(Abstract of the MA thesis published at the University of Vienna, May 2000)

Reality is often a little harsher than the poet would have it. In real life, the 'subtle charm of beautiful pronunciation' is attributed to some accents rather than others, and along with such thinking, inferences are made about the speakers using the accents.

The purpose of the field study presented in this paper was to record some of the inferences generally made about, and resulting attitudes towards, speakers of Southern American English (i.e. the version(s) of American English spoken in the Southern States).¹ The study was conducted over a two-month period at four different universities/colleges in the states of Vermont and Tennessee; the informants were all U.S. undergraduate students.

The theoretical framework for this investigation is provided by the social psychological approach to language attitude study, as adapted most notably by Ellen B. Ryan and Howard Giles in their 1982 book Attitudes towards Language Variation.² In this approach, language attitudes, (i.e., generally, attitudes directed towards language as a referent), can be defined as "any affective, cognitive or behavioural index of evaluative reactions towards different language varieties or their speakers" (Ryan - Giles - Sebastian 1982: 7, my italics). The focus is "upon the individual and his/her display of attitudes toward ingroup and outgroup members as elicited by language..." (Ryan - Giles - Sebastian 1982: 2). This is why according to this approach the main interest lies in speaker evaluation studies - i.e. studies where informants are in one form or another asked to rate speaker samples, thus yielding evaluative reactions, namely those elicited by language. The present study, too, is based on the principle of speaker evaluation.

The social psychological approach to the study of language attitudes also holds that members of speech communities do not have a single unitary attitude towards two contrasting language varieties, but rather that, among other things, the context/setting of the evaluation is a vital factor in the display of attitudes and thus in the speaker evaluation (cf. Giles - Ryan 1982: 219; Smit 1994: 53-58; and esp. Cargile e.a. 1994): "[t]he extent to which language variety A is preferred over language variety B depends upon the situation in which the assessment is made" (Giles - Ryan 1982: 219). Simply put, different 'priorities' in the line of language prestige and/or expression of group solidarity apply in different contexts. Thus, to avoid ambiguity of results and the drawing of undue conclusions, it is necessary to choose and closely define a very specific situational setting for any language attitude study.

The setting chosen for the present study is a job interview situation in sales. This decision was deemed on the one hand to give the study a pragmatic quality - the main reason for language attitude research today being its applicability to real life situations with regard to language problems (cf. Smit 1994: 54) - and on the other, to lend the necessary plausibility to the set-up for the informants, who were told that they should act as personnel managers in a hiring company, evaluating salesjob applicants. Using defining parameters identified by Giles and Ryan (1982: 219-220) as well as results of previous studies (cf. Kalin 1982, Shields 1979), one can identify a salesjob interview as a setting that stresses language status/ language prestige and group-centeredness/ impersonality (as it does not generally build on the intimacy between two people). It is thus a rather formal setting; this implies that speech could be rather carefully monitored by the judges (cf. Cargile e.a. 1994:225), a fact which might disfavor a 'minority' language.

The tool applied to elicit language attitudes in the given context is sort of a 'classic' throughout the paradigm:³ it is an adapted form of the so-called matched guise technique as introduced by Wallace Lambert and colleagues in the 1960's (cf. i.a. Lambert 1967). In the version used here, four different speakers were recorded using their very own language variety: two with a 'neutral' accent, male and female (i.e. an accent that could not really be regionally placed), and two with a Southern/Tennessee accent, also male and female. Voices were selected in matching pairs as to pitch and quality in order to avoid too much divergence apart from the one in accent. All of the speakers were recorded reading the same text - a neutral one-minute piece about sales and salespeople.

For the evaluation, the informants were provided with a questionnaire containing a rating grid of semantic differential scales. The rating grid was also so designed as to match the sales context. It contained 21 attribute items in the form of opposite pairs: likeable - not likeable, educated - uneducated, trustworthy - not trustworthy, polite - impolite, intelligent - not intelligent, friendly - unfriendly, honest - dishonest, sociable - unsociable, ambitious - not ambitious, self-confident - not self-confident, helpful - not helpful, determined - wavering, reliable - unreliable, leadership qualities - no leadership qualities, sense of humor - no sense of humor, industrious - lazy, open-minded - not open-minded, sharp - slow, good manners - bad manners, successful - not successful, outgoing - shy. This list was compiled as a common denominator of mainly two paradigms: first, the qualities deemed necessary in a salesperson,⁴ and secondly, common Southern stereotypes (as previously assessed in a 'content analysis of societal treatment' of the South and Southern American English).

The informants were asked to place their marks on a 5-point scale between the poles according to the degree they believed an attribute to be true for a speaker. The list was complemented by three 'summarizing' statements ("This speaker would make a good salesperson", "I would employ this speaker in my company as a salesperson", "I would like to get to know this speaker on a personal basis"), for which the same rating scales were used.

In the questionnaire, the four rating grids for the speakers were followed by a fifth, similar section asking the informants to use the same grid to describe their picture of a perfectly successful salesperson, the 'Ideal Salesperson', in order to provide a sort of 'standard' measure against which to compare the speaker ratings.

With Giles and Ryan's demand for methodological eclecticism (1982: 223) in mind, the speaker evaluation core of the field study questionnaire was complemented by a second, more cognitively oriented part that contained mostly closed questions leading from matters of American regional accents in general into the particular of Southern American English.

A third and final section sought to record the relevant informant biographical data in view of an ensuing statistical evaluation of the questionnaires. This also allowed for a careful selection process to obtain a very homogeneous group of informants, as it was judged necessary for a study of this limited scope. The informants were all U.S. undergraduate students, males and females in comparable parts, aged 18-24, and all native to one of the two test regions selected - New England and Tennessee, representing, in a simplification, the 'North' and the 'South'. The population was all white ('Caucasian'), for the simple reason that it was felt that in a minority/black population an investigation of Southern American English might be prone to call up touchy history-related issues of race or racism, the handling of which would have been entirely beyond the scope of such a small study as the present.

The final population consisted of 291 students: 141 from New England, 150 from Tennessee; 122 male, 169 female. The fact that students should be used at all in studies relying on employment opportunity settings has been justified by Rudolf Kalin (1982: 158/159), who observed that many students are in fact future employers who would soon be making real hiring decisions, and that in a number of comparative studies the responses given by students and those by actual employment interviewers were very similar. The only difference to emerge was that student judges tended to be somewhat more lenient than actual job interviewers.

The necessary parameters for the field study in place, the results and outcome of the statistical data evaluation shall now be presented.

At the core of the analysis are comparisons of mean values. The original ratings on the 5-point scales were encoded using values from 5 to 1 - higher ratings being those closer to the positive adjective pole (educated, intelligent, etc.). The mean values were then calculated and compared using the statistical tools of Levene's and T-Tests. The cut-off level for statistical significance was set at .05, with p=.01 delimiting high statistical significance.

At the outset, the 21 attribute items of the rating grid were subjected to a factor analysis using the Principal Component Analysis. An eigenvalue of greater than one was adopted as criterion of extraction (Kaiser's criterion). The process yielded three factors, which were then rotated employing the varimax method. These factors allowed for the attributes to be drawn together in three groups: one that could be entitled competence (sharp, successful, determined, educated, leadership qualities, intelligent, ambitious, industrious, self-confident), one of personal integrity (honest, trustworthy, polite, good manners, reliable, likeable, helpful, open-minded), and one of social attractiveness (outgoing, sense of humor, sociable, friendly).

A first working hypothesis for the study predicted that the Southern speakers would do worse in the overall speaker evaluation than the 'neutral' speakers. This hypothesis was quite distinctly confirmed in the general outcome. Split up according to factors, the results for the competence cluster were most explicit: both 'neutral' speakers consistently ranked before the Southerners with high statistical significance. The 'neutral' male ranked before his female counterpart; with the Southerners, the opposite occurred, the Southern female surpassing her male counterpart. In the personal integrity category, ratings were rather level, only the Southern male speaker consistently came in last. The third cluster under the heading of social attractiveness presented a very different picture, and a rather interesting one: it was the one instance where the Southern accent did not lower the scores for its speakers, but rather gave them a realistic chance to pull even with the 'neutral' speakers. In the case of the Southern female, it even allowed her to take the overall lead.

As has been mentioned before, three so-called summarizing statements concluded the rating grid for each speaker. They referred to how good a salesperson the informants believed a speaker to be, if they would hire them as such, and whether they would feel any incentive to get to know a speaker better personally. The first two statements were thus directly (sales-) 'performance'-related. Highest scores were once more achieved by the 'neutral' speakers, the 'neutral' male having the edge over his female counterpart. In view of the earlier results, this leads to the overall conclusion that a good performance in sales is seen as directly related to competence rather than social attractiveness or personal integrity; this, despite the communicative component of transactions in selling. In that sense, it could not be astonishing that the Southerners should lose ground here; but it does seem a little astonishing that more personal and social aspects were ultimately disregarded by the informants here. Yet again, the Southern female still did better than her male counterpart.

The ratings for the third, 'sympathy'-related statement formed a category entirely apart from the former two statements. Both female speakers retained a slight edge over the males, with the Southern female again in the lead.

As can be gleaned from the picture given so far, another original working hypothesis based on results of previous studies (cf. e.g. Van Antwerp - Maxwell 1982) must remain unconfirmed as such: i.e., the assumption that female speakers would be rated lower than males. Though for the 'neutral' speakers the competence and performance-related ratings, and even the social attractiveness scores, did in fact establish the predicted overall hierarchy, with the male speaker ranking higher than the female in the majority of cases, only 'losing' to her in the 'sympathy'-ratings of summarizing statement #3 ("get to know on a personal basis"), and both pulling even for personal integrity, the general outcome is entirely different for the Southern speakers. In no instance did the Southern male speaker receive higher scores than his female counterpart; in no instance could he close the gap, even if pulling even with one or both of the 'neutral' speakers (e.g. with both on social attractiveness, with the 'neutral' male on the 'sympathy' score of summarizing statement #3). Contrary to previous results, therefore, this outcome suggests that the female speaker with the Southern accent tended to profit from her ‘combination’ rather than being hurt by it; consistently so in comparison with the Southern male, and in terms of social attractiveness and 'sympathy' scores even in comparison with both 'neutral' competitors. This is one of the most salient findings of the present study.

All in all, though the Southern female's competence and summarizing 'performance' ratings were unaffected by her high social attractiveness and 'sympathy' scores, it should not be excluded that, in a real life job-interview situation, a potential 'country-boying' charm, as is often popularly attributed to Southern women, once tapped, might actually turn out to be a compensation for other perceived shortcomings. Further investigation into actual behavioral consequences of language attitudes towards Southern speech would thus promise to be very interesting. At least, what the present results point out is that any similar study of language attitudes towards Southern accents must by all means take the variable of speaker's sex into account, to avoid distortions.

An analysis of the correlations among speakers using Pearson's r showed some strong analogous relationships between the ratings of the two Southern speakers. This pointed to the fact that the speaker evaluation as such was not done at random, and confirmed the Southern accent to have been picked up as a salient parameter in the informants' assessment, as expected at the outset. That the Southern accent of both speakers was actually recognized by a wide majority of informants was verified in a set of 'control' questions about the speakers' origin ("Where in the USA do you think these speaker come from?") at the beginning of the second part of the questionnaire.

The evaluation of the 'ideal salesperson', though in its own 'control' function legitimating another aspect of the analysis, namely the assumption that higher ratings equaled better ratings at all times, brought a slight discrepancy with it: personal integrity and social attractiveness scores were here emphasized over competence, in contrast to the speaker evaluations in the grid and in the summarizing statements. Tentative explanations could pick up on the more cognitively oriented aspect of the 'ideal salesperson' evaluation, and a possible influence of 'social desirability' considerations on the informants' part.

Subsequent to the analysis of the overall results from the speaker evaluation, the body of data was broken down into samples according to different independent grouping variables gleaned from the informants' biographical data. A set of five grouping variables was subjected to statistical testing: informants' region of origin (New England, Tennessee), informants' sex, their parents' origin, informants' travel experience, and time spent with friends/relatives in or from the respective other region.

Sampling according to 'parents' origin' and 'time spent with friends/relatives' did not give rise to any statistically significant developments at all. 'Travel experience' gave mere hints at a possible influence on language attitudes regarding traveling to the respective other region, which seemed to enhance social attractiveness-ratings while apt to decrease perceived competence, for both groups of informants. Further testing would be needed to get to the bottom of these findings.

As it turned out, sampling according to informants' sex yielded only minor insights, namely that female informants generally tended to give higher scores, and that male speakers (especially the 'neutral' male) at times would receive an 'opposite sex' bonus in social attractiveness- and 'sympathy'- related scores. As predicted in another working hypothesis, then, informants' origin proved to be the most salient of all grouping variables. Yet, even here, the differences recorded were not as clear-cut and numerous as originally expected, altogether departing not too far from the overall picture.

In this line, a last working hypothesis had basically predicted that Southern speakers would do better when rated by Southerners and worse when rated by Northerners. But the results of the sample analysis showed, rather surprisingly, that in terms of competence, Southern informants were far from more 'generous' towards their peers; rather, they were outright 'stricter', lowering their scores vis-à-vis the Northern informants'. In the personal integrity and social attractiveness evaluations, scores did get equaled out between speakers in the Southern informant sample (as opposed to the New England sample), the Southern speakers catching up with the 'neutral' speakers in the Tennessee ratings, but the Southern speakers received no such strong boost as to be given an edge over their 'neutral' counterparts. This same picture is reflected in the 'performance'-related summarizing scores ("good salesperson"/ "hire in my company"). Once more, only the Southern female could slightly profit in the 'sympathy' score ("get to know..."). For the Southern male speaker, this also means that in his ratings no evidence of any 'covert prestige' phenomenon (cf. Trudgill 1972) could be traced, contrary to other studies (e.g. Luhman 1990).

The influence of group solidarity on the speaker evaluation was therefore simply overrated in the last working hypothesis for the present study. However, both the virtual setting (salesjob-interview) and the real-life setting (university/college) in which the present language attitude assessment was done were rather highly status-stressing, as opposed to solidarity-stressing. In a different set-up, more of a group solidarity among Southerners might come to bear. This, too, would be a profitable subject for further investigation.

What would still follow from the outcome of the present set-up as it stands is the confirmation that Southern American English is generally associated with low status and non-standardness, as its speakers fail to 'perform' in the context given here. In other words, in as status-dominated a setting as the present, 'neutral' accents just fit the expected language variety profile better than Southern accents. And, if +status is associated with +standardization, as is usually the case (cf. Cargile e.a. 1994: 226), negative marks for Southern American English can also be taken to confirm what was outlined i.a. in numerous studies by Preston (e.g. 1997), which is that in the 'default' definition of a 'standard' in the United States,⁵ what 'Standard American English' decidedly is not is Southern American English. On the other hand, as the results also suggest, what seems to come close to 'standardness' in the U.S. is in fact a 'neutral', 'deregionalized' accent as used by the respective 'neutral' speakers in the study (cf. also Wolfram - Schilling-Estes 1998: 12).

The results from the second, complimentary part of the questionnaire largely confirm what the outcome of the speaker evaluation has suggested so far. Majorities of informants respectively agreed in their responses that a regional accent would indeed make a difference in a salesperson working for a nationwide corporation, with most of them saying that the difference would be a negative one; when asked, they agreed that a Southern accent could be an impediment in the salesjob market, and that they could think of other situations, too, where a Southern accent might seem inappropriate or disadvantageous. Informants saying they would not consider advising salesjob applicants to unlearn their accent were in the minority as opposed to those who said they would, or might under certain circumstances.

Throughout, the Southern informants appeared more pessimistic or disillusioned than the Northerners with respect to the prestige of regional accents. For example, only 26.7% of the Tennessee informants indicated they would definitely not advise a salesperson to unlearn their accent. Yet two thirds of the Tennesseans also said that, on a more affective level, they actually liked Southern speech - as opposed to only 47.5% of the New Englanders saying they did (with 34.8% relativizing that they might do so under certain conditions).

The informants assessed a Southern accent in general to be 'cute', but not 'awkward', nor 'beautiful', 'cool', 'too slow', or 'ridiculous'. In contrast to Tennesseans, New Englanders tended to associate the accent with non-standardness, and said it was rather 'amusing', which Tennesseans also rejected.

When asked whether they believed there was one generally acceptable and desirable U.S. 'standard', more than half of the informants answered in the negative (two thirds of Tennesseans). Seen in relation to the responses to the question about 'unlearning a regional accent', and the general assessment of the 'neutral' and Southern speakers in Part I of the questionnaire, this once more corroborates the premise that 'standardness' in the U.S. is not perceived as an emulation of one particular language variety or form of speech, but as the avoidance of regional features (such as speaking Southern); this is another important finding of the present study. That the Southerners themselves have picked up this notion seems to be one more piece of evidence for a general latent and pervasive linguistic insecurity on the Southerners’ part, as confirmed in their speaker ratings.

Other findings, gleaned from a sort of synthesis of the essay answers to the questions posed in Part II of the questionnaire ("Explain your answer"), suggested that knowledge of Southern stereotypes, such as a lack of education/intelligence or a general friendliness, is probably widespread in U.S. society as represented by the informant population here. What also came out is the notion that linguistically, at least, the South (i.e. a region of some 25% of the population)⁶ and the non-South constitute a clear dichotomy, which means that what is effective and appropriate in the one place is not at all so in the other: one in five students volunteered the opinion at some point in the questionnaire that a regional accent would be most effective or, as it were, least harmful in its region of origin. Within the South, however, as the speaker evaluation has made evident, this does not necessarily mean that too much unconditional linguistic solidarity can be expected. Lippi-Green (1997: 213) suggests, though, that Southerners exhibit insecurity about their language, and themselves subscribe to criticism of it, primarily when in direct contact with a Northern (or, probably, any more 'prestigious') 'opposite' - thus, further studies would have to show if in a more 'protected', distinctly Southern environment, the cards would not be dealt differently in terms of accent evaluation.

In fact, the research perspective, with respect to language attitudes towards Southern American English, seems exceedingly wide, and many answers are still to be found, or, at least, to be double-checked. Further investigations along similar lines as the present, and expanding its scope, could thus study the effects of different Southern accents in a given setting (as opposed to the single Tennessee accent used here), or the impact of race issues on language attitudes, of using other dialect features instead of mere accents, and, of course, of all kinds of different formal and informal settings and/or set-ups. Studies in attitude strength over time would also be called for (cf. Petty - Krosnick 1995) - for the present, suffice it to say that the seemingly institutionalized character of the common Southern stereotypes through the media and popular culture actually suggests that attitudes based on these generalizations are rather strong and durable, constantly tilting the power balance in favor of the non-South.

In short, the core findings of this present study, as it stands, are the following: language attitudes towards Southern American English are rather negative in comparison with a 'neutral' accent - for male speakers more so than for females. In a salesjob-interview situation, having a Southern accent is a first strike against the applicant. In a way, Southern speech seems therefore a likely imminent subject for deliberations of the American Equal Employment Opportunity Commission.⁷ Positive associations of Southern speech cannot compensate for the negative impressions called up. Generally, a Southern accent is considered low-status and non-standard. The subordination process concomitant with this stigma (cf. Lippi-Green 1997: 68), i.e. the devaluing of the 'non-mainstream', has proved successful in a superregional (national) context, as the Southerners themselves subscribe to it.

How to change such a picture? In the short run, further studies on the subject of regional variation in the U.S. might contribute to increase public awareness of the issue.

In the long run, it would help to teach the next generation(s) more respect towards linguistic variety; in the U.S. just like anywhere else around the world.

1 For a delimitation of American regional dialects refer to Carver (1987) and especially to the Phonological Atlas project directed by William Labov e.a. at the University of Pennsylvania: http://www.ling.upenn.edu/phono_atlas/NationalMap/NationalMap.html

Speaker ratings – overall results (mean values):

p values:

statistical significance at p<.05 ; highly significant at p£ .001