implications of sociolinguistic findings for phonological theory william labov, university of...

Implications of sociolinguistic findings for

phonological theory

William Labov, University of Pennsylvania

18th Manchester Phonology Meeting

• The implications of phonological conditioning for the identification of underlying forms and the feed-forward model

• The implications of the regularity of sound change for phonological models of perception and production

The first study of internal factors governing linguistic variationProportion of –t,d clusters deleted for 11 members of the Jets in single interviews

Labov, Cohen, Robins and Lewis 1968

Lexically regular variation: coronal stop deletion in English. Two basic constraints on the simplification of /t,d/ clusters in English: the effect of a following consonant vs. a following vowel (a vs. b) and the effect of grammatical boundary (c vs. d)

Monomorphemic Past tense

Proportion deleted of monomorphemic –t,d clusters before consonants and vowels for eleven members of the Jets.

0 5 10 15 20 25 30 35 40 450.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

_K_V

Number of tokens

Prop

ortio

n de

lete

d

Effect of following consonant on realization of past tense –ed in spontaneous speech of 112 African American struggling readers [N=722]

Lateral Nasal Fricative Stop Vowel0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Varb

rul w

eigh

t for

abs

ence

Following segment

Where is variation located?

Variable morphological insertion of the past tense morpheme {d} vs.Invariant morphological insertion of the past tense morpheme {d}Regressive assimilation to /t,d/Variable deletion of /t,d/

Proposal: Postlexical phonetic conditioning implies morphological invariance. Conversely, variable morphological insertion implies the absence of postlexical conditioning

Question:If linguistic variation can look up a derivation, why not look down?

“do not insert segment X if by so doing we create an impermissible cluster at the post-lexical level?”

AAVE copula: Variable insertion vs. variable contraction and deletion

əz

z Contraction

Deletion z

0 əz

0

əz z

0

Insertion1 Insertion2

Phonetic conditioning of contraction and deletion for two adolescent groups in South Harlem [from Labov, Cohen Robins and Lewis 1968]

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

_K _V _K _V

Varbul weights

ContractionDeletion

Cobras Jets

K__ V__ K__ V__

Jo’s talkingJen talking

Jen talking

Jo’s talking

Phonotactic effects on contraction and deletion

Jo is talking Jen is talking

CVVC CVC

Jo’s talking Jen’s talking

CVC CVCC

Jo talking Jen talking

CVC CVC

Contraction

Deletion

CV subject CVC subject

Variable AAVE copula and auxiliary is the result of successive contraction and deletion of an underlying form /iz/

Lexical entry V => /iz/

Stress assignment [+str] => [-str]

Vowel reduction /iz/ => /əz/

Contraction: /əz/ => (z)

Deletion: /z/ => (0)

əz

Əz z

0

Contraction

Deletion z

Absence of phonological conditioning of word-final /s/ for AA groups in South Harlem [from Labov, Cohen, Robins & Lewis 1968]

0

10

20

30

40

50

60

70

80

90

100

_K _V _K _V _K _V _K _V

Percent absence

T-Birds (8)Aces (4)Cobras (9)Jets (17)Oscar Bros. (6)Lames (20)Adults (8)

Group Single Group Single

So. WC

nice dog (ni’ dog)

nice apple

he’s cold

he’s out

Type of Recording:Group Session vs Single Interviews

Monomorphemic Verbal {s}

Adults

Logistic regression analysis of verbal {s} in the spontaneous speech of 58 African-American struggling readers, Philadelphia, 2001

(Note that all blue represent non-significant differences; only red are significant)

0

0.2

0.4

0.6

0.8

1

Consonant

Vowel

Consonant

VowelPauseverbdoes

he, she, itNoun phrase

Atlanta

PhiladelphiaCalifornia

Grade 2Grade 3Grade 4FemaleMale

Proportion/weight of absence

ProportionVR weight

Preceding segment

Following segment

Verb

Subject Pronoun/ Noun phrase

Region

Grade

Gender

Hypercorrection as evidence of morphological variation

He can goes out (13, T-birds)I don’t know how to gets no girls. (13, Jets)He’d knows that (15, NYC)

Third singular {s}

Past tense {d}

* He can tried to do it.* He will passed me.

Establishing the locus of variation

Systematic phonological conditioning of a variable implies invariance at the morphosyntactic level and absence of phonological conditioning implies variability at the morphosyntactic level

Effect of

following

Morphological

segment

variation

English -t,d deletion +

-

AAVE copula +

-

Spanish (s) +

-

English (REL) -

+

AAVE verbal {s} -

+

AAVE possessive -

+

Convergent evidence from the study of reading errors

A study of reading errors in an Individualized Reading Program, 2001-2

When to intervene in oral reading?

To decide when to correct an oral reading, the tutor should distinguish between deviations from standard pronunciation, and failure to identify the meaningful elements in the text.

How can this be done?

Absence of verbal /s/ in oral reading: Jason P., age 7

Hey Black doesn’t eat cats, not even one

He just likes to growl and watch them run.grow which

True error

True error

CorrectReading

Absence of verbal /s/ in oral reading: Jason P., age 7

Hey Black doesn’t eat cats, not even one

He just likes to growl and watch them run.lose

Clearerror

Absence of verbal /s/ in oral reading: Jason P., age 7

Hey Black doesn’t eat cats, not even one

He just likes to growl and watch them run.licked

Clearerror

Absence of verbal /s/ in oral reading: Jason P., age 7

Hey Black doesn’t eat cats, not even one

He just likes to growl and watch them run.like

Potential error

grow which

True error

True error

The Semantic Shadow Hypothesis: An error in the identification of a given morpheme increases the probability of errors in the decoding of the following text.

Frequency of following errors for clear errors and correct reading by dialect feature

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

_CC1

Verbal {s}

Possessive {s}{ed}clusterother {ed}

plural

copula {s}Irregular past

ch_ sh_ br_

Clear errors

Correct

Dialect feature

A potential error type (e.g., omission of verbal {s}) may be treated as a true reading error if the rate of following errors is significantly greater than the rate for correct readings.

Interpreting potential errors

Frequency of following errors for clear errors, potential errors and correct reading by dialect type [N=567]

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

_CC1

Verbal {s}Possessive

{s}

{ed}clusterother {ed}

plural

copula {s}Irregular

past

ch_ sh_ br_

Clear errors

Potential errors

Correct

Potential errors

Frequency of following errors for clear errors, potential errors and correct readings by dialect type for African American readers [N=238]

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Homovoiced_CC (264)Heterovoiced

_CC (37)

{ed}cluster (69)other {ed} (72)Verbal {s} (62)Possessive {s}

(195)plural (73)copula {s}(180)

Irregular past

ch_ (5)br_ (64)

sneaked /snuck

(9)

Error rates

Clear errorsPotential errorsCorrect

W. Labov and B. Baker. What is a reading error? In press. Applied Psycholinguistics.

• The implications of phonological conditioning for the identification of underlying forms and the feed-forward model

• The implications of the regularity of sound change for phonological models of perception and production

• The implications of phonological conditioning for the identification of underlying forms and the feed-forward model

• The implications of the regularity of sound change for phonological models of perception and production

The Neogrammarian viewpoint

Every sound change, inasmuch as it occurs mechanically, takes place according to laws that admit no exception. --Ostoff and Brugmann 1878

Sound-change is merely a change in the speakers’ manner of producing phonemes and accordingly, affects a phoneme at every occurrence, regardless of the nature of any particular linguistic form in which the phoneme happens to occur. . . The whole assumption can be briefly put into the words: phonemes change.

--Bloomfield 1933:353-4

The lexical diffusion viewpoint

The phonetic law does not affect all items at the same time: some are designed to develop quickly, others remain behind, some offer strong resistance and succeed in turning back any effort at transformation. --Gauchat (cited in Dauzat 1922)

We hold that words change their pronunciations by discrete, perceptual increments (i.e., phonetically abrupt) but severally at a time (i.e., lexically gradual)

--Wang and Chen 1977:150.

A proposed resolution

Regular sound change is the result of a gradual transformation of a single phonetic feature of a phoneme in a continuous phonetic space.

Lexical diffusion is the result of the abrupt substitution of one phoneme for another in words that contain that phoneme.

The lexically gradual view of sound change is incompatible, in principle, with the structuralist way of looking at sound change.

--Chen and Wang 1957:257.

But--

The feed-forward model

In fluent, mature speakers, the phonetic implementation system is a modular, feed-forward system, reflecting its nature as an extremely practiced and automatic behavior. . . The model is feed-forward because no arrows go backwards, from articulatory to phonological encoding, or from the phonological encoding to the lexical level. It is modular because no lexeme information can influence the phonetic implementation directly, bypassing the level of phonological buffering.

--Pierrehumbert 2002, “Word-specific phonetics”

The intersection of the exemplar model and the lexical diffusion view of sound change

Obviously, this treatment is not confined to lenition; any systematic bias on the allophonic outcome would incrementally impact high frequency words at a greater rate than low frequency words. In short the model is applicable to any Neogrammarian sound change, by which I mean sound changes which get started in the phonetic implementation and eventually sweep through the vocabulary.

--Pierrehumbert 2002

The fronting of /ow/ in North America (from ANAE Ch. 12)

Distribution of /ow/ vowels for all of North America. [N=8313].Vowels before /l/ are shown in black [N=1577].

Absence of fronting of Vw in vowel system of Alex S., 42, Providence, RI TS 474.

Fronting of all Vw in the vowel system of Danica L., 37, Columbus, OH, TS 737.

34 most frequent /ow/ words in the Brown corpus with Telsur frequenciesBrown Telsur F2 MEAN F2 SD

no 2201 348 1497 214home 639 547 1066 176go 347 626 1386 237coat 313 43 1302 230sofa 227 6 1282 168both 218 730 1202 214know 179 683 1409 239most 153 1160 1215 220old 145 660 1016 175goal 137 60 1017 110coke 136 4 1368 191phone 101 54 1112 191goat 84 6 1427 243pole 79 18 932 110boat 72 165 1293 208coast 66 61 1321 201donut 66 1171 161over 63 1236 1195 200Polish 59 19 992 135road 57 197 1327 195Minnesota 57 13 1282 195gold 52 60 1009 120mostly 48 44 1196 207doe 37 1 1438 238ago 37 1387 220fold 31 7 971 182ocean 27 34 1403 278cold 26 171 989 143notice 25 59 1360 277bowl 23 79 1000 126low 21 174 1235 128toast 19 248 1376 219nose 17 60 1535 147soda 3 406 1336 182

Fronting of /ow/ for words before /l/ and others for all of North America and for the Southeast (South and Midland). Words selected by stepwise regression analysis at p <.001 level as ahead of phonological prediction, light blue; behind, yellow.

800

900

1000

1100

1200

1300

1400

1500

1600

1700

polePolish

coldgoldbowlgoalfold oldhome

low oversofaboatcokecoastoceannose

no

F2 in Hz

All __lSE__lAllSE

Is home a lexical exception to the fronting of /ow/?

N F1 F2

/ow/ 5950 616 1304

/owl/ 2576 575 1010

home 775 669 1068

Oklahoma 14 589 1045

homebody, etc. 28 641 1037

Omaha 10 655 1119

hoe 26 621 1233

Phonetic effects of onset /h/ and coda /m/ on fronting of /ow/

10001100120013001400520

540

560

580

600

620

640

660

680

F2

F1

/owl/ 2,976

Oklahoma (14)

Home cmpd (28)

home (775)

Omaha (10)

hoe(26)

/ow/(5,050)

Significant regression coefficients (p < .01) of F2 of /ow/ in the Southeastern region.

Significant regression coefficients (p < .01) of F2 of /ow/ in the Southeastern region.

Significant regression coefficients (p < .01) of F2 of /ow/ in the Southeastern region.

Significant regression coefficients (p < .01) of F2 of /ow/ in the Southeastern region.

Significant regression coefficients (p < .01) of F2 of /ow/ in the Southeastern region.

Conclusion

Regular sound change is governed by persistent, stable and predictable phonetic effects on all members of a phonemic category, but may be accompanied by minor, unstable and unpredictable lexical effects on individual lexical items

What kind of phonological theory is consistent with studies of linguistic change and variation?

A theory that describes speech production as the selection and linearization of abstract categories.

A theory that can incorporate probabilities into the rules and constraints governing production and perception.

A hierarchical system in which decisions made at a given level are independent of decisions made at a lower level.

A theory which predicts when the unit of sound change is the phoneme and when it is the word stem.

An architecture that accesses and uses lexical information at the final stages of speech production.

Absence of morphosyntactic segments in the spontaneous speech of 399 struggling readers, grades 2-4, by ethnicity/language, 2001

Possessive /s/ Verbal /s/ Copula /s/ syllabic -ed td clusters0

10

20

30

40

50

60

70

African AmericanLatino (Eng)Latino (Span)Euro-American

Perc

ent s

egm

ent a

bsen

ce

www.ling.upenn.edu//~labov

Distribution of no [N=348, yellow] and know [N=630, blue] in F1/F2 space

Absence of morphosyntactic segments in the spontaneous speech of 399 struggling readers, grades 2-4, by ethnicity/language, 2001

Possessive /s/ Verbal /s/ Copula /s/ syllabic -ed td clusters0

10

20

30

40

50

60

70

African AmericanLatino (Eng)Latino (Span)Euro-American

Perc

ent s

egm

ent a

bsen

ce