowel in the imperfective
The back vowel in the infinitive only arises after Surface Palatalization

Armed with this idea, we can tackle other cases of unexpected softness, even when there are no alternations
- [mʲaso] ‘meat’ ← /minso/
- [lʲubʲitʲ] ‘love.INF’ ← /leubitʲ/

Tackling unexpected hardness

In Polish, some /e/-initial suffixes trigger Surface Palatalization of non-velars; they usually trigger 1VP or 2VP for the velars

Suffix	Non-velars	Velars
LOC.SG	pas-ie [ɕ] ‘belt’	rzec-e [t͡s] ‘river’
V stem	łys-ieć [ɕ] ‘go bald’	droż-eć [ʐ] ‘become dearer’

When they do not, they do trigger Surface Palatalization of velars

Suffix	Labials	Coronals	Velars
INS.SG	tłum-em ‘crowd’	pas-em ‘belt’	krok-iem [kʲ] ‘step’
ADJ.DAT.SG.M	grub-emu ‘fat’	bos-emu ‘barefoot’	wielk-iemu [kʲ] ‘big’
ADJ.GEN.SG.M	grub-ego	bos-ego	wielk-iego

In effect, Polish seems to provide ample evidence that just like there is a distinction between ‘softening’ and ‘non-softening’ versions of the high unrounded vowel (traditional i vs. y), there is an entirely parallel ‘softening’ and ‘non-softening’ version of the mid unrounded vowel. This leads us to postulate an underlyingly back /ɤ/ for the latter case (Rubach 1984).

What about diachrony?

This kind of move is not so well supported by diachrony. Some of the non-palatalizing e’s are indeed historically back vowels that merged with *e — more on Thursday. But most of the time, this case in Polish comes from contraction where the deleted vowel was back: *grubu-jemu > grubēmu. What does this tell us about the diachronic argument in favour of the general approach?

How far can we go?

Our examples so far have been from two kinds of languages:

Russian or Polish, where [i] and [ɨ] both exist on the surface, and the analysis consists in elaborate reshufflings of the distribution
Czech or BCMS, where the hard-soft contrast exists as coronal vs. palatal and there is no surface [ɨ]. If we try, we can look for alternative analyses, for example by coalescence with [j]

Possible analysis without underlying front-back contrasts
	Morpheme	Labials	Coronals	Velars	UR
Czech	INS.SG	dub-em ‘oak’	hrad-em ‘city’	rok-em ‘year’	/-ɛm/
Czech	LOC.SG	dub-ě [bj]	hrad-ě [ɟ]	roc-e [t͡s]	/-jɛ/
BCMS	ACC.SG	rib-u ‘fish’	crt-u ‘line’	drag-u ‘bay’	/-u/
BCMS	INS.SG	krvlj-u [vʎ] ‘blood’	smrću [t͡ɕ] ‘death’		/-ju/
BCMS	NOM.SG.M	grub-i ‘rough’	tvrd-i ‘hard’	jak-i ‘strong’	/-i/
BCMS	CMP	grublj-i [bʎ]	tvrđ-i [d͡ʑ]	jač-i [t̪͡ʃ̪]	/-ji/

Bulgarian plurals

Two types of [i] in Bulgarian
Gender	Number	Labials	Coronals	Velars	UR
Feminine	SG	riba ‘fish’	rana ‘wound’	dъga ‘arc’	/-a/
Feminine	PL	ribi [b(ʲ?)]	rani [n(ʲ?)]	dъgi [gʲ]	/-ɨ/
Masculine	SG	zъb ‘tooth’	elen ‘deer’	vъlk ‘wolf’	/-∅/
Masculine	PL	zъbi [b(ʲ?)]	eleni [n(ʲ?)]	vъlci [t͡s]	/-i/

Bulgarian demonstrates at least two types of [i] suffix.⁵ Their behaviour is typical in the sense that one does not trigger major place changes of velars, and the other one does (2VP in this case, but there also 1VP triggers). What happens to the other consonants, though?

⁵ In fact, more than two

Recall that the hard/soft distinction in Bulgarian is neutralized before [i ɛ]: velars are soft in this position (a neutralizing alternation) and the non-velars are ‘semi-soft’, traditionally interpreted as phonemically hard.

Scatton (1975) explicitly appeals to the parallel with Russian in treating the ‘less palatalizing’ [i] as underlying /ɨ/ (with something like Post-Velar Fronting and Surface Palatalization of velars to get the pattern) and the ‘more palatalizing’ [i] as /i/ (with Velar Palatalization rules but nothing happening for other consonants). Bulgarian does not have surface [ɨ] at all!

Scatton, Ernest A. 1975. Bulgarian phonology. Columbus, OH: Slavica Publishers.

Kiparsky, Paul. 1968. How abstract is phonology? Bloomington: Indiana University Linguistics Club.

Gussmann, Edmund. 1980. Studies in abstract phonology (Linguistic Inquiry Monograph 4). Cambridge, MA: MIT Press.

Rubach, Jerzy. 1984. Cyclic and lexical phonology: The structure of Polish (Studies in Generative Grammar 17). Berlin; New York: Mouton de Gruyter.

Rubach, Jerzy. 1993. The lexical phonology of Slovak. Oxford: Clarendon Press.

Further, /ɨ/ and /i/ in this analysis have the exact same effects on preceding consonants. This means we likely have not ‘Post-Velar Fronting’ of /ɨ/ but actually absolute neutralization where /ɨ/ → [i] in all cases. In post-SPE generative phonology, this eventually became extremely problematic (Kiparsky 1968), but these kinds of analysis remained current in generative approaches to Slavic (Gussmann 1980; Rubach 1984; Rubach 1993).

Summary

The standard analysis of Slavic in early generative phonology assumes highly abstract representations and complicated derivations that largely reproduce diachrony
The aim is to capture as many generalizations as possible within the phonological component
The distinction between soft and hard consonants can be derived from the backness of following vowels in underlying representations
This is seen as economical: storage is expensive, computation is cheap
Conversely, differences in morphemes’ behaviour with regard to palatalization are encoded in the featural make-up of their segments