Charm or Harm: Effect of Passage Content on Listener Attitudes toward American English Accents

Abstract

This study examined how passage content influences attitudes towards American English Accents. Participants listened to passages differing in topic content spoken in an American Southern English or Standard American English accent. Although Southern-accented speakers were rated higher in sociality, but lower in status, than standard-accented speakers, sociality ratings varied as a function of passage topic only for Standard-accented speakers. Linguistic content appeared most likely to influence listeners' attitudes when preexisting assumptions based on regional accent were absent.

Differences in how individuals and groups speak, whether with accents or dialects, appear to profoundly influence how individuals and groups of individuals interact (Baugh, 2003; De la Zerda & Hopper, 1979; Matsuda, 1991; Squires & Chadwick, 2006). Listeners attribute certain characteristics to speakers, such as whether they are friendly, smart, or trustworthy, just by hearing them speak (Demirci & Kleiner, 1999; Lambert, 1967; Preston 1999; Edwards 1982; Sebastian & Ryan, 1985). For example, Giles, Henwood, Coupland, Harriman, and Coupland (1992) found that when producing a standard British accent, a speaker was judged as more competent but less benevolent than when using a non-standard British accent. Lambert (1967) found that both English and French Canadian students rated the more “prestigious” English Canadian speakers to have “positive” attributes such as being more attractive, intelligent, and dependable relative to French Canadian speakers.

The present study was designed to examine the extent to which these types of attitudes can change depending on aspects of a communicative situation. We focused on one particularly salient factor that could potentially influence listeners' attitudes -- message content or meaningful topic of spoken materials. Relatively less attention has been paid to systematic evaluations of how what is being said might influence judgments of speaker attributes arising from differences in speaking style, accent, or dialect. Our goal was to determine if and how the content of speech interacts with accent to determine listeners' judgments of speaker attributes.

To that end, the present experiment focused on attitudes toward two varieties of American English – an American Southern accent, one of the many regional varieties of speech in the United States, and Standard American English.¹ We focused on the use of accents rather than dialects in order to isolate the effects of pronunciation on listener attitudes apart from differences in factors such as vocabulary usage and syntactic construction.

The two accents were chosen in part because each can be identified from differences in pronunciation alone (e.g., Clopper & Pisoni, 2004) and particular assumptions or stereotypes appear to be associated with the Southern dialects (Cooke-Jackson & Hansen, 2008). Preston (1999) found that students identified the South as a distinct dialect region in comparison with other regions of the United States and that Southern speech styles were considered more friendly, but less correct than standard speech styles. These and other similar findings (e.g., Giles et al., 1992; Luhman, 1990) suggest that regional accents, and Southern accents in particular, significantly influence our perception of others and may be indicative of stereotyping.

Although relatively less is know about the impact of passage content on judgments of speakers with regional accents, context in general, in the form of information related to the subject (e.g., location, occupation) appears to influence judgments (Wittenbrink, Judd, & Park, 2001; Barden, Maddux, Petty, & Brewer, 2004). With respect to linguistic content, language users appear to use a speaker's word choices to infer the competence of a speaker (Nath, 2007). Levin, Giles, and Garrett (1994) found that the perceived formality of lexical items based on word origins (Latinate versus Germanic) influenced listeners' judgments of speakers' personality traits. The current study extends this finding by examining whether the overall theme or topic of the speech sample will influence speaker attribute judgments and influence judgments differently depending on whether the accent is standard or non-standard.

In the experiment, participants listened to two passages that varied along several meaning dimensions, but primarily differed in being rated as more typical of the South (hunting and cooking) or not typical of the South (medicine and investment). The passages were read by both standard and Southern-accented speakers. Following each passage, participants completed an attitude questionnaire. Ratings of individual attributes were predicted to vary with type of accent. Individuals speaking a Southern accent were expected to be rated as lower status and more friendly than individuals speaking a standard accent. In addition, however, it was predicted that accent ratings would vary depending on passage topic. Speakers were expected to be rated differently both in terms of their status and their overall friendliness depending on the passage content or topic.

Methods

Participants

Participants were 64 undergraduate students (51 female, 13 male) at Emory University. Participants were either paid $10 or received research credit for their participation. A variety of regions were represented, including the North (n = 6), Northeast (n = 19), South (n = 19), Midwest (n = 3), Southwest (n = 3), and West (n = 5).

Stimulus materials

Passages

Two passages were constructed to be consistent with typical Southern activities (hunting and cooking). The hunting passage described how to load a gun and the cooking passage described preparing a soufflé. Two passages were constructed to be inconsistent or neutral with respect to Southern stereotypes (medical and investment). The medical passage described how to perform an appendectomy and the investment passage described short selling. All passages were matched on number of words (between 220 and 230) and sentences (between 14.5 and 15) and on reading difficulty (between 58 and 67 on the Flesch Reading Ease scale and 8.5 on the Flesch-Kincaid Grade Level scale; Flesch, 1948).²

Twelve raters reported on a scale of 1 (not at all) to 5 (very) how characteristic passage topics were for seven American regions (West, Northeast, North, Northwest, Midwest, South, and Southwest). Because the passages necessarily varied on other attributes as well, ratings of other aspects of passage meaning were also collected. A separate group of 33 raters judged how formal, rural, and characteristic of a leisure activity each of the passages was on a scale of 1 (not at all) to 5 (very). On the basis of these ratings, medicine and investment constituted the non-Southern or “neutral” topics while hunting and cooking constituted the Southern topics. Southern topics were rated as significantly more characteristic of the South (t(11) = 3.94, p = .01) than neutral topics. Southern topics were also rated as less formal (t(32) = 3.51, p = .01), more rural (t(32) = 11.87, p = .01), and more leisure (t(32) = 18.65, p = .01) than the neutral topics.

Recording

Standard- and Southern-accented male and female speakers recorded each of the four passages. The two standard-accented speakers (56 and 53 years old) were both from Cleveland and had been raised in Ohio. The two Southern-accented speakers (52 and 48 years old) were from Anderson, South Carolina and were raised in Tennessee and South Carolina. All speakers graduated from college and three held postgraduate degrees.

Pilot tests were performed to confirm that the speakers' accents were associated with the intended regions and that the passages were equally comprehensible. Fifteen raters were presented with the first three sentences of each passage over headphones and asked to rate on a scale of 1 (not likely) to 5 (very likely) how likely it was that each speaker was from each of the seven regions (see above). Standard-accented speakers were rated as being significantly less likely to be from the South (t (14) = 5.73, p < .01) than Southern-accented speakers. Likewise, Southern-accented speakers (M = 4.39) were rated as being significantly more likely to be from the South (t (14) = 12.76, p < .01) than from other regions.³

A separate group of 12 participants read each passage and answered a set of 10 questions to assess baseline comprehension of the passages. Independent means t-tests using a Bonferroni corrected alpha level of .008 confirmed that the passages did not differ significantly from one another in baseline comprehensibility.

Attitudes Assessment

An attitudes assessment was developed to evaluate judgments about the speaking style of each talker. The scale consisted of 22 adjectives formatted and adapted from Giles et al. (1992) and Preston (1999). The scale was intended to assess a range of attitudes about the speaker and speaking style. The attitudes of most interest for the current experiment were characteristics of status and competence such as intelligent, smart, competent, Good English and well-educated, and characteristics of sociality such as friendly, sociable, trustworthy, sympathetic, and nice. These characteristics have been found in the previous literature to differ significantly depending upon dialect or accent (Edwards 1982; Giles et al., 1992; Lambert, 1967; Preston 1999; Sebastian & Ryan, 1985).

Procedure

Participants were randomly assigned to one of four conditions (Southern speakers/Southern passages; Southern speakers/non-Southern passages; standard speakers/Southern passages; standard speakers/non-Southern passages). In each condition, participants were presented two passages auditorily over headphones. Each group listened to both a male and female speaker from the same region reading a different passage with the same regional content. Across participants, order in which the passages were heard and which speaker produced which passage was fully counterbalanced. After each passage, participants answered a set of comprehension questions to insure they were attending to passage content and then pressed a button on the computer to continue to the attitudes assessment. Participants rated each speaker (on a scale from 1 to 7) on each of the 22 adjectives. Ratings were collected on-line with each adjective pair being presented one at a time. Presentation of passages and the collection of responses were controlled on-line by E-Prime experimental software (Schneider, Eschman, & Zuccolotto, 2002).

Results

Ratings were collapsed across individual passages to yield means for Southern-typical and non-Southern content passages. Because the male and female speakers were not rated significantly differently, ratings were collapsed across the two Southern and the two standard talkers as well.

Factor Analysis

A factor analysis (principle components analysis with varimax rotation) was performed on the attitude ratings. Two factors accounted for 54% of the variance and represented the least number of factors accounting for most of the variance, according to a Scree test. The first factor corresponded to a construct that has been characterized as status or intelligence in previous work (34.82% variance, eigenvalue = 7.7) with attributes such as intelligent (0.86), competent (0.81), and well-educated (0.88) loading highly on this factor. The second factor corresponded to what has been described as solidarity in previous research (19.19% variance, eigenvalue = 4.2) with attributes such as sociable (0.74), cheerful (0.78), and friendly (0.85) loading highly on this factor. The differential loading of subsets of the rating scale suggests that raters were evaluating speakers along at least two dimensions corresponding to their status or competence and to their solidarity or sociality.

Effects of passage content and accent on judgments of speaker attributes

Preliminary examination of speaker gender and the region of origin of our listeners indicated no significant main effects or interactions with other variables and, as such, neither variable was considered further. In addition, few consistent effects of passage order were found and no interactions with passage content or speaker accent were found. As a consequence, this variable was also not considered further.

In order to evaluate the extent to which passage content and accent type influenced listeners' ratings of speaker attributes, a series of 2 × 2 analyses of variance (ANOVAs) were conducted using each of the 22 adjective ratings as a dependent measure. Passage content (Southern typical versus non-Southern) and accent type (Southern versus Standard) were between-subjects factors.

Passage Effects

When the speakers were reading the non-Southern content passages, they were rated as significantly more intelligent (F(1,60)= 7.83, p < 0.01), educated (F(1,60) = 17.49, p < 0.01), important (F (1,60) = 7.06, p = 0.01), richer (F(1,60) = 19.06, p < 0.01), and as having better English (F (1,60) = 8.19, p < 0.01) than when reading Southern content passages. Each of these attributes was associated with the status factor derived from the factor analysis.

Speaker Effects

Speakers with a standard accent were rated significantly more intelligent (F (1,60) = 4.14, p = 0.05), arrogant (F (1,60) = 5.47, p = 0.02), smarter (F (1,60) = 4.49, p = 0.04), better educated (F (1,60) = 5.02, p = 0.03), and as having better English (F (1,60) = 12.90, p < 0.01) than Southern-accented speakers, regardless of passage type. These attributes all appeared to be associated with status.

Speakers with a Southern accent were rated as more amusing (F (1,60) = 8.16, p < .01), friendlier (F (1,60) = 11.44, p < .01), more polite (F (1,60) = 3.65, p = 0.06), and nicer (F (1,60) = 2.89, p < .01) than standard-accented speakers, regardless of passage type. Southern-accented speakers were also rated as significantly more sociable (F (1,60) = 24.98, p < .01) and more cheerful (F (1,60) = 11.48, p < .01), although these main effects were mediated by significant interactions between accent and passage type.

Interactions

The analyses revealed significant interactions between passage content and speaker accent for ratings of sociability (F(1,60) = 6.24, p = .02), likeability (F(1,60) = 5.77, p = .02), and cheerfulness (F(1,60) = 4.33, p = .04). Follow-up means comparisons showed that standard-accented speakers were considered significantly more sociable (t(30) = 2.58, p < 0.02), likeable (t(30) = 2.19, p = 0.04), and cheerful (t(30) = 2.03, p = 0.05) when producing the Southern passages than when producing the non-Southern passages. Judgments of Southern-accented speakers did not vary with passage content (all p's > .05).

Status and Sociality

Two overall 2×2 ANOVAs were conducted using a status or competence measure that collapsed across ratings with high loading (> .40) on Factor 1 and a sociality or friendliness measure that collapsed across ratings with high loadings (> .40) on Factor 2.

For the composite measure of status, significant main effects, but no interactions, were found for passage content (F(1,60) = 7.71, p < .01) and speaker accent (F(1,60) = 4.53, p = .04). For both content and accent, ratings of status were at or above the midpoint of the scale regardless of passage content. However, non-Southern content passages were rated as higher status than Southern-themed passages. Likewise, standard-accented speakers were rated as higher status than Southern-accented speakers.

In the analysis of the sociality measure, a significant interaction was found (F(1,60) = 5.74, p = .02) as well as a main effect for speakers (F(1,60) = 14.22, p < .01). Follow-up means comparisons showed that Standard speakers were rated as significantly higher in sociality when the passage had Southern (or rural, casual, or leisure) content than when the passage had non-Southern content (t(30) = 2.02, p = 0.05), but no difference was found for Southern speakers (p > .05).

Discussion

Both passage content and speaker accent influenced listeners' ratings of speaker attributes. Southern passages and Southern-accented speakers were rated as lower in status attributes and higher in sociality attributes than non-Southern passages and standard-accented speakers. However, for sociality attributes, ratings varied as a function of passage content for Standard, but not Southern, speakers.

Consistent with previous research, the two factors that emerged from the speaker attribute ratings broadly corresponded to the constructs of “status” and “solidarity” and were differentially associated with the Standard and Southern accented speakers. There is considerable evidence at this point that non-standard dialects or accents are judged as higher in sociality or solidarity while standard dialects or accents are judged as higher status (Demirci & Kliener, 1999; Edwards, 1982; Giles et al., 1992; Lambert, 1967; Preston, 1999). Much of the previous research had focused on dialects, however, in which changes in lexical choice or sentence structure could have led to differences in attitude judgments (Lambert, 1967; Levy & Cook, 2006; Luhman, 1990; Demirci & Kliener, 1999; Purnell, Idsardi & Baugh, 1999; Squires & Chadwick, 2006). The present finding confirms that the pronunciation of speech sounds themselves, or phonological variation, impacts listener's attitudes.

Significant differences in attitudes also appeared as a function of passage content. Passages were judged differently based on what the topic was, regardless of the accent type. In the case of status, this finding indicates that in some cases, message content may override speaker accent in the evaluation process and is consistent with previous work demonstrating the influence of lexical formality on listeners' ratings of intelligence and competence (Levin et al., 1994; Nath, 2007). However, the current finding extends previous work to the domain of passage content.

Crucially, passage content and accent type were not entirely independent. Passage content and accent type not only interacted for three of the individual attitude measures, but also for the overall composite measure of sociality. Attitudes toward standard-accented speakers changed depending on passage topic, while attitudes towards Southern-accented speakers tended to remain more stable across passage topics. One possible explanation may be that standard-accented speech does not necessarily elicit strong associations with sociality attributes. As a consequence, ratings of these associated traits were more flexible. In contrast, because Southern-accents are closely associated with attributes such as sociality, their ratings remained relatively more stable across conditions. In the case of a “marked” accent such as Southern English, the accent may be quickly used to generate initial expectations or biases about speaker traits. When the accent is considered standard, no specific expectations may be created, at which point other aspects of the communicative situation, such as content, may influence judgments about speaker attributes.

Although the current results strongly suggest significant effects of passage content and accent type on listeners' attitudes, there were limitations to the study. One concerns the range of passage topics that was examined. Certainly, the passages differed on a variety of dimensions other than whether each activity was typical of Southern regions. Future work will need to isolate the role of perceived typicality of content for a certain region from the myriad other ways in which passages can differ. Our results are also limited with respect to types of standard and non-standard accented speech that were used. Our standard speakers came from one particular region and our Southern speakers were representative of only a certain type of Southern speech. Although the results are consistent with previous work on attitudes toward Southern-accented speech in general, conclusions about the extent to which our findings can be generalized to other varieties of standard and non-standard speech should be drawn with caution.

In conclusion, this experiment examined attitudes toward common standard and non-standard American regional accents. Within this context, judgments of speaker attributes were influenced by a talker's particular speaking style as well as by passage content. The content of spoken language may act as one kind of linguistic context that frames the listener's overall assessment of characteristics of that speaker. Future research will need to take into account not only properties of a particular speaking style in explaining social evaluations in speech, but also properties of the communicative and linguistic setting in the form of conversational topic or content.