You are on page 1of 17

Language Variation and Change, 20 (2008), 67– 83. Printed in the U.S.A.

# 2008 Cambridge University Press 0954-3945/08 $16.00


doi: 10.1017/S0954394508000069

Regional differences in perceiving vowel tokens on Southerness,


education, and pleasantness ratings
VA L E R I E F R I D L A N D
University of Nevada, Reno

ABSTRACT

This study investigates the role of regional dialect experience on the social awareness
of synthesized vowel tokens to regional in-group and out-group members. For the
study, speakers from Reno, NV, were given the same perception test used in a
previous study in Memphis, TN. Comparing the Reno results to those found in
Memphis, the study examines whether differences in regional vowel norms affect
how Westerners rate Southern-shifted and non-Southern-shifted vowel variants on
Southernness, education, and pleasantness scales. The study also looks at how
Reno raters interpreted shifted back vowel variants, found productively in their
local community, compared to front vowel shifts found exclusively in the South.
Finally, the paper explores how the results suggest that regional dialect exposure
attunes listeners to attend to different aspects of vowel quality than those outside
the region. In examining how regional dialect experience affects listener
recognition and evaluation of local and nonlocal vowel norms, the paper begins to
explore how much the production/perception relationship is mediated by speakers’
participation in locally constructed and defined speech communities.

Previous
research: 美 This paper presents a sociolinguistically grounded examination of the role of
国东南 regional dialect experience in shaping speakers’ evaluation of vowel quality.
Through the administration of the matched guise vowel perception test using
insiders synthesized vowel tokens, an earlier study (Fridland et al., 2004) investigated
whether Southerners from Memphis, TN, found vowel formant positions
associated with the changes affecting Southern dialects (the Southern Vowel
Shift [SVS]) as more Southern sounding than nonshifted variants and whether,
due to the apparently divergent nature of vowel changes across regions, front
vowel positioning was more salient as a regional identifier than back vowel
positioning. A follow-up study (Fridland et al., 2005) measuring how
Memphians perceived speakers of these variants on pleasantness and education
scales suggested that the greater the recognition of the variant as Southern, the
lower its ratings on pleasantness and education scales. Overall, results from this
work suggest that the ability of participants to accurately rate differences

This research has been supported by grants from the National Science Foundation Linguistics Program
BCS#0132145 and BCS#0001725.

67
https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press
68 VA L E R I E F R I D L A N D

between vowel variants depends on whether the local community speech norms
involve those particular variants and whether those variants are shared with
other regions. Vowel dynamics also appeared to be a factor in whether
Memphians recognized shifted variants as Southern, with long glided vowels
increasing a vowel’s potential for Southern association. The study clearly
indicated that some shifts emerge as more salient regional markers than others,
even when there is also widespread productive use of other localized shifts in the
community. Both dialect exposure and mechanical linguistic factors played a
role in which shifts were most readily assigned social significance.
Such results bring up the question of how speakers from outside the South
would perform on the same test. Without similar dialect exposure, would
美国西部 listeners show less sensitivity to the “localness” of variants in their ratings and
outsiders would different linguistic experiences alter the type of linguistic cues relied on
to make such decisions? To approach these questions, speakers from Reno, NV,
were given the same synthesized vowel token test used in the previous study.
Comparing the Reno results to those found in Memphis, the study examines
whether differences in regional vowel norms affect how Westerners rate
Southern-shifted and non-Southern-shifted vowel variants on Southernness,
competence, and solidarity scales. The study also looks at how Reno raters
interpreted shifted back vowel variants, found productively in their local
community, compared to front vowel shifts found exclusively in the South.
Finally, the paper explores how the results suggest that regional dialect exposure
attunes listeners to attend to different aspects of vowel quality than listeners from
outside the region.

VOWEL CHANGES IN THE U.S.

Although American dialects generally share the same vocalic inventory, they differ
substantially in terms of the phonetic range in which vowel tokens are realized
within these prescribed categories. Much work in the variationist paradigm has
focused on describing and instrumentally measuring the productive changes
affecting the vowels in a variety of American dialects. A recent wealth of such
work has lead to a very clear picture of regional differences and similarities,
including some fairly dramatic shifts in the relative position of vowels in all
three major dialect regions—the North, South, and West. (See work by Clarke
et al., 1995; Eckert, 1988, 2000; Evans, 2001; Feagin, 1986; Fridland, 2000,
2001, 2003a, 2003b, 2006; Gordon, 1997; Labov, 1991, 1994, 2000; Labov
et al., 2006; Thomas, 1997, 2001).
The changes affecting front vowels in the three main dialect areas appear to be
creating greater regional differentiation, with the high and mid-front system
effectively moving in opposite directions across dialects. Acoustically, the
formant structure of the realigned front vowels is associated with changes along
both the first and second formants, F1 and F2, respectively (roughly, along the
height and advancement dimensions). In the South, the most prominent change

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 69

in the front system is the reversal in acoustic position of the /ey/ (bait) and /e/
(bet) class, with the nucleus of /ey/ appearing lowered and centralized and /e/
becoming higher and peripheral relative to /ey/. The position of /iy/ (beat) and
/i/ (bit) is also similarly affected in some Southern dialects. In the Western and
Northern systems, the front tense system retains its position at the front edge of
the acoustic vowel space (although other vowel shifts affect and differentiate
these two regions).
Contrasting the regionally distinguishing front vowel system, a number of
similar changes affecting the back vowels, namely the fronting of /uw/, /u/,
and /ow/, have been reported for all major dialect regions (Ash, 1996;
Fridland, 2001, 2006; Hagiwira, 1997; Labov, 1994; Luthin, 1987). In all
these dialects, /uw/ and /u/ are the first back vowels to be affected by
change, with /ow/ showing some shift in more advanced (and generally
younger) systems, though shift in this class is much less common. The
linguistic contexts for fronting also appear quite uniform (Fridland & Bartlett,
2006). In general, compared to the divergent regional trends affecting front
vowels, the pressure toward back vowel fronting is more globally affecting
American dialects.
Based on these convergent and divergent tendencies across regional dialects, the
current study measures how these differences in regional dialect experience affect
how listeners perceive and evaluate regionally unique vs. regionally shared vowel
shifts. Certainly, a number of studies have suggested that speaker dialect affects
vowel categorization tasks. Early studies such as Willis (1972) and Janson
(1986) showed that dialect experience indeed influences speakers’ perception of
vowel categories. Janson found differences in /a:/-/o:/ phoneme classification
across two generations within the same community in Sweden, suggesting that
changes in progress could be discerned by perceptual tests. Willis found
differences in phoneme categorization for speakers from Buffalo, NY, compared
to speakers from Fort Erie, ON, for the /e/-/æ/, /æ/-/ /, and /o/-/u/
oppositions. The current study, rather than focusing on whether phoneme
classification is affected by participation in regional shifts, attempts to get at the
social meaning, if any, such shifts carry and whether there is a difference in the
recognition of global (shared) vs. localized changes.
This research joins a growing body of work examining the social meaning
assigned to vowel differences. Niedzielski (1999) used recordings of a Detroit
speaker’s raised /aw/ variant that she alternately introduced to raters as
produced by either a Canadian or Michigan speaker to show that speech
stereotypes play a role in the selection of the /aw/ variant listeners believed they
had heard. Clopper and Pisoni (2004) used sentences solicited from young men
from regionally diverse dialect areas of the U.S. to measure how well listeners
were able to identify the sentence producer’s regional affiliation. They then
acoustically measured how certain stereotypical regional variants such as /ay/
monophthongization were correlated with the sentences rated most accurately by
region, suggesting that those variants are most responsible for triggering regional
association. These studies suggest that speakers do use phonetic cues in

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


70 通过perception解释social meaning VA L E R I E F R I D L A N D

classification of sounds and social traits, revealing that we are perceptually aware
of this phonetic variation and its related social embedding and use it to organize
our experience. Labov and Ash (1997) examined the role of regional sound
changes on listener identification of Southern vowels using real speech data from
young speakers in Birmingham, AL. In the controlled vowel identification and
gating tests, they found speakers surprisingly bad at correctly identifying what
words they heard, even when they were from the same dialect as the target
speaker. However, Birmingham listeners did show greater ability than nonlocals
to change their initial misidentification using context as cue. Such results
suggested their possible acoustic range for a particular vowel was more fluid,
having been exposed to advanced and less-advanced tokens of vowels involved
in the sound change. In addition to those cited here, Thomas (2002) provides a
comprehensive overview of other recent studies that explored the importance of
perception to sociolinguistic research.
In general, this previous work attempts to better understand how vowel
differences become social resources for speakers and the role such social factors
may play in the process of language change. These types of studies recognize
the importance of examining not just speech production but speech perception
from a sociolinguistic perspective and point to an emerging and promising area
of research. The move toward integrating the study of perception with work
measuring the productive patterns found in local communities might begin to
successfully address some of the larger questions about the origin and diffusion
of sound change that have preoccupied sociolinguists over the last several
decades. Within this context, the current paper explores how perception research
might be used to provide insight into the formation of the meaning-making
processes used by speakers to approach the task of unraveling the social import
of the speech surrounding them.

PREVIOUS WORK IN MEMPHIS

The earlier research in Memphis consisted of three parts: a production study, a


perception study, and a folk dialectology study. The production study (Fridland,
2000, 2001, 2003a, 2003b) measured the participation of Memphians in the
SVS, the series of phonetic changes affecting the high and mid-front vowels in
Southern speech. For Memphians, this shift is most prominently realized in the
/ey/ and /e/ classes, with complete reversal the norm for most speakers, both
White and Black. High front vowels retain their traditional relationship, although
they are much more closely aligned than is found elsewhere. While /ey/
centralization remained fairly robust in the Memphis system for all speakers, /e/
peripheralization was much less vigorous, particularly among younger speakers.
A later study (Fridland & Bartlett, 2006) examined the degree to which
back vowel fronting had affected the Memphis system. This research showed
an extreme shift present for Memphians in both the /uw/ and /u/ classes.
Shift in the mid-back /ow/ class was much less advanced, with younger

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 71

speakers showing significantly more shift in /ow/ than older speakers (F[32] ¼
4.69, p , .05). This research provided the description of community norms on
which the subsequent perception study was based.
The perception study (Fridland et al., 2004, 2005) was designed to determine the
degree to which Memphis speakers recognized these shifts occurring in their
regional vowel system as local variants. Results showed Memphians were more
likely to recognize different degrees of shift in the front vowels as signaling
increasing Southernness than different degrees of shift in the back vowels, even
though Southerners have been reported to be quite advanced in back vowel fronting
(Labov, 1994, 2000; Thomas, 2001). Vowel length and diphthongization also
appeared to be a factor, with short front vowels showing the lowest amount of
perceptual recognition. For example, the /i/and /e/ tokens were the only token
types to show accuracy rates less than chance. Memphians performed much better
on vowels with more complex and longer trajectories, suggesting that such
information was useful in interpreting the social information carried by vowel variants.
However, structural differences did not fully account for the ratings given to the
different vowel classes involved, as subjects did not consistently rate all long
vowels with the same degree of accuracy. Results also suggested that the
recognition of variants seemed to be tied to community vowel norms, with
listeners showing greater accuracy in recognizing shifts found uniquely as
Southern variants compared to shifts found in many U.S. dialects, such as
fronted high back vowels. Participants’ performance on the “Southernness” test
suggested that local norms are perceptually available to listeners, with acoustic
and structural effects mediating the recognition of such norms. A follow-up
study (Fridland et al., 2005) measured how Memphians perceived these same
Southern-shifted vs. nonshifted variants on pleasantness and education scales.
Results from this task suggested that the more the token had been shifted toward
SVS norms, the lower the token was ranked on education and pleasantness.
This research seems to suggest that local dialect experience did play a role in how
participants perceived the vowel variants on a number of social dimensions. To
further assess the impact of productive dialect patterns on listeners’ recognition
and evaluation of local and nonlocal speech samples, this study administers the
same test outside the region, in an area that clearly involves different dialect
exposure than that in the South. Running the same studies in a region that does not
share some of these local changes can lend insight into whether this earlier study
really showed any local advantage in perception and how these shifted variants are
comparably evaluated by those not using them productively. Reno speakers, in
contrast to Memphis speakers, maintain the traditional alignment of the high and
mid-back classes, with /iy/and /ey/ remaining along the front periphery of vowel
space. (J. H., a young Reno native, provides a sample Reno system in Appendix I.)
Although not as advanced as most Memphians, Reno speakers share the tendency
toward back vowel fronting, most extensively in the /uw/ and /u/ classes. So, by
administering the perception test performed in Memphis in a region showing both
distinct and similar trends, we can examine how much local dialect experience
affected Memphians’ recognition of shifted vowel variants.

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


72 VA L E R I E F R I D L A N D

METHODOLOGY

The current study replicates the study performed in Memphis using the same
stimulus and testing materials. An abbreviated methodological description
follows. Full description of synthesis procedures and study methodology appears
in Fridland et al. (2004, 2005).
The perception study was administered to 230 subjects native to Reno, NV. As
the study participants were recruited at a university, the vast majority of participants
were between the ages of 18 and 25, similar to the Memphis administration. The
study was designed as a matched guise test using synthesized vowel tokens
manipulated roughly by intervals of 25 to 50 Hz on vowel height frequencies
(F1) and vowel front/back advancement frequencies (F2), respectively.
Participants were played two of these slightly altered versions of the same
monosyllabic word presented in a two-word set, with each member of the pair
having been altered along the first formant and/or second formant dimension.
Each token pair presented two versions of the same vowel in three possible
contrasts along a shifted Southern to non-Southern continuum: two Southern-
shifted variants, with one version exhibiting a greater degree of the same shift, one
Southern- shifted and one non-Southern variant, or two non-Southern variants,
with one variant more traditionally positioned than the other variant. Target values
for synthesized tokens were based on the frequency ranges for vowel trajectories
observed in earlier work on the realization of the SVS (Fridland, 2000, 2001,
2003a, 2003b) and on the mean frequency ranges provided for American speakers
by a number of different researchers, summarized in Kent and Read (2002).
After hearing the token pair, the participants were asked to determine which
pronunciation of the word was the most “Southern” sounding. In calculating the
results, participants were considered accurate in their selection of the most
Southern token when they selected the token exhibiting the most Southern shift
or the least non-Southern shift.
Following this part of the study, participants were asked to rate a subset of these
tokens on education and pleasantness scales. The /ey/, /e/, /uw/, and /ow/
tokens were selected out of the larger set used for the Southernness study, as
these tokens are the most commonly or uniquely affected vowels in the Southern
system. This test used a three-point scale from one (least) to three (most). Paired
comparisons t-tests were run within each regional sample and independent t-tests
were used to compare the results across the regional samples.

RESULTS AND DISCUSSION

Reno participants’ Southerness ratings


Table 1 presents the results by descending mean score within the Nevada sample.
Clearly, the /ey/, /e/, /ow/, and /iy/ classes are the most salient to Reno raters
as recognizable Southern markers. Although Reno participants performed
significantly better than chance in recognizing Southern variants of /u/ (t[230] ¼

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 73

TABLE 1. Reno raters’ descending


mean accuracy scores for
selecting the most Southern-
shifted token in each token pair

Nevada

/ey/ .85 (.19)


/E/ .71 (.25)
/ow/ .71 (.17)
/iy/ .64 (.17)
/U/ .60 (.25)
/uw/ .49 (.28)
/I/ .46 (.27)

Note: ( ) Indicates standard deviation.

6.17, p , .01), the /uw/ and /i/ classes were accurately rated just at or less than
chance, respectively. In general, Southern-sounding front vowels were more easily
recognized than Southern-sounding back vowels, with the exception of the /ow/
class. The increased awareness of shifted variants in this class, however, is not
surprising considering such shift, although it does mildly affect the Nevada
system, seems to be less advanced than shift in either high back vowel (and, as
will be discussed later, similar /ow/ salience was found for Memphis speakers).
Although /uw/ and /u/ shift in an analogical pattern (Labov, 1994), it is less
clear that shift in /ow/ involves the same underlying drift mechanism, and it is
predominately found in younger speakers’ systems, suggesting an incipient shift
compared to that found in the high back vowels. In addition, close proximity to
California, where /ow/ fronting carries an association with “Valley girl” talk may
also make /ow/ shift more salient than the more progressed and widely
distributed shifts affecting other back vowels. In this task directing their attention
to perceiving Southern speech patterns, any marked shift, as /ow/ appears to be,
may be heard as Southern to Reno listeners. On the other hand, shift in the /uw/
and /u/ class is widespread productively in the Reno area, explaining the
difficulty in picking out Southern- vs. non-Southern-soundig variants.
Why the /i/ shift was difficult, however, is not clear, particularly because all the
other front vowels were quite salient for participants. Vowel trajectory information
does not appear to make much of a difference in terms of Nevadans’ ability to
recognized Southern-sounding variants, as there was no real pattern to the results
by vowel length or gliding. If short vowels were somehow less accessible to
listeners, then /e/ and /u/, which show accuracy rates significantly better than
chance, t(230) ¼ 12.51, p , .01 and t(230) ¼ 6.17, p , .01, respectively, should
also show similarly low accuracy rates. These results suggest that, in fact, Reno
respondents are fairly accurate in recognizing Southern-sounding vowel
realizations, regardless of vowel complexity. Moreover, these vowel distinctions are
particularly salient for front vowels and the /ow/ vowel. What Nevadans are using

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


74 VA L E R I E F R I D L A N D

TABLE 2. Mean scores by token pair shift-comparison type

Shift-comparison involved Mean N Std. deviation

Southern to Southern (S2S) .779 231 .194


Southern to non-Southern (S2N) .619 231 .107
Non-Southern to non-Southern (N2N) .634 231 .318

as an internal guide in making Southern judgments is less clear. It is doubtful their


ratings are a result of much exposure to Southern speakers, as few of the subjects
report much travel experience outside the West. In addition, no significant
differences were found in rating trends between participants who had traveled
outside of the West and those who had not. Instead of Southern exposure, Reno
participants must be relying on some type of internal model of vowel space that
maps the fit with the acoustic vowel position most familiar to them, that is, that
found in their own dialect. Vowel variants that sound closest to their own would be
a less likely candidate for a Southern token than variants that are less familiar.
When looking at the results (Table 2) in terms of what type of contrast was
involved in the token pair, Reno raters appeared to be better at selecting the most
Southern-shifted token in token pairs where the contrast between the two
pronunciations was a matter of degree of Southern shifting (e.g., a mild Southern
shifting compared to more extensive Southern shifting). In fact, they performed
significantly better when such pairs where involved than when the token pairs
involved either Southern to non-Southern pronunciations or non-Southern to
non-Southern pronunciations (Table 3). Such results are interesting as it would
seem that token pairs involving Southern-shifted tokens compared to nonshifted
tokens would make the Southern articulation more salient. However, in contrast,
hearing two tokens exhibiting different relative degrees of the same shift actually
made it easier for Reno participants to pick up on the more shifted token. Such
results are not so surprising when formant normalization effects such as those
found in Lagefoged and Broadbent (1957) are considered. In that study, the
researchers found that speakers were able to meaningfully perceive low-level
variation in vowel quality by using internally constructed default settings if
given no additional information to help them analyze the acoustic range used by
a speaker. However, when played a phonetically altered unambiguous sentence
before hearing a target sentence, participants seemed to analyze the acoustic

TABLE 3. T-tests comparing Reno raters accuracy by shift-


comparison type

t df Sig. (2-tailed)

Pair 1 S2S–S2N 11.376 230 .000


Pair 2 S2S–N2N 5.785 230 .000
Pair 3 S2N –N2N 2.652 230 Not sig.

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 75

TABLE 4. Reno education ratings by vowel class ad


regional shift type

Mean N Std. deviation

EYSOED 1.390 231 .374


EYNOED 1.967 231 .470
ESOED 1.787 231 .458
ENOED 1.857 231 .503
UWSOED 2.303 231 .451
UWNOED 2.272 231 .421
OWSOED 2.050 231 .439
OWNOED 2.059 231 .362

range used for vowels exhibited in the priming sentence to determine the proper
vowel quality of the selected vowels in the target sentences. It appears that
something similar may be at work here, with participants using the relative
degrees of shift to determine more and less extreme vowel differences.

Reno participants’ education and pleasantness ratings


Table 4 shows the Reno results for rating perceived speaker education levels for a
subset of the tokens used in the Southerness test (/ey/, /E , /uw/, /ow/). Southern
shifted tokens (SOED) receive lower ratings on education, but only when they
involved the front vowel classes. When Southern shifted back vowels were
compared to their nonshifted (NOED) back token counterparts, education ratings
did not seem to be affected. For the front vowels, however, education ratings
were pulled down by Southern shifting. This rating difference was significant for
the /ey/ class (t[230] ¼ 17.11, p , .01) and near significance for the /e/ class
(t[230] ¼ 1.95, p , .053), but neither back vowel class showed any significant
difference in ratings between Southern- and non-Southern-shifted tokens.
Compared to education ratings, pleasantness ratings were more consistent across
vowel class by regional variant type (Table 5). Reno raters found all Southern-shifted
tokens, regardless of front or back association, significantly less pleasant than

TABLE 5. Mean pleasantness ratings by vowel class and


vowel shift type

Mean N Std. deviation

EYSOPL 1.581 231 .543


EYNOPL 1.906 231 .485
ESOPL 1.726 230 .441
ENOPL 1.792 231 .521
UWSOPL 1.997 231 .522
UWNOPL 2.128 231 .466
OWSOPL 1.945 231 .476
OWNOPL 2.200 231 .502

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


76 VA L E R I E F R I D L A N D

TABLE 6. Mean pleasantness ratings by vowel subsystem


and regional shift type

Mean N Std. deviation

FRNTSOPL 1.641 230 .404


FRNTNOPL 1.847 230 .364
BACKSOPL 1.967 230 .381
BACKNOPL 2.164 231 .391

nonshifted tokens, although they still found back vowels significantly more pleasant
than front vowels, regardless of shift type (Tables 6 and 7). As can be seen in Table 8,
this difference in ratings was also significant within vowel classes with the exception
of the shifted /e/ tokens, which, although not significant, still received lower mean
scores for Southern-shifted variants. Clearly, less familiar or nonlocal variants are not
as well received on solidarity scales regardless of how they fared on competence
scales. In fact, comparing mean scores in Tables 4 and 5, most tokens received
lower pleasantness scores than education scores regardless of shift type,
suggesting that even the less educated sounding tokens did not receive the benefit
often accrued by nonstandard variants on solidarity scales.

Memphis and Reno participant comparison for Southerness


ratings
Table 9 compares descending mean scores for accuracy for the Memphis and Reno
samples, respectively. Memphis participants were accurate at identifying which

TABLE 7. Comparison of pleasantness ratings by vowel subsystem


and regional shift type

t df Sig. (2-tailed)

FRNTSOPL–FRNTNOPL 27.420 229 .000


FRNTSOPL–BACKSOPL 28.524 229 .000
FRNTNOPL– BACKNOPL 29.827 230 .000
BACKSOPL–BACKNOPL 27.948 230 .000

TABLE 8. Comparison of pleasantness ratings by vowel class and


regional shift type

t df Sig. (2-tailed)

EYSOPL– EYNOPL 28.589 230 .000


ESOPL– ENOPL 21.771 229 .078
UWSOPL– UWNOPL 23.812 230 .000
OWSOPL–OWNOPL 28.084 230 .000

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 77

TABLE 9. Descending Southernness test mean accuracy


scores for Memphis run compared to Reno run

Memphis group Reno group

Vowel class Mean Vowel class Mean

/ey/ .84 (.21) /ey/ .85 (.19)


/iy/ .67 (.19) /E/ .71 (.25)
/ow/ .62 (.23) /ow/ .71 (.19)
/uw/ .54 (.28) /iy/ .64 (.17)
/U/ .51 (.24) /U/ .60 (.25)
/E/ .49 (.27) /uw/ .49 (.28)
/I/ .39 (.23) /I/ .46 (.27)

word was the more Southern-shifted token at an overall rate of 58% (t[140] ¼
10.02, p , .01), significantly better than chance, with accuracy rates for
individual vowels varying from 84% to 39%. Surprisingly, Reno participants
were significantly better overall than Memphians at identifying the more
Southern-shifted token in the test as a whole (t[437] ¼ 5.51, p , .01),
performing at an overall rate of 63% and they also had a slightly smaller (and
higher) range for rating individual vowel classes, from 85% to 46%.
Looking at their comparative results, some similarities and contrasts emerged in
the two regional samples’ rating trends. First, the vowel subsystem appears to make a
difference in rating accuracy for both groups. As in Memphis, Reno raters also
appeared to respond to the salience of front vowels as Southern markers, with
greater accuracy identifying Southern variants in the front subsystem compared to
the back subsystem as a group (t[230] ¼ 4.04, p , .01). Both the Memphis and
Reno participants performed best when hearing the /ey/ vowel pairs, with both
groups showing an accuracy rate over 80%. In addition, the short front vowel /i/
was the most difficult for listeners in both groups, where they both performed
significantly below chance, t[212] ¼ 5.67, p , .01 and t[230] ¼ 2.51, p , .05
for Memphis and Reno raters, respectively. The two regional groups were
statistically identical in determining which member of the two-token pair was
more Southern for the /iy/, /ey/, and /uw/ classes, but were significantly
different in which token was heard as more Southern in the /i/, /e/, /u/, and
/ow/ classes. Surprisingly, it was Nevada speakers that were significantly better
at picking out the most Southern variant in all four of these classes (Table 10).

TABLE 10. Memphis to Reno comparisons

Comparative accuracy in selecting SVS shifted variants by vowel class

/I/ Southern accuracy t(442) ¼ 2.44, p , .05


/E/ Southern accuracy t(441) ¼ 28.507, p , .01
/U/ Southern accuracy t(439) ¼ 23.223, p , .01
/ow/ Southern accuracy t(439) ¼ 24.178, p , .01

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


78 VA L E R I E F R I D L A N D

Interestingly, most of these significant differences between the groups revolved


around short vowel classes. A strong influence on Memphians’ ability to select the
most Southern-shifted token appeared to be vowel length and gliding, with longer
more dynamic vowels serving more often as salient cues to a token’s potential for
making Southernness. Reno raters, on the other hand, did not appear to show more
sensitivity to long tense vowels, as they showed comparatively high rates of
accuracy for short vowels, with both /e/ (t[230] ¼ 12.51, p , .01) and /u/
ratings (t[230] ¼ 6.17, p , .01) significantly above chance and short vowel
accuracy rates dispersed across their ratings range. When comparing the
Southern pronunciations, Reno raters seem to be using their own vowel formant
ranges as the default or prototype, making F1 and F2 positions outside that
range more salient, rather than length and gliding. Such a pattern-matching
approach to speech perception has been widely discussed, particularly by
Massaro (1987), as the Fuzzy Logical Model of Perception (FLMP). Particularly
important in such a model is the integration of multiple cues that are used to
arrive at a decision in the processing and categorizing of speech input, with all
features contributing to the ultimate outcome but to differing degrees. Such a
model that assumes that a variety of factors influence participants’ input
processing suggests that such cues or features could have relative rankings of
importance to different participants, depending on experience. Thus, the reliance
of Memphians but less so of Reno raters on length and gliding, common features
in Southern speech, becomes not so surprising. Reno participants, with little first
hand experience, rely more on F1 and F2 difference from their own dialect
(which presumably provides their prototype experience).
Clearly, Memphis and Reno raters are using different cues in interpreting what
sounds Southern and what does not, with vowel dynamics a much more important
cue for Memphians. Still, for both groups, formant differences in the front tense
vowel subsystem are the most marked for Southern/non-Southern association,
suggesting that these vowels in particular serve as indicators for Southernness
for groups inside and outside the South. Shift in the /ow/ class, occurring less
often and predominately in young speakers’ systems, also carries Southern (or
nonstandard) salience. On the other hand, fronted high back vowels, a shift
shared by both the Memphis and Reno speakers, are less often recognized as
markers of Southernness by both groups than these other vowels. However, shift
in the short front subsystem only signals a Southernness association for
Nevadans. Memphians’ recognition of Southern variants appears to occur only
for tense vowels, suggesting that, for Memphians, duration and gliding make
vowels relevant as Southern variants, leading them to look for salient formant
differences.
Looking at their comparative performance based on token-pair contrast type,
both groups were more accurate in selecting the most Southern pronunciation
when the token-pair contrast involved mild/moderate Southern shift to more
extreme shift (Table 11). For all contrast types, Reno participants performed
significantly better than Memphians (Table 12), echoing their generally more
accurate ratings by vowel class discussed above. These results lead to two

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 79

TABLE 11. Memphis to Reno comparisons for overall shift-type


categories

Token-pair contrast type Memphis Nevada

Southern to Southern Shifts .67 (.23) .78 (.19)


Southern to non-Southern .58 (.12) .62 (.10)
Non-Southern to non-Southern .52 (.31) .63 (.32)

TABLE 12. T-test comparisons for Memphis to Reno results

Contrast type Memphis/Reno Comparison

Southern to Southern shifts t(442) ¼ 5.60, p , .01


Southern to non-Southern t(438) ¼ 3.61, p , .01
Non-Southern to non-Southern t(441) ¼ 3.83, p , .01

observations: (a) formant normalization effects are at work here regardless of


regional background and (b) having default settings for the vowels that differed
from the test dialect (e.g., Western norms) aid speakers in recognizing formant
contrast. In other words, Nevadans perform better at selecting the most
Southern-shifted vowels because they can use distance from their own vowel
system to help guide them. This strategy results in Reno participants’ better
overall performance, particularly when rating the short vowel subsystem. Using
their own system as a guide, Reno participants were fairly good at picking out
the contrasting Southern variant in short and long vowels alike, whereas
Memphians, using local norms for Southernness as a guide, performed well only
on the long tense vowels, in which shifts were more often heard as Southern
features.

Comparison of education and pleasantness ratings


Tables 13 and 14 provide increasing mean scores by vowel class for education and
pleasantness ratings for the Memphis and Reno groups, respectively. The mean
education score ordering was identical for both Memphis and Reno raters, with
/ey/ class tokens sounding the least educated and /uw/ class tokens sounding

TABLE 13. Increasing education mean scores by vowel class

Memphis means Reno means

ey 1.61 (.32) ey 1.58 (.33)


E 1.88 (.41) E 1.81 (.40)
ow 2.02 (.30) ow 2.05 (.33)
uw 2.29 (.39) uw 2.28 (.36)

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


80 VA L E R I E F R I D L A N D

TABLE 14. Increasing pleasant mean scores by vowel class

Memphis means Reno means

E 1.72 (.43) ey 1.68 (.45)


ey 1.73 (.45) E 1.75 (.40)
ow 1.92 (.38) ow 2.07 (.42)
uw 1.98 (.44) uw 2.08 (.42)

the most educated. Front vowels were clearly heard as less educated than back
vowels. There was more contrast between the groups in terms of pleasantness
scores, but both groups still found front vowels less pleasant than back vowels.

CONCLUSION

Based on the findings reported above, it appears the salience of Southern speech
and Southern speech markers has traveled far outside the South. Both Reno and
Memphis raters show higher rates of recognition for Southern formant positions
when the vowels involved the long front vowels and the mid-back vowel /ow/.
The high back vowels showed lower rates of Southernness accuracy in both the
Reno and Memphis samples. Regional uniqueness seems to be at work here as
the vowel shifts found in the South but not shared by Reno speakers showed
higher recognition rates than shifts in the /uw/ and /u/ classes that are more
widespread.
Although raters in both regions showed several of the same trends in terms of
what sounded most Southern, Memphis and Reno raters appear to use different
criteria to help them judge the Southernness of tokens. For Reno raters, distance
from their own system seems to determine which classes were most salient, and
vowel dynamics did not appear to be much of a factor. On the other hand, for
Southerners, vowel trajectory appeared to play a very important role in denoting
Southernness. It is not that longer more dynamic vowels are necessarily more
likely to shift, as short front vowels also have shifted variants, but that shifts
affecting tense vowels are more likely to be attended to in terms of this social
marking, and shifts affecting short vowels appear off the social radar. Clearly,
dialect stereotypes that recognize a Southern drawl feature make reference to this
perceptual acuity for longer diphthongal vowels. What is surprising is that Reno
raters and Memphis raters do not rely on vowel dynamics to the same degree and
that Reno raters are actually more attuned to formant shifts that are unique to the
South, as evidenced by their greater accuracy in placing Southern shifted /e/
and /i/ tokens. In fact, their overall greater accuracy in this task, regardless of
vowel class, suggests that, when forced to base decisions on internalized norms,
speakers perform better when the dialects involved are different from their own.
Based on the results of this study, it appears that a number of different factors
both mechanical (vowel dynamics, type of formant change) and social (shift

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 81

uniqueness, dialectal difference) influence the meaningfulness of vowel change


with the aspects most salient varying by community. In addition, local dialect
experience does not necessarily increase speakers’ ability to recognize their own
productive norms. Such results suggest that researchers must be careful when
making generalized claims about what features are salient markers of various
social facts such as gender, region, or ethnicity and the degree to which speakers
attend to them. Clearly, community and speaker group differences go well
beyond altering only productive resources to in fact even altering what linguistic
cues are relied on in making perceptual distinctions on speech input.

REFERENCES
Ash, Sharon. (1996). Freedom of movement: /uw/-fronting in the Midwest. In J. Arnold et al. (eds.),
Sociolinguistic variation: Data, theory and analysis: Selected papers from NWAV 23 at Stanford.
Stanford: CSLI Publications. 3– 25.
Clarke, Sandra, Elms, Ford, & Amani, Youssef. (1995). The third dialect of English: Some Canadian
evidence. Language Variation and Change 7:209–228.
Clopper, C. G., & Pisoni, D. B. (2004). Some acoustic cues for the perceptual categorization of
American English regional dialects. Journal of Phonetics, 32, 111 –140.
Dipaolo, Marianna, & Faber, Alice. (1990). Phonation differences and the phonetic content of the tense-
lax contrast in Utah English. Language Variation and Change 2:155– 204.
. (1995). The discriminability of nearly merged sounds. Language Variation and Change
7:35 –78.
Eckert, Penelope. (1988). Adolescent social structures and the spread of linguistic change. Language in
Society 17:183– 208.
. (2000). Linguistic Variation as Social Practice: The linguistic construction of identity in Belten
High. Malden, MA & Oxford, UK: Blackwell.
Evans, Betsy. (2001). Dialect contact and the Northern Cities Shift in Ypsilanti, Michigan. Dissertation:
Michigan State University.
Feagin, Crawford. (1986). More evidence for vowel change in the South. In David Sankoff (ed.),
Diversity and Diachrony. Amsterdam & Philadelphia: John Benjamins. 83–95.
Fridland, Valerie. (1998). The Southern Vowel Shift: Linguistic and social factors. Ph.D. dissertation,
Michigan State University.
. (2000). The Southern Vowel Shift in Memphis, TN. Language Variation and Change
11:267– 285.
. (2001). Social factors in the Southern Shift: Gender, age and class. Journal of Sociolinguistics
5:233–253.
. (2003a). Network strength and the realization of the Southern Vowel Shift among African-
Americans in Memphis, TN. American Speech 78:3–30.
. (2003b). Tide, tied and tight: The expansion of /ai/ monopthongization in African-American
and European-American speech in Memphis, TN. Journal of Sociolinguistics 7:279–298.
. (2004). The spread of the cot/caught merger in the speech of Memphians: An ethnolinguistic
marker? Paper presented at LAVIS III (Language and Variation in the South II) Conference.
Tuscaloosa, AL, October 10–13, 2004 (NWAV).
Fridland, Valerie, & Bartlett, Kathryn. (2006). The social and linguistic conditioning of back vowel
fronting across ethnic groups in Memphis, TN. English Language and Linguistics 10:1– 22.
Fridland, Valerie, Bartlett, Kathryn, & Kreuz, Roger. (2004). Do you hear what I hear? Experimental
measurement of the perceptual salience of acoustically manipulated vowel variants by Southern
speakers in Memphis, TN. Language Variation and Change 16:1– 16.
. (2005). Making sense of variation: Pleasantness and education ratings of regional vowel
variants. American Speech 80:366– 387.
Gordon, Matthew. (1997). Urban Sound Change Beyond City Limits: The Spread of the Northern Cities
Shift in Michigan. Ph.D. dissertation, University of Michigan.
Hagiwara, R. (1997). Dialect variation and formant frequency: The American English vowels revisited.
Journal of the Acoustical Society of America 102:655– 658.

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


82 VA L E R I E F R I D L A N D

Janson, Tore. (1986). Sound change in perception: An experiment. In John Ohala and Jeri Jaeger (eds.),
Experimental Phonology. Orlando: Academic Press. 253– 260.
Kent, Ray, & Charles, Read. (2002). The Acoustic Analysis of Speech. Canada: Delmar.
Labov, William. (1991). The three dialects of English. In Penelope, Eckert (ed.), New Ways of Analyzing
Variation. New York: Academic Press. 1– 44.
. (1994). Principles of linguistic change: Internal factors. Malden, MA: Blackwell.
. (2000). Principles of linguistic change: Social factors. Malden, MA: Blackwell.
Labov, William and Sharon Ash. (1997). Understanding Birmingham. In Cynthia, Bernstein, Thomas,
Nunnally and Robin, Sabino (eds.), Language Variety in the South Revisited. Tuscalossa, Alabama:
University of Alabama Press. 508– 573.
Labov, William, Ash, Sharon, & Boberg, Charles. (2006). Atlas of North American English. Berlin:
Mouton de Gruyter.
Ladefoged, Peter, & Broadbent, D. E. (1957). Information conveyed by vowels. Journal of the
Acoustical Society of America 29:98–104.
Luthin, Herbert. (1987). The story of California (ow): The coming-of-age of English in California. In
Denning, K. M. et al. (eds.), Variation in language: NWAV-XV at Stanford (Proceedings of the 15
Annual New Ways of Analyzing Variation Conference). Stanford: Department of Linguistics,
Stanford University. 312–324.
Massaro, Dominic W. (1987). Speech perception by ear and eye: A paradigm for psychological inquiry.
Hillsdale, NJ: Lawrence Erlbaum Associates Publishers.
Niedzielski, Nancy. (1999). The effect of social information on the perception of sociolinguistic
variables. Journal of Language and Social Psychology 18:62– 85.
Thomas, Erik. (1997). A Compendium of vowel plots. A publication of the North Carolina Language and
Life Project. Raleigh, NC: Department of English, North Carolina State University.
. (2001). An acoustic analysis of vowel variation in New World English. Publication of the
American Dialect Society 85, Durham, NC: Duke University Press.
. (2002). Sociophonetic applications of speech perception experiments. Americn Speech
77:115–147.
Willis, Clodius. (1972). Perception of vowel phonemes in Fort Erie, Ontario, Canada, and Buffalo,
New York: An application of synthetic vowel categorization tests to dialectology. Journal of
Speech and Hearing Research 15:246– 255.

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press


REGIONAL DIFFERENCES IN PERCEIVING VOWEL TOKENS 83

APPENDIX I
Sample Reno Speaker system (J. H., a younger White female)

https://doi.org/10.1017/S0954394508000069 Published online by Cambridge University Press

You might also like