universität tübingen seminar für sprachwissenschaftcebert/teaching/10phonphon/handout08.pdf ·...
TRANSCRIPT
1
Autosegmental Phonology
Phonetics k Phonology
(Hall, Chapter 6)
Christian Ebert
Universität Tübingen
Seminar für Sprachwissenschaft
2
Linear Phonology
generative phonology as introduced by (Chomsky & Halle, 1968) is linear.
representations are linear sequences of feature matrices
+cons-son-voice-cont-nasCOR+ant
-consDORS+back-high+low-tense+long
+cons-son+voice-cont-nasDORS
/t/ /a:/ /g/
3
Linear Phonology
features are properties of single segments
segments are ordered linearly (as in a chain)
within a segment, features are unordered and 'simultaneous'
problem:not all features seem to fit into this picture
examples are tone (≈ distinctive use of pitch), quantity (≈ duration of phonological units) and intonation (≈ pitch variation in speaking).
since these features are independent of single segments and may be related to multiple segments they are called suprasegmental
4
Tone
in many languages of the world pitch is decisive for determination of the meaning of a word
these tonal languages comprise african languages, american indian languages and east-asian languages
tone means that phenomenon that distinguishes the meaning of lexical elements by their pitch
tone must be kept apart from intonation, which means the variation in pitch during speaking („Sprachmelodie“):
Peter schläft Peter schläft
5
Tone
tonal language show up to five different tone levels (noted as 1...5)
there are two types of tones:
register tonesremain on one tone level; may be described by a sequence of two identical numbers (11, 22, 33, 44, 55)
contour toneschange in level during articulation; may be described by a sequence of two or more non-identical numbers (eg. 51, 214, etc.)
the actual pitches that correspond to different tone levels are idiosyncratic to a language and cannot be compared across languages
usually vowels carry a tone, rarely sonorants, too
6
Tone
the IPA provides different diacritical symbols to describe tones
another option: a vertical line shows the range of levels from 1 to 5; a horizontal line indicates the actual level of the tone
register tones in the IPA:eå ó 55 extra-higheæ ô 44 higheç õ 33 middleeè ö 22 loweé ÷ 11 extra-low
example from the west-african language Nupe:
[beæ] „to come“ [beç] „to add to“ [beè] „to resemble sth.“
IPA
7
Tone
contour tones are noted as follows:
e< ø risingeì ù fallingeí ú high risingeî û low risingeñ ü rising fallingeò ý falling rising
with the number and vertical bar notation, many more contour tones can be noted
IPA
8
ToneIPA
example from Mandarin:[ó ma] 55 „mother“ [ú ma] 35 „hemp“[ý ma] 214 „horse“ [ù ma] 51 „to rant“
example from Hmong (spoken in Laos, Vietnam):[ó po] 55 „ball-like“ [po] 53 „female“[õ po] 33 „pancreas“ [po] 24 „to throw“[ö po] 22 „thorn“ [poÐ] 42 „father's father“
[po?] 31 „to see“examples from Cantonese:
[ó si] 55 „poem“ [si] 21 „time“[õ si] 33 „to try“ [si] 24 „to cause“[ö si] 22 „thing“ [si] 23 „city“
9
Toneas a distinctive feature
in the system of generative phonology tone could in principle be expressed by features, eg.
[+H] high tone [-H] low tone
[+R] rising tone, ...
-cons+son+cont-nasDORS+back-high+low+H
/aæ/-cons+son+cont-nasDORS+back-high+low-H
/aè/-cons+son+cont-nasDORS+back-high+low+R
/a</
10
Tone
in the chadic language Margi (spoken in Nigeria) on observes the following variation wrt. the suffix /aær/ (indicating definiteness):
(a) after stem-final consonant:[saæl] [saælaær] „man“[kuèm] [kuèmaær] „meat“
(b) after stem-final vowel with high tone:[?æmæ] [?æmjaærè] „water“[kuæ] [kwaærè] „goat“[taæguæ] [taægwaærè] „horse“
(c) after stem-final vowel with low tone:[tè] [tja<rè] „morning“[huè] [hwa<rè] „grave“[uæ?uè] [uæ?wa<rè] „fire“
as a distinctive feature
11
Tone
observations and rules:
the high tone of the suffix-initial vowel changes to a rising tone after stem-final vowel with low tone:
R1: [+H] > [+R] / [-cons, -H]+ __
stem-final [i] and [u] change into corresponding semi-vowels [j] und [w] (before suffixes with initial vowel), see (b) and (c):
R2: /i u/ > [j w] / __ +[-cons]
as a distinctive feature
12
Tone
This yields correct derivations:
/saæl+aærè/ /kuæ+aærè/ /huè+aærè/ UR
--- --- â rule R1
saæl+aærè kuæ+aærè huè+a<rè
--- â â rule R2
[saælaærè] [kwaærè] [hwa<rè] SR
as a distinctive feature
13
Autosegmental Phonology
these rules correctly derive the observed data; but they do not explain, why the high tone of the suffix-initial vowel changes to a rising tone after stem-final low tone vowel
autosegmental phonology (Goldsmith, 1976) gives such an explanation
basic Idea: features are arranged in independent tiers (dt. „Schichten“)
segmental features occur on the segmental tier, tone features on an (independent) tone tier
on each tier, the segments are arranged linearly, but the segments on different tiers are connected only by association linies
in this sense, these representations are non-linear/non-concatenative
Basics
14
we start with one binary feature [±H] on the tone layer and we write H for [+H] and L for [-H]
-cons+son+cont-nasDORS+back+low
/aæ/+cons+son+cont-nasCOR+ant+apic
/r/-cons+son+cont-nasDORS-back+high-low
/ è/
H L
segmental tier
association lines
tone tier
Autosegmental PhonologyBasics
15
a segment on the segmental tier that is associated with a tone is called tone bearing unit (TBU)
what segments can function as TBUs depends on the language;eg. in Margi only vowels can be TBUs
contour tones are represented by association of one TBU with more than one tone:
/aì/
[a]
H L
/a</
[a]
L H
segmental tier
association lines
tone tier
Autosegmental PhonologyBasics
16
Autosegmental Phonology
in autosegmental phonology the Margi data can be explained:
Reminder: /uæ?uè+aærè/ > [uæ?wa<rè]
Example 1: Margi
/u ? u + a r i/
H L H Lthe 'glide' rule R2 for changing /i u/ into [j w] works only on the
segmental tier:
/i u/ > [j w] / __ +[-cons]
the underlying autosegmental representation is as follows:
17
application of the 'glide' rule R2 yields the following result
remember: only vowels (and not semi-vowels) can be TBUs in Margi; hence the low tone feature looses its anchor and becomes a floating tone
At this point, the floating tone is associated with the next possible TBU:
/u ? w + a r i/
H L H L
/u ? w + a r i/
H L H L
Autosegmental PhonologyExample 1: Margi
18
in addition to the 'glide' rule R2 we only need another rule that specifies what happens with floating tones:
[-cons]
L Hrule R1 states that a floating low tone is to be associated with the
following high tone vowel
this illustrates tone stability:the tone remains, although its TBU is gone; it's reassociated with another TBU
Autosegmental PhonologyExample 1: Margi
R1:
19
Autosegmental Phonology
Another example for the analysis of contour tones as sequence of tone features: tone assimilation in Margi
the underlying suffix for the infinitive /+na/ does not carry any tone; in combination with the stem it receives the last tone of the stem:
/saæ+na/ > [saænaæ] „to loose“/ndaèl+na/ > [ndaèlnaè] „to throw“/n@<+na/ > [n@<naæ] „to plan“
this assimilation can be explained, if contour tones are represented as sequences of two tone features and if it is assumed that the last feature spreads
Example 2: tone assimilation
20
/sa+na/
H
> /sa+na/
H
/ndal+na/
L
> /ndal+na/
L
/n@+na/
HL
> /n@+na/
HL
assimilation rule:
[-cons][-cons]
[αH]
here, two elements on the segmental tier are associated with one feature on the tone tier
again we see: tone features are not segmental
Autosegmental PhonologyExample 2: tone assimilation
21
in many languages the number of felicitous tone patterns is limited
eg. in the Bantu language Etung we find four tones: H, L, HL, LH
The distribution of these tones in Etung is limited:
one syllable: kpaæ, kpeè, naì, no<
two syllables: oæbaæ, eèkaæt, oèboì, oædaè, aæbo<*aìboæ, *a<boæ
three syllables: eækuæeæ, oækpuègaè, bèsoæNeæ, aædèmbaæ*aædèmbaì, *aædìmbaè, *aìdèmbaè, *aædæmbaè, *aèdèmbaæ
No Crowding Constraint, No Crossing Constraint and the OCP
Autosegmental Phonology
22
these data can be explained in an autosegmental analysis if we assume that there are certain underlying tone patterns
these are associated with TBUs according to universal association conventions
tone patterns in Etung: L H L H H L L H L H L Hassociation conventions (after Goldsmith, 1976):
associate...(a) ... tone features with TBUs 1-on-1 from left to right,(b) ... remaining TBUs with the last tone feature,(c) ... floating tone features with the last TBU
No Crowding Constraint, No Crossing Constraint and the OCP
Autosegmental Phonology
23
Examples: /kpa/
H
/na/
H L(a) (a) (c)
in many languages the number of tone features that can be associated with one TBU is limited → no crowding constraint
in Etung, this number is two; hence the following is disallowed:
* /na/
L H Lin tone languages without contour tones this number is one
No Crowding Constraint, No Crossing Constraint and the OCP
Autosegmental Phonology
24
Examples: /oba/
H
/oda/
H L(a) (a) (a)
the no crossing constraint is implicit in the association conventions
no crossing constraint: association lines must not cross
(b)
/abo/
H L(a) (a) (c)
H
* /oda/
H L(a) (a)
the no crossing constraint is an implication of the assumed linearity of the single tiers
No Crowding Constraint, No Crossing Constraint and the OCP
Autosegmental Phonology
25
in the case of /oæbaæ/ one could think of two alternative tone patterns:
/oba/
H
/oba/
H Hthe right alternative is ruled out by a general constraint, called the
Obligatory Contour Principle (OCP):
Adjacent identical tones are disallowed
the OCP is the reason for assuming the postulated underlying tone patterns, instead of eg. H H or L L H
(see exercise 2. for what goes wrong without the OCP)
No Crowding Constraint, No Crossing Constraint and the OCP
Autosegmental Phonology
26
Summary
some phenomena such as tone are suprasegmental
a non-linear autosegmental analysis, where features are located on different independent tiers, is capable of explaining observed data much better than a traditional linear analysis
next to segmental rules, there are association conventions and association rules, which describe how features are associated and how they spread
in each tonal language the maximal number of tone features per tone bearing unit is limited
the no crossing constraint, the no crowding constraint and the OCP restrict the distribution and association of features across tiers.
27
Exercise 8
(1) Use the tone patterns of Etung and the association conventions on p. 22 to derive the tone distribution for eækuæeæ, bèsoæNeæ, und aædèmbaæ.
(2) Consider the following inflected form from Margi (cf. p.10ff):
/laèguè+aærè/ > [laègwaærè]
Provide the underlying autosegmental representation and derive the surface form using the 'glide' rule R2 (on p.16) and the association rule R1 (on p.18)
Assume that the OCP would not hold and that the underlying tone pattern for the stem was L L. Which incorrect derivation would one get for /laèguè+aærè/ ?