Skip to navigation – Site map

HomeAnglophonia30Some sociolinguistic evaluations ...

Some sociolinguistic evaluations of performances of the California Vowel Shift: a matched-guise study

Pierre Habasque

Abstracts

Sociolinguistic studies (Eckert: 2000; Labov: 1966; 1972) have shown that some allophones can carry social meaning. The use of a given allophone can indeed inform about the age, gender, or social class of a speaker. The aim of this paper is to report on the findings of a study investigating some social meanings attributed to performances of the California Vowel Shift (CVS), a chain shift that affects almost all vowels of American English, which undergo a counterclockwise rotation. The CVS is regularly used in parodic performances on television to portray clueless, vapid, shallow characters, which suggests that this phenomenon might be associated with potentially stigmatizing social meanings. To test this hypothesis, a quantitative perceptual study was conducted with 123 students of California State University, Northridge. They were asked to rate four voices: that of one male and one female speaker pronouncing the same sentence with their standard accent and while performing the CVS. The rating of the voices by participants was done with a seven-point semantic differential scale to test for the perception of six variables. Results show that the perception of friendliness, pleasantness, education, and competence is heavily affected by the presence of the performance of the CVS, suggesting that performances of the chain shift are indeed given social meanings by participants. The perception of the other variables, age and gender, do not seem to be significantly influenced by the performances, however. It is argued that the acoustic properties of some of the vowels of this chain shift may index stigmatizing traits for the participants. Four hypotheses are explored to explain these results: the idea that participants may share a common ideology regarding the CVS because of parodic media representations they may have been in contact with, the principle of sound iconicity, a “linguistic uncanny valley,” or a form of linguistic misogyny.

Top of page

Full text

1. Introduction

1.1 Phonetics and Social Meaning

  • 1 Similarly, the use of standard” (non-allophonic) phonemes can also inform about the speakers back (...)

1Sociolinguistic studies have shown that some phonetic realizations of a phoneme can carry social meaning, which “[…] accrues specifically to concrete sounds – to phonetic elements – and not to the phonological structures in which those sounds participate.” (Eckert & Labov: 2017, 2). It has indeed been shown that the use of an allophone can inform the hearer about, for example, the age, gender, race or social class of a speaker1 (Coates: 2016; Eckert: 2000; Labov: 1966; 1972; 1990; Purnell et al.: 1999).

2Hearers may also ascribe more arbitrary social meanings to phones, which can be embraced, ignored or stigmatized. The relation between phonetics and social meaning depends on phonological expectations. Eckert & Labov (2017, 15) give two examples of this phenomenon:

People frequently say, in rejecting a certain social variant, that it ‘sounds so ugly.’ However, it is not the sound that is so evaluated, but rather the use of that sound as the particular allophone representing a certain phoneme. Thus, the use of high ingliding [i:ə] in man or bad is stigmatized in the Middle Atlantic States, but no one reacts in the same way to the use of the same sound in idea. The use of a medial voiced flap in brother as [brʌɾɚ] is widely rejected as uneducated and vulgar speech, but the same sound in utter is accepted as normal or as a regional variant.

3The goal of this paper is to explore the meaning attributed to the realization of phones by studying arbitrary social perceptions of performances. Since “in English, continuous variation in the phonetic realization of vowel allophones is the most heavily employed resource for the construction of social meaning” (Eckert & Labov: 2017, 15), it was decided to focus on a chain shift involving multiple vowels, which is the reason why the California Vowel Shift was chosen for this study.

1.2 The California Vowel Shift (CVS) and its Representations

4The California Vowel Shift (henceforth: CVS) was first studied in the late 1980s and has since been thoroughly described. It affects almost all vowels of American English (figure 1), most of which undergo a counterclockwise rotation (Podesva: 2011). More specifically, the open(-mid) back vowels [ɔ] and [ɑ] tend to merge (Hall-Lew: 2009; Kennedy & Grama: 2012). Other vowels are fronted, including the (near)close back vowels [ʊ] and [u] (Hagiwara: 1997; Hall-Lew: 2009; Hinton et al.: 1987; Podesva: 2011), which can merge when followed by the phoneme /l/ (Guenter: 2000), as well as the open(-mid) back vowel [ʌ] (Hinton et al.: 1987; Hagiwara: 2006). The vowel [ʊ] may also diphthongize into [ɪʊ], and the diphthong [oʊ] may triphthongize into [ɛoʊ] (Eckert: 2008a). The front vowels [ɪ], [ɛ] and [æ] lower and move back (Hagiwara: 1997; Eckert: 2008a). These vowels may be affected by adjacent nasal consonants: when followed by the velar nasal [ŋ], [ɪ] closes and moves forward, the same happens with [æ] when it is followed by a nasal consonant (Holland: 2014; D’Onofrio: 2015a), and Kennedy & Grama (2012) state the same phenomenon occurs with the vowel [ɛ].

5In terms of the relationship between vowel movements, the CVS is primarily a push chain (Kennedy & Grama: 2012, 51):

[…] the chain shift itself is not a pull chain in progress initiated by the merging of lot and thought, but a push chain precipitated by the lowering of kit. Due to the lowering of kit in the vowel space, dress and subsequently trap are also incited to lower and retract. The final part of the chain is then the retraction of lot, which, given the current data set, does not seem to be fully instantiated in all of our speakers.

Figure 1: The California Vowel Shift(adapted from Eckert, 2008a)

Figure 1: The California Vowel Shift (adapted from Eckert, 2008a)

6These shifts are not necessarily specific to California: Labov et al. (2006) note that the [ɔ]/[ɑ] merger occurs in other western US accents for example. It is the co-occurrence of these vowels shifts that makes the CVS unique (Shafer: 2012; Holland: 2014).

  • 2 More specifically, Hinton et al. (1987) found that [u], [ʊ] and [o] were more front and less rou (...)

7In terms of social distribution, the CVS seems to be favored by younger speakers2 (Hinton et al.: 1987; Kennedy & Grama: 2012; Wolfram & Schilling: 2006) who also tend to live in urban areas (Hinton et al.: 1987; Podesva: 2015). Features of the chain shift are also particularly prevalent in female speakers (Hinton et al.: 1987; Kennedy & Grama: 2012; Labov et al.: 2006; Holland: 2014), which is actually typical for chain shifts (Labov: 1972).

  • 3 The term “persona” is meant to be understood in line with Cooper’s definition (1999, 124): Pers (...)
  • 4 Though the term used to point to a Californian stereotype (“Valley” originally referred to the S (...)
  • 5 It should be noted that the CVS can paradoxically be associated to an educated, formal type of s (...)

8Specific vowel movements of the CVS such as [æ] backing (D’Onofrio: 2016), or the chain shift as a whole (Villarreal: 2016) may trigger various social meanings, including, unsurprisingly, being Californian. It can also be associated to Californian personae3, including surfers (Kennedy & Grama: 2012) and “Valley Girls4” (Bucholtz et al.: 2007; D’Onofrio: 2015a; D’Onofrio: 2015b; D’Onofrio: 2016; Eckert: 2008a; Eckert & Mendoza-Denton: 2006; Hinton et al.: 1987; Podesva: 2011; Villarreal: 2016; Wolfram & Schilling: 2006). This term was popularized by the eponymous 1982 hit song by Frank and Moon Zappa in which the latter performs a character, the Valley Girl, a stereotypical white urban teenage girl who is affluent, privileged, and somewhat vacuous and idiotic5. The CVS can also index a fun, laid-back and carefree attitude (Podesva: 2011), and a “drama queen” or “divapersona (Eckert: 2008b; Podesva: 2011).

9It has also regularly been used in mainstream media to represent and perform (stereo)typical Californian personae, or characters who are supposed to be Californians, similarly to Moon Zappa’s character. Research papers on the CVS almost never fail to mention this. For example, Pratt & D’Onofrio (2017) note that the CVS is heavily used and caricatured (to the point that characters can sometimes barely be understood) on Saturday Night Live’s skit The Californians, and Podesva (2011) mentions movies and television shows in which Californian characters (namely Jeff Spicoli in Fast Times at Ridgemont High, Bill Preston & Ted Logan in Bill and Ted's Excellent Adventure, and Cher Horowitz in Clueless) use it as well.

10The folk-linguistic perception of the CVS is therefore likely to be tied to representations and performances of the chain shift, which may or may not depict actual speakers’ use of the shift.

1.3 The Present Study

  • 6 Only the gender of the speakers was taken into account in the present study, and not the gender (...)

11The aim of the present study is to test potential social meanings naive listeners may attribute to performances of the CVS by comparing them to that of a standard pronunciation. More specifically, three research questions are explored: Is the perception of performances of the CVS consistent among listeners? Is the performance of the CVS perceived as a prestige form or is it stigmatized? Do social meanings attributed to the CVS correlate with the perceived gender of the speaker6?

12The study was conducted on the campus of California State University, Northridge (henceforth: CSUN) with 123 participants. The methodology is similar to the one used in perceptual dialectology studies (such as Preston: 1986; 1996; Fought: 2002). Participants were asked to rate four voices (that of one male and one female speaker) pronouncing the same sentence with their standard accent and while performing the CVS. The two speakers were therefore asked to pronounce a sentence with their standard accent, and then to pronounce the same sentence while performing the CVS. The rating of the voices by participants was done with a seven-point semantic differential scale to test for the perception of six variables.

2 Material & Method

2.1 Context of the Study

13The study was conducted, as part of a PhD dissertation, between the end of April and early May 2019 with students attending CSUN. This university is situated in the heart of the San Fernando valley, North of the city of Los Angeles. Since all participants were living in California at the time of the study (most of them were also born in California), it was assumed that their perception of the CVS would not be heavily affected by its association with California but rather by other social meanings.

14According to the university (CSUN Profiles: 2018) 38,716 students attended the institution in 2018, including 34,900 undergraduates. 44.9% were male, 55.1% female and the average age was 23.6. The student population was also ethnically diverse: 50.8% of students identified as Latino/Latina, 22.1% as White, 10.4% as Asian, 4.6% as African-American.

  • 7 Each professor decided whether they would incentivize students.

15Participants were recruited with an oral announcement in class. Some students7 were incentivized to take part in the survey with extra credits. Volunteers were transferred a link to the survey, which was entirely administered online via Google Forms. It was possible to take part at any time during the two-week span of the study.

16The title of the survey, “How Do You Feel About Language Use,” was purposely vague as not to inform participants about what linguistic marker was tested. They were informed that the aim of the survey was to collect their subjective perception of voices, though. It was therefore clearly specified that there were no “right” or “wrong” answers. They were asked to report their gender, age, native language(s) and state of residence at the beginning of the survey.

2.2 Sampling

17Participants to the study were all attending one or more linguistics classes at CSUN, including for example: Language Acquisition; Sociolinguistics; Syntax; Language, Gender and Identity; Language and the Law; Language(s) in California. This potential selection bias is addressed in the “Limitations of the Study” section of this paper.

18In total, 157 people participated in the survey. Participation was only effective when the questionnaire was entirely filled out, which means that if students quit the survey before submitting results, all responses would be lost.

1934 participants were excluded from analysis since they reported English was not their first language. It was decided that taking the perception of these students into account would bias overall results since the performances of the CVS might not be indexical of social meanings to them. Therefore, only the responses of the remaining 123 participants are reported in this paper.

20The gender repartition of the sample is as follows: 80 women (65.0%), 39 men (31.7%), 4 participants did not define as either (3.3%). The median age is 23.6 (Med: 21; σ: 8.5). 104 people (84.6%) were born in California, 7 (5.7%) elsewhere in the US, 4 (3.3%) in another English country and 8 (6.4%) in a non-English-speaking country, though English is (one of) their native language(s). 38 participants (30.9%) say they only speak English and 85 (69.1%) say they have two native languages (including English).

21A sample of 123 participants represents 0.3% of the entire student population attending CSUN. It was therefore judged necessary to test for the representativity of the sample.

22Concerning gender, a Chi-Square goodness of fit test was performed to examine the relation between the proportion of males and females in the sample compared to the overall CSUN population X2 (3, N = 38835) = 7.03, p = .008. Compared to the entire CSUN student population, females are slightly over represented in the sample but the disparity is small (φc < .3) according to statistical standards (Cohen 1969: 79-81).

23The average age of the sample was the same as the average age of the entire population, suggesting the sample is representative as far as this variable is concerned.

24The proportion of students coming from California (as opposed to anywhere but California) in the sample is representative of the CSUN population as well. A Chi-Square goodness of fit test examined the relation between the proportion of students born inside/outside California in the sample compared to the origin of first-time CSUN freshmen X2 (3, N = 4900) = 44.92, p = < .001. No significant difference was found between the sample and the overall CSUN population.

25Since all answers given by participants whose native language is not English were excluded from analysis, the sample intentionally over-represents native English speakers.

26Finally, since this survey was conducted as part of a PhD in France, it had to comply with French laws regarding data collection, which forbid the collection of data regarding the race or ethnicity of participants. It is therefore impossible to assess if the sample is as ethnically diverse as the entire CSUN population.

2.3 Tasks8

  • 8 Besides the ones presented in this paper, the participants completed other tasks, which are not (...)
  • 9 Four other voices were also used, but since they do not pertain to the perception of the CVS, th (...)

27The perception of the CVS was tested using the matched-guise technique (Lambert et al.: 1960). Participants listened to two voices doing two different accents (i.e. there were 4 audio samples to listen to)9 pronouncing the same sentence. One voice was that of a male speaker with a standard General American accent, another was that of the same male speaker performing the CVS, and the two other voices were that of a single female speaker, with a General American accent and performing the CVS. The order in which the participants heard the voices was the same for all and was assigned randomly: voice #1 was the female voice with a General American accent; voice #2 was the male voice with a General American accent; voice #3 was the female voice performing the CVS; voice #4 was the male voice performing the CVS. Participants were instructed to wear headphones and to take part in the survey in a quiet environment.

28After listening to each voice, they were asked to rate” it; instructions were as follows:

After each recording, you are asked to report in the table below how you feel about the person you just heard. Don’t overthink this, just tick what feels right. You may want to hear the recording more than once.

29In order to rate their perception of a voice, participants used a semantic differential scale (Osgood et al.: 1957), a variant of the Likert scale (Likert: 1932) with two opposite adjectives on each end.

Figure 2: Semantic differential scale used for the “age” variable

Figure 2: Semantic differential scale used for the “age” variable

30An odd-numbered scale allowed participants to feel neutral about a variable by ticking number 4. This was explicitly stated to them (figure 2).

  • 10 These include: unreliable/trustworthy; self-involved/selfless; rural/urban; poor/rich; tense/rel (...)

31The perception of six variables was tested (figure 3). Originally, six others10 were to be tested as well, but the survey might have taken too much time to complete and participants might have felt discouraged, which is why it was decided to narrow it down to six variables.

Figure 3: Variables and corresponding pairs of adjectives used

Variable

Adjectives used

Age

Old

Young

Pleasantness

Unpleasant

Likable

Friendliness

Unfriendly

Friendly

Education

Uneducated

Educated

Gender

Masculine

Feminine

Competence

Incompetent

Competent

  • 11 Though results show that the perception of age is indeed not heavily affected by the presence of (...)

32The variables “pleasantness, friendliness, education, competence” were chosen to assess whether participants ascribe prestige or stigma to the CVS. The “gender” variable was used to test whether prestige or stigma correlate to the perception of masculinity or femininity. The “age” variable was meant to act as a control group as it was believed it would not correlate to either prestige11, stigma, or gender.

33No question was directly asked about the CVS, either before or after having taken the survey. Asking participants if they were biased against it before taking the survey could have helped determine whether their beliefs and their actual evaluations coincided, but participants might have been over-attentive to it. It was also decided not to ask questions about it after they had completed the tasks so as to make completing the survey as quick as possible. Since the CVS has not gained prominent folk-linguistic popularity, it was also believed that most participants would be unfamiliar with the phrase, even though they might have heard the shifted vowels.

2.4 Selection of voices

  • 12 Since the purpose of the performances of the CVS was that they be biased imitations of the chain (...)
  • 13 The audio quality of the female voice was not as clear as the male voice (see the “Limitations o (...)

34The voices the participants heard were the ones of a male speaker and a female speaker whose native language is English. Both were born in the United States and have a General American accent. They were therefore asked to pronounce a sentence with their regular accent, and then to perform the CVS while pronouncing the same sentence12. The male speaker was recorded with a Lavalier microphone; the female speaker recorded herself with a smartphone13.

  • 14 [æ] raising is due to the presence of the following nasal phone [n].
  • 15 [ɪ] fronting is due to the presence of the following velar nasal phone [ŋ].

35The sentence these two speakers recorded (and that the participants heard) is: “The wolf is standing in front of the den.” This sentence was chosen for two main reasons. First, it is semantically neutral, and is unlikely to trigger specific social meanings relating to age, location, social class, or gender. The second reason is that all vowels of this sentence (except [ə]) are affected by the CVS phonetically (figure 4). Vowel movements include [ʊ] centralizing (“wolf”), [ɪ] lowering (“is; in”), [ɛ] lowering (“den”), [ʌ] raising (“front), [æ] raising14 (“standing”), and [ɪ] fronting15 (“standing”).

Figure 4: Expected phonetic differences between General American and CVS vowels

The

wolf

is

stan

ding

in

front

of

the

.den.

General American

ə

ʊ

ɪ

æn

ɪŋ

ɪ

ʌ

ə

ə

ɛ

CVS

ə

ʌ

ɛ

ɛɪn

ɛ

ɛ

ə

ə

æ

36To have the best recordings possible for the participants to listen to, both speakers recorded multiple versions with their native General American accents, and multiple versions with their best impressions of the CVS.

37Since only four versions were needed for the study though (one General American version and one CVS version for the male speaker, and one General American version and one CVS version for the female speaker), a selection was made between competing recordings.

38This process had to take into account slight phonetic variations. For example, the male speaker, in both cases using his General American accent, sometimes over articulated, as in version A (which features two minor groups and an aspirated [tʰ] sound in “front”). After a few tries, the speaker produced more prototypical General American recordings, such as version B, which features a regressive assimilation (unvoiced [z̥] in “is”) and a voiced [t̬] sound in “front.

39A. [‖ðə  ˈwʊɫf  |  ɪz  ˈstæ̃n.dɪŋ  ɪn  ˈfrʌntʰ  əv  ðə  ˈdɛn‖]
B. [‖ðə  ˈwʊɫf ͜  ɪz̥  ˈstæ̃n.dɪŋ ͜  ɪn  ˈfrʌnt̬ ͜  əv  ðə  ˈdɛn‖]

40Since coarticulation may affect vowel production (Farnetani & Recasens: 2010), the quality of each vowel was not rigorously the same between these General American versions. It was therefore necessary to choose recordings which would serve as the General American voices (one male voice, one female voice) and recordings which would serve as the CVS performances.

41In order to do so, the first two formants of each vowel (henceforth: F1 and F2) were measured with Praat (Boersma & Weenink: 2017). Formant values were automatically calculated by Praat and were measured in the middle of each vowel. When using standard settings of the software, the values given by Praat did not always fit the standards given by Labov et al. (2006), which is an issue that has been previously mentioned (for example by Styler: 2017). Most reliable formant values were obtained by showing the first six formants for the male voice and the first three for the female voice. If obvious errors remained, the value was manually extracted from the spectrogram with the cursor.

  • 16 Here, the phrase “prototypical versions” refers to the versions of the recording in which vowels (...)

42To find the most prototypical versions16 (which would serve as the General American voices), all recordings were compared to each other with a statistical tool: the Z score. This tool measures the position of a data point compared to the population average. To compare all versions in terms of vowel quality, the value of F1 for each vowel was compared to the value of F1 for all other vowels of the same word in other recordings. For example, the value of F1 and F2 for each [ʊ] sound produced by the female speaker was compared to her average value of F1 and F2. The same was done with all other vowels. At the end of this process, the recording that was kept for the study, that which participants heard, features the most prototypical vowel formants for the female speaker.

  • 17 Vowel qualities being labeled “least prototypical” is the result of statistical analysis, not li (...)

43Similarly, the Z score helped determine which recordings had the least prototypical vowel qualities17 by comparing their first two formants. These recordings, which served as the CVS performances are necessarily imperfect in terms of vowel quality, though. Both speakers did their best to perform CVS vowels, but even after multiple tries, some vowels do not completely correspond to what is expected of the CVS (figure 5). For example, there is no [ɪ] fronting when followed by [ŋ], [ʌ] is not raised in the male voice and [æ] is not raised in either voices. However [ɪ] is lowered (when not followed by a nasal phone) in both voices, [ɛ] is clearly lowered in the male voice and [ʊ] centralized in both voices.

Figure 5: General American vowels (blue) and CVS vowels (red) of voices used in the study

Figure 5: General American vowels (blue) and CVS vowels (red) of voices used in the study

44It was impossible to directly manipulate formant values and retain a natural-sounding voice, so it was not possible to correct imperfections (this issue is also addressed in the “Limitations of the Study” section). Besides, it was not possible to use corpus data from actual speakers of the CVS as the present study required the very same sentence be pronounced with and without the CVS in the same recording conditions by two different speakers of different genders.

2.5 Statistical Testing

45For each variable, male voices were compared to each other, and female voices were compared to each other.

46The mean score voices received was sometimes extremely dissimilar, but was sometimes almost the same. The question was therefore: are the mean scores attributed to each voice statistically significantly different from each other, or are the two voices rated similarly despite slight differences in mean scores? In other words, are the differences between mean scores attributed to voices due to chance alone, or are they due to an actual difference in perception by participants?

47To answer this question, Student’s T tests (Gosset: 1908) were performed. This type of test helps determine the probability that two data sets (in this case scores attributed to two voices) belong to the same population by comparing their mean score and standard deviation.

48An alpha level of .05 was used for all tests discussed in this paper and results presented as statistically significant below have a statistical power above 0.8, calculated with G*Power (Faul et al.: 2007) with a post hoc analysis. The interpretation of effect sizes is drawn from Cohen (1969) and Sawilowsky (2009). All results were rounded to the nearest decimal point.

3 Results

3.1 Mean Scores, Statistical Significance and Effect Sizes

49The mean scores attributed to voices differ according to the gender of the speaker, the variable tested, and the presence of a performance of the CVS (figure 6). Some voices are rated similarly for some variables (for example “age”) while ratings differ drastically for others (“education, competence”). As has been mentioned, both the mean score and the standard deviation of data sets must be taken into account to assess whether differences between mean scores are due to chance alone or due to an actual difference in perception by participants.

Figure 6: Mean scores attributed to voices for each variable

Figure 6: Mean scores attributed to voices for each variable

50The male voice with a General American accent (henceforth: Male Gen. Am.) was compared to the male CVS performance (Male CVS) on the six variables. The same was done between the two female voices (Female Gen. Am., Female CVS). Therefore, in total, there could have been twelve differences (one difference per variable, times six variables, times two speakers). Ten statistically significant differences were found (figure 7), some of which have a larger effect size than others.

Figure 7: Interpretation of statistical significance & effect size

Variable

Statistically significant difference

Effect size

Competence

Female Gen. Am.

is more competent than

Female CVS

1.78 (Very large)

Education

Female Gen. Am.

is more educated than

Female CVS

1.43 (Very large)

Education

Male Gen. Am.

is more educated than

Male CVS

1.09 (Large)

Competence

Male Gen. Am.

is more competent than

Male CVS

1.07 (Large)

Pleasantness

Female Gen. Am.

is more pleasant than

Female CVS

0.90 (Large)

Pleasantness

Male Gen. Am.

is more pleasant than

Male CVS

0.87 (Large)

Friendliness

Female Gen. Am.

is friendlier than

Female CVS

0.56 (Medium)

Friendliness

Male Gen. Am.

is friendlier than

Male CVS

0.55 (Medium)

Age

Female Gen. Am.

is older than

Female CVS

0.45 (Small)

Gender

Female Gen. Am.

is more feminine than

Female CVS

0.38 (Small)

Age

Male Gen. Am.

is no different than

Male CVS

n/a

Gender

Male Gen. Am.

is no different than

Male CVS

n/a

3.2 Interpretation of Results

51The presence of a performance of the CVS has very little effect on the participants’ perception of age. The female voice with the CVS is perceived as slightly younger than the General American female version, but the size of this effect is among the smallest in the study. Though mean scores attributed to the male voices are not strictly identical, no statistically significant difference was found between them. This means that the idea that the perception of both voices is the same, in terms of age, cannot be rejected. More generally, both speakers, no matter whether the CVS performance is present or not, are perceived as neither particularly old nor young (all mean scores gravitate towards the central value: 4). This might have to do with the fact that the male speaker was in his thirties at the time of recording, and the female speaker in her twenties. The respective age of the speakers was therefore rather close to the participants’, whose median age is 23.6.

52Regarding gender perception, the two male voices are unsurprisingly rated as rather masculine (1.71 and 2.09), and the two female voices as rather feminine (6.46 and 6.08). The CVS performance has little impact on the perception of gender, though. The difference in the perception of the two male voices is not statistically significant, so it cannot be argued that the CVS performance makes the male voice sound less masculine. However, the CVS performance does make the female voice sound slightly less feminine, but though statistically significant, the size of this phenomenon is the smallest observed in the study.

53Performances of the CVS affect the perception of friendliness more straightforwardly. Both General American voices are rated above average, and both CVS voices below average. The male voice with the CVS is perceived as less friendly than the General American male voice, and this effect is medium. The same happens with the female voice, and the size of the effect is also medium, though very slightly bigger than for the male voice.

54The perception of pleasantness is also affected by performances of the CVS. General American voices are rated as rather pleasant (above average) but CVS voices are rated as rather unpleasant. Both the male CVS voice and female CVS voice are rated as less pleasant than their General American counterparts, and the effect is large for both, but the difference is slightly larger for the female voice.

55The same happens with the perception of education. Both General American voices are rated as rather educated (5.46 for the male voice and 4.93 for the female voice) but while performing the CVS, both speakers are perceived as less educated. The phenomenon is much bigger for the female speaker. According to statistical standards, the female speaker is perceived as very largely less educated when performing the CVS, whereas the male speaker is perceived as only largely less educated when performing it.

56The perception of the level of competence follows the same pattern. General American voices are rated as rather competent (both score above 5.5 on average), but the CVS performances are rated as much less competent. The effect for the female voice is very large according to standards (it is actually the largest observed in the study) whereas the effect for the male voice is only large.

57The results therefore show that the presence of the CVS performances marginally affect the perception of age and gender, but more clearly change the perception of friendliness, pleasantness, education and competence. Each time, the size of the effect is bigger, or much bigger for the female voice. This means that a voice with an imitation of the CVS is in any case rated as less friendly, less pleasant, less educated and less competent, but this perception is always more pronounced for the voice of the female speaker.

4 Discussion

4.1 Possible Causes of Results

  • 18 Ascribing negative social meanings to an accent perceived as non-standard is not specific to the (...)

58It has been shown that performances of the CVS are given social meanings by participants. They seem to ascribe to the chain shift rather negative meanings18, including lack of education, pleasantness, competence and friendliness. The question then becomes: what may cause these specific social meanings to be attributed to performances of the CVS? Four possible explanations can be discussed.

  • 19 It would be reasonable to believe participants’ perceptions of actual Californian speakers diffe (...)

59First of all, participants may ascribe negative social meanings to the performances of the CVS simply because such ideologies are present in the representations of speakers using it in mainstream media. Such speakers may be presented as rather unintelligent, superficial and generally unappealing to the audience (Pratt & D’Onofrio: 2017). Such ideologies may circulate and unconsciously enter the participants ’belief systems. However, the fact that all participants lived in California at the time of the study suggests this is probably not the sole factor, though19.

60Secondly, the perception of the performances of the CVS may be linked to the principle of sound iconicity (Nobile: 2014), which is a proposed phenomenon where “certain openings of the mouth are associated with meanings, with ideas” (Gómez Milán et al.: 2013, 4). The CVS, which features a greater opening of the jaw, may index a form of stupidity or vulgarity, which are ideologies that have been linked to such jaw position (Bourdieu: 1982). The principle of sound iconicity has also been demonstrated in studies on what is sometimes called the “bouba/kiki” effect (Köhler: 1929; Ramachandran & Hubbard: 2001), where a rounded shape is more likely to be given the name “bouba” by participants (due to the presence of a bilabial consonant, and back vowels, one of which is rounded), and a spiky shape “kiki” (the close front vowel and the voiceless velar plosive being unconsciously associated to spikiness and sharpness).

61Thirdly, this perception might have to do with a form of extrinsic linguistic misogyny. Gender does play an important role in the perception of performances of the CVS, but there seems to be a double standard at play. The performance of the chain shift does not significantly change the perception participants have of the gender of a speaker, but it strongly interacts with the perception participants have of a speaker perceived as female. In other words, a performance of the CVS does not significantly make a speaker sound either more masculine or feminine, but a female speaker performing the CVS is perceived more negatively than a male speaker performing the CVS. Such performances may be negatively perceived not because they intrinsically index femaleness, rather, the negative perception of these performances of the CVS may be strengthened when a female speaker is performing. In other words, the perception of a linguistic phenomenon, the CVS, may therefore be misogynistic (i.e.: directed at women), but not because of an intrinsic association with femaleness, but because of an extrinsic cause: the performer being a woman.

62Fourthly, the very design of the study might be responsible for the results. It may be possible that since the speakers produced unconvincing performances of the CVS (because they were not fully acquainted with it), they created a sort of “linguistic uncanny valley” which might have confused participants. The term “uncanny valley” was first used in the field of robotics by Mori (1970) to describe the fact that the more a being resembles a human, the more uncanny and repulsive its imperfections become. Mori theorized that beings who do not resemble humans at all elicit positive reactions, those who resemble imperfect humans are repulsive, but those who tend to perfectly mimic humans are liked, thus creating a “valley” in perception.

Figure 8: The uncanny valley

Figure 8: The uncanny valley

This figure is taken from Mathur & Reichling (2016), and is itself adapted from Mori (1970).

63It may be possible that this concept applies to linguistics as well. Instead of measuring an individual’s reaction to imitations of a being, the reaction would be that of a hearer reacting to an imitation of their accent. If the hearer believes both them and the speaker have the same accent (because they actually do or because the performance of the speaker is convincing), the hearer may respond positively to the stimulus. If the performance of the hearer’s accent is unconvincing, they may be repulsed by it (because they feel offended, or because they feel something is off) thus creating a valley in their perception.

Figure 9: Proposition for the “linguistic uncanny valley”

Figure 9: Proposition for the “linguistic uncanny valley”

64In robotics, the uncanny valley occurs when most — but not all — features of the being resemble that of a human. The participants in the study presented here may have thought that the sample they heard met most of their linguistic expectations, for example in terms of word stress, intonation, syntax or semantics, but failed to meet their expectations regarding one key linguistic feature, the vowel system, thus creating a “valley,” explaining why negative social meanings were attributed to CVS performances. Maybe the performance of the female speaker was simply less convincing than the male speaker’s, thus explaining the dissimilarity in participants’ perception.

4.2 Limitations of the Study

65This last point leads us to discuss the limitations of the study. As has been mentioned, the selection of participants is biased in some aspects. Though the sample is representative of CSUN students, this population is not representative of the entire population of Los Angeles, California, or the United States as a whole. The participants might have rated the CVS performances more favorably than the general population for two reasons. All participants were students, and therefore younger than the US population. Since the CVS tends to be used by younger speakers, the participants might have been less likely to stigmatize the shift. In other words, they might have unconsciously ascribed some sort of prestige to voices featuring performances of the CVS. Besides, participants all attended one or more linguistics classes, some of which might have made students realize that arbitrary social meanings can be attributed to linguistic markers. The results presented in this study might therefore undermine the sizes of the effects described.

66The study was entirely administered online. For that reason, it is not possible to know whether participants actually took the time to listen to the voices and rate them with care. Since some students received extra credits for the participation in the survey, those especially might have rushed the listening process.

67The four recordings participants listened to also have limitations. Since both speakers voluntarily altered their vowel productions, and since it was not possible to manipulate vowel formants with Praat, speakers might have created the aforementioned “linguistic uncanny valley”. Besides, the fact that the female speaker recorded herself on her smartphone, resulting in lower audio quality, might have interfered with results.

68The matched-guise technique has been criticized. Since the exact same sentence is repeated multiple times, participants may tend to pay a disproportional amount of attention to how the message is delivered as opposed to the informational content itself (Lee: 1971). Others have also pointed out that the semantic content of a sentence itself can also directly correlate to the perception of the dialect used (Houck & Bowers: 1969).

  • 20 Participants were indeed never led to believe some stimuli were performances.

69Finally, though the study was anonymous, participants may also have been unwilling to report on their actual perception of the performances of the CVS20, but rather on whatever they believed was non-offensive, i.e. a “political correctness effect” (Bucholtz et al.: 2008).

5. Conclusion

70Performances of the California Vowel Shift can induce negative sociolinguistic evaluations in terms of pleasantness, friendliness, education or competence. This effect could be due to a shared perception between participants, which might have been propagated by parodic representations of the chain shift in the media. Since all participants lived in California at the time of the study though, this seems unlikely. Maybe the principle of sound iconicity played a part in their perception as well, more open vowels being unconsciously associated with stupidity or vulgarity by participants. They might also have felt a linguistic uncanniness when hearing the stimuli, thus creating a valley in perception. The fact that negative evaluations are even more negative when the performance is that of a female speaker could also point to a form of linguistic misogyny. Replicating this study with several male speakers and several female speakers may help understand if the performances of the latter are systematically more negatively perceived than men’s.

71Though this study focused on performances of the CVS, it cannot be ruled out that a hearer could stigmatize an actual CVS user because of negative perceptions in terms of pleasantness, friendliness, or competence, especially if the hearer is unacquainted with the chain shift. This could mean that a heavy CVS user could theoretically face rejections or discriminations because of their accent, for example on the job or housing market, in a higher education institution, or even while developing personal relationships. On the other hand, hearers using the CVS themselves may perceive the speaker in a positive way, the chain shift thus acting as a marker of inter-speaker solidarity because of a feeling of shared identity and ingroupness.

Top of page

Bibliography

Boersma, Paul and David Weenink. Praat: Doing Phonetics by Computer. [Software]. Version 6.0.36. Online: http://www.praat.org/, 2017.

Bourdieu, Pierre. Ce que parler veut dire. Paris, Fayard, 1982.

Bucholtz, Mary, Nancy Bermudez, Victor Fung, Lisa Edwards and Rosalva Vargas. “Hella Nor Cal or Totally So Cal? The Perceptual Dialectology of California.” Journal of English Linguistics 35(4) (2007): 325–352.

Bucholtz, Mary, Nancy Bermudez, Victor Fung, Rosalva Vargas and Lisa Edwards. “The Normative North and the Stigmatized South: Ideology and Methodology in the Perceptual Dialectology of California.” Journal of English Linguistics 36 (2008): 62–87.

Coates, Jennifer. Women, Men and Language a Sociolinguistic Account of Gender Differences in Language. Routledge Linguistics Classics (3rd edition), 2016.

Cohen, Jacob. Statistical Power Analysis for the Behavioral Sciences. Taylor & Francis, 1969.

Cooper, Alan. The Inmates are Running the Asylum. Sams - Pearson Education, 1999.

CSUN Profiles. Profile 2018. Online: https://www.csun.edu/sites/default/files/profile_2018_0.pdf, 2018.

DOnofrio, Annette. “Perceiving Personae: Effects of Social Information on Perceptions of TRAP-backing.” University of Pennsylvania Working Papers in Linguistics, 21(2), Selected Papers from New Ways of Analyzing Variation (NWAV) 43, 2015a.

DOnofrio, Annette. “Persona-based Information Shapes Linguistic Perception: Valley Girls and California Vowels.” Journal of Sociolinguistics 19(2) (2015b): 241–256.

DOnofrio, Annette. Social Meaning in Linguistic Perception. [PhD Thesis]. Stanford University, 2016.

Eckert, Penelope and William Labov. “Phonetics, Phonology and Social Meaning.” Journal of Sociolinguistics 21(4) (2017): 1–30.

Eckert, Penelope and Norma Mendoza-Denton. “Getting Real in the Golden State (California).” In: Wolfram, Walt and Ben Ward (Eds.). American Voices. MA: Wiley-Blackwell, 2006: 139–143.

Eckert, Penelope. “Getting Emotional about Social Meaning in Variation.” New Ways of Analyzing Variation [Conference] 37, Houston, TX, 2008b.

Eckert, Penelope. “Where Do Ethnolects Stop?International Journal of Bilingualism, 12 (2008a): 25–42.

Eckert, Penelope. Linguistic Variation as Social Practice: The Linguistic Construction of Identity in Belten High. Language in Society 27. [Oxford Malden]: Blackwell Publishers, 2000.

Farnetani, Edda and Daniel Recasens. “Coarticulation and Connected Speech Processes.” In: Hardcastle, William J., John Laver and Fiona E. Gibbon (Eds.). The Handbook of Phonetic Sciences. Blackwell Handbooks in Linguistics. Wiley, 2010.

Faul, Franz, Edgar Erdfelder, Albert-Georg Lang and Axel Buchner. “G*Power 3: A Flexible Statistical Power Analysis Program for the Social, Behavioral, and Biomedical Sciences.” Behavior Research Methods, 39, 175-191. Online: http://www.gpower.hhu.de, 2007.

Fought, Carmen. “California Students’ Perceptions of, You Know, Regions and Dialects?” In: Long, Daniel and Dennis Preston (Eds.). Handbook of Perceptual Dialectology Volume 2. Amsterdam Philadelphia (PA.): J. Benjamins, 2002.

Gómez, Milán Emilio, Oscar Iborra, María José de Cordoba, V. Juarez-Ramos, María Ángeles Rodríguez Artacho and José Luis Rubio. “The Kiki-Bouba Effect A Case of Personification and Ideaesthesia.” Journal of Consciousness Studies, 20(1-2) (2013): 84-102.

Gosset, William Sealy. (“Student”). “The Probable Error of a Mean.” Biometrika, 6(1) (1908): 1-25.

Guenter, Josh. “What Is English /l/ Really?” In: Proceedings of the Twenty-Sixth Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on Aspect. University of California, Berkeley: eLanguage, 2000.

Hagiwara, Robert. “Dialect Variation and Formant Frequency: The American English Vowels Revisited.” Journal of the Acoustic Society of America 102 (1997): 655–658.

Hagiwara, Robert. “Vowel Production in Winnipeg.” The Canadian Journal of Linguistics / La Revue Canadienne de Linguistique 51(2) (2006): 127–141.

Hall-Lew, Lauren. Ethnicity and Phonetic Variation in a San Francisco Neighborhood. Stanford, CA: Stanford University dissertation, 2009.

Hinton, Leanne, Birch Moonwomon, Sue Bremner, Herb Luthin, Mary Van Clay, Jean Lerner and Hazel Corcoran. “Its Not Just the Valley Girls: A Study of California English.” Proceedings of the Thirteenth Annual Meeting of the Berkeley Linguistics Society (1987): 117–128.

Holland, Cory. Shifting or Shifted? The State of California Vowels. University of California Davis, 2014.

Houck, Charles L. and John Bowers. Waite. “Dialect and Identification in Persuasive Messages.” Language and Speech, 12(3) (1969): 180–186.

Kennedy, Robert and James Grama. “Chain Shifting and Centralization in California Vowels: An Acoustic Analysis.” American Speech, 87(1) (2012): 39–56.

Köhler, Wolfgang. Gestalt Psychology. New York: Liveright, 1929.

Labov, William, Sharon Ash and Charles Boberg. The Atlas of North American English: Phonetics, Phonology, and Sound Change: a Multimedia Reference Tool. Mouton de Gruyter, 2006.

Labov, William. The Social Stratification of English in New York City. Cambridge New York Melbourne [etc.]: Cambridge University Press, 1966.

Labov, William. Sociolinguistic Patterns. Basil Blackwell. Oxford, 1972.

Labov, William. “The Intersection of Sex and Social Class in the Course of Linguistic Change.” Language Variation and Change, 2(2) (1990): 205-254.

Lambert, Wallace, Richard Hodgson, Robert Gardner and Stanley Fillenbaum. “Evaluational Reactions to Spoken Languages.” Journal of Abnormal and Social Psychology, 20(1) (1960): 44–51.

Lee, Richard R. “Dialect Perception: A critical Re-evaluation.” Quarterly Journal of Speech, 57(4)(1971): 410–417.

Likert, Rensis. “A Technique for the Measurement of Attitudes.” Archives of Psychology, 22 (140) (1932). New York.

Mathur, Maya B. and David B. Reichling. Navigating a social world with robot partners: A quantitative cartography of the Uncanny Valley.” Cognition 146 (2016): 2232.

Mori, Masahiro. “The Uncanny Valley.” Energy, 7(4) (1970): 3335.

Nobile, Luca. “L’iconicité phonologique dans les neurosciences cognitives et dans la tradition linguistique française.” CILF (conseil international de la langue française). Formes de l’iconicité en langue française: vers une linguistique analogique. Le Français Moderne - Revue de linguistique Française, 82(1) (2014): 131169.

Osgood, Charles, George Suci and Percy Tannenbaum. The Measurement of Meaning. Urbana, IL: University of Illinois Press, 1957.

Podesva, Robert. “The California Vowel Shift and Gay Identity.” American Speech, 86(1) (2011): 32-51.

Podesva, Robert. “California Dialects.” [Lecture]. Stanford EFS Lectures, 2015.

Podesva, Robert, Lauren Hall-Lew, Jason Brenier, Rebecca Starr and Stacy Lewis. “Condoleezza Rice and the Sociophonetic Construction of Identity.” In: Hernandez-Campoy, Juan Manuel and Juan Antonio Cutillas-Espinosa (Eds.). Style-shifting in Public: New Perspectives on Stylistic Variation. Amsterdam: John Benjamins, 2012: 65–80.

Pratt, Teresa and Annette DOnofrio. “Jaw Setting and the California Vowel Shift in Parodic Performance.” Language in Society, 46(3) (2017): 283-312. Cambridge University Press.

Preston, Dennis. R. “Five Visions of America.” Language in Society, 15(2) (1986): 221–240.

Preston, Dennis R. “Where the Worst English is Spoken”. In: E. Schneider (Ed.). Focus on the USA. (1996): 297-360. Amsterdam: John Benjamins.

Purnell, Thomas, William Idsardi and John Baugh. “Perceptual and Phonetic Experiments on American English Dialect Identification.” Journal of Language and Social Psychology, 18(1) (1999): 10-30. Sage Publications.

Ramachandran, Vilayanur and Edward Hubbard. “Synaesthesia: A Window into Perception, Thought and Language.” Journal of Consciousness Studies, 8(12) (2001): 3–34.

Sawilowsky, Shlomo. “New Effect Size Rules of Thumb.” Journal of Modern Applied Statistical Methods, 8(2) (2009): 597–599.

Shafer, Scott. “Do Californians Have an Accent?” [Radio show]. KQED, 2012.

Styler, Will. Using Praat for Linguistic Research. Online: http://wstyler.ucsd.edu/praat/UsingPraatforLinguisticResearchLatest.pdf, 2017.

Villarreal, Dan. The Construction of Social Meaning: A Matched-Guise Investigation of the California Vowel Shift. [PhD Thesis]. University of California, 2016.

Wolfram, Walt and Natalie Schilling. American English: Dialects and Variation. John Wiley & Sons (2006, ed. 2015).

Top of page

Notes

1 Similarly, the use of standard” (non-allophonic) phonemes can also inform about the speakers background.

2 More specifically, Hinton et al. (1987) found that [u], [ʊ] and [o] were more front and less rounded in the 1980s than they were in the 1950s, but that the [ɔ]/[ɑ] merger had already begun in California the 1950s.

3 The term “persona” is meant to be understood in line with Cooper’s definition (1999, 124): Personas are not real people […]. They are hypothetical archetypes of actual users. Although they are imaginary, they are defined with significant rigor and precision.”

4 Though the term used to point to a Californian stereotype (“Valley” originally referred to the San Fernando Valley, North of Los Angeles, hence the association of the stereotype with the CVS), its reference has since expanded to include teenage girls of other states.

5 It should be noted that the CVS can paradoxically be associated to an educated, formal type of speech (D’Onofrio: 2016; Podesva et al.: 2012), i.e. a persona antithetical to that of a Valley Girl.

6 Only the gender of the speakers was taken into account in the present study, and not the gender or the participants themselves. This choice was made because dividing participants into gendered subsets would have required even more participants be involved in the study in order to provide statistically significant results.

7 Each professor decided whether they would incentivize students.

8 Besides the ones presented in this paper, the participants completed other tasks, which are not discussed here because they do not pertain to the perception of the CVS. These tasks aimed at assessing the perception of the High Rising Terminal contour, vernacular functions of like, and the Valley Girl persona.

9 Four other voices were also used, but since they do not pertain to the perception of the CVS, they are not discussed here.

10 These include: unreliable/trustworthy; self-involved/selfless; rural/urban; poor/rich; tense/relaxed; not Californian/Californian.

11 Though results show that the perception of age is indeed not heavily affected by the presence of the CVS in the male voice, and not affected at all in the female voice, the possible correlation between this variable and prestige is discussed in the “Limitations of the Study” section of this paper.

12 Since the purpose of the performances of the CVS was that they be biased imitations of the chain shift, they were not submitted to actual speakers of the CVS in order to assess their credibility before the study was conducted.

13 The audio quality of the female voice was not as clear as the male voice (see the “Limitations of the Study” section of this paper).

14 [æ] raising is due to the presence of the following nasal phone [n].

15 [ɪ] fronting is due to the presence of the following velar nasal phone [ŋ].

16 Here, the phrase “prototypical versions” refers to the versions of the recording in which vowels have formant values most similar to the speaker’s average. Therefore a prototypical version is to be construed as prototypical for the said speaker.

17 Vowel qualities being labeled “least prototypical” is the result of statistical analysis, not linguistic prescriptivism on the part of the author.

18 Ascribing negative social meanings to an accent perceived as non-standard is not specific to the CVS, as Preston (1996) has shown.

19 It would be reasonable to believe participants’ perceptions of actual Californian speakers differ from that of over-the-top fictional Californian characters.

20 Participants were indeed never led to believe some stimuli were performances.

Top of page

List of illustrations

Title Figure 1: The California Vowel Shift (adapted from Eckert, 2008a)
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-1.png
File image/png, 59k
Title Figure 2: Semantic differential scale used for the “age” variable
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-2.png
File image/png, 27k
Title Figure 5: General American vowels (blue) and CVS vowels (red) of voices used in the study
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-3.png
File image/png, 42k
Title Figure 6: Mean scores attributed to voices for each variable
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-4.png
File image/png, 28k
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-5.png
File image/png, 19k
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-6.png
File image/png, 18k
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-7.png
File image/png, 18k
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-8.png
File image/png, 19k
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-9.png
File image/png, 18k
Title Figure 8: The uncanny valley
Credits This figure is taken from Mathur & Reichling (2016), and is itself adapted from Mori (1970).
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-10.png
File image/png, 19k
Title Figure 9: Proposition for the “linguistic uncanny valley”
URL http://journals.openedition.org/anglophonia/docannexe/image/3556/img-11.png
File image/png, 39k
Top of page

References

Electronic reference

Pierre Habasque, Some sociolinguistic evaluations of performances of the California Vowel Shift: a matched-guise studyAnglophonia [Online], 30 | 2020, Online since 20 December 2020, connection on 20 January 2022. URL: http://journals.openedition.org/anglophonia/3556; DOI: https://doi.org/10.4000/anglophonia.3556

Top of page

Copyright

Licence Creative Commons
Anglophonia – French Journal of English Linguistics est mis à disposition selon les termes de la licence Creative Commons Attribution - Pas d'Utilisation Commerciale - Pas de Modification 4.0 International.

Top of page
Search OpenEdition Search

You will be redirected to OpenEdition Search