Do You Speak Science? Dialect and its Role in Research Training

Abstract

Mastery of communication skills is critical to success in research careers. In this article we report on a study that focuses on a potential barrier to research career success that is virtually unstudied in academic interventions in higher education, specifically the possible effects of using a non-standard variety of English spoken by the student or trainee, often known as ‘dialect’. The term ‘dialect’ is often used as a blanket term denoting any variety of a given language that is associated with a specific region, ethnicity, or social status. More specific terms used in sociolinguistics are ‘dialect’ for regional varieties (Southern, Midwestern, Appalachian) and ‘sociolect’ for ethnic, cultural, and class-based varieties. Here, we examine how these varieties influence scientific communication (SciComm) in the academic research training environment as well as the related perceptions of both trainees and their mentors. As part of a larger study, we surveyed 124 mentor/trainee dyads over 4 time points; for the present analysis, we included only those dyads where trainees reported a language variety (n = 139 English speakers or bilingual, (n=16)) and related to perceived comfort within the research environment as reported by respondents endorsing a non-standard home variety of English compared to their counterparts. We found that those trainees raised speaking non-standard dialects or sociolects (‘home dialect’) were more likely to report discomfort in the research environment compared to their counterparts who were raised speaking Standard Academic English (SAE) (regardless of ethnicity or social class). The speakers of non-standard varieties were not necessarily using those varieties in the research environment. The number of trainees reporting discomfort was higher than the number of trainees identifying as belonging to a minority ethnic group or first-generation status. In the same study, we evaluated matched mentors’ perceptions of trainee performance and motivation in SciComm activities and found that mentors (without knowing the trainee’s self-reported dialect or sociolect) perceived higher barriers to mentoring non-standard variety speakers on 3 items. Given these preliminary findings, we suggest that language variety may play an important role in mentor-mentee interactions in the research environment and speculate that it could be a more sensitive indicator of trainee’s perceived and actual discomfort in academia than race-ethnicity or first-generation status. We discuss implications for both mentors and trainees, as well as potential interventions.

3. Introduction

3.1. Sociolinguistics

Linguistics is the analysis of language and languages. It encompasses phonology (study of speech sounds), analytical grammar (morphology, syntax), historical linguistics (change and development over time, language families), sociolinguistics (Wolfram & Schilling-Estes, 2006), and much more. For the purposes of this work, we will focus on sociolinguistics, which is the study of how people use language in a social context, including the use of varieties of a single language (Labov, 1966). ‘Language variety’ is a term of art that is familiarly known as ‘dialect’ or sometimes ‘accent’. From a sociolinguistic perspective, all varieties of all languages comprise internally consistent, naturally occurring patterns; the patterns used by an individual depend on the social and regional context (Labov, 1966). All language varieties encompass unique patterns and norms of pronunciation, vocabulary, grammar, tempo, intonation, and communication style. Because these aspects are used systematically and perceived by other speakers of that variety as normal, they are, by definition, not “mistakes” or “underdeveloped.” A sociolinguist, thus, would not consider the English sentence ‘Ah ain’t sure’ as “mistaken” in any sense and certainly not an indicator of intelligence or character. The sentence ‘Not oh sure am’ would, on the other hand, be considered mistaken or nonsense, since no group of English speakers shares and recognizes this usage. A sociolinguist, in other words, follows a descriptivist, non-judgmental view of language.

In society in general and academia in particular, the variety of English known as Standard Academic English (SAE) is considered the “correct” and preferred way to speak English, because it follows the patterns of grammar, usage, and pronunciation shared by a particular group of people and is sometimes called the ‘prestige dialect’. This variety is usually identified as the one used by national newscasters. However, the patterns are largely arbitrary, i.e., they reflect precedence and preference of certain people in certain contexts. Most societies around the world recognize a prestige dialect and one or more stigmatized dialects and hold strong views on the inherent superiority of the prestige dialect. Sociolinguists refer to this perspective as prescriptivism. Purnell, Idsardi, and Baugh (1999) suggest that this commonly shared prescriptivist viewpoint can be detrimental to how we perceive speakers of stigmatized dialects, who have less exposure to the norms of the prestige dialect and may or may not choose to acquire it. Societies everywhere are sensitive to dialect as a marker of class membership. While SAE may be the accepted language style of English-speaking academia, it is not “correct” for all contexts of speaking English and is not structurally ‘superior’ to other varieties. In the descriptivist understanding, claiming this would be similar to claiming that the song of a white-throated sparrow is superior to the song of a house sparrow.

As we grow, we develop a repertoire of language patterns that allows us to adapt to different social and professional contexts. For example, someone from a small Midwestern town may have a regional dialect (i.e., language variety based on geographic region) that they adapt once they move to a city in the South or the upper East Coast. Alternatively, another person may grow up speaking a sociolect (a language variety informed by membership in a socially-defined category, such as socioeconomic status (SES), ethnicity, or urban/rural background) that is different from the language variety that they use in their work environment (Lippi-Green, 1997). Contrary to our preconceptions, dialects and sociolects do not represent slang, poor grammar, a “half” language, or a measure of intelligence (Lev-Ari & Keysar, 2010). Importantly, although language variety is not reflective of intellectual ability, in nearly every society with an educated class, people tend to associate them. That we make such associations is almost entirely out of awareness for most people in daily life. Other studies have shown that we unconsciously profile people based on their language variety, and, as a consequence, treat them differently (Baugh, 2000; Purnell et al., 1999). According to social psychologists, the dominant group tends to judge dialect and sociolect speakers as less credible, trustworthy, or knowledgeable, regardless of the content of their speech (Lev-Ari & Keysar, 2010).

3.2. Academia, the scientific research environment, and language variety

Students entering the academic environment who arrive without having learned SAE in the home gradually acquire SAE through social exposure and through direct instruction. Because speakers of a language intuitively associate language variety with unrelated personal characteristics (Baugh, 2000; Cameron et al., 2013; Cooper, 1986; Lev-Ari & Keysar, 2010, p.; Lippi-Green, 1997; Preston, 1997; Tannen, 1994; Valdes, 1988), speakers of lower-prestige varieties of English are often misjudged by speakers of SAE, such as faculty and administration. As a student moves through the academic ranks, she typically becomes proficient at SAE, knowing that it is critical for acceptance into the group, whether she chooses to use it or not. This process does not always occur seamlessly. An illustrative example is the following quote from Khan (2011, p. 102), made by an African-American high school student of working-class background studying at an ultra-elite private school, speaking to her teacher:

“I do really well here,” she began, “… but I did really badly at first. And I thought it was because I was stupid or something. Like I didn’t get it. But that was the thing, I didn’t get it. I wasn’t dumb.… [But] until I learned how to talk like you wanted me to talk, how to write how you wanted me to write, I was dumb.”

Once a speaker has mastered and uses the language variety of a particular group, she is more likely to be accepted as a member of the group. This passage highlights how the process of language usage and adaptation forms a foundational part of a person’s identity, sometimes out of awareness of the speaker herself, as well as the instructor’s possibly inaccurate assessment of her academic ability.

Research training at the tertiary level requires yet another phase of linguistic socialization, from SAE to scholarly scientific style. Changes in the demographics of both mentors and trainees over recent decades have brought many more traditions and styles of discourse into the picture (Aitchison & Lee, 2006), increasing the complexity of imparting scientific style to trainees by mentors who are themselves of highly diverse backgrounds and linguistic traditions.

Research trainees, in particular, strive to develop their identity as a scientist, an important factor in persisting in STEM fields (Chemers, Zurbriggen, Syed, Goza, & Bearman, 2011; Estrada, Woodcock, Hernandez, & Schultz, 2011; Hurtado, Cabrera, Lin, Arellano, & Espinosa, 2009; Nettles & Millett, 2006; Remich, Naffziger-Hirsch, Gazley, & McGee, 2016; Schulz, 2016).

Learning to speak “like a scientist” is a part of this process (Remich et al., 2016; Thiry & Laursen, 2011). By the time students reach graduate school, dialect or sociolect speakers are typically “fluent” in SAE, but research training at the graduate level and beyond requires a much more specialized variety of English. Scientific English must be mastered in order for the trainee to publish and formally present their research and from there to gain a career research position.

As part of a broader study of how trainees acquire and mentors teach scientific communication (SciComm) skills, we sought to determine whether doctoral and post-doctoral trainees’ linguistic background played a role in the development of SciComm skills and in their interactions with their mentors.

4. Methods

4.1. Quantitative survey

We created a longitudinal, dyadic trainee/mentor survey based on a previous cross-sectional survey of SciComm skill development (Cameron et al., 2013), grounded in a modified version of Social Cognitive Career Theory (SCCT) (Lent, Brown, & Hackett, 1994, 2000). SCCT emphasizes the role of social cognitive variables in career choice and pursuit, particularly self-efficacy beliefs, outcome expectations, interest, and goals, which were of interest to our study.

Data were collected over 4 time points during an 18-month period. The complete survey included over 250 items for trainees and over 150 items for mentors. Trainees and mentors were aware of each other’s participation but did not see each other’s responses. The majority of topics covered procedural details of how trainees acquired scientific communication skills, including scholarly writing, formal presentation, and spontaneous research conversations, as well as social-cognitive constructs such as self-efficacy, outcome expectations, and contextual factors for SciComm skill development. Mentors were asked about their mentoring practices, style, and attitudes about trainee SciComm skill development.

Our analysis detailed below focuses on the items related to language variety and discomfort (for trainees) and barriers to SciComm mentoring (for mentors). Results of the survey concerning aspects unrelated to language variety are currently being analyzed.

4.2. Participants

Eligibility criteria for the mentor included (1) being a researcher who is currently the primary mentor or has mentored within the last 3 years and (2) having or having had at least one research trainee at the doctoral or postdoctoral level. Mentors were recruited nationally via e-mail messages or invitation cards sent by a Scientific Communication Skills Study project coordinator. Of the 208 candidates who showed initial interest, 185 met the eligibility criteria (i.e., had a matched trainee who responded) and agreed to participate by providing online informed consent.

Specific mentees paired to each mentor (nominated by the mentor) were then contacted to take the trainee survey. The research team selected one of the nominated trainees and issued an invitation to participate in the survey. Selection was partially based on achieving sufficient participants for demographic categories of gender and race/ethnicity. Mentors were not informed which of their trainees was invited and declined to minimize concerns of trainees. If the invitation was declined, the team invited the next trainee on the list. If the invitation was accepted, the trainee completed informed consent, and the mentor was informed of the identity of the trainee so that the dyad could be created and respond appropriately. Each of the four surveys (for both mentors and their matched mentees) took about 20 minutes to complete. Data were collected from the autumn of 2014 through the spring of 2016; the four waves of the survey were issued over a period of approximately 18 months. Mentors received up to $30 per questionnaire if they filled out at least 90% of the questions (a total of $120 in gift cards for the entire study). Mentees received up to $40 per questionnaire if they completed at least 90% of the questions. To encourage mentor retention in the study, trainees received an additional $10 and $30 if their mentor completed the questionnaire at the 1^st and 4^th time points, respectively. In total, mentees who completed the questionnaire at all 4 time points and whose mentors completed the questionnaire at all 4 time points received a total of $200 in gift cards for the entire study. All participants signed an informed consent document before participating in the study.

4.3. Trainee items and procedures

In addition to basic demographic information, we collected information about participant language variety at Time 1. Prior to asking about dialect, we separated native English speakers (L1, n = 123), bilingual speakers (n = 16), and English-language learners (L2, n = 139). (Items 1–2, Appendix A: Trainee Survey Questions). Afterward, we asked native speakers of English or bilingual English/other speakers to select the regional or social language variety that they spoke growing up from a list of 9 choices, which included ethnic sociolects, regional dialects, and Standard English (Item 3 of Appendix). Trainees who claimed a native language other than English did not answer these questions, with the exception of Spanish-English (S-E) bilingual individuals who had resided in the US for more than 10 years. The list of varieties included regional dialect (Southern, Texan, Cajun, Appalachian, Northeastern, and Midwestern) and ethnic sociolect categories (Spanish-influenced English, Black English/Black English Vernacular/African-American, and Asian-influenced English). An ‘Other’ category was provided which could be filled in with the participant’s self-identified variety.

We then asked all English-speaking and bilingual (S-E) participants to rate their discomfort on the following items: feeling “judged unfavorably based on the way [the trainee] speaks English” (Item 4, Time 1), level of comfort in “speak[ing] home style of English to mentor or other senior scientists” (Item 5, Time 1), frequency of “feeling uncomfortable at work or school because of the way you speak English (Item 7, added at Time 2) in agreement with “reconsider[ing] my goal of pursuing a research career based on the writing and speaking needed to be successful” (Item 6, Time 1). Notably, none of the discomfort items assumed that the trainee was using their home language variety in the research environment.

4.4. Mentor items

In addition to surveying the trainees about their language variety and level of comfort communicating in a research environment, we also surveyed their mentors about the barriers they faced with their trainees (see Appendix B: Mentor Survey Questions). We have created and validated scales for SciComm mentoring practices specific to three modes of SciComm (writing, presenting (rehearsed), and speaking (unrehearsed; e.g., asking questions or speaking up in meetings)) (C. B. Anderson et al., 2016). A more detailed description of the item development and psychometric properties of the measures is discussed elsewhere (C. Anderson, Chang, Lee, Baldwin, & Cameron, 2016).

Mentors were not aware of the trainee’s self-identified language variety during the course of the study. The results of this survey will specifically focus on perceived barriers (Items 13, 17, and 24) related to “language differences between trainee and mentor that make it difficult to mentor in SciComm skills,” “[trainee’s] lack of ease using professional academic language,” and “trainee attitude or personality problems (e.g., shy, arrogant),” respectively. Mentor-reported trainee barriers were measured at Time 2 and again at Time 4. All items were measured on a 5-point Likert-type scale (1 = a very insignificant barrier to 5 = a very significant barrier) and were pretested and revised prior to their administration.

Regarding statistical analysis, a series of two-way repeated measures analysis of variance (ANOVA) using IBM SPSS statistics 24 was conducted to compare the mean differences among three groups of trainees (sociolect, dialect, and SAE speakers) for 4 discomfort items and to examine the interaction between time (Time 2 and Time 4) and trainee language variety group on the data collected for perceived mentor barriers. To see which specific groups within each time were significantly different from each other, post hoc tests were conducted (p < 0.05 indicates a significant difference).

5. Results

5.1. Sample description

Demographics of respondents are shown in Table 1 and represent those who were engaged nationally in biomedical research training. Of the 185 initial dyads (includes L1 and L2 trainees) eligible to participate, we had a total of 139 L1 trainees and 44 L2 trainees (missing n = 2) complete Time 1 of the survey. L2 trainees were excluded from this study.

Table 1:

Demographics of trainee study population (at Time 1) by language status, gender, Hispanic/Latino, race/ethnicity, and generational status × rank

		Rank
		Doctoral students	Postdoctoral fellows	Total
Language status	English speaker or bilingual (N=16)	108	31	139
Gender	Female	70	21	91
	Male	38	10	48
Hispanic/Latino	Hispanic	11	6	17
Race/Ethnicity	American Indian	1	1	2
	Asian or Asian-American	12	2	14
	African-American	17	4	21
	White	60	18	78
	Other or More than one race	7	0	7
	Missing or don’t know	11	6	17
Generational status	First-generation to attend college	27	8	35
	Legacy	80	23	103
	Missing or don’t know	1	0	1

Trainees who grew up speaking a language other than English (L2, N=44) were excluded from this analysis.

For the mentor data, we had a total of 131 mentors of L1 trainees who completed Time 1 of the survey (mentors of L2 trainees excluded from this analysis, n = 40). A total of 124 matched mentor/trainee dyads, 30 unpaired trainees, and 15 unpaired mentors completed the survey at Times 1–4.

5.2. Trainees’ responses

Trainees’ reported home language variety by race/ethnicity is summarized in Table 2.

Table 2:

Demographics of trainee study population by language variety × ethnicity

		Trainee Race/Ethnicity
		Latino/a	Asian-Am	African-Am	White/Anglo	Other	Total
Language variety	Ethnic sociolect	9	0	5	0	0	14
	Regional dialect	1	12	2	19	1	35
	Standard Academic English (SAE)	6	5	13	58	8	90
	Total	16	17	20	77	9	139
	Frequency Missing = 47 (L2 or dialect status not specified)

Ethnic sociolect = Black English/Black English Vernacular/African American Vernacular English; Spanish-influenced English (or “Spanglish” or “Tex-Mex”); Asian language-influenced English or mixture of English and an Asian language

Regional dialect = Southern, Texas, Cajun, or Appalachian accent; Northeast urban accent (Boston, New York); Midwestern accent (Chicago, Wisconsin, Minnesota)

Trainees’ home language variety showed higher percentages of discomfort in the research environment than did generational status (first-generation or legacy) or race/ethnicity (under-represented minority (URM) vs well-represented groups). Figure 1 summarizes trainee responses (% within sub-group) to discomfort items, by language variety and by generational status. Due to small sample sizes, the results are presented as bubble plots – where larger bubbles indicate greater percentages within a subgroup – to explore potential trends for future studies. Refer to Appendix C: Tables of Counts/Percentages for Discomfort Questions for percentage and counts for each item by generational status, race/ethnicity, and language variety.

Figure 1:

Discomfort items by generational status, under-represented minority (URM) status, and language variety. Data reported as percentage of total within sub-groups (first-generation, legacy, URM, well-represented, ethnic sociolect, regional dialect, and Standard Academic English (SAE)). A larger circle indicates a larger percentage within a particular sub-group. For a full table of percentages and counts for each bubble, see Appendix C: Tables of Counts/Percentages for Discomfort Questions.

For Item 1A, “uncomfortable speaking home style of English to mentor or other senior scientists,” we observed that ethnic sociolect and regional dialect speakers had two distinct clusters of trainees: those who reported that they were comfortable or very comfortable and those who reported they were uncomfortable or very uncomfortable; on the other hand, SAE speakers generally reported complete comfort. Similar trends to SAE speakers were observed regardless of generational status or race/ethnicity.

For Item 1B, “uncomfortable at work or school because of the way trainee speaks English,” we found that ethnic sociolect speakers again had two distinct clusters (never/rarely or most/all of the time), while all other language varieties, races/ethnicities, and generational statuses reported never/rarely (i.e., complete comfort) in general.

Due to small sample sizes, we condensed Item 1C, “judged unfavorably because of the way I speak English,” into two groups: those who responded “definitely not/probably not,” and those who responded “definitely/very probably/probably.” Again for this item, we observed two clusters for the ethnic sociolect speakers and regional dialect speakers, but SAE speakers reported disagreement with this statement. URM trainees also reported higher levels of discomfort (comparable to ethnic sociolect and regional dialect speakers) than did well-represented groups.

Lastly, trainees responded to Item 1D, “thinking about the writing and speaking necessary to be successful makes me reconsider my goal of pursuing a research career,” with slightly different trends from the previous items. Again, ethnic sociolect appeared to have the most distinct clusters (either “strongly disagree/disagree” or “agree/strongly agree”), but these clusters were also seen for the other groups. Regional dialect, URM, and first-gen trainees showed a similar distribution of responses, whereas legacy, well-represented, and SAE trainees aligned similarly. For this item in particular, all trainees reported a higher level of discomfort regarding the writing and speaking necessary for pursuing a research career.

5.3. Mentors’ responses to perceived barriers to mentoring trainees in SciComm

The mentor survey included three sections regarding perceived barriers to mentoring trainees in SciComm (Figure 2); these sections were organized according to barriers specific to the mentor, barriers specific to the institutional environment, and barriers specific to their paired mentee, with multiple items following the stem “How significant are the following [institutional, personal, trainee] barriers to you in mentoring your trainee in scientific communication skills?” Within the mentee barriers category, three items revealed meaningful differences in the mentors’ perceptions of their mentee according to the language variety endorsed by the trainee. Although mentors were not aware of the mentee’s reported language variety, they reported higher barriers in mentoring sociolect speakers. The items identified included Item 2A: “[mentee’s] lack of ease using professional academic language,” Item 2B: “language differences between the trainee and mentor that make it difficult for me to mentor them in oral and written communication,” and Item 2C: “trainee attitude or personality problems (i.e., shy or arrogant).” From Time 2 to Time 4 (approximately 9 months), these barriers became more pronounced for all three items (Figure 2).

Figure 2:

Mentor responses (n =131 at Time 1, with ~27% attrition by Time 4) about their paired mentees, sorted by mentee language variety (ethnic sociolect, Standard Academic English (SAE), and regional dialect). Data are reported as means. 1=“A very insignificant barrier,” 5= “A very significant barrier.” Main effects of time or group are reported for p < 0.05 by repeated measures ANOVA.

Figure 2 shows the results of the repeated measures ANOVA by trainee language variety and time. For all items (Figure 2A–C), there were no interaction effects between time and language variety, but the effect of time was significant (p ≤ 0.001), indicating that mentors’ responses about trainee barriers changed (i.e., increased) over time. In addition, for two items (Figure 2A and C), group effects of language variety were significant. That is, regardless of time points, mean differences in these two items were significantly different (p = 0.001) among the three trainee language varieties (sociolect, dialect, and SAE). More specifically, for the item asking “lack of ease using professional academic language” (Figure 2A), mentors of sociolect trainees reported higher barriers than mentors of regional dialect or SAE trainees (p = 0.001). In addition, for the item asking about trainee attitude or personality problems (Figure 2C), mentors of sociolect trainees reported higher barriers than mentors of regional dialect or SAE trainees (p = 0.001). Mentors of regional dialect trainees also reported higher barriers than mentors of SAE trainees (p = 0.001).

Effects were not observed for other items regarding trainee barriers, such as trainee’s lack of skill in writing or speaking, denial of deficits in SciComm, or disregard for feedback.

6. Conclusions and Discussion

6.1. The experience of speakers of non-standard language varieties in the research environment

These preliminary findings suggest that language variety, like race, ethnicity, or generational status, can impact trainees’ level of comfort in the research environment and may affect how mentors perceive them. From trainee discomfort items on our survey, language variety was more likely to be associated with discomfort than race/ethnicity or status as a first-generation college student. Ethnic sociolect speakers were somewhat more likely to convey discomfort than regional dialect speakers. What is clear is the prevalence of two distinct sub-groups of dialect and sociolect speakers: those who experienced discomfort and those who did not regarding their home style of English.

We observed that a higher percentage of dialect and sociolect speakers perceived themselves to be “judged unfavorably at work or school” because of the way they speak English. Furthermore, without knowledge of their language variety (or discomfort), their mentors attributed greater barriers to SciComm mentoring for trainees who grew up speaking a sociolect. Taken together, these findings suggest that even if trainees use Standard Academic English in the research environment, their home language variety may play an important role in their sense of comfort in the academic workplace as well as in their mentors’ assessment of their abilities and attitudes.

Recent research on the concept of ‘discomfort’ in the academic setting has suggested that undergraduate students of under-represented ethnic groups who experience discomfort in an academic environment are less likely to persist, and that mentoring can play an important role in influencing persistence (Aguinaga & Gloria, 2015; Gloria, Castellanos, & Herrera, 2015). Discomfort in the academic research environment because of the way the trainee speaks or writes in English may be similar to discomfort included in the cultural congruity scale established by Gloria and authors (Aguinaga & Gloria, 2015; Gloria et al., 2015), where Latina/o community college students reported that “[they] feel [they] have to change [themselves] to fit in school.” Although our analysis specifically focused on discomfort of doctoral and postdoctoral biomedical sciences trainees with different language varieties, we found our results to be consistent with the literature on discomfort for URM students. All trainees, regardless of language variety, reported increased discomfort with the item “thinking about the writing and speaking necessary to be successful makes me reconsider my goal of pursuing a research career,” highlighting the important but frequently neglected role of linguistic development in research training and career choice.

6.2. How can mentors approach SciComm mentoring for their diverse trainees?

Underlying both the scholarship and implementation of recent approaches to effective mentoring is an increased emphasis on mentor agency (Abedin et al., 2012; Berk, Berg, Mortimer, Walton-Moss, & Yeo, 2005; Byars-Winston, Branchaw, Pfund, Leverett, & Newton, 2015; Estrada, Eroy-Reveles, & Matsui, 2018; Pfund et al., 2014; Pfund, Pribbenow, Branchaw, Lauffer, & Handelsman, 2006; Remich et al., 2016), a mentor’s awareness and application of the possibilities, responsibilities, and opportunities afforded by their role, with one of their key responsibilities being to socialize the trainee into a disciplinary community of practice (Golde, 2005; Lave & Wenger, 1991; Morton, 2012, Thiry 2011). Mentors can support their trainees in becoming authentic junior members of the discipline by providing frequent opportunities for legitimate peripheral participation, i.e., not just assigning exercises for practice or observing professional activities and interactions from the sidelines, but taking guided steps toward performing them. A major pathway for entering the community of practice is achieving competence in SciComm, especially discipline-specific competence, which cannot be taught in a generic classroom or workshop. Ultimately, the traditional practice of discipline-specific SciComm pedagogy is founded on apprentice-style learning between the mentor and the trainee (Aitchison & Lee, 2006; Maher, Feldon, Timmerman, & Chao, 2014).

Yet mentors often report that training their mentees in SciComm can be fraught with challenges and frustrations (Aitchison, Catterall, Ross, & Burgin, 2012; Cameron et al., 2013; Kranov, 2009). Mentors from this study reported increasing barriers to SciComm mentoring over time for all trainees, regardless of language variety. A possible explanation for the perceived increase in barriers is that trainees are likely to be writing, publishing, and presenting more in the later stages of their training, for example writing dissertation proposal and chapters and publishing research articles based on the dissertation, reporting results of the culmination of experiments, etc. Mentors who have some awareness of the impact of language variety on trainees and on their own mentoring may be able to use those insights to mentor SciComm skill development more responsively.

6.3. Study strengths

To our knowledge, this is the first examination of long-term dialect effects past the K-12 level as well as of the influence of language variety in personal relationships. Most studies of language variety in subjects past the childhood stage are experimental or quasi-experimental and test instantaneous reactions of subjects listening to a variety being spoken. In contrast, the results of this study suggest “shadow” effects of dialect even when the home dialect is not necessarily used in the work environment. In other words, trainees do not consciously attribute academic discomfort to their home language variety, despite the salient role it plays in their identity and sense of belonging.

In addition to studying long-term dialect effects on trainees, we also looked at their mentors’ behaviors and perceptions. One advantage to investigating mentor-trainee pairs was that we were able to track the effects of mentoring attitudes and behaviors in specific relationships over time. While sociolect speakers disclosed greater discomfort in the academic research environment, mentors reported, for some categories, significantly higher barriers to mentoring this group of trainees (without knowing that they identified as sociolect speakers). These observations raise questions about the role of language variety in bias in academia aside from gender and visually-observed race or ethnicity.

6.4. Study limitations

Because of small sample sizes for dialect and sociolect speakers, it was challenging to distinguish statistical differences across dialect, sociolect, and SAE groups of trainees, although the size of our mentor sample allowed us to identify statistically significant differences for mentors’ practices and attitudes. We sometimes experienced reluctance to responding to items about dialect or sociolect from our trainee participants. We found that these questions sparked sensitive or confused reactions from trainees (e.g., “Can you define how each of these is different from Standard English?” and “I’m Black but I’m not sure what is Black informal English,” “why am i not allowed to skip questions, that violates my right as a participant”). This response to reporting language variety may have resulted in identification of fewer dialect/sociolect speakers within our participant population than actually participated. In future work, we plan to provide more explanatory background for these questions to guide survey participants and also to allow the selection of both a regional dialect and an ethnic sociolect by the same participant. While these are preliminary findings, we have identified language variety as an important topic that warrants further investigation regarding identity and persistence as a biomedical scientist.

6.5. Implications and interventions

If these findings are reproduced, language variety may prove to represent a more specific indicator of sociocultural discomfort or perceived lack of belonging in the research environment than race, ethnicity, or parents’ education level and as such a possible area for intervention. In order to support trainees’ persistence in pursuing research careers and their entry into the research community of practice, mentors and administrators would benefit from learning about the role of language variety in social interaction and individual development; such learning could be part of a more general training of mentors working with diverse trainee groups. Trainees as well would benefit from learning about language variety vis-à-vis cultural diversity. The goal of such learning is not to change or suppress the trainee’s usage of language but to surface and discuss possible out-of-awareness assumptions of both mentor (such as unconscious bias), and of trainee (such as imposter syndrome).

As a step to address this and other issues of SciComm skill acquisition and mentoring, our group has initiated a mentor-training workshop program for mentors which addresses language variety along with other areas of SciComm mentoring, funded by the National Institute of General Medical Sciences (R25 GM125640). As a research component of this study, we will collect additional information on trainee language variety and level of comfort in the research environment, as well as on mentor perceptions and attitudes, in order to reinforce our analysis. As we continue to identify and address diverse trainee needs, we seek to answer the following questions:

What do people think about and feel when they think about language variety?

Is dialect or sociolect a marker of bias or disadvantage?

How do we express trust and respect when talking about language variety?

Appendix A: Trainee Survey Questions

Dialect/sociolect questions

• 1
  Is English your primary language?

  1. Yes
  2. Bilingual
  3. No
  4. If you answered “Bilingual” or “No” specify other language:
• 2
  Did you speak English fluently before the age of 14?

  1. Yes
  2. No
     At this point, non-Native English speakers (L2, N=44) were excluded from the study.
• 3
  Would you say you grew up speaking one of the following regional or social dialects of English?

  1. Black English/Black English Vernacular/African American Vernacular English
  2. Spanish-influenced English (or “Spanglish” or “Tex-Mex”)
  3. Noticeable Southern, Texas, Cajun, or Appalachian accent
  4. Noticeable Northeast urban accent (Boston, New York)
  5. Noticeable Midwestern accent (Chicago, Wisconsin, Minnesota)
  6. Asian language-influenced English or mixture of English and an Asian language
  7. Standard English without noticeable accent, grammar, or vocabulary differences
  8. Other:_________
  9. English is not my primary language

Discomfort questions:

• 4
  If you are a native English speaker, have you been judged unfavorably because of the way you speak English?

  1. I did not speak English growing up
  2. [1: Definitely not – 5: Definitely]
• 5
  How comfortable would you be speaking your home style of English to your mentor or other senior scientists?

  1. [1: Completely comfortable – 5: Very uncomfortable]
• 6
  Thinking about the writing and speaking necessary to be successful makes me reconsider my goal of pursuing a research career.

  1. [1: Strongly disagree – 5: Strongly agree]
• 7
  How often do you feel uncomfortable at work or school because of the way you speak English?

  1. [1: Never – 5: All of the time]

Appendix B: Mentor Survey Questions¹

How significant are the following issues with the TRAINEE YOU ARE MENTORING IN THE CURRENT STUDY barriers in your mentoring scientific communication skills?

(Very insignificant barrier, insignificant, neither significant nor insignificant, significant, very significant barrier)

• 13. Language differences between trainee and mentor that make it difficult for me to help them in writing and oral communication.

• 15. Trainee’s weakness in basic English writing skills (poor skill in sentence structure, word usage, grammar)

• 16. Trainee’s weakness in basic English speaking skills (problems with pronunciation, word usage, grammar)

• 17. Lack of ease using professional academic language

• 18. Refusal of available education resources (e.g., will not utilize seminars, classes, software, books, etc., to improve writing or speaking)

• 19. Insufficient effort (i.e., trainee is not willing to put in the necessary time and work to write or speak well)

• 20. Inappropriate expectations of me as a mentor (e.g., trainee over-rely on me to edit and correct their work)

• 21. Denial of their deficits (e.g., trainee believes their writing and speaking skills are adequate)

• 22. Dread of negative feedback (i.e., trainee does not like criticism of their writing and oral presentations)

• 23. Inefficiency or mismanagement of time and effort in preparing writing projects or presentations

• 24. Trainee attitude or personality problems (e.g., shy, arrogant)

• 25. Disregard for my feedback on their writing

• 26. Disregard for my feedback on their speaking and presentations

Appendix C: Tables of Counts/Percentages for Discomfort Questions

	Uncomfortable speaking home style of English to mentor or other senior scientists
	Completely/ comfortable	Neither	Very/ uncomfortable	Total
First-generation	26 (74%)	3 (9%)	6 (17%)	35
Legacy	86 (83%)	7 (7%)	10 (10%)	103
URM	30 (77%)	4 (10%)	5 (13%)	39
Well-represented	81 (84%)	6 (6%)	10 (10%)	97
Ethnic sociolect	8 (57%)	2 (14%)	4 (29%)	14
Regional dialect, e.g., Southern	25 (71%)	1 (3%)	9 (26%)	35
Standard English	80 (89%)	7 (8%)	3 (3%)	90

	Uncomfortable at work or school because of the way trainee speaks English
	Never/ Rarely	Some of the time	Most/ All of the time	Total
First-generation	26 (84%)	2 (6%)	3 (10%)	31
Legacy	79 (90%)	5 (6%)	4 (5%)	88
URM	28 (85%)	3 (9%)	2 (6%)	33
Well-represented	77 (91%)	3 (4%)	5 (6%)	85
Ethnic sociolect	8 (73%)	1 (9%)	2 (18%)	11
Regional dialect, e.g., Southern	26 (81%)	4 (13%)	2 (6%)	32
Standard English	74 (95%)	1 (1%)	3 (4%)	78

	Judged unfavorably because of the way trainee speaks English
	Definitely/ Probably Not	Definitely/ Very/ Probably	Total
First-generation	24 (73%)	9 (27%)	33
Legacy	78 (78%)	22 (22%)	100
URM	23 (64%)	13 (36%)	36
Well-represented	77 (81%)	18 (19%)	95
Ethnic sociolect	7 (54%)	6 (46%)	13
Regional dialect, e.g., Southern	15 (45%)	18 (55%)	33
Standard English	79 (90%)	9 (10%)	88

	Thinking about the writing and speaking necessary to be successful makes me reconsider my goal of pursuing a research career
	Strongly/ Disagree	Neutral	Strongly/ Agree	Total
First-generation	22 (63%)	5 (14%)	8 (23%)	35
Legacy	71 (69%)	15 (15%)	17 (17%)	103
URM	24 (62%)	6 (15%)	9 (23%)	39
Well-represented	68 (70%)	13 (13%)	16 (16%)	97
Ethnic sociolect	6 (43%)	3 (21%)	5 (36%)	14
Regional dialect, e.g., Southern	21 (60%)	4 (11%)	10 (29%)	35
Standard English	65 (72%)	12 (13%)	13 (14%)	90