Prosodic Domains and Prosodic Structure

Prosodic domains and prosodic structure

1. Paradigmatic and syntagmatic distinctions

Distinctive features encode paradigmatic distinctions between lexical items. Some contrasts between words are fairly local e.g. "sip" vs. "ship". Thus, some features (autosegments) are fairly localized:

e.g.	s	ip	vs.		ip
	\|			\|
	[+anterior]			[-anterior]

Some contrasts are not local i.e. some features are less localized, e.g. Coeur d'Alène diminutive glottalization:

mar-marím-EntEm-ilts	vs.	m'-m'ar'-m'ar'ím-En'tem'-il'ts
\___________________/		\__________________________/
not glottalized		glottalized
`they were treated one by one'		`they little ones were treated one by one'

But there are also syntagmatic distinctions between successive units. For example:

i) Stress:

À pa là chi có la

strong/weak strong/weak strong/weak Distinctions between syllables

weaker stresses stronger stress Distinctions between stressed syllables

ii) Vowel disharmony, e.g. in Turkish (Roca and Johnson p. 155):

km lejin

\_____/
[+back] \_____/
[-back]

`evening' `during'

There are three varieties of such prosodic features:
i) those associated with the edges of a domain (section 2);
ii) those associated with the head of a domain (section 3);
iii) those associated with both the head and the edges: these are consequently associated with a span (section 4).

2. Prosodic features associated with the edges of a domain.

2.1. Russian. Maximum 1 voicing distinction in consonant clusters. Doesn't respect word boundaries:

Morphemic boundary:	/gorod+k+a/	=>	[gorotka]	`little town'
Clitic boundary:	/mtsensk# bi/	=>	[mtsenzgb]	`if Mcensk'
Word boundary:	/mtsensk## bil/	=>	[mtsenzgbl]	`it was Mcensk'

Sonorants e.g. [m] are not distinctively voiced, and they are transparent to voicing assimilation:

iz # mtsensk+a	=>	[is mtsenska]	`from Mcensk'
ot # mzd+i	=>	o[d mzd]	`from the bribe'

Domain of voicing is C⁺i.e. sequences of one or more Cs, irrespective of syllable or word boundaries.

2.2. French juncture: a) anticipation of [+voice]:

b) anticipation of [-voice]:

2.3. English nasal place assimilation:

	a) Within a morpheme	b) Across two morphemes
bilabial	swamp	impress
labiodental	triumph, envy	invade, inflammable
dental	plinth	tenth, enthrall
alveolar	wince	insert
postalveolar	winch	insure
(pre)palatal	onions	unusual
velar	think	encode

2.4. "Liaison sounds" e.g. French; English linking /r/, /w/, /j/; glottal stop in word-initial position.

3. Prosodic features associated with the head of a domain.

3.1. English. Maximum of one voicing distinction in consonant clusters e.g. Syllable initial: tip vs. dip, trip vs. drip (voicing is not distinctive in /r/), strip (no contrast). Syllable final: bet vs. bed, built vs. billed (voicing is not distinctive in /l/), colts vs. cold[z].

Dissimilar values of [±voice] are OK across word-boundaries, e.g. bankbalance.

Sub-syllable domains (the head of the onset and coda are shown in bold):

		Syllable
		/	\
		/	Rime [±voice]
	/		/		\
Onset [±voice]			Nucleus		Coda [±voice]
/	\|	\	/	\	/	\|	\
(C)	C	(C)	V	(V)	(C)	C	(C)
\| (s)	\| t d :	\| (r) :	\| i e a :	\| j w r	\| (l) :	\| t d :	\| (s) (z) :

3.2. Thai (Henderson 1948)

Although consonant clusters /kr/, /kl/, /kw/, /k^hr/, /k^hl/, /k^hw/, /tr/, /pr/, /pl/, /p^hr/, /p^hl/ are found syllable-initially in Thai, only unreleased //, /k/, /t/, /p/ and nasals //, /n/, /m/ occur syllable finally. "It will be seen that plosion, aspiration, affrication, friction, [distinctive] voice, and /r/ or /l/ are properties of the syllable initial only, and mark the beginning of a syllable whenever they occur. These features may, therefore, be regarded as belonging to the prosodic system, while what is common to both syllable parts, initial and final, may be postulated as the consonant system."

4. Prosodic features associated with both head and edges (a span).

4.1. English vowel harmony:

woman
[wUm

n]
\ /
[+back]
[+round]

vs.

women
[wImIn]
\ /
[-back]
[-round]

4.2. Kirgiz vowel harmony:


	[-back]		[+back]
[-round]	bir + inti	`first'	att + nt	`sixth'
	be + inti	`fifth'	djrma + nt	`twentieth'

[+round]	yt + ynty	`third'	on + untu	`tenth'
	tørt + ynty	`fourth'	toguz + untu	`ninth'

5. Prosody as function, not form

"Almost any type of `sound' may have prosodic function, and the same `sound' may have to be noticed both as a consonant or vowel unit and as a prosody". (Firth 1948)

6. Syllable structure and prosody

6.1. Onset vs. Rime.

In quantity-sensitive stress systems, stress falls on heavy syllables. Whether a syllable is heavy or light depends on the "size" of the rime alone, not the onset. (Usually: but cf. Everett and Everett 1984.)

Latinate English words Structure of stressed rime

ab.strác.ted
ab.stáin
ab.strúse
a.brúpt
a.cró.stic
am.bi.dé[k.s]trous VC
VVC
VVC
VCC
V (cf. ácrobat)
VC

Stress tends to fall on -VV(C) or -VC(C) syllables (rimes). In words with only short vowels and no consonant clusters e.g. arídity, definítion, stress is assigned according to distance from end of word (typically 2 or 3 syllables), or by prosodically specific factors e.g. -ic tends to take stress on the preceding syllable e.g. exótic. Note (1) that in e.g. eléctric a VC rime and -ic's preferred stress placement are in agreement. In -ícity, distance from the end of the word overrides -íc's exceptional stress requirement. (2) in word-final VVC and VCC we can leave off the final C and the syllable is still heavy.

6.2. Nucleus vs. coda.

Vocalic alternations such as keep ~ kept, come ~ came affect the nucleus only, illustrating its independence from the coda. Consonantal metathesis (e.g. ask ~ `aks', wasp < O.E. wæps) provides evidence that the consonants of the coda form a constituent. Also, stop epenthesis in sonorant + fricative clusters e.g. pul^tse, min^tce =/= mint+s.

6.3. Syllable vs. appendix.

The set of word-final consonant clusters contains the set of intervocalic clusters. There may be extra coronal consonants at the end of a word leading to clusters that are not permitted word-medially e.g. breadths. (Neither dths nor even dth occurs word-medially.) See Goldsmith (1990:127).

Such extra consonants have no effect on Latinate (quantity-sensitive) stress in English, e.g. they do not make a word-final syllable heavy e.g. démagogue+s, oríginate+s, cómpliment+s.

7. Metrical structure

7.1. Bounded vs. unbounded feet.

In Latinate words of English, the main stress is always one of the last three syllables of the Latinate stress domain:

[<--stress no further back than here.

e.g. Hà ma mè li dán the mum

One way of accounting for the position of non-final but end-limited stress is to propose the existence of polysyllabic units (feet) of which the first syllable is stressed and the others unstressed:

pre(sént) 1-syllable final foot

(árid) 2-syllable final foot

(président) 3-syllable final foot

Before the final foot, feet are usually disyllabic e.g. (Hà ma)(mè li)(dán the mum).

7.2. The exceptionality of word-final ternary feet can be accounted for by regarding the final syllable as extra-metrical. Just as word-final coronal consonants may be extrasyllabic and stress-neutral, so too with such extra-metrical syllables. This makes the remaining feet maximally binary: (Hà ma)(mè li)(dán the) <mum>, cf. (À pa)(là chi)(có la).

7.3. Three notations:

a) Metrical grids

* Head of leftmost foot is more strongly stressed

* * * Stressed syllables: heads of feet

* * * * * * One star per syllable (rime projections)

A pa la chi co la

No. of stars 2 1 2 1 3 1

b) Metrical trees with s/w labelled nodes:

PrWd PrWd

/ | \ / | \

F_w F_w F_s F_w F_w F_s

/ \ / \ / \ / \ / \ / | \

_s _w _s _w _s _w _s _w _s _w _s _w _[+em]

A pa la chi co la Ha ma me li dan the mum

c) Bracketted grids

(. . *)

(* .) (* .) (* .)

* * * * * *

A pa la chi co la

References

Everett, D. and K. Everett (1984) Syllable onsets and stress placement in Pirahã. In Proceedings of the Third West Coast Conference on Formal Linguistics, pp. 105-116.

Firth, J.R. (1948) Sounds and Prosodies. Transactions of the Philological Society. Reprinted in J.R. Firth (1957) Papers in Linguistics 1934-1951. Also reprinted in F.R. Palmer, ed. (1970) Prosodic Analysis.

Henderson, E.J.A. (1948) Prosodies in Siamese: a study in synthesis. Reprinted in F.R. Palmer, ed. (1970) Prosodic Analysis.

Goldsmith, J.A. (1990) Autosegmental and Metrical Phonology. pp. 103-190.

Homework

1) Read Roca and Johnson chapters 11 and 13.
2) Do the exercise on Catalan on pp. 318-319.
3) Write down the metrical representation of the words in the Catalan data set using a) grids, b) trees and c) bracketted grids. What difficuly arises in b) and c)?
4) Read Trubetzkoy, N. S. (1939) Grundzüge der Phonologie. (Translated into English in 1969 as Principles of Phonology.) Part II: The theory of delimitative elements. (This is only about 25 pages.)

À	pa	là	chi	có	la
strong/weak		strong/weak		strong/weak		Distinctions between syllables
weaker stresses				stronger stress		Distinctions between stressed syllables

Latinate English words		Structure of stressed rime
ab.strác.ted ab.stáin ab.strúse a.brúpt a.cró.stic am.bi.dé[k.s]trous		VC VVC VVC VCC V (cf. ácrobat) VC

					[<--stress no further back than here.

e.g.	Hà	ma	mè	li	dán	the	mum

pre(sént)	1-syllable final foot
(árid)	2-syllable final foot
(président)	3-syllable final foot

					*		Head of leftmost foot is more strongly stressed
	*		*		*		Stressed syllables: heads of feet
	*	*	*	*	*	*	One star per syllable (rime projections)
	A	pa	la	chi	co	la
No. of stars	2	1	2	1	3	1