Contrast and underspecification

Plan for today

• Underspecification and phonological behaviour
• Issues with lexical contrast and underspecification
• The contrastive hierarchy

Why underspecification?

Contrastive behaviour and underspecification

• We encountered underspecification on Monday to reflect predictable aspects of lexical specification
• We also mentioned ‘linking’ and Structure Preservation as fallbacks for when the phonology tries to do something weird
• We now look at the positive case for underspecification

A simple analysis of final devoicing

The Czech consonant inventory

Manner	Labial	Coronal	Palatal	Dorsal
Stop	p b	t d	c ɟ	k ɡ
Affricate		t͡s	t͡ʃ d͡ʒ
Fricative	f v	s z	ʃ ʒ	x ɦ
Nasal	m	n	ɲ
Rhotic		r r̝
Approximant		l	j

NOM.SG	GEN.SG	Gloss	NOM.SG	GEN.SG	Gloss
xlat	xladu	‘could’	mlat	mlata	‘hammer’
ʒlap	ʒlabu	‘manger’	xlap	xlapu	‘man’
mraːs	mraːzu	‘frost’	ɦlas	ɦlasu	‘voice’
tvaːr̝̥	tvaːr̝ɛ	‘cheek’	lɦaːr̝̥	lɦaːr̝ɛ	‘liar’
kr̝ɛn	kr̝ɛnu	‘horseradish’	dɛn	dnɛ	‘day’
dar	daru	‘gift’	t͡sar	t͡sara	‘czar’

A first attempt:

[-syl] → [-voi] / _#

Does this work?

A better attempt

[-syl -son] → [-voi] / _#

Sure, but observe…

• The [-son] segments are exactly the ones that contrast in [±voi]
• As noted earlier, /r/ is the ‘non-nasal non-lateral non-fricative approximant’
• By contrast, ‘stop’ covers both [t] and [d] — we need to specify voicing to narrow it down

A third attempt

Since sonorants are predictably voiced, we are justified in positing a redundancy rule to fill in the [±voi] value. However, unlike the redundancy rules of the lexicon, it must come after the final devoicing rule.

Rule	da[r[ ]voi]	xla[d[+voi]]	xla[p[-voi]]
+voi → voi / _#		xlat
[ ]voi +son → +voi	da[r[+voi]]

More evidence for underspecification

Voicing assimilation in Czech

Prevocalic	Preconsonantal	Gloss
plateb	pla[db]a	‘payment’
hudba	hudeb	‘music’
matek	matka	‘mother’
sladit	sla[tk]ý	‘sweet’

[-syl] → [αvoi] / _[-syl αvoi] … with Structure Preservation

Not so fast…

Prevocalic	Preconsonantal	Gloss
bydel	bydlo	‘livelihood’
vyder	vydra	‘otter’
světel	světlo	‘light’
sester	sestra	‘sister’

Sonorants are voiced, but do not trigger voicing assimilation!

This agrees with our findings yesterday that presence/absence of structure corresponds to phonological activity!

A final wrinkle

Rule	/t[v[ ]voi]á[ř[ ]voi]/	/plat[b[+]voi]a/	/xla[d[+]voi]
αvoi → αvoi / _[αvoi]	not applied in /tv/!	pladba
[ ]voi → +voi / [v ř]	t[v[+voi]]á[ř[+]voi]
+voi → -voi / _#	tvář̥		xlat
[ ]voi → +voi / [+son]

Contrastive underspecification

Final devoicing again

• [-syl] → [αvoi] / _[αvoi]

Compare with

• [-son] → [αvoi] / _[-son αvoi]

All the [-son] clause is doing is singling out segment that don’t have a contrastive [±voi] specification. That seems like a hell of a coincidence.

What’s redundant anyway?

One analysis

Feature	p	b	m
voi	-	+
nas	-	-	+

What about this?

Another analysis

Feature	p	b	m
voi	-	+	+
nas		-	+

Here, /p/ is the only voiceless phoneme, so it is sufficient to specify it as [-voi]. We no longer need the [-nas]

How do we decide?

A hypothesis

Both are fine: this is a point of cross-linguistic variation

The Successive Division Algorithm

For a short(ish) description of the approach, see Dresher (2015); for the full-length treatment, Dresher (2009)

Dresher, B. Elan. 2015. The motivation for contrastive feature hierarchies in phonology. Linguistic Variation 15(1). 1–40. doi:10.1075/lv.15.1.01dre.

Dresher, B. Elan. 2009. The contrastive hierarchy in phonology. Cambridge: Cambridge University Press.

• Take an inventory and assign a + or - value for some feature to every segment in that inventory
• Within each subinventory, repeat the procedure with a different feature
• Once a subinventory consists of one segment, that segment is uniquely specified: stop and do not add any more features to it
• The order of features is not universal

[nas] » [voi]: Czech

[voi] » [nas]

Prediction

Under [voi] » [nas], both voiced obstruents and sonorants have active voicing

• Île de Groix Breton (data from Ternes 1970; analysis by Krämer 2000; Hall 2009)
  • unačypaš ‘a crew’ + baːk ‘boat’ → unačypaž baːk
  • trizek ‘thirteen’ + miːs ‘month’ → trizeɡ miːs

Ternes, Elmar. 1970. Grammaire structurale du breton de l’Île de Groix (dialecte occidental). Heidelberg: Carl Winter Universitätsverlag.

Krämer, Martin. 2000. Voicing alternations and underlying representations: The case of Breton. Lingua 110(9). 639–663.

Hall, Daniel Currie. 2009. Laryngeal neutralization in Breton: Loss of voice and loss of contrast. In Frederic Mailhot (ed.), Proceedings of the 2009 annual conference of the Canadian Linguistic Association.

Fun with the contrastive hierarchy

Cross-linguistic variation: Ifẹ Yoruba

Based on Sandstedt (2018)

Sandstedt, Jade J. 2018. Feature specifications and contrast in vowel harmony: The orthography and phonology of Old Norwegian height harmony. Edinburgh: The University of Edinburgh PhD thesis.

Variation in feature ordering → variation in phonological behaviour

Ifẹ Yoruba ATR harmony again

ATR		RTR
òɡùrò	‘spurtle’	ɔrúkɔ	‘name’
eúrò	‘bitter-leaf’	ɛ̀lùbɔ́	‘yam flour’
oríwo	‘boil, tumour’	ɔdídɛ	‘parrot’
èbúté	‘harbour’	ɛúrɛ́	‘goat’

• [+hi] vowels are transparent to ATR harmony
• The hierarchy is [hi] » [ATR]
• [+hi] vowels lack [ATR] specifications and remain inert

Cross-linguistic variation: Standard Yoruba

Ifẹ Yoruba	Standard Yoruba	Gloss
ɔrúkɔ	orúkɔ	‘name’
ɛ̀lùbɔ́	èlùbɔ́	‘yam flour’
ɔdídɛ	odídɛ	‘parrot’
ɛúrɛ́	ewúrɛ́	‘goat’

• Same inventory, but [+hi] vowels initiate a new harmonic span
• [ATR] » [hi]

What have we learned from this?

• ‘Contrast’ is defined at the level of the system
  • Not on pairwise comparison
  • Not on a priori markedness considerations

• ‘The same’ phonological unit can have different representations in different languages

• The presence of a particular phonetic property (like [+ATR] in Ifẹ Yoruba high vowels) does not guarantee associated phonological behaviour

Underspecification and variation

Persistent underspecification

• We are now considering an architecture where underspecification is not just for the lexicon, but for the phonology too
• How does this relate to phonetics?

A hypothesis

Lack of phonological specification is associated with phonetic variability

This is actually a hypothesis developed in the phonetic literature (e.g. Keating 1988a; Keating 1988b), albeit often without an explicit theory of what counts as contrastive.

Keating, Patricia. 1988a. The window model of coarticulation: Articulatory evidence. UCLA Working Papers in Phonetics 69.

Keating, Patricia. 1988b. Underspecification in phonetics. Phonology 5(2). 275–292. doi:10.1017/S095267570000230X.

Languages with no laryngeal contrast

• Hyman (2008), a candidate universal:

Hyman, Larry M. 2008. Universals in phonology. The Linguistic Review 25(1–2). 83–137.

All languages have voiceless stops

As a descriptive universal, it is falsified by languages like Yidiɲ that have a single series of stops described as [b d ɟ ɡ]

As an analytical universal, it is a statement about a theoretical object — so what are the stops of Yidiɲ?

Phonetic variation and underspecification

• Kakadelis (2018): three languages with no laryngeal contrast
  • Bardi: persistent voicing and manner variation in all stops
  • Sierra Norte de Pueblo Nahuatl: variable voicing in all stops, lenition in velars
  • Arapaho: no voicing, manner lenition of labials

Kakadelis, Stephanie M. 2018. Phonetic properties of oral stops in three languages with no voicing distinction. New York, NY: Graduate Center, City University of New York PhD thesis.

Conclusion: these languages have the same system of contrast, but different phonetics, so contrast does not matter

An alternative

Based on (Iosad)

Iosad, Pavel. Why the search for rarities must take phonology seriously. In Cormac Anderson, Shelece Easterday & Natalia Kuznetsova (eds.), Rarities in phonetics and phonology: Evolutionary, structural, typological and social dimensions. Berlin: Language Science Press.

Bardi contrastive hierarchy

Sierra Norte de Pueblo Nahuatl contrastive hierarchy

Arapaho contrastive hierarchy

On the importance of featural analysis in typology, see Lass (1984); Vaux (2009). For applications of the contrastive hierarchy in typological analysis, see Dresher, Oxford & Harvey (2018); Youssef (2021). For more examples of contrastive hierarchies and synchronic variation, see Natvig (2018); Purnell, Raimy & Salmons (2019)

Lass, Roger. 1984. Phonology: An introduction to basic concepts. Cambridge: CUP.

Vaux, Bert. 2009. The role of features in a symbolic theory of phonology. In Eric Raimy & Charles Cairns (eds.), Contemporary views on architecture and representations in phonology (Current Studies in Linguistics 48), 75–97. Cambridge, MA: MIT Press.

Dresher, B. Elan, Will Oxford & Christopher Harvey. 2018. Contrastive feature hierarchies as a new lens on typology. In Larry M. Hyman & Frans Plank (eds.), Phonological typology (Phonetics and Phonology 23), 273–311. Berlin: De Gruyter. doi:10.1515/9783110451931-008.

Youssef, Islam. 2021. Contrastive feature typologies of Arabic consonant reflexes. Languages 6(3). 141. doi:10.3390/languages6030141.

Natvig, David. 2018. Contrast, variation, and change in Norwegian vowel systems. Madison, WI: University of Wisconsin — Madison PhD thesis.

Purnell, Thomas C., Eric Raimy & Joseph C. Salmons. 2019. Old English vowels: Diachrony, privativity, and phonological representations. Language 95(4). e447–e473. doi:10.1353/lan.2019.0083.

Contrastive hierarchies and sound change

Contrast shift

• We have seen that the same inventory could be described in terms of different contrastive hierarchies, and thus different patterns of predicted phonological behaviour

A proposal

Covert reinterpretation of featural specification is a possible type of historical change

Dresher, Harvey & Oxford (2014): ‘contrast shift’

Dresher, B. Elan, Christopher Harvey & Will Oxford. 2014. Contrast shift as a type of diachronic change. North East Linguistic Society (NELS), vol. 43, 103–116.

Anglo-Frisian Brightening

• Traditional picture: PGmc /æː/ > PWGmc /aː/ > OE, OFris /æː/

• Motivation
  • PGmc /æː/ is uncontroversial
  • PWGmc /aː/ is on the basis of back reflexes in OHG, for example
  • AF /æː/ is securely attested

• Hogg (1992): the changes are driven by contrast

Hogg, Richard M. 1992. A grammar of Old English. Vol. 1. Oxford: Blackwell.

Proto-Germanic long vowels

Height	Front	Back
High	iː	uː
Mid	eː	oː
Low	æː	aː

• /æː/ is contrastively front
• WGmc /aː/ merges with /oː/

Proto-West-Germanic long vowels

Height	Front	Back
High	iː	uː
Mid	eː	oː
Low	æː

• There is only one low vowel /æː/: frontness is noncontrastive
• This may mean that it has a broader range of phonetic realizations

Anglo-Frisian long vowels

Height	Front	Back
High	iː	uː
Mid	eː	oː
Low	æː	ɑː

• OE, OFris: /ai/ > /aː/ (PGmc stainaz > OE stān)
• The ‘changes’ of /æː/ involve not rules of fronting and backing but the phonetic realization of the long vowel in a changing system of contrast

Formalizing contrast shift

Proto-West Germanic contrastive hierarchy

• This is consistent with the Proto-Germanic phonological system:
  • /a/-umlaut: lowering of /i u/ to /e o/: change in [±hi]
  • Raising of /e/ to /i/ before /i/ (and sometimes /u/): change in [±hi]

Anglo-Frisian

Anglo-Frisian contrastive hierarchy

Extension of [±bk] contrast to [+lo] branch (‘cloning’)

Implementing the shift

Another contrastive hierarchy for Anglo-Frisian

• Promotion of [±hi] so that it becomes relevant to [+bk] vowels
• Necessarily, this demotes [±lo]

Cloning and new vowels

We can now clone the [+rd] branch to accommodate [-bk +rd] vowels, which phonemicize at this stage

Now with front rounded vowels

Summary

• Modified Contrastive Specification allows to carefully formalize traditional insights into the role of contrast in diachronic change

Warning

I’m not saying anything about the mechanism of this change, or claiming that the contrast system is causing these changes!

Conclusion

Modified Contrastive Specification allows us to make explicit some key insights

• Precise scope of lexical contrast
• Link between presence of structure and phonological activity
• Link between phonological inactivity and phonetic variation
• Diachronic change

Oh and by the way…

The alleged problems for the phoneme identified by Halle can be solved with the contrastive hierarchy (Dresher & Hall 2020).

Dresher, B. Elan & Daniel Currie Hall. 2020. The road not taken: The sound pattern of russian and the history of contrast in phonology. Journal of Linguistics 57(2). 405–444. doi:10.1017/s0022226720000377.