Abstract
Those who use non-indigenous accented speech often experience prejudice and discrimination, and in the United States, those speaking with Spanish accents are likely to be impacted. In research on speaker perceptions by type of accent, however, the gender of the speaker or of the perceiver has received less attention. In the present study, the impact of accent (North American- vs. Spanish-accented English), gender of speaker, and gender of rater on perceptions of competence were investigated in a sample of U.S. undergraduates. Participants heard a recording by either a male or female speaker who spoke in English with either a North American or Spanish accent. As hypothesized, Spanish-accented speakers were more likely to be judged negatively, female speakers were more likely to receive negative assessments, and male participants were more likely to show bias related to accent. The neglect of gender in the study of accent bias is discussed.
Perceived membership in out-groups, such as racial/ethnic minorities or immigrants, can lead to bias and discrimination (Park-Taylor et al., 2008). There is also evidence that accented speech is associated with out-group membership (Ng, 2007) and that people are good at recognizing when language is being spoken with an accent (Lindemann, 2003). Furthermore, those who use accented speech that is not indigenous to the area in which they reside often experience prejudice (e.g., Berk-Seligson, 1984). Specifically within the United States, Latino Americans speaking Spanish-accented English have been perceived relatively less favorably in research assessing perceptions of job applicants (Hosoda, Nguyen, & Stone-Romero, 2012; Purkiss, Perrewé, Gillespie, Mayes, & Ferris, 2006). Very little attention has been given, however, to the effects of either the gender of the speaker or of the perceiver on accent biases. Researchers and theorists in the multicultural and intersectionality arenas agree that gender is a significant identity that needs attention in combination with other social identities (e.g., Parent, DeBlaere, & Moradi, 2013; Reid, Lewis, & Wyche, 2014; Shields, 2008; Sue, Bingham, Porsché-Burke, & Vasquez, 1999). The purpose of this study is, therefore, to examine the effects of accent and gender of both speaker and participant on perceptions of competence and intelligence.
When we hear a person speaking, we immediately associate the particular style of language that we perceive with an identity for the speaker (Bergman, Watrous-Rodriguez, & Chalkley, 2008). It only takes 10 to 15 seconds for this assessment to be made (Gill, 1994). Depending on the identity we assign, we may impose a variety of biases, both positive and negative. For example, the use of Asian-accented English can elicit an association of perceived intelligence. This may be caused by the U.S. stereotype that suggests Asian individuals excel in technical and scientific areas (Carlson & McHenry, 2006; Madon et al., 2001). In addition, the use of British-accented English is sometimes associated in the United States with refinement and prestige (Carlson & McHenry, 2006), although U.S. students may still view such accents less favorably than they do North American-accented English (Gill, 1994). In general, the use of accents other than North American English in the United States tends to elicit a negative bias (e.g., Bergman et al., 2008; Ryan, Carranza, & Moffie, 1977).
Some research suggests that this negative reaction is caused by a perceived threat to the homogeneity of a culture. For example, those who speak with accents other than North American-accented English in the United States are often perceived as bringing outside cultural influences into the country (Cobas & Feagin, 2008). Neuliep and Speten-Hansen (2013) found that the higher the level of ethnocentrism in listeners, the more negative the perceptions of speakers with a non-native accent. Even those who speak with non-indigenous accents often rate the use of indigenous accents as more correct and appropriate (Clément, Noels, & Deneault, 2001; Ryan & Carranza, 1975). The desires to identify with the majority group may lead those with out-group accents to embrace the indigenous accents (Clément et al., 2001). Persons who self-identified as speaking with accents not native to the United States also rated themselves as lower on a measure of sense of belonging to the United States (Dovidio, Gluszek, John, Ditlmann, & Lagunes, 2010; Gluszek & Dovidio, 2010a).
The way people use language is not only associated with perceived cultural membership but also is used to identify other perceived characteristics of the user. For example, research involving Costa Rican Spanish has shown that phonological variables can cause differences in perceived levels of socioeconomic status (Berk-Seligson, 1984). The degree of perceived femininity in voices within gender has also been shown to affect U.S. undergraduates’ ratings of the speaker in a stereotypic direction (Ko, Judd, & Blair, 2006). Subsequent research by Ko, Judd, and Stapel (2009) showed that job applicants with voices rated as less feminine were judged higher on competence by undergraduate students in the Netherlands compared with those with voices rated as more feminine. In a review of the literature on use of the English language and attitudes, Cargile and Bradac (2001) concluded that there is much evidence showing a bias in English-speaking cultures against non-native English speakers on characteristics such as intelligence, competence, and social class (see also Carlson & McHenry, 2006; Gluszek & Dovidio, 2010b). On attractiveness, in contrast, Cargile and Bradac found mixed evidence, with some studies showing a negative bias and others not.
Both lines of research just reviewed suggest that characteristics such as competence may be more likely to produce biases when the speaker has an accent or sounds feminine. Competence and similar characteristics are stereotypically associated with males, and often referred to as instrumental or agentic traits. Two popular personality measures containing an instrumental trait scale are the Bem Sex-Role Inventory (Bem, 1974) and the Personal Attributes Questionnaire (Spence, Helmreich, & Stapp, 1975). Both measures contain items that can be characterized as reflecting high status and competence, such as feels superior, dominant, assertive, and independent (Gerber, 2009). Spence and Buckner (2000) showed that the instrumental items were still being perceived as more characteristic of men than women by U.S. undergraduates of any gender more than 20 years after the scales were first developed (see also Gerber, 2009). These findings suggest the possibility that female speakers may be more likely to be perceived negatively than male speakers when using accented speech and being assessed on instrumental characteristics.
Research has also shown differences between women and men in ways that might result in perceiver gender effects. Among German high school students, boys tended to overestimate their actual cognitive abilities, whereas girls tended to underestimate these abilities (Steinmayr & Spinath, 2009). This gender bias is also present when persons make judgments of others. The phenomenon has been found across many cultures and is attributed in part to male hubris and female humility (von Stumm, Chamorro-Premuzic, & Furnham, 2009). On measures of attitudes, men in the United States are more likely to endorse the use of stereotypes (Carter, Hall, Carney, & Rosip, 2006) and also report consistently less egalitarian attitudes than do women on explicit measures of sexism (e.g., Glick & Fiske, 1996; Moradi & Parent, 2013; Swim, Aikin, Hall, & Hunter, 1995). There also is a general tendency for White men to be more likely to express racist attitudes than are White women (Spanierman, Beard, & Todd, 2012). These gender associations suggest that male perceivers may be more likely to show stereotypic biases in favor of indigenous and male speakers than would female perceivers.
The use of gender in any form as a factor in the study of accent-based bias is relatively rare. The tendency to use male speakers only (e.g., Bradac & Wisegarver, 1984; Cargile, 1997; Cargile & Giles, 1998; Gill, 1994; Hosoda et al., 2012; Neuliep & Speten-Hansen, 2013; Purkiss et al., 2006; Ryan & Bulik, 1982; Ryan & Carranza, 1975; Ryan & Sebastian, 1980) has been noted by Hosoda, Stone-Romero, and Walter (2007) as well as others. Similar, but less commonly encountered, limitations include using only female speakers (Carlson & McHenry, 2006; Weyant, 2007), and either testing only one gender of participant or not testing for gender of participant effects, perhaps because of limitations in the population being sampled (e.g., Hosoda et al., 2007; Purkiss et al., 2006; Riches & Foddy, 1989; Ryan & Carranza, 1975; Ryan et al., 1977; Weyant, 2007). In one study, where both male and female non-indigenous accented speakers were used, participants were paired with same-gender actors (Lindemann, 2002).
Very few studies are available, therefore, in which both gender of speaker and participant are directly assessed. Podberesky, Deluty, and Feldstein (1990) had U.S. undergraduate students rate persons speaking English with either North American English, Spanish, or Asian accents on three dimensions (what they called competence, personal integrity, and social attractiveness). The pattern of findings was complex and not consistent with other work showing negative biases toward those speaking with Spanish accents (cf. Cargile & Bradac, 2001; Hosoda et al., 2012; Purkiss et al., 2006; Weyant, 2007). Female speakers were rated as both more competent and more attractive, and male participants rated Asian-accented speakers as more competent than did female participants when rating indigenous or Asian-accented speakers. Hosoda and Stone-Romero (2010) compared North American-, Japanese-, and French-accented speakers using a mock job application scenario. Speaker gender was not associated with any of the manipulation checks or dependent measures. Participant gender was varied but apparently either not analyzed or not significant; a variety of accent-related biases were detected, however, toward persons with Japanese and French accents.
In addition, studies that examine bias toward Spanish-accented English, including in relation to gender, are also limited. Besides the research cited above, an analog study was conducted by Arroyo (1996) to determine whether psychotherapists had any bias in treating Hispanic clients. The same person, a woman, was rated by therapists as having a poorer likely outcome when speaking with a Spanish accent than when not. Anderson and Smith (2005) found that Latina professors were rated as warmer when they were described as less strict. The available studies point to biases against persons speaking Spanish-accented English, but the role of gender is not clear.
In addition to the problems associated with understanding gender of speaker and gender of participant effects when those factors have not been consistently studied or varied, there is another methodological issue to consider: the use of relatively more or less obtrusive measures or manipulations. Some argue that more obtrusive measures will result in social desirability responding, in which groups who typically suffer from discrimination will be rated more highly or at worst similar to majority group members (e.g., Hosoda & Stone-Romero, 2010). A recent review of studies on employment discrimination by Stone, Hosoda, Lukaszewski, and Phillips (2008) showed, however, that larger effects were detected when measures were obtrusive versus unobtrusive, although the tendency may have been more pronounced when judgments being made were based on more ambiguous information.
In the present study, therefore, we assessed the effects of gender of speaker and type of accent on ratings of competence, using a relatively more ambiguous scenario. Male and female undergraduate students listened to a recording of a speaker reading a passage from a newspaper article. The speaker was either male or female, and spoke with either North American- or Spanish-accented English. The participants rated the speaker’s accent, gender, and several aspects of competence. The dimension of competence was selected as one more likely to be impacted by both gender and accent, compared with others such as likeableness. We assessed both communicative competence, which Hosoda and Stone-Romero (2010) felt should be particularly impacted by variations in accents, and more general competence, perceived speaker knowledge, and understanding of the material being read. The accent comparison, between North American-accented English and Spanish-accented English, was chosen because previous research has suggested that Spanish accents may be especially likely to be perceived less favorably by North American English-speaking samples on the competence dimension than are other types of accents (e.g., Cargile & Bradac, 2001; Lindemann, 2005), in addition to the relative neglect in the literature of an important U.S. ethnic group. Finally, we used a more obtrusive measure, in which participants gave their judgments about the speaker’s accent and gender before doing the rest of the ratings.
Intersectional and multicultural theorists agree that gender (among other identities) is important to consider when a person belongs to multiple social categories that may also differ in power or status within a culture (e.g., Purdie-Vaughns & Eibach, 2008; Reid et al., 2014; Shields, 2008; Sue et al., 1999). It is still not clear, however, exactly what the impact of the categories may be in combination (e.g., Penner & Saperstein, 2013; Warner & Shields, 2013). Thus, our hypotheses are offered somewhat tentatively with the understanding that there may be interactions not easily predicted from previous work. We hypothesized first that those speakers with Spanish accents would be rated less competent in terms of communication and knowledge than would those with North American English accents. Second, we hypothesized that bias against those with Spanish accents would be more likely to be observed when the speaker was also female. Third, we hypothesized that the two biases expected above would be more likely to occur when the participant was male.
Method
Participants
Students from a small campus of a state university in the Mid-Atlantic region of the United States participated in the study. Of the 80 persons who volunteered, 39 were female (49%) and 41 were male (51%). Participants were asked to indicate whether they were between 18 and 25 or ≥26, and virtually all (98%) were in the former category. One person did not answer the age question. With regard to race and ethnicity, 48 persons responded that they were White (60%), 20 African American (25%), six Hispanic (7.5%), two Asian (2.5%), two multiracial (2.5%), and two Other (2.5%).
Materials and Procedure
Four recordings of actors reading a few paragraphs from The New York Times article, “Biofuels Deemed a Greenhouse Threat” (Rosenthal, 2008), were used. The recordings were loaded into an Apple iPod MP3 player. Each recording was approximately 1 minute 30 seconds in length.
Two of the actors were native residents of the state in which the study was conducted. One was a female aged 53, and one was a male aged 28. They were employed as high school teachers. Their voices were used to create North American-accented English recordings. To create Spanish-accented English recordings, actors who were born in Latin American countries were used. The female actor, age 39, was born in Colombia and moved to the United States at the age of 24. She was also employed as a high school teacher. The male actor, age 41, was born in Chile and moved to the United States at the age of 25. He was currently employed as a repairperson, but had prior experience with acting and reading prose in group settings. Although the latter two speakers were not from the same Spanish-speaking country, previous research has indicated that undergraduate students cannot distinguish sub-categories of Spanish accents (Podberesky et al., 1990). There are, however, some age differences as well among the actors. The manipulation checks reported below, therefore, will be important to consider in judging the relative similarity of the speakers in regard to the accent condition.
The design used in this study was a 2 (participant gender: female, male) × 2 (speaker gender: female, male) × 2 (actor accent: North American English, Spanish) between-subjects factorial design. Following institutional review board (IRB) approval, the experiment was conducted in five sessions over a period of 12 days by five undergraduate experimenters (two male, three female). One experimenter asked participants in high-traffic areas of the campus whether they were willing to participate in a study. Those who agreed were taken to a more private area of the building where another experimenter read the oral recruitment script and gave prospective participants an implied consent form to read. The number of students who declined to participate was not officially tallied, but only a few declined. Participants were then randomly assigned to one of the four cells of the design. Neither the experimenter nor the participant had foreknowledge of which recording would be heard. After listening to a recording by means of a headphone and the MP3 player, participants were asked to complete the questionnaire. After finishing the survey, participants were offered a piece of candy as payment and were given a debriefing statement. All participants who initially agreed to participate did so and completed the study.
On the questionnaire, participants were first asked to rate the degree to which the recording was spoken with an accent (from 1 = none detected to 5 = distinct accent). Those who rated the speaker as having an accent were asked if they could identify the speaker’s country or indicate if they could not tell. They were next asked the gender of the speaker (female, male, or could not tell). Four questions were aimed at the competence dimension. To assess perceptions of communication competence, participants were asked to rate their ability to understand what the speaker was saying (from 1 = all to 5 = none) and their ability to understand the words in the article as read by the actors (from 1 = all to 5 = none). To assess more general competence (knowledge), participants rated the speaker’s ability to understand what they were reading (from 1 = full understanding to 5 = not at all), and the knowledge of the speaker on current events (from 1 = very knowledgeable to 5 = not at all knowledgeable). Finally, participants were asked for demographic information on their own age, gender, and ethnicity.
Scoring
Items on assessing communication competence and general knowledge were reverse scored so that higher scores indicated greater communication competence and higher knowledge. Correlations among these four items were examined. The two items that concerned the speaker’s communication competence were highly correlated, r(78) = .86, p < .001, as were the two items that concerned the speaker’s knowledge, r(78) = .78, p < .001. As a result, each of the two pairs of items was averaged. The internal consistency reliabilities for the respondents’ understanding of the actor scale and the perception of the actor’s understanding and knowledge scale were .93 and .87, respectively.
Even after averaging the two items on communication competence, the responses of the sample were still highly skewed to the left, with 42.5% of the sample using the highest rating for both items in the scale. As will be seen below, these high ratings were clustered in the North American English accent condition. In order to analyze the data, a categorical variable was formed by collapsing average ratings from 1 to 4 into lower communication competence and 5 into high.
Results
Manipulation Checks
To check on the effectiveness and comparability of the recordings, we first looked at the rated gender of the speaker. Almost all participants correctly identified the gender of the speaker (78/80 or 97.5%). The two who did not were a female participant in the North American English/male speaker condition and a male participant in the North American English/female speaker condition.
The other preliminary analysis used the dependent variable of rated accent (scale from 1 to 5, with higher = more accent) to check on the comparability of the speakers within gender and accent conditions. Table 1 displays the means and standard deviations for the analysis. A 2 (participant gender: female, male) × 2 (speaker gender: female, male) × 2 (speaker accent: North American English, Spanish) between-subjects ANOVA produced only one significant effect, a large main effect for speaker accent, F(1, 71) = 243.7, p < .001, d = 3.66. Actors with North American English accents were rated as having virtually no accent (M = 1.7) compared with those with Spanish accents (M = 4.5). Particularly relevant to the tests of the hypotheses concerning gender was that there were no main effects or interactions with speaker or participant gender. Accent ratings for the male and female speakers in each accent condition by participant gender were comparable.
Mean Rated Accent Level as a Function of Accent Condition, Speaker Gender, and Participant Gender.
Note. Possible scores range from 1 to 5, with a higher value indicating a greater level of perceived accent. One participant did not answer the question.
Participants were also asked an open-ended question about whether they thought the person was speaking with an accent, and if so, to identify from which country, or to indicate they were not sure. Many participants in the North American English condition did not answer this question, presumably because they did not consider North American English to be accented. Only two persons in the North American English conditions named a location outside of the United States (one said Europe and one said South America). Among those in the Spanish-accented English conditions, 13 (32.5%) said they did not know, could not tell, or left the question blank. Among those who did name an accent, 12 (44.4%) were correct. Among those incorrect, only one person said there was no accent, with the rest of the accents attributed to various other regions of the world. These rates of correct identification appear to be lower than reported in other studies (e.g., Carlson & McHenry, 2006; Hosoda & Stone-Romero, 2010), although in these other studies participants were not explicitly encouraged to indicate they were uncertain. Nonetheless, accuracy in identifying the speaker’s country of origin was not related to the perception of the degree of accent, or any other measures.
Speaker Characteristics
To test the three hypotheses, the effects of speaker accent, speaker gender, and participant gender were examined for the communication competence scale and the knowledge scale.
The communication scale, which had to be collapsed into a categorical high/low measure, was analyzed using a 2 (participant gender: female, male) × 2 (speaker gender: female, male) × 2 (speaker accent: North American, Spanish) × 2 (communication competence: high, low) log-linear analysis. The frequencies for this analysis are shown in Table 2. The first hypothesis, that speakers with Spanish accents would be rated as less competent than would speakers with North American accents, was supported. There was an interaction of speaker accent and communication competence, χ2(1) = 20.28, p < .001. Among those who heard a speaker with a North American accent, 65% rated the speaker as high in communication competence, compared with 20% who heard a speaker with a Spanish accent, d = 1.10.
Frequency of Rated Communication Competence (High, Low) as a Function of Accent Condition, Speaker Gender, and Participant Gender.
Note. High = rating the speaker as 5 on the communication competence scale; low = rating the speaker as 1 to 4 on the communication competence scale.
There was also an additional, unexpected interaction of participant gender and communication competence, χ2(1) = 9.46, p = .002. Male participants were more likely to rate the speakers as high in competence (56%) than were female participants (28%), d = 0.65.
The second scale focused on the perceptions of the speakers’ knowledge (see Table 3). A 2 (participant gender: female, male) × 2 (actor gender: female, male) × 2 (actor accent: American English, Spanish) between-subjects ANOVA on the knowledge scale produced two significant interactions. Supporting the second hypothesis, there was a Speaker Gender × Speaker Accent interaction, F(1, 72) = 5.38, p = .023. Simple effects tests indicated that the female speaker with the Spanish accent was rated as less knowledgeable (M = 3.1) than was the female speaker with the American English accent (M = 4.1), p = .002, d = 1.0.
Mean Rated Knowledge as a Function of Accent Condition, Speaker Gender, and Participant Gender.
Note. Possible scores range from 1 to 5, with a higher value indicating a greater level of perceived knowledge.
Supporting the third hypothesis was a Speaker Accent × Participant Gender interaction, F(1, 72) = 4.7, p = .042. Simple effects tests indicated that male participants rated Spanish-accented speakers as less knowledgeable (M = 3.2) than North American-accented speakers (M = 4.2), p = .004, d = 0.94.
Subsumed by the interactions was a main effect for speaker accent, F(1, 72) = 4.54, p = .037, showing that speakers with a North American accent were rated as more knowledgeable than were speakers with a Spanish accent.
Participant Ethnicity
As there was a bias detected against speakers with Spanish accents, we also conducted the above analyses without including the seven participants who said they were Hispanic (six who cited only Hispanic, and one who was multiracial). The patterns of results were the same for non-Hispanic participants.
Discussion
The three hypotheses were supported by our results. First, those speakers with Spanish accents were rated less competent in terms of communication and knowledge than those with North American English accents. Second, an increased bias against those with Spanish accents was observed on the knowledge measure when the speaker was female. Third, an increased bias against those with Spanish accents was observed on the knowledge measure when the participant was male.
The finding that speakers with Spanish accents were rated as less competent on two dimensions than speakers with North American English accents is consistent with previous research (e.g., Bergman et al., 2008; Ryan et al., 1977). Reviews of the literature on accent bias also agree that Spanish accents in the United States are prone to elicit bias (e.g., Cargile & Bradac, 2001). Hearing non-indigenous speech may activate negative schemas or stereotypes about ethnic groups and create concerns about communication difficulties (e.g., Gluszek & Dovidio, 2010a, 2010b; Sebastian & Ryan, 1985). Thus, the first finding established a consistency with prior work that can be used to assess the results from the second and third hypotheses tests.
The second finding of an increased bias against female Spanish-accented speakers compared with female North American English-accented speakers was contrary to the findings of Podberesky et al. (1990). The Podberesky et al. study was also, however, inconsistent in other ways with previous research, in that the results failed to show any bias against speakers with Spanish accents. What is more notable has been the neglect of gender as a factor in the study of accent bias across several studies. Although Hosoda et al. (2007) and Hosoda and Stone-Romero (2010) were among the few researchers who did vary the gender of the speaker in their research, they also were not studying Spanish accents. It is possible, for example, that the effects of speaker gender vary according to the specific accent in question. As noted earlier, intersectionality theories suggest that combinations of identities may result in different outcomes than examining one social identity in isolation (Parent et al., 2013; Purdie-Vaughns & Eibach, 2008; Warner & Shields, 2013).
There are other possible reasons for the inconsistencies in the prior findings. Stone et al. (2008) argued that recent employment discrimination research showed more bias when measures were obtrusive but based on ambiguous information. The studies reviewed above tended to be less obtrusive in their manipulations (Hosoda & Stone-Romero, 2010; Hosoda et al., 2007; Podberesky et al., 1990). In terms of ambiguity of qualifications, which would be lower if resumes or grade point averages (GPAs) were presented, Hosoda and Stone-Romero (2010) did provide resumes whereas the other two studies were not in a job context and simply asked for evaluations of the speaker. The present study employed a more obtrusive measure and provided no objective information about the speakers, such as a resume.
The last finding of an increased bias against those with Spanish accents when the participant was male is suggested by previous research (Carter et al., 2006) showing that men in the United States are more likely than women to endorse the use of stereotypes. This interaction was not, however, further modified by speaker gender, nor did men show any bias against female speakers. Male and female participants shared the bias against the female speaker with the Spanish accent, similar to Ko et al.’s (2009) study in which there were no gender of participant effects, thus indicating that both the male and female undergraduates showed a tendency to rate persons with less feminine voices higher on competence. It is possible that gender of the rater or observer may be a less crucial factor than the gender of the speaker.
The limitations to the study include the following. First, because there was no manipulation of obtrusiveness or ambiguity within the study itself, it is not possible to conclude that those factors account for the differences between the present and prior research. Second, there are limitations to using different speakers to represent accent and gender, in that there are other differences among individuals that can impact perceptions. Most recent research on accents, however, has used this procedure because it is believed to present more natural forms of speaking, aside from the difficulties inherent in finding individuals who can realistically switch accents (Hosoda & Stone-Romero, 2010; Hosoda et al., 2007; Podberesky et al., 1990). Finally, the ability to detect more complex interactions may have been hindered by the sample size.
A question and thus a potential limitation concern the generalizability of the present procedure to real-life contexts. Clearly, the present scenario is not directly applicable to employment situations where resumes and other information would be presumably available (Stone et al., 2008), and also where names as well as voices would presumably be available (Purkiss et al., 2006). A better analogy might be to teaching, where communication competence would be considered important and where detailed qualifications may not be available to the persons forming the impressions (i.e., students), although presumably at least some of the students would know the name of the instructor in addition to gender and accent. Gill’s (1994) study showed that U.S. undergraduates rated teachers with North American English accents more favorably than they did teachers with British or Malaysian accents. Although not an assessment of accents per se, Hamermesh and Parker (2005) found lower instructor ratings for women, minority, and non-native English-speaking faculty. Gluszek and Dovidio (2010b) concluded, in reviewing studies on graduate teaching assistants, that when rating persons with non-indigenous accents, prejudicial attitudes could be evoked based on perceptions alone.
It is, therefore, difficult to assess the relative impacts of various factors when so little attention in general has been given to how gender may interact with accents. Intersectionality theory research into the multiple factors that could impact these biases could open further doors for discussion. Gender (including identity), sexual orientation, race, ethnicity, linguistic abilities as well as cultural issues are some of the interconnected issues of both the participants and raters that should be considered (Reid et al., 2014; Shields, 2008). Furthermore, this lack of attention to gender is not limited to the accent literature. In the review of racial discrimination studies in the occupational domain published between 1991 and 2007 (Stone et al., 2008), a search of the 28 studies listed in Table 1 (Stone et al., 2008, pp. 245-248) shows only four in which gender was varied along with race. In regard to Spanish-speaking individuals or persons of Latino backgrounds, there has been much less attention given in general to prejudice and discrimination toward Latinos (Anderson & Smith, 2005; Dovidio et al., 2010; Hosoda et al., 2012; Purkiss et al., 2006), but what information is available suggests that the gender of both participants and those being judged also needs to be considered. In addition to the overall higher average incomes of Whites compared with Latinos (Dovidio et al., 2010), Latinas continue to lag behind men and White women in income. For each of the 30 years from 1979 to 2008, the median earnings for Latinas were lower than both the median earnings of White women and the median earnings of Latino men (U.S. Department of Labor, U.S. Bureau of Labor Statistics, 2009).
Thus, there is clearly a need for future research examining how gender of both perceiver and person being evaluated may interact with types of accents across situations that vary in their gender stereotypic nature and in their degree of ambiguity and cueing of gender and ethnicity. One important avenue for future research may be to use more implicit measures of bias. The study of prejudice has moved from the older more blatant measures, such as the Attitudes Toward Women Scale (Spence, Helmreich, & Stapp, 1973), to the less obvious measures, such as Modern Sexism (Swim et al., 1995), to the current emphasis on implicit associations (e.g., Greenwald, Poehlman, Uhlmann, & Banaji, 2009). Weyant (2005) recently developed an implicit measure, which showed that respondents held a stereotype of Hispanics as less intelligent than non-Hispanic Whites. Implicit measures may help answer some of the questions about the inconsistencies in past research, but more attention to the intersections of gender and other prejudices is clearly needed.
Footnotes
Acknowledgements
The authors thank Christopher Hobbs, Yidi Li, Shaneka McDonald, Dominique Millsapp, Melanie Nelson, April Rhodes, and four anonymous actors for their assistance in various aspects of this study.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.