1. Introduction
1.1 Appalachian English
Appalachian English (AppE) is a dialect spoken in the Appalachian region of the United States which spans 13 states along the eastern coast of the country. There is a current debate among linguists concerning whether AppE is a dialect or language, but for the purposes of this experiment, AppE will be considered as a dialect of English. Due to its expansive size, the Appalachian Region has been subdivided into 5 regions: North, North Central, Central, South Central, and South (University of South Carolina).
Due to geographic isolation, there are many varieties of AppE; communities separated by a single mountain may display divergent linguistic evolution. This point is being made as a caution moving forward. It is easy, as with other dialects, to believe in a hegemonic or ‘standard’ Appalachian dialect, but this is a reductive viewpoint and can lead to linguistic bias which is the subject of this experiment. Compared to other dialects in the United States, AppE has a closer tie to older forms of English. This, again, is a consequence of the geographic isolation of AppE speakers since early colonial periods.
The Appalachian region has been a locus of exploitation throughout history; particularly in natural resources such as coal (Smith, 2020). Coal companies would enter the Appalachian region to remove coal from both mines and mountain tops through strip mining. The labor force was made of usually poor Appalachian men who were willing to work in dangerous conditions. The coal companies paid these men very little compared to the value of their labor and the value of the coal that was removed from the region (Smith, 2020). The profits of the coal industry were not shared with the people of Appalachia and the effects of this are still prevalent today. 39 of the top 50 counties in the United States with the lowest median household income are in states that are a part of the Appalachian region; with Kentucky alone comprising 13 (Wikimedia Foundation, 2024).
1.2 Perceptions of Appalachian English
While perceptions of AppE vary across the United States, it has been found that attitudes towards speakers of AppE are generally negative, with perceptions of lack of intelligence and untrustworthiness (Cramer, 2018). These perceptions often thrive in more affluent communities; specifically upper-class urban and suburban areas. AppE speakers are often subject to the pejorative “hillbilly” among other less pleasant terms. (Hazen, 2023). Other perceptions include the idea that AppE speakers are racist and incestuous. These biases against Appalachian English create immense tension between the Appalachian region and the rest of the U.S as well as among Appalachian communities of different socioeconomic statuses.
1.3 Guiding Question and Hypothesis
Past literature has focused on how AppE speakers have experienced linguistic bias as a whole, however, there is a gap in the literature pertaining to the material consequences of these biases. This experiment chiefly aims to answer the question: when dialect is varied, will there be a change in the perceived intelligence and trustworthiness of the individual seeking a business loan? Informed by previous literature, we hypothesize that AppE speakers will receive allocations of loans at a lower rate than Standard American English (SAE) speakers.
2. Methods
2.1 Participants
A total of 100 participants were recruited via an online application as well as standard snowballing method. The ages of participants range from 18 – 54 and the distribution of men and women was close to 50%. (49% men and 51% women). Participants come from a diverse set of linguistic backgrounds, however, this is limited to backgrounds within the United States as all participants currently reside in the U.S. Furthermore, the socioeconomic status of the participants varies greatly with annual incomes ranging from $20,000 to $100,000.
2.3 Random Allocation Game
2.3.1 Game setup
Each participant was placed alone in a room with a standard table and chair. In front of the participant are two opaque jars with a slit in the lid in which a coin can be inserted. The jars are labeled Group A and Group B respectively. The jars are opaque to prevent the player from keeping track of the total allocation while playing the game. Behind each jar was a speaker that was used to play voice recordings during the experiment. Participants were given colorful plastic coins to allocate during each round.
Figure 1. An aerial view of the game space.
2.3.2 Game Procedure
Possible biases towards AppE speakers were measured through the use of a Random Allocation Game (RAG). A RAG is a one player game in which during each round the player is asked to make a decision of how to allocate a certain resource, however, unlike standard allocation games, participants in RAGs are then asked to flip a coin and allocate resources instead to the group assigned to either heads or tails.
Each participant completed 40 rounds of the random allocation game. The participants were told that they were playing the role of a loan officer and they would be hearing requests for a small business loan. During each round a single voice recording was played from each speaker behind jars A and B. Each voice recording was of an individual requesting a loan for their small business. The businesses were very similar as to not allow business type to be an additional variable. After hearing each voice recording once, the player was asked to think about which individual they wanted to allocate the loan to. After this, the player was then asked to flip a coin and allocate the loan to the corresponding group (Heads = Group A and Tails = Group B).
Each group consisted of voice recordings of 20 individuals. Group A consisted of 10 AppE speaking men and 10 AppE speaking women. Group B consisted of 10 SAE speaking men and 10 SAE speaking women. The voice recordings each round were paired by gender (women with women and men with men). This measure was taken to control the interaction of the variables of gender and dialect so the scope of this experiment would remain an appropriate size. The participants listened to each set of voice recordings twice (non-consecutively) for a total of 40 rounds. After each participant completed the 40 rounds, the amount of coins in each jar was counted by the researchers.
3. Results
While this experiment was not able to be performed it would be remiss to not suggest possible results informed by past research.
According to the rules of RAGs we would expect the distributions of loans to each group to be 50% or very close to this proportion. This is due to the fact that a coin flip is what should ultimately determine each allocation. In practice, however, we find games in which the difference between the actual allocation and the expected 50% is statistically significant (Kundtová Klocová et al., 2022). This suggests that participants may cheat and choose to not follow the rules of the random allocation game; the implications of which will be discussed later.
We can surmise that there will be instances of cheating during our trials, but most likely the occurrence rate will be low. Past literature suggests that while the occurrence rate of cheating may be low, there may be a slight favoring towards Group B.
3.1 Visualization
Below are two figures that work to visualize the data that will be obtained when this experiment will be completed. Figure 2 represents the distribution of allocation that we expect when the player follows the rules and only allocates based off of the random coin flip. A coin flip has a binomial distribution, so we expect the distribution of unbiased games to be a bell curve with the peak at 50% (20 coins) allocation to Group A. This is unlikely to be the bell curve that we will actually observe after running the experiment.
Figure 2. Visualization of Allocation That is Unbiased
What we will most likely find is a bell similar to Figure 3. This bell is skewed to the left, which suggests that more players unfairly withheld allocation of coins to Group A. It is important to note that this bell curve is simply to represent data that might be gathered from this experiment; the true bell curve may be much less biased.
Figure 3. Visualization of Allocation that is Biased
4. Discussion
A RAG was chosen for this experiment for the occurrences of cheating that comes with the game. The randomness of the coin flip before each allocation provides a sense of plausible deniability for each participant; if a participant has a bias against a certain outgroup (in this case AppE), then the participant may be more inclined to break the rules and allocate more resources to a perceived ingroup. This allows us to analyze biases better than in a normal allocation game. In a normal allocation game, the participants may feel pressured to allocate a certain way and that was not the desired outcome of the game for this experiment. Dialect discrimination often occurs behind the guise of anonymity, so we wanted to recreate this environment in our game design.
The results from this RAG have many implications beyond just economics, but it is helpful to begin there and then extrapolate further. In this experiment, participants were asked to play the role of a loan officer which is an economic authority. This places the biases shown against AppE speakers in a new light as past literature has investigated the perceptions of AppE speakers as a whole while this experiment has questioned perceptions in certain realms of life. This gives us a clearer idea of the real-world consequences of linguistic biases. The results of this experiment suggest that the impact of linguistic biases may go beyond just linguistic bullying.
It is accurate to say that AppE speakers are perceived as uneducated, but it is more impactful to have evidence of how AppE speakers are perceived in different situations. This experiment has focused specifically on how AppE speakers may be discriminated against in the context of economics. The results of this study implicate that AppE speakers may have more difficulty with loan acquisition and negotiation with banks than SAE speakers. This creates a cycle in which AppE speakers are unable to break out of the poverty that affects many parts of the Appalachian region. Due to their dialect, AppE speakers are perceived as poor and uneducated and, as was seen in our experiment, this leads to these speakers receiving less allocation of economically advantageous resources such as loans and credit. This then empowers the notion that AppE speakers are poor and incapable of managing wealth that originally fueled the lack of allocation. A visualization of this process can be seen in Figure 4.
Figure 4. Visualization of the cycle of biases against AppE Speakers
The perceptions of AppE speakers revealed in this economic game can be extrapolated to other scenarios in which an authority is making a decision. For example, AppE speakers may have more difficulty in the job application process or during college admissions. This again inflates the issues that many AppE speakers are unable to make change in their current socioeconomic status which then fuels the negative perception of the dialect.
5. Conclusion
There are many more aspects of this apparent bias towards AppE that need to be explored further. Particularly, the effect of gender of the speaker on allocation as well as the socioeconomic status of the allocator. Bias is a multifaceted issue that requires extensive research to begin to fully understand. The research presented here makes an effort to add to the existing literature of perceptual dialectology while attempting to focus on certain aspects of real life interaction that may lead to material disadvantages for AppE speakers. The perception cycle of AppE speakers has deep historical roots and change of this cycle will be a major challenge. Change will require an intentional, impactful, and multifaceted approach. These biases towards AppE run so deep that many people do not realize that they are being biased. The future is bright, however, as America’s cultural consciousness has been shifting to question linguistic biases such as these.
References
Cramer, J. (2018). Perceptions of Appalachian English in Kentucky. Journal of Appalachian Studies, 24(1), 45–71. https://doi.org/10.5406/jappastud.24.1.0045
Hazen, K. (2023, August 17). Combatting stereotypes about Appalachian dialects. SAPIENS. https://www.sapiens.org/language/appalachian-dialects-stereotypes/
Kundtová Klocová, E., Lang, M., Maňo, P., Kundt, R., & Xygalatas, D. (2022). Cigarettes for the dead: Effects of sorcery beliefs on parochial prosociality in Mauritius. Religion, Brain & Behavior, 12(1–2), 116–131. https://doi.org/10.1080/2153599x.2021.2006286
Smith, E. (2020, October 9). Human rights in the Appalachian region of the United States of America: An introduction. UAB Institute for Human Rights Blog. https://sites.uab.edu/humanrights/2020/10/13/human-rights-in-the-appalachian-region-of-the-united-states-of-america-an-introduction/
University of South Carolina. Where is Appalachia? | Southern Appalachian English. (n.d.). https://artsandsciences.sc.edu/appalachianenglish/node/783 Wikimedia Foundation. (2024, December 15). List of lowest-income counties in the United States. Wikipedia. https://en.wikipedia.org/wiki/List_of_lowest-income_counties_in_the_United_States
