PERCEPTION, LANGUAGE REGARD, and SOCIOPHONETICS

1
PERCEPTION, LANGUAGE REGARD, and SOCIOPHONETICS
Dennis R. Preston Michigan State
University preston_at_msu.edu
2
Social and stylistic stratification of (ing) in
the random sample of the Lower East Side of New
York City adults N81
Source Labov 1966
3
Social and stylistic stratification of (ing) in
the random sample of the Lower East Side of New
York City adults N81
Source Labov 1966
4
Social and stylistic stratification of (ing) in
the random sample of the Lower East Side of New
York City adults N81
Source Labov 1966
5
Social and stylistic stratification of (ing) in
Norwich, England.
Trudgill 1974
6
Percent /in/ in (ing) in the Philadelphia
neighborhood study grammatical conditioning
--Labov 2001, Fig. 3.4
7
Historical development of (ing), 10th-20th C.
10th century Verbal noun Participle
-inge -inde loss of
final shwa /i?g?/
/ind?/ loss of final C
/i?g/ /ind/ 15th century
fixed orthography
-ing
/i?/ /in/ social
stratification 19th century
20th century stable social /i?/
/in/ stylistic variation
(favored in nouns ) (favored in
verbs)
8
The transmission of stylistic and grammatical
constraints on the acquisition of (ing) in a King
of Prussia family
Curt Kay C., David, 7
parents
sonSTYLE less formal (narrative) .72 .69
more formal (other) .38 .38 GRAMMATICAL FORM
progressive (Im working on it) .61 .73
participle (A man working for you) .43 .30
nominal (Working is bad for you) .17 .20
--Labov, The child as linguistic historian, 1989
9
The Audition Test
An actor is reading for the part of a
construction worker in a Philadelphia play. Here
are two trials in his audition for the part.
Would you please rate each one on the following
scale by putting a check in one box. Thank you.
Perfectly Very natural
unnatural 1 2 3 4
5 6
7 /_______/_______/_______/_______/_______/_______
/_______/
10
Two versions of the Audition Test
I. Husband to wife Look honey, I know I was
supposed to be paintin the ceiling tonight. But
they had me workin since six in the morning on
the god damned federal building. We were fixin
the wiring on the west wall, and I was hanging
onto the pipe railing all day. My back is killin
me.
II. Husband to wife Look honey, I know I was
supposed to be painting the ceilin tonight. But
they had me workin since six in the mornin on
the god damned federal buildin. We were fixing
the wirin on the west wall, and I was hanging
onto the pipe railin all day. My back is killing
me.
11
Distribution of /in/ and /i?/ in the two versions
of the Audition test
Participle I II Noun I II painting in i
? ceiling i? in working in in morning i? in fixing
in i? building i? in hanging i? i? wiring i? in
killing in i? railing i? in
/in/ 80 20 0 100
12
Pilot results of the Audition Test N10
t test, p lt .00001
13
(No Transcript)
14
(No Transcript)
15
Actual results
16
Race and Residence in the San Francisco area
(Baugh 2007)
17
Success rate for apartment showings by area and
guise (Baugh 2007)
18
The Philadelphia Story I gotta get out of the
house. house as h??s and as
hæ?s N70 Graff, Labov, and Harris 1986
19
(No Transcript)
20
He darted out about four feet before a car, and
he got hit hard. We didnt have the heart to
play ball or cards all morning Labov 1966
21
The ratings 1 TV personality 2 Executive
secretary 3 Receptionist 4 Switchboard
operator 5 Salesgirl 6 Factory worker 7 None
of these
22
(No Transcript)
23
(No Transcript)
24
JIMS BED WAS GRIMEY. ?i?mz be?d w?z
gra?m? ONCE THE GOOD SHEETS WERE NEW. w?ns d?
g?d ?e?ts w? nj?? NOW THEYRE
OLD, FADED, AND SOILED. næ? da?? ???d
fa?d?d æn s???d
25
Imitation high and mid front vowels (Evans 2002)
26
Identification results (Evans 2002)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
48 voice samples (Clopper 2004)
31
Sites where voice samples could be placed
(Clopper 2004)
32
Perfect results (chance is 17 in a six-way test)
33
Actual results (overall 26) (Clopper 2004)
34
(No Transcript)
35
1 2 3 4 5 6 7
S
N
Plichta and Preston 2005
36
Results Step Mean Region
1 2.8490 1. Saginaw 2
3.1684 2. Coldwater 3 3.8733
3. South Bend 4 4.8924 4.
Muncie 5 5.9948 5. New Albany
6 6.5781 6. Bowling Green
7 7.0174 7. Nashville 8.
Florence 9. Dothan
37
(No Transcript)
38
Male, EA, 19, Chicago (Preston 1996)
39
1
i
i
11
u
?
2
10
U
3
e
9
ç
o
4
E
8
ACTUAL
7
v
CANON
6
A
5
Q
HYPER
Experimental vowels, Niedzielski 2000
40
(No Transcript)
41
(No Transcript)
42

• What the next part of this presentation will NOT
  discuss
• 1) Slips of the tongue and ear
• 2) Interference of delivery manner or noise
• 3) Modal interference
• 4) Social identity interference

43

• What it WILL deal with
• Study 1) The comprehension of single-word tokens
  on the basis of a variety of phonetic and
  phonological characteristics
• Study 2) The comprehension of tokens in carrier
  phrases on the same bases

44
THE LINGUISTIC SETTING
Vowel movement in the Northern Cities Shift
45
Study 1) The comprehension of single-word tokens
(Preston 2005) In most previous work single
word tokens have been very well understood.
46
Overall comprehension results for Peterson
Barney, 1952 (percent correct)
47
Peterson and Barney,1952 General American
(/e/ and /o/ added from Stevens, 1998)
48
Overall comprehension results for Hillenbrand et
al., 1989 (percent correct)
49
Hillenbrand et al.,1989 female vowel mean scores
50
Peterson and Barney (left) Hillenbrand et al.
(right)
51
Overall comprehension results for Cutler et al.,
2004 (percent correct)
52

• PROBLEMS
• Peterson Barney
• a) Huge sample overnormalized
• b) Little control for region of speakers or
• hearers
• 2) Hillenbrand et al.
• a) Phonetically trained hearers
• b) Speakers carefully screened
• 3) Cutler et al.
• a) Mismatch between speaker (Iowa) and
• hearers (Florida University students)

53
This research seeks to overcome these problems
but will look at a vowel system undergoing rapid
and dramatic change. The main focus of this
research will be to see the degree to which
groups of local speakers can accurately identify
single word tokens of advanced vowels from the
NCS. Three other questions, however, will also
be addressed
54

• Question 1
• Do locals have an advantage in understanding
  local norms, and does local reflect such
  characteristics as sex, age, status, urbanity,
  ethnicity?
• Previous research (Evans, et al., 2001) shows
  southeastern Michiganders to be more advanced in
  the NCS if they are
• a. Young b. European American
• c. Female d. Middle class
• e. Urban f. Non-immigrant (non-Appalachian)

55
Question 2 When vowels are misunderstood, is
it in the direction of the pre-shifted token
(Labov and Ash 1997)? Isnt it more likely that
a mistake will be made that reflects the new
position of the sounds in the emerging system?
56
Figure 7. Selected elements of the
NCS Heres a part of the NCS. Suppose a listener
hears a shifted version of pet. It has moved to
the territory formerly occupied by pat. But
pat has also moved and pot has come into the
vacated pat territory. Will hearers be likely
to hear pet as pat (the old system) or pot
(the new system)?
57

• Question 3
• What other factors might influence comprehension?
• a) Order of change in the NCS shift
• b) Phoneme class history
• c) Magnitude of shift
• d) Increased (decreased) proximity to
• other items in the system
• e) Changed phonetic characteristics
• f) Different perceptual characteristics

58
METHODOLOGY 20 words spoken by young E-A women
from urban southeastern Michigan were played
(twice) bag, cut, big, can, bond, bed (bud),
hawk, done, sock, tin, hot, caught, pat, Ben
(bun), dawn, bed (bad). boot, beet, bait, and
boat were included but showed little
misunderstanding (.96, .98, .98, .99). In
contrast, the best comprehension rate for the NCS
tokens was .85 (for short a, i.e. æ). The
respondents wrote down the word they heard, and
were judged correct if the intended vowel was
indicated (e.g., pig was judged correct for big).
59

• GROUPS, young respondents (15-30) only
• A. European American
• 1. EAUSUrban southeastern MI (N70)
• 2. EAYUrban southeastern MI
• immigrants from Appalachia (N2)
• 3. EARSRural southeastern MI (N17)
• 4. EARMRural mid-MI (N39)
• B. African American
• 5. AAUSUrban southeastern MI (N24)

60
DISTRIBUTIONAL CHARACTERISTICS All groups were
relatively evenly distributed by sex but not by
status. The five demographic groups were entered
as an independent variable, along with sex,
social status, and the six individual
vowels. The dependent variable was accuracy of
comprehension of the vowel.
61
Figure 8. Vowel positions in the current study
62
Figure 9 Overall comprehension results
Preston, 2005 (percent correct)
63
Factor Groups Weight
Correct/Total Group EAUS .54 .71 795/1121
EAY .535 .70 23/33 EARS .50 .68 185/272
EARM .46 .65 305/569
AAUS .42 .62 249/404 Sex
Female .n.s. .69 961/1400 Male .n.s. .66 596
/899 Status Middle n.s. .69 1399/2041
Working n.s. .61 158/258 Vowels short
æ .70 .85 366/432 short o .67 .83 357/431
wedge v .64 .80 231/287 short
e .49 .69 298/429 short i .24 .42 122/288
open oh .24 .42 183/432
Overall results by independent variable
groups (NB Newer work on Arab Americans in
Dearborn MI shows a .62 correctness rate Mexican
Americans in Lansing MI have .65.)
64
Figure 10 GoldVarb weights for groups
65

• QUESTION 1 DO LOCALS HAVE AN ADVANTAGE
• IN COMPREHENDING ADVANCED TOKENS? YES.
• AND CAN LOCAL REFER TO OTHER
• DEMOGRAPHIC FEATURES? YES.
• Features which prefer comprehension
• Ethnicity European American
• Area urban (southern MI)
• Sex female (n.s.)
• Status middle (n.s.)
• Immigrant non-immigrant (n.s.)

66
QUESTION 2 ARE VOWELS MISUNDERSTOOD AS
PRE-SHIFT OR POST-SHIFT ITEMS? For example,
when short e backs and lowers, it will be
misunderstood as short æ in the pre-shift
interpretation but as short o in the post shift
interpretation.
67
VOWELS ARE MISUNDERSTOOD AS PRE-SHIFT.
Error rates.
68
QUESTION 3 What about the individual vowels?
GoldVarb and percentage results for vowels
69
Figure 11 Experimental (yellow) and PB vowels
(white)
70

• Assumptions Items which will be better
  understood if they have
• 1. An older position in the NCS shift history
• 2. Uncomplicated phoneme class histories
• 3. Phonetic clues
• 4. A distinct acoustic space
• 5. Moved a shorter distance in the shift
• 6. Not introduced processing difficulty (F1 vs F2)

71
Criteria for vowel comprehension rates
72

• How do these results compare with previous
  studies? (Labov and Ash, 1997, Rakerd and
  Plichta, 2003)?
• 1. Is there a local advantage? Yes The young,
  EA southeastern MI group was best.
• 2. Mistakes are made in the direction of the
  former system. Yes, in every case.
• 3. Local subjects more advanced in the shift
  themselves are better at comprehension. Yes,
  but

73
Study 2) The comprehension of words in carrier
phrases (Plichta, Preston Rakerd 2005) Do
carrier phrases give hearers an advantage in
understanding words with tokens of
change-in-progress sounds (Labov and Ash 1997,
Plichta and Rakerd 2003). What is the role of
specific items in carrier phrases? This
experiment focuses on the role of NCS vowels in a
carrier phrase promoting (or demoting)
understanding of another vowel.
74
Labov and Ash Crossdialectal findings for NCS item
75
Labov and Ash Crossdialectal findings for SVS item
76

• There are difficulties with these studies
• 1. Other tokens In Labovs work, the carrier
  phrases contain other tokens of the NCS and
  Southern Vowel Shift (SVS).
• a. NCS test had and sandals occur with
  socks.
• b. SVS test I and knew occur with guy.
• 2. Same tokens
• a. NCS test The socks vowel does not occur.
  
• b. SVS test I occurs, the same vowel as in
  guy.
• 3. Real words The misunderstanding is a word in
  the NCS test (socks) but not in the SVS test
  (gah).
• 4. Clues The phrase in the NCS test gives a clue
  to the identity of the word the SVS test does
  not.

77
Plichta and Rakerd resynthesized positions for
the vowel of sock in a fronting tolerance
experiment.
78
Results for non-NCS respondents.
79
Results for NCS respondents.
80
In Plichta and Rakerd, the resynthesized tokens
of sock were embedded in the following carrier
phrases

• In addition to examples of the sock vowel, each
  phrase contains other several words with NCS
  vowels.

81
Plichta and Rakerd avoid some of the problems of
Labov and Ash 1) Misunderstood words are real
words sock-sack 2) The carrier phrase gives
no clue to the identity of the item. But it
is still not clear if the retention of the sock
understanding farther along the F2 dimension was
enhanced by 1) the presence of NCS vowels in
general, or 2) the presence of the vowel that
matched the test item in particular.
82

• This experiment is based on the same 7-step
  manipulation of the NCS vowel hot. We selected
  three young (20 year old) male speakers from
  Minnesota who showed no evidence of the NCS.
• They read the following phrases
• Katie heard Bob say the word hot
• Katie heard dad say the word hot
• Katie heard her boss say the word hot
• In each phrase the word hot occurs, but there is
  only one other NCS word (dad, Bob, or boss).

83
We resynthesized these readings as
follows Speaker 1 is always NCS
shifted. Katie heard Bob say the word hot Katie
heard dad say the word hot Speaker 2 is mixed
boss was shifted dad not. Katie heard her boss
say the word hot Katie heard dad say the word
hot Speaker 3 was never NCS shifted. Katie
heard Bob say the word hot Katie heard her boss
say the word hot
84
Resynthesis of Bob
85
Resynthesis of dad
86
Here are some examples of the stimulus sentences
the respondents heard with hot at step
1 Speaker 1- shifted dad Speaker 2 - shifted
boss unshifted dad Speaker 3 - unshifted
Bob Each stimulus was played with the word hot
added at the end in each of its seven modified
steps. Since each of the three speakers used two
carrier phrases the total was 6 X 7 42.
87

• The demographic variables for the 44 respondents
  were 1) Sex
• 2) Age by decades, teens through 70s
• 3) Region NCS or non-NCS
• The experimental variables were
• 1) Speaker-token
• Speaker 1 a) Bob-shifted b) dad-shifted
• Speaker 2 c) boss-shifted d) dad-unshifted
• Speaker 3 e) bob unshifted f) boss
  unshifted
• 2) Level the 7-step fronting used in Plichta
  and
• Rakerd.
• The data were analyzed with GoldVarb II
  selection of hot was the applications value.

88
Table 5. Weights are shown only for significant
factor groups the Apps column indicates the
number of times the stimulus was judged to be the
hot vowel, i.e., the number of times the rule
applied. S and U indicate NCS shifted and
unshifted tokens, respectively.
Weight Apps Total SEX Male 0.386 78 329 42
0 Female 0.560 86 689 812 AGE Teens 0.331 77 86
112 20s 0.485 83 488 588 30s 0.651 88
172 196 40s 0.507 84 94 112 50s 0.487 79
178 224 60s 1 resp 55 15 28 70s KO 100
42 42 LEVEL 1 0.868 98 172 176 2 0.716 94
166 176 3 0.619 91 161 176 4 0.475 86
151 176 5 0.334 77 136 176 6 0.299 74
131 176 7 0.161 57 101 176 AREA NCS ns 81.6
457 560 no NCS ns 83.5 561
672 TOKEN S-BOB/BOSS 0.543 85 522
616 U-BOB/BOSS 0.457 81 496 616 S-DAD ns 80.5
276 343 U-DAD ns 82.2 282 343
89
Figure 19 Weights for the hot-hat continuum
90
Figure 20 Weights for sex
91
Figure 21 Weights for age (with for older
respondents)
92
Figure 22 Percentages for region (n.s.).
93
Figure 23 Weights for speaker-tokens (n.s.)
94
Figure 24 Weights for shifted unshifted Bob
boss.
95

• Conclusions to Study 2
• Perception of items in changing systems
• may be more related to demographic features such
  as age and sex than to region
• 2. Women have a considerable advantage in hearing
  such forms.
• 3. Young adults and middle-age persons have an
  advantage in hearing such forms.
• 4. A carrier phrase gives a listener an advantage
  in hearing such forms only if
• it has an example of the test item itself or
• an adjacent item.

96

• SO WHAT?
• Speech science
• Sociolinguistics
• Linguistic Theory
• Language change

97

• 1. Speech science
• Although there is recognition that vowel
  perception is less categorical, the prevailing
  view does not recognize the considerable
  gradience uncovered in these more recent
  perception experiments.

98

• 2. Sociolinguistic theory
• In a recent thorough review of the history and
  current status of the notion speech community,
  Patrick does not mention the possibility that
  norms of perception might play as big a role as
  those of production and evaluation.

99

• 3. Phonological theory
• The findings here suggest that system prototypes
  might be organized around idealized ones, ones
  which disallow the overwhelming influence of
  numerous exemplars, perhaps particularly in the
  perceptual system.

100

• 4. Language change
• Perhaps outliers and seeds of change are rather
  conservatively approached by listeners, who
  appear to anchor their abilities in a previous
  system.

101
CONCLUSION Perception/comprehension does not
simply mirror production. The tacit assumption
that it does and the failure to take the fact
that it does not into consideration may derail
several approaches to our enterprise.
102
Thank you very much