**By: Jiah Hwang**

**Author’s Note**

As I spent my childhood traveling between Korea and America, I noticed a striking difference between the two countries. It wasn’t the food, the looks, or the culture. In Korea, if foreigners were attempting to speak Korean, Koreans would try to speak in broken English on their behalf or kindly try to decipher their Korean. However, in America, I cannot put a number on how many times my parents had received looks that non-verbally told them they were less-than simply because of their Korean accents. The same went in the classroom — if someone had a slight Hispanic, Korean, Chinese, or Indian accent, my white friends would try to trade the same looks at me, none of which were reciprocated. Shockingly, regional accents from the same country receive similar treatment as well. So why do certain accents inherently change how others see you, and how does it change? I will condense and review the research article from the School of Health Sciences, Stockton University, written by Amee P. Shah, which covers how the average perception changes based on the accent from statistical testing.

**Reviewed Research Article: “Why are Certain Accents Judged the Way they are? Decoding Qualitative Patterns of Accent Bias” (**__https://files.eric.ed.gov/fulltext/EJ1230374.pdf__**) **

**Summary**

The testing procedure for determining how different accents are perceived involves various participants. There were 55 native English-speaking participants in the test from 16-66 years old with the roles of Listeners, all Caucasian. The testing consisted of recording 7 accents as stimuli for the Listeners, with the following accents: Brooklyn, Atlanta, London, Cleveland, Glasgow, Berlin, and Hispanic Tex-Mex, which are considered representatives of New York, Southern, English, Midwest, Scottish, German, and Hispanic accents. One speaker imitated and recorded those accents in this experiment.

The 55 Listeners evaluated 6 out of the 7 recorded accents, with the rating done on a Likert scale, 1 to 5, where 1= low and 5=high for each of the following attributes: Arrogance, Friendliness, Honesty, Intelligence, Pleasantness, and Socioeconomic Status.

**Results**

New York accent: high in Arrogance, low in Friendliness, Honesty, Pleasantness, and Friendliness, and lowest in Intelligence and Socioeconomic status. Keywords used: “‘pushy’” and “‘rude’” with connections to the Italian mafia.

English accent & German accent: high in Arrogance, low in Friendliness, high in Intelligence, low in Pleasantness, and high in Socioeconomic status. (The English accent was perceived as “sophisticated” and “haughty,” while the German was seen as “cold” or “untrustworthy.”)

Southern accent & Scottish accent: low in Arrogance but high in Friendliness, Honesty & Pleasantness, and medium in Intelligence & Socioeconomic status. (a **bimodal pattern** for the Southern accent appeared, keywords consisting of “friendly” while some said “liar,” and more. There were mixed gender perceptions to the Scottish accent, with some references to media characters.)

Hispanic accent: overall low averages in each category sum up to low in Intelligence, Pleasantness, Socioeconomic status, and medium in Friendliness. Comments guessed the origin of the speaker, while some characteristics mentioned “dull,” or “street smart, maybe slutty,” and more references to the media.

Midwest accent: lies somewhat in the middle of the other accents, with comments ranging from “normal” to “bland” and “annoying.”

**Unpaired t-tests** were conducted to determine whether the difference between the perception of the Listeners for these different accents was statistically significant.

A factor that the author took into account was the fact that the majority of the respondents were born in Ohio, where the Midwest accent was quite prevalent. That may have contributed to the Midwest accent having moderate averages in positive attributes and the lowest ratings for negative attributes. The following results were all compared to the Midwestern accent for this reason

Arrogance: The difference in the score for the Midwestern accent is statistically significantly different from New York, English, and German accents.

Friendliness: The t-test tables score for the differences of all but the Hispanic accent is statistically significant from the Midwest accent.

Honesty: The mean score on the German accent was significantly different from the Midwestern accent.

Intelligence: The mean scores for English, Scottish, and German accents were significantly higher than those for the Midwestern accent.

Pleasantness: All but English and Hispanic accents were found to be statistically different from the Midwest accent.

Socioeconomic status: The Southern, English, and German were statistically different from the Midwestern accent.

**Review**

First, for those who aren’t familiar with the key statistical terms mentioned in the research above, please reference the following key I made below and consult the articles I cited for this information in the Bibilography.

Moving onto my thoughts on this research article, I believe some questions I had about the method of testing became more prevalent as the paper went on, the major two being:

Why not choose a wider variety of Listeners?

Most of the Listeners were from Ohio, and the author of the research articel even addressed that this may cause regional bias as the speakers would see the Midwestern accent as the “norm” due to the region being their hometown. In order to make the research completely impartial and without sample bias, I would improve this by picking a certain number of people belonging in the most common ethnicities in the U.S. workplace or public school and putting them into groups to see the average ethnic reactions to each of the 7 accents. In this research, the Listeners were all Caucasian, which I did not see any importance in and feel as if it may have caused more confusion as they should not be considered as the middle ground for this testing. The purpose of this test was to decide patterns in speech perception, and if that were the case and if Caucasian perceptions were the only view studied, I would specify the article to be “Why are Certain Accents Judged the Way they are? Decoding Qualitative Patterns of Accent Bias for [Caucasians]” and further the research by performing the same experiment with the smae number of variables on other ethnic groups.

Respondents or Listeners?

In the beginning of the research article, where it states, “Participants” under “METHODOLOGY,” the author states that there are a total of “55 native English-speaking participants” as Listeners for each of the accents. However, later in the same paragraph, the author brings up “Of the 44 respondents, 21 respondents were men (47%) and 23 respondents were women (53%),” which made me question if there was a second group to judge the 7 accents. Under the section “Data Analysis,” the author again corrects it to “In this study 55 respondents were asked to evaluate six different attributes of seven accents,” which made me wonder if she had miswritten “55” to “44” earlier and was choosing to switch between the titles “Listeners” and “Respondants” to reference the participants judging the 7 accents. I would prefer to have consistent titles to reference the volunteers throughout the whole article in order to avoid any confusion and obviously change the typo of the number of participants.

**Conclusion**

Despite those initial questions regarding the process of experimentation, the main takeaway from this research paper to answer my initial question was that the perception of accents is greatly influenced by the stereotypical characters we see in the media — mainly movies, shows, history, and the news. The main examples of this include the New York (connections to the Italian mafia seen on shows and throughout history), German (wars and military leaders in history portrayed through various films), and Hispanic accents (exaggerated and heavily stereotyped on TV programs, connected to lower-class and education) and the comments guessing the characteristics of the speaker. The general pattern you can see from the results is that European accents are seen as more sophisticated and intimidating than the rest. According to Amee P. Shah, these thoughts and perceptions are mostly based on “preconceived notion[s] on part[s] of cultural groups, established by influences and messages from the media and popular culture.” This explains the prejudice the English-speaking Listeners held for certain accents compared to others, even when they had the same level of unintelligibility. “However, existing theories of social identity and [preconceptions] do not explain the findings here of why some outgroups received negative ratings and others received positive ones; the influence of media and cultural buy-in to some of the propaganda is still unclear and worthy of further study.”

**Statistics Key**

What does it mean for results to be **statistically significant**?

Statistical significance is a determination that a relationship between two or more variables is caused by something other than simple chance, and therefore, proving the null hypothesis. It is expected that a p-value of 5% or lower is considered statistically significant.

→ Null Hypothesis: A statistical claim that there is no difference between two samples in a data set.

→ Alternative Hypothesis: A statistical claim that there is a difference between two samples in a data set.

→ P-value: The probability of getting the observed sample result (or one even more dramatic), if the null hypothesis is true, is the P-value. The lower the p-value, the greater the statistical significance.

What is an **unpaired t-test**?

An unpaired t-test compares the averages (means) of two unrelated samples to determine if the null hypothesis is true or not. In an unpaired t-test, the dependent variable is normally distributed and is measured on a rising scale, such as the ratio or interval scale.

The two groups must also have equal standard deviations, which is explained further down.

→ Normal distribution: Normal distribution is a probability distribution where the mean is 0 and the standard deviation is 1. It shows that the data near the mean occurs more frequently than data that is far from the mean.

→ Ratio scale: In a ratio data scale, there are equal distances between any adjacent points on the scale. The ratio scale has an absolute zero point, like height; if your height was 0, you would be nothing.

→ Interval scale: In a ratio data scale, there are equal distances between any adjacent points on the scale. The interval scale has an arbitrary zero point, like temperature; if we were to see 0 degrees, we wouldn’t assume that there is no temperature at all.

What is a **bimodal pattern**?

Data distributions in statistics can vary, with some common ones being the normal distribution. The bimodal distribution has two peaks, which shows that you have a divided sample group. An example of this may be grades; one group of students gets many As while the other gets Fs.

What is the** standard deviation**?

Standard deviation is a statistic that measures the deviation of each point in the data set in relation to its mean. It is calculated by the square root of the variance. The further away the data points are from the mean, the higher the standard deviation.

→ Variance: Variance measures how far each number is from the mean in a data set.

Bibliography

“Bimodal Distribution: What is it?” *Statistics How To*, Statistics How To, https://www.statisticshowto.com/what-is-a-bimodal-distribution/. Accessed 6 May 2023.

Boyle, Michael. “What Is Variance in Statistics? Definition, Formula, and Example.” *Investopedia*, Dotdash Meredith, https://www.investopedia.com/terms/v/variance.asp. Accessed 6 May 2023.

Chen, James. “Normal Distribution: What It Is, Properties, Uses, and Formula.” *Investopedia*, Dotdash Meredith, https://www.investopedia.com/terms/n/normaldistribution.asp. Accessed 6 May 2023.

Gleichmann, Nicole, and Sarah Whelan. “Paired vs Unpaired T-Test: Differences, Assumptions and Hypotheses.” *Technology Networks*, 14 February 2020, https://www.technologynetworks.com/informatics/articles/paired-vs-unpaired-t-test-differences-assumptions-and-hypotheses-330826. Accessed 6 May 2023.

Hargrave, Marshall. “Standard Deviation Formula and Uses vs. Variance.” *Investopedia*, Dotdash Meredith, https://www.investopedia.com/terms/s/standarddeviation.asp. Accessed 6 May 2023.

Kenton, Will. “Statistical Significance: What It Is, How It Works, With Examples.” *Investopedia*, Dotdash Meredith, https://www.investopedia.com/terms/s/statistically_significant.asp. Accessed 6 May 2023.

Shah, Amee P. “Why are Certain Accents Judged the Way they are? Decoding Qualitative Patterns of Accent Bias.” *Advances in Language and Literary Studies*, vol. 10, no. 3, 2019, p. 12. *U.S. Department of Education*, https://files.eric.ed.gov/fulltext/EJ1230374.pdf. Accessed 6 5 2023.

“What is the difference between ordinal, interval and ratio variables? Why should I care? - FAQ 1089.” *GraphPad*, Dotmatics, https://www.graphpad.com/support/faq/what-is-the-difference-between-ordinal-interval-and-ratio-variables-why-should-i-care/. Accessed 6 May 2023.

## Comments