2024 – Page 11 – Languaged Life

Roses are Red, Violets are Blue. You’re in Love with my Man? Guess my Voice Will Lower Too.

Kelly Eun, Isabelle Filen, Adeline Villarreal, Sylvia Le

Engaging in conversation with the man you like may lead you to feel all sorts of emotions. Maybe your heart starts racing, you find yourself laughing at every little thing he says, or you possibly say things you wouldn’t normally say. These are all very common character changes we may go through during these types of situations, but have you ever wondered if speaking to the man you like could also cause changes to your pitch? Our group conducted a sociolinguistic study in order to determine if a woman’s pitch altered while in conversation with a man of her interest, especially within the competitive environment of a dating show such as The Bachelor. With this objective in mind, the three longest running contestants were selected in order to analyze whether there was a possibility of pitch modulation while in one-on-one conversations with the bachelor. Praat was used to input data to find pitch means, as well as to discover if pitch change actually occurred.

Introduction and Background

Some of us may have a preconceived idea in regard to what happens to a woman’s voice when she talks to the man she likes, right? We could ask our friend, parent, or boyfriend to do an impression of a woman flirting with a guy and we would most likely hear something incredibly high-pitched and giddy with lots of giggling and hair twirling. This is a commonly held phenomenon in which a woman’s pitch raises in scenarios where she is interacting with a potential romantic partner (Re et al., 2012).

However, a study done in the UK (Pisanski et al., 2018) found that women on speed dates actually demonstrated a decrease in pitch value when the men they were interested in were in high demand, even highlighting how lower-pitched voices were more favorable than their higher-pitched counterparts. Why did they find a decrease in pitch? This is precisely what our group sought to uncover through our research, as we also asked ourselves…why?

To better understand typical traits associated with pitch, we discovered that higher-pitched voices are commonly indexed with fertility, youth, and femininity (Apicella and Fienberg, 2008)—all the traits men seem to absolutely swoon over, right? However, there might be some other traits men seem to be entranced by as well, such as sexual explicitness, seductiveness, dominance, and authority (Klofstad et al., 2012); such traits are also associated with a lower-pitched voice (Fraccaro et al., 2013).

Thus, we hypothesized that when we look at the average pitch values of our target population in our targeted scenario (which will be uncovered in our methodology discussion), the average pitch values of the contestants will be lower when they feel insecure of their position in the competition and higher when they feel more secure. The question we asked and pursued the answer to was the following: Does a lower pitch truly index the contestants’ desire and motive to assert their dominance through embodying a seductive, authoritative character who deserves not only a spot in the competition, but also the bachelor’s heart?

Methods

For our study, we first selected three women from Season 22 of The Bachelor (labeled as Contestants A, B, and C) who would remain for the longest duration, resulting in the selection of the winner, runner-up, and the “third place” contestant. We specifically chose these women because it allowed us to collect consistent data from the same three contestants throughout the whole season.

To analyze Contestants A, B, and C’s pitch across the entirety of Season 22, we divided the season into three stages: beginning (episodes 1-4), middle (episodes 5-7), and end (episodes 8-11). Pitch refers to the frequency of the sound waves made by a voice’s vibration. For instance, higher frequencies indicate higher pitch.

We then collected two utterances (e.g. brief comments) made by each of the three contestants per stage, for a total of 18 utterances (6 for each woman). To avoid the pitch being influenced by other voices, we only recorded utterances from one-on-one conversations between each of the contestants and the bachelor. The first utterance (Utterance A) was made by recording what the contestant had said before she received assurance from the bachelor, and the second (Utterance B) was made by recording what the contestant had said after. For the sake of this project, we not only considered comments such as “I like you too” or “I love you” as assurance from the bachelor, but we also took into consideration that a kiss could also be perceived as non-verbal assurance.

Afterwards, we separately inputted each utterance recording from each contestant and stage into Praat to determine if there was a change in pitch before and after Contestants A, B, and C received assurance from the bachelor. Praat is a software tool used in linguistic research to examine speech; in this specific study, we used it to analyze pitch. Rather than maximum pitch, the mean pitch values for Utterance A and Utterance B were used in order to account for possible pitch peaks and outliers that could arise and skew the pitch results. This would include a rise in pitch to indicate the end of a question or a potential fall of a declarative statement. So, utilizing the means of pitches provided a bigger picture.

Results and Analysis

Throughout the course of Contestant B’s run on the show, our group found a consistent increase in pitch, as evident by the table below.

Table 1: Comparison of Contestant B’s Pitch Values

Our data for pitch mean is measured in hertz (Hz), the standardized unit of frequency equal to one cycle per second. Sound frequency is determined by how waves oscillate while they travel to our ears and when the oscillation of frequency waves is higher, we hear a higher pitch.

The table shows a positive difference between Utterance B Mean Pitch and Utterance A Mean Pitch in hertz in all three stages of the competition (beginning, middle, and end), indicating that the pitches of Contestant B became higher after assurance from the bachelor. Based on our background research, this likely indicates that Contestant B became more relaxed after assurance during her conversations with the bachelor.

Additional research was conducted to find that Contestant B was the eventual winner of Season 22 of The Bachelor, meaning that she was proposed to by the bachelor in the final episode. Our group predicts that this likely contributed as a factor for the increase in her pitch, as she was one of the frontrunners of the competition.

However, it is also important to note that it is difficult to make definite inferences about how Contestant B was feeling confidence-wise as the competition progressed. For example, it is possible that she could have felt more relaxed, knowing that there were less women competing against her; contrastingly, it is also possible that the stakes became higher for her, and she felt more possessive or competitive with the remaining women.

Another important factor in the data for Contestant B is that specifically in episodes 2 and 7, Contestant B received a rose during a date from the bachelor, which is a rare occurrence as roses are typically offered during elimination “Rose Ceremonies.” The offer of this rose in addition to the lead-up to it could also have had an effect on Contestant B’s confidence, further playing a part in her pitch raise as an indicator of her ease.

When it comes to Contestants A and C, both of these women demonstrated instances where the changes of their pitch decreased (i.e. their pitches before getting a response from the bachelor were higher than their pitches after), as opposed to the increase we were anticipating. Let’s first take a look at Contestant A, whose middle and end pitch means had a decreasing comparison.

Table 2: Comparison of Contestant A’s Pitch Values

Looking at Table 2, we can see how Contestant A’s beginning stage pitch comparison was in alignment with our hypothesis. When she started her one-on-one conversation from that segment of the episode, her Utterance A had the mean pitch of 184.0992024 Hz. After the bachelor chimed in, the mean pitch of her reply was higher than before at 193.5689707 Hz. We found that this could be due to Contestant A feeling reassured by what the bachelor had said, which could have further contributed to her feeling less concerned about her placement in the competition and not as worried about having to modulate her pitch or portray a sensual appeal.

Let’s now take a look at Contestant A’s middle stage pitch comparison, which contrasts the comparison result of her beginning stage. While she had begun the conversation from this segment with a mean pitch of 177.2157487 Hz, this dropped to a mean pitch of 167.5482242 Hz after the bachelor had his turn of speaking to her. With cases like this where the contestant’s pre-assurance mean pitch value is higher than their post-assurance value, it may have been because what the bachelor had said to the contestants contributed to their feelings of uncertainty with both their relationship and place in the competition. By responding to the bachelor with a lower-pitched voice, it would be more likely to project more authority, which could strengthen their relationship with the bachelor and assert their reason to remain on the show.

Moving onto Contestant C, we can see in Table 3 that she also had two of her three pitch mean comparisons being a decrease like Contestant A.

Table 3: Comparison of Contestant C’s Pitch Values

Interestingly enough, whereas Contestant A’s mean pitch comparisons throughout the stages went from an increase to two decreases, Contestant C’s progression was in the reverse order of Contestant A’s and went from two decreases to an eventual increase by the end of the season. This development could possibly be due to Contestant C gradually reaching a place in her relationship with the bachelor that solidified her confidence, while Contestant A may have felt more hesitation with hers. It is important to note that in all of these utterances, although we used our best judgment to decipher the scenarios and contestants’ reactions, there may have been other factors outside of the ones we targeted that contributed to the pitch outcomes of these women.

Discussion and Conclusion

Upon gathering and analyzing our results, our group was able to determine that while our data somewhat aligned with our aforementioned hypothesis, the results may not have been as consistent as we previously imagined. With the conclusion of our research, we then began to account for potential limitations to the big picture we desired to paint as a result of our findings.

For example, we encountered a scenario with Contestant A in the end stage of the competition where her pitch began high. According to the scope of our project, this should indicate that her confidence level was high. However, after receiving an assuring statement from the bachelor, her pitch actually dropped, meaning that she felt less confident by the end of the interaction. An example of this scenario is shown below in Figure 1.

Figure 1 – Contestant A’s End-Stage Interaction Documentation with the Bachelor from Episode 11.

To account for this limitation, we attempted to think outside the realm of our project and dig a little deeper as to what factors could have led to this particular utterance not aligning with our hypothesis, and we may have gotten our answer from looking into something almost all of us have probably done at least once in our lives: lied.

Let’s say your boyfriend made you upset and finally mustered up the courage to ask the dreaded question: “Are you okay?” You look him dead in the eyes, clearly not okay, and say something along the lines of “I’m fine” or “I’m just tired.” Now, are you really fine or just tired? Absolutely not, but for reasons unbeknownst to everyone but the universe itself, you evaded the truth. Similarly, this could have been what we saw in our data as well.

Therefore, even though we used our best judgment to select a scenario in which the contestant received verbal assurance, when we consider the limited scope of our sociolinguistic research, we may not have been able to account for the internal psychological factors that pertain to a woman’s verbal expression of having accepted that assurance. Thus, as we saw in Contestant A’s final utterance, we presumed that her response of “I’m as ready as I’ll ever be” indicated to the listener that she has been assured and is feeling ‘ready’ to meet the bachelor’s family for the first time. Ideally, this should have been accompanied by a higher pitch to index her confidence and security in the competition. However, we can now presume that Contestant A’s statement did not actually reflect what she was feeling inside and she may have felt as though she was nowhere near ‘ready’ to meet the bachelor’s parents, thus explaining why her confidence level and (consequently) the mean pitch were lower by the end of her utterance, which actually does align with our hypothesis! If we had not calculated the mean pitch of this utterance or discovered an unexpected decrease that misaligned with our hypothesis, we would have not been able to think beyond the realm of our research and propose an internal factor that may have impacted what we found in our results. So, what our group initially identified as a suspected failure may not have been a failure at all.

Furthermore, our group looked beyond the bindings of our research and constructed further possible iterations of our project that could be used to bridge the gaps in both previous research and our project on pitch modulation and its indexicality. These iterations could dive into many possible examinations of pitch modulation such as mental/physical well-being, age of the target population, the presence of vowel shifting, and even examining a scenario in which the woman contrarily is not interested in the man she is interacting with. The inclusion of all of these external and, as we know now, internal factors could certainly come together and paint a clear and concise picture of this phenomenon, as it is one that certainly cannot be attributed to just one definitive variable.

References

Apicella, C. L. & Feinberg, D. R. (2008) Voice pitch alters mate-choice-relevant perception in hunter-gatherers. Proceedings: Biological Sciences, 276(1659), 1077–1082. https://doi.org/10.1098/rspb.2008.1542

Fraccaro, P. J., O’Connor, J. J. M., Re, D. E., Jones, B. C., DeBruine, L. M., & Feinberg, D. R. (2013) Faking it: Deliberately altered voice pitch and vocal attractiveness. Animal Behaviour 85(1), 127–136. https://doi.org/10.1016/j.anbehav.2012.10.016

Klofstad, C. A., Anderson, R. C., & Peters, S. (2012) Sounds like a winner: Voice pitch influences perception of leadership capacity in both men and women. Proceedings: Biological Sciences, 279(1738), 2698–2704. http://www.jstor.org/stable/41549338

Pisanski, K., Oleszkiewicz, A., Plachetka, J., Gmiterek, M., & Reby, D. (2018) Voice pitch modulation in human mate choice. Proceedings: Biological Sciences, 285(1893), 20181634-20181634. https://doi.org/10.1098/rspb.2018.1634

Re, D.E., O’Connor, J. J. M., Bennett, P. J., & Feinberg, D. R. (2012) Preferences for very low and very high voice pitch in humans. PloS one, 7(3), 1-8. https://doi.org/10.1371/journal.pone.0032719

confidence, dating shows, Media, vocal pitch

Speak of the Devil: How Popular Film Antagonists Use Language

Sociolinguistics

Sarah Belew, Jacques Gueye, Kaley Phan, Boyi Zheng

In this study, we analyze the linguistic behavior of antagonists in psychological thriller movies in order to understand/index the features in their language that make them “creepy”. We chose 4 different films to view: Misery (1990), Silence of the Lamb (1991), The Lovely Bones (2009), and Gone Girl (2004). From these films, we analyze how abnormality is constructed using subtle linguistic behaviors of word choice, intonation, and sociolect. We theorize that abnormality in the character’s linguistic traits is rooted in deviation of their demographic’s language pattern and what is considered appropriate for social interaction, or “creepy.” As a result, we find that the antagonists are aligned in their sociolects, word choice patterns of calling the antagonist’s name with great frequency, and that female antagonists have a “(Rise)-Rise-Fall” pattern in prosody. These marked patterns come together to create vivid, memorable characters that are unmistakably creepy.

Introduction and Background

What is creepiness? McAndrew and Koehnke identified the study of creepiness as a major gap in the scientific literature (2016). Their study identified creepiness correlates, such as gender identity (“being male”), physical traits (“having greasy hair”), and behavioral patterns (“stood too close to your friend”), concluding that creepiness is detected when people perceive an ambiguity of threat, often interpreted as a sexual threat among female participants.

We are unsatisfied with this approach. By eliciting people’s understanding of creepiness, perhaps we receive nothing more than a “creepiness profiling” mechanism, echoing the findings of Implicit Association Test studies conducted on racial stereotypes (Oswald et al, 2013). Are certain racial groups perceived as possessing more ambiguity of threat? Can women ever be creepy? Since our social life is riddled with implicit biases, this strand of study demands a more experimental approach that challenges our biased perception. Scripted Fictional Content (SFC) has the transformative potential in understanding the depth of this human emotion. The four psychological thriller films that we watched for our studies, including Misery (1990), Silence of the Lamb (1991), The Lovely Bones (2009), and Gone Girl (2004) demonstrated that creepiness does not rely on stereotypes and non-normativity known to our daily life. The mismatch between language stereotypes, creepiness assumptions, and the language performance confirm that actors often combine verbal strategies that run counter to expectations of norms but are not abnormal in a caricatural sense. This confirms the third-wave sociolinguistic assumption of the speaker’s agency (Eckert, 2012, Bell and Gibson, 2011). Overall, we find that actors do not achieve creepiness by piecing together socially neutral elements in the context of the movie plot and the character, which come together to deliver the creep.

Methods

In this study, we use sociophonetic and behavioral patterns to describe the technique of character building. We examined over eight hours of audiovisual media to develop a corpus study, manually coding all the speech of antagonists for marked patterns. We also listen to the other characters to qualitatively compare, so that we may develop a sketch of each antagonist including sociophonetic and rhetorical strategies. For the quantitative analysis, we pick creepiness correlates in their coded context to establish a statistical distribution per film.

In conducting this pilot study, we take the third-wave linguistics assumption that speakers consciously use sociolinguistic elements. Since there is a gap in the literature, we only landed on prosody, register (word choices), and sociolect as our relevant variables after preliminary viewing. Many previous studies demonstrated that affective prosody allows great intra-speaker speech variety, and plays a critical role in emotion expression (Frick 1985). On the other hand, socio-prosody study of inter-speaker variation is a field at its early development (Holliday 2021). Nevertheless, the evidence in our corpus convinced us that creepiness correlates, such as the markedness of the (rise)-rise-fall prosody has both intra-speaker patterns (reserved for narrative manipulation), and inter-speaker patterns (indexing non-standard dialects, feminine speech).

For our most nuanced and productive variable, the (rise)-rise-fall pattern, we have provided an edited video clip for demonstration purposes. Our coding of this pattern has followed the following rule-of-thumb criteria: (1) well-defined rising accent (L+H*) followed by a falling accent (H+L*) and a low boundary tone (L%), often the result of emphasizing the penultimate word (Beckman & Ayers 1997) (2) exclude rising in questions, unmarked rising in enumeration and long words (3) usually occurs toward the end of an utterance (4) characteristically violating the Strong-Weak alternation in modal rhythm (Kiparsky 2014).

Here is an edited clip from Gone Girl (2004), where Amy Dunne uttered 50 instances of characteristic (rising)-rising-falling speech in the course of a 7-minute monologue.

https://drive.google.com/file/d/1Tcc106ftXEQYeOe08oTVmTule9zlkP40/view?usp=share_link

Results and Analysis

1. Word Frequency Analysis

The following word clouds (Figure 1) visualize how antagonists in our corpus exhibited two discourse behaviors that made their interlocutor and the audience uncomfortable. Amy Dunne (Gone Girl), Anne Wilkes (Misery), and Hannibal Lector (Silence of the Lamb) all have the name of the protagonist as their most frequent item, showing their unhealthy obsession with them. They all engaged in extensive manipulation of the protagonists, calling them by name to draw their attention. This article by Michigan State University elaborates on some of the psychology behind using someone’s name, including manipulation. George Harvey (The Lovely Bones) is the only character that does not share this pattern, particularly because of a lack of dialogue between the two (although he does refer to the protagonist by her last name occasionally). The second behavior is the use of words that may appear manipulative. These words include “know”, “think”, “see”, etc. Under different contexts, these words are used in dialogue to guide the protagonist’s thinking and actions. For example, consider the following line in “Misery” by Annie:

ANNIE: I know you didn’t mean it when you killed her, and now you’ll make it right.

In this context, the antagonist is telling the protagonist what she believes he is thinking, in a way that he has no choice but to accept. With the other high frequency words, it is common that the antagonist used it in the same way: to make controlling statements or pure assumptions. Phrases such as “you know”, “I think you” and “you see”, were very frequent ways that these words were being used.

Figure 1: Word Clouds for each Antagonist

2. Prestigious, Psychopathic Dialects and Rhetorics

In the three antagonist-centered psychological thrillers (excluding The Lovely Bones), we see antagonists being portrayed as highly intelligent, socially prestigious characters, only to be revealed as vile, manipulative psychopaths. They are Harvard-graduate writers, head nurses, and renowned psychologists, all having high social standing. Based on the story setting and the voice acting, we qualitatively judge Amy to have virtually no accent for her background in New York, and Anne Wilkes to leave a Mid-American housewife impression but no recognizable Bakersfield phonetic feature as well. Hannibal Lector, on the other hand, speaks with the highly prestigious Mid-Atlantic English, plus Hollywood Golden Age paralinguistic sound quality: consistent nasal leakage, strong twang (forte consonant articulation), drawl (prolonged vowels), seldomly enunciating rhotics. These patterns are close to Katherine Hepburn, credited as the actor that popularized this accent (Wang 2014). However, this style of speech has been long obsolete, and the decision to act in such an accent, combined with a strange melodic quality resembling Truman Capote, the actor Anthony Hopkins has truly created an uncanny accent.

On the level of rhetoric, all three of these antagonists use elaborate rhetorical strategies. Hannibal is known for his cruel puns (‘having a friend for dinner’, implying cannibalism). As a writer, Amy uses her rhyming puzzles and vulnerable writing style to play the innocent, bubble-headed victim of a domestic murder, leaving the world lamenting her disappearance and his husband being hated by everyone. Even Anne Wilkes, the least articulate of them, uttered rhymed lines in a confession of love for Paul, showing a mastery of poetics:

The rain, sometimes it gives me the blues./ When you first came here, I only loved the writer part of Paul Sheldon. But now I know I love the rest of him, too. / I know you don’t love me. Don’t say you do./…. You’ll never know the fear of losing someone like you… /That’s very kind of you, but I bet it is not altogether true.

George Harvey, the antagonist of The Lovely Bones, has to be analyzed as an exception. Being a protagonist-centered film, the picture showed him as a more marginalized, conventionally creepy character. His speech patterns always sound like an act, given that there is little exposure to the nuances and complexity of his persona. He uses underprivileged elements preferred by men, sometimes exhibiting free in~ing, /s/~/ʃ/ alternation (Hall-Lew et al, 2021, Labov, 1966). Conspicuously, the alternation seems to happen when he is confronted by somebody else and putting on pretension. If we accept this theory and suggest that George does not exhibit his authentic speech pattern, it does not pose a problem to our thesis.

3) Prosody Distributional Analysis in the Sociolect Context

In watching Gone Girl, Lovely Bones, and Misery, we find the distribution of the (Rise)-Rise-Falling pattern to be precisely patterned with three social factors: gender, region of origin, and positions of power.

What first drew our attention was the clip played in the method section, in which Amy Dunne displayed an impressive array of the said pattern – contributing fifty out of 194 total occurrences of the pattern in the film within seven minutes. When we look at the setting of Gone Girl, in suburban Missouri, we can make sense of this pattern as a regionalism for most of the occurrences, since most characters, except for Nick, Amy, and Margo, display a non-standard accent. Perhaps due to the cultural marginality of the American Midwest, we do not find the renowned monophthong /o/ instead of /əʊ/ (‘b-oh-t’ for BOAT [boːt]), /æ/ heightened to be /e/ (‘behg’ for BAG [beg]), but instead recognize the vowel chain shift nascent to the American south (/a/ for /aɪ/, /ei/ for /ɪ/, and so on) (Labov et al, 2005). As in the previous analysis, the character Amy is designed as a native New Yorker with high social standing, often showing snobbery and resenting the simple life in the Missouri countryside. The (Rise)-Rise-Fall pattern feels quite out of place in her standard American repertoire of sounds, and we are inspired by the clip to say that the voice actors use this strategy to show a dominance over the narrative – for the first half of the film, she has set herself up as the narrator, but is later to be revealed as a psychopathic liar who is obsessed with manipulating herself and others. This agrees with the common perception of the prosody pattern- salient in TV announcements, radio shows, and advertisements. When the audience saw the run-away Amy put up the facade of a New Orlean housewife, she experienced another spike of (Rise)-Rise-Fall patterns, only that this time it is contextualized by a clear southern accent. This time, her interlocutor speaks with the same southern accent and similar rise-fall patterns. This gives us the final confirmation that the pattern we discovered is indeed marked and regionally conditioned.

The same findings are confirmed in Misery, even though a complete count of the relevant token has not been possible due to copyright reasons, we have an observation that exactly follows the above: Anne Wilkes speaks a normative accent, and only uses the (rise)-rise-fall pattern to fulfill the function of manipulation, infamous scenes such as the iconic “I’m your number one fan”. Otherwise, the same pattern only occurred in the conversation between two southern speakers, Sheriff Buster and his wife.

Should this prosody pattern also be found among male antagonists? We do not find this to be the case, but the statistics from the three films we coded (excluding Silence of the Lamb) all speak to the fact that (Rise)-Rise-Fall indexes female identity. In all three films, male utterances of the phrase make up less than 25% of the total tokens, and characters who consistently speak this way are all decidedly Southerners. The Lovely Bones showed the same consistent pattern, with the prosody of Suzie’s mom functioning as a signal for a controlling character who commands her children to follow rules – similar to the manipulative personalities we saw.

Table: Occurrences of the Rise-Rise-Fall pattern in *Gone Girl* and *The Lovely Bones*

Figure 2: The chart shows that Amy has the most occurrences of (Rise)-Rise-Fall of any other character by a wide margin.

Figure 3: The chart shows that Suzie has the most occurrences of (Rise)-Rise-Fall of any other character.

Discussion and Conclusion

Our research has yielded a small contribution to the understanding of the interactions between language and media, by providing insights into the intricate ways in which language and presentation can be used to create a sense of creepiness or psychopathy in characters. By exploring the techniques employed by the antagonists in the four films we analyzed, we have uncovered a diverse range of patterns that contribute to the development of creepy characters.

It is important to acknowledge that our study represents a small sample size and may not be generalizable to all media or contexts. However, with our final results, we can contribute to a deeper understanding of this popular media phenomenon. We hope that our research will provide further insights into this topic, as there is much more to be learned about the nature of language in the media and our relationship with both.

At the end of our project’s journey, our findings have uncovered the intricate ways in which language and presentation can be used to create a sense of creepiness or psychopathy in characters. The identification of common patterns in the techniques used in the four films we studied is a crucial step in the advancement of our understanding of this complex phenomenon.

In conclusion, our research has provided valuable insights into the complex phenomenon of creepiness in the media. Our analysis of language and presentation techniques used to create creepy characters can contribute to a deeper understanding of human psychology and social behavior. Our study also highlights the importance of representation and diversity in media and has practical applications for filmmakers. We hope that our research will inspire further investigations into this fascinating area of study.

References

Bauman, R., & Briggs, C. L. (1990). Poetics and performances as critical perspectives on language and social life. Annual review of Anthropology, 19(1), 59-88.

Beckman, M. E., & Ayers, G. (1997). Guidelines for ToBI labeling. The OSU Research Foundation, 3(30), 255-309.

Bell, A., & Gibson, A. (2011). Staging language: An introduction to the sociolinguistics of performance. Journal of Sociolinguistics, 15(5), 555-572.

Buckland, W. (2021). Feminism, narrative, authorship: (Gone Girl and Orlando). In Narrative and Narration: Analyzing Cinematic Storytelling (pp. 65–80). Columbia University Press. http://www.jstor.org/stable/10.7312/buck18143.9

Cao, H., Beňuš, Š., Gur, R. C., Verma, R., & Nenkova, A. (2014). Prosodic cues for emotion: analysis with discrete characterization of intonation. Speech prosody (Urbana, Ill.), 2014, 130–134. https://doi.org/10.21437/SpeechProsody.2014-14

Clopper, C. G., & Smiljanic, R. (2011). Effects of gender and regional dialect on prosodic patterns in American English. Journal of phonetics, 39(2), 237-245.

Demme, J. (Director). (1991) The silence of the lambs [Film]. Strong Heart Productions.

Fincher, D. (Director). (2014). Gone girl [Film]. Twentieth Century Fox.

Jackson, P. (Director). (2009) The lovely bones [Film]. Dreamworks Pictures.

Jacquemet, M. (1999). Conflict. Journal of Linguistic Anthropology, 9(1/2), 42–45. http://www.jstor.org/stable/43102422

Holliday, N. (2021). Prosody and sociolinguistic variation in American Englishes. Annual review of linguistics, 7, 55-68.

Hall-Lew, L., Moore, E., & Podesva, R. (Eds.). (2021). Social meaning and linguistic variation: Theorizing the third wave. Cambridge: Cambridge University Press. doi:10.1017/9781108578684

Kiparsky, P., & Youmans, G. (2014). Rhythm and meter: Phonetics and phonology, Vol. 1. Academic Press.

Labov, W. (2006) The social stratification of English in New York City—Google Books. (n.d.). Retrieved June 19, 2023, from https://www.google.com/books/edition/The_Social_Stratification_of_English_in/bJdKY0mZWzwC?hl=en&gbpv=1&dq=william+labov+1966+social+stratification&pg=PA3&printsec=frontcover

Labov, W., Ash, S., & Boberg, C. (2008). The atlas of North American English: Phonetics, phonology and sound change. Walter de Gruyter.

Leistedt, S. J., & Linkowski, P. (2014). Psychopathy and the cinema: Fact or fiction? Journal of Forensic Sciences., 59(1), 167–174. https://doi.org/10.1111/1556-4029.12359

Reiner, R. (Director). (1990) Misery [Film]. Castle Rock Entertainment.

Schegloff, E. A., Jefferson, G., & Sacks, H. (1977). The preference for self-correction in the organization of repair in conversation. Language, 53(2), 361. https://doi.org/10.2307/413107

Langlotz, A. (2017). 17. Language and emotion in fiction. In 17. Language and emotion in fiction (pp. 515–552). De Gruyter Mouton. https://doi.org/10.1515/9783110431094-017

McAndrew, F. T., & Koehnke, S. S. (2016). On the nature of creepiness. New Ideas in Psychology, 43, 10–15. https://doi.org/10.1016/j.newideapsych.2016.03.003

Oswald, F. L., Mitchell, G., Blanton, H., Jaccard, J., & Tetlock, P. E. (2013). Predicting ethnic and racial discrimination: A meta-analysis of IAT criterion studies. Journal of Personality and Social Psychology, 105, 171–192. https://doi.org/10.1037/a0032734

Stephenson, V. L., Wickham, B. M., & Capezza, N. M. (2018). Psychological abuse in the context of social media. Violence and Gender, 5(3), 129–134. https://doi.org/10.1089/vio.2017.0061

Wang, C. (2014). A Sociophonetic analysis of American theater speech as exemplified by Katherine Hepburn’s filmography. https://scholarship.tricolib.brynmawr.edu/handle/10066/13761

Werner, V. (2022). Pop cultural linguistics. In Oxford Research Encyclopedia of Linguistics. https://doi.org/10.1093/acrefore/9780199384655.013.999

Wolter, Keely (2022) FULL INTERVIEW: Accent and dialect coach Keely Wolter talks about dat dere “Midwestern accent.”—YouTube. (n.d.). Retrieved June 19, 2023, from https://www.youtube.com/watch?v=vC4gyH4nhIw

accents, film antagonists, Media, prosody, sociolect

Gender Bias in Celebrity Interview Questions: Topic Study at the Oscars 2023

Sociolinguistics

Sofia Duffy, Lauren Nemeh, Audrey Tseng, Venus Vu

Red carpet interviews at award shows are often a hot topic that circulates on the internet and are viewed by millions of people. This being said, the quality and nature of what is said in the interviews can wildly influence the viewers. Previous research and social movements have shown that female celebrities who participate in these interviews are more likely to receive questions and comments related to their appearance compared to their male counterparts. With this in mind, we aimed to investigate gender bias in celebrity interviews through analyzing interview clips of the 2023 Oscars red carpet interviews. Specifically, we examined if there were differences in the theme of questions asked (word choice) and the quality of interviewers’ voice (tone) when interviewing male celebrities versus female celebrities.

Introduction

Celebrity interviews on the red carpet hold a prominent position within the fabric of award shows, captivating global audiences as they eagerly seek glimpses into the lives of their beloved stars. However, within this realm, a disconcerting disparity often emerges in the nature of questions posed to male and female celebrities, serving as a reflection of deep-seated gender biases and the perpetuation of societal expectations. In our study, we set our sights on the precise queries directed at celebrities during the 2023 Oscars, embarking on an investigation to unravel the discrepancies in how interviewers engage with male and female celebrity interviewees. Through a meticulous analysis of interviewers’ word choice and tone, our objective is to cultivate a comprehensive understanding of the pervasive gender biases that permeate these high-profile interactions. In doing so, we aim to unearth patterns that mirror societal gender biases and shed light on the implicit biases harbored by the interviewers themselves. By undertaking this critical exploration of language use in celebrity interviews, we contribute to the ongoing discourse surrounding gender equality in the realm of entertainment and challenge the perpetuation of gender stereotypes on the illustrious red carpet.

Background

Celebrities hold a unique position in society due to their high-profile status and influence. The media coverage they receive significantly shapes public perceptions, particularly among impressionable children and teenagers who often look up to their favorite stars. Red carpet events are particularly important because they are televised occasions with substantial social significance. Among these events, the red carpet at the Oscars stands out as the most prestigious. Celebrities are asked a series of questions, and these interactions are recorded and broadcasted worldwide.

A study from 1956, “Age and Sex in the Interview” by Benny et al., offers valuable insights into the intersectional relationship between age and sex. Their analysis focuses on responses and utilizes post-interview surveys to gauge participants’ perspectives. Their research reveals a consistent pattern of communication inhibitions between individuals of the same age but different sexes, particularly among young men and young women. These findings underscore the complexity of interview interactions and the influence of societal expectations.

The Oscars red-carpet pre-show in question, with its apparent glamor and spontaneity, hides a reality that is heavily scripted and choreographed, where questions and reporters are carefully selected to meet industry demands and viewer expectations. Beneath the glamor of red-carpet lies a troubling tradition of objectifying the female body as a spectacle. Over the years, women have been typecast as the sole consumers of topics involving beauty and fashion, influenced by the rise of the beauty industry and photography (Lundén, 2021, p. 212). This has perpetuated a questionable tradition of objectification, reinforcing societal beliefs about gender roles. The red carpet has become a platform where women’s appearances are harshly criticized, focusing on mean-spirited evaluations of their bodies rather than celebrating their talent and accomplishments. Additionally, while women often bear the brunt of fashion-related questions, this does not exempt men from scrutiny. Fashion discourse has expanded to include men, but the discernible imbalance in how men and women are judged persists.

However, the emergence of the #askhermore movement in 2015 reflects a growing resistance to the recurring practice of asking women the superficial “what are you wearing” question on the red carpet. The movement encourages reporters to delve deeper and ask female celebrities more substantial questions beyond their clothing choices. In our research, we will concentrate on the questions asked to both male and female celebrities, shedding light on patterns that mirror societal gender biases and expected gender performance.

In the context of award shows, red carpet celebrity interviews hold great significance as they attract thousands of viewers seeking insights into the inner lives of these celebrities. However, the quality of questions posed to male and female celebrities differs significantly, with women often subjected to shallower inquiries about their appearance rather than their careers. The central focus of our research is to investigate the disparities in the treatment of male and female celebrity interviewees during the 2023 Oscars. Specifically, we aim to analyze the word choice and tone utilized by interviewers in their interactions with these individuals. By examining these linguistic aspects, we seek to gain a deeper understanding of the gender biases that manifest in celebrity interviews. Our research question revolves around whether interviewers hold implicit gender biases, leading to a tendency to ask female celebrities more superficial questions compared to their male counterparts. Through this inquiry, we aim to contribute to the broader conversation on gender equality within the entertainment industry and challenge the perpetuation of gender stereotypes on the red carpet.

Project Design

For our data sample, we selected 16 video clips from the 2023 Oscars red carpet. Since we are focusing on the gender difference of the celebrity, we selected 8 interview clips where the interviewee is a male celebrity and 8 interview clips where the interviewee is a female celebrity. To reach our sample size, we gathered video clips from three media sources: Extra TV, Vanity Fair, and E! News. In total, we analyzed 49 minutes and 53 seconds of interviews. The time varied from interview to interview, from an interview that was 41 seconds long to one that was 4 minutes and 49 seconds. We aimed to analyze whether there are differences in the interviewer’s tone and word choice based on the gender of the celebrities being interviewed. To do so, we first transcribed everything that was said by the interviewers. We omitted what the celebrities said to better focus on the style of the questions and comments and color coded the content according to these following categories:

Word Choice (categorizing questions/comments according to theme)
- Red: career related questions/comments (e.g. film projects, experience working with co-stars)
- Blue: personal related questions/comments (e.g. family, love life)
- Green: fashion related questions/comments (e.g. outfit, makeup)

Tone (categorizing questions/comments according to quality of interviewer’s speech)
- Yellow: serious tone
- Purple: playful tone

For the purpose of our study, we treated each categorization as mutually exclusive. Next, we counted how many instances each category was used in each interview and added them up according to the gender of the celebrity. We compared the number of instances between the two genders to see if there was any relationship between them. We also placed our transcripts into a word counter in an attempt to see if any significant words were used more or less in one gender over the other. In our collection and analysis of the data, we disregarded the gender of the interviewer as we solely wanted to focus on the gender of the celebrity in relation to the type of question asked.

Based on previous knowledge, we hypothesized that:

Female celebrities will be asked more questions related to their personal life and fashion.
Male celebrities will be asked more questions related to their career.
Interviewers will use a more playful tone when interviewing female celebrities.
Interviewers will use a more serious tone when interviewing male celebrities.

Data and Results

Before discussing the findings of our study, here are some examples of the types of questions/comments interviewers asked and how we categorized them. Each example is pulled from a different celebrity interview.

Figure 1: Examples of questions/comments asked to celebrity interviewees, categorized based on both gender and tone and word choice. Each question/comment is pulled from a different celebrity interview.

Here are our findings, separated into two bar charts. One chart is organized based on questions/comments that were categorized for tone, and the other is organized based on questions/comments that were categorized for word choice.

Figure 2: Bar chart representing data, with interviewer’s tone on the x-axis and total frequency in 16 interviews on the y-axis. Categorized based on celebrity’s gender.

Figure 3: Bar chart representing data, with interviewer’s word choice on the x-axis and total frequency in 16 interviews on the y-axis. Categorized based on celebrity’s gender.

In terms of word choice categorization, female celebrities were asked twice as much about fashion compared to men, as we recorded 10 instances for women compared to 5 for men. Male celebrities were asked career related questions twice as much as women, as we recorded 18 instances for men compared to 9 for women. We originally predicted that female celebrities would be asked more questions related to their personal life, which was proven to be incorrect as the results were roughly equivalent. Women were asked five personal questions compared to four for men.

In terms of tone, we found that women were asked twice as many questions in a serious tone compared to men (9 vs 4). However, we originally predicted that interviewers would use a more serious tone when interviewing male celebrities. We also found that male and female celebrities were spoken to in a playful tone at roughly the same frequency, with 20 instances for men compared to 18 for women.

When we put the data into a word counter, which we separated by the gender of the celebrity, we did not find any significance in the density of words used for male versus female celebrities.

Discussion and Conclusion

As we previously discussed, while we did find slight differences in the types of questions that were asked to the male versus female celebrities in our data recording, our data was not all-revealing nor conclusive. While we did find that men were asked more career-related questions whereas women were asked more fashion-related questions, we are not able to use our data to conclude that this is true for red carpet interview events in general. Our data consisted of 8 male celebrity interviews and 8 female celebrity interviews, which is not a complete representative of red carpet interviews as a whole. Further research could be done looking at a variety of different events. It would be interesting to see how/ if the results change depending on if it is a music awards event, premiere event, awards event, small event, etc. Each red carpet could be researched as an independent study, to then later compare the data. This comparison would offer insight into how the different events influence the types of questions being asked to male versus female celebrities. Alternatively, a study could be done by taking a set amount of interviews from each red carpet in order to have an accurate data sample of red carpets as a whole. This would offer insight into the trends of the questions male celebrities versus female celebrities are being asked on red carpets as a whole. Moreover, in our study we only analyzed the questions that were asked to the celebrities and did not take into account their responses. Future studies could benefit from also analyzing the responses.

It is also important to note that for the purpose of our study, we approached gender in a binary way through only analyzing the interviews of female and male celebrities. However, gender is not binary, and there is plentiful research that could be done on how transgender, non-binary, and other identifying celebrities are approached by interviewers on red carpets. In particular, we believe that it would be insightful to research and analyze the red-carpet interviews of transgender celebrities before and after transitioning to see if there are any differences in the questions that the same individual is asked. One of our male celebrities, Austin Butler had an interesting response to being asked a fashion question about his tuxedo. He responded with “What story am I telling you? I mean this Saint Laurent and I don’t know what story I’m telling you,” (Butler, 2023). This response, and others, offer insight into the history and the “why” matters of the questions that are asked on red carpets and therefore would be a useful data addition to a project like ours.

In addition to what we have already mentioned, in reflecting back on our project we have identified some factors that could improve our research. First, we did not account for the length of the interviews we were analyzing. Therefore, if there are three career-related questions in a 2 minute video, and none in a 30 second interview, it would be an inaccurate comparison. In future studies, we would better control the interview lengths. Additionally, we did not control who the interviewer was. Reflecting back, we realize that the motive and employer of the interviewer could play a large role in the types of questions that are asked. In future studies, we would definitely control who the interviewer is. Doing so would also open up an opportunity for more research through analyzing and comparing the types of questions that interviewers from different companies and magazines ask. It is evident that our research serves as a good starting point and reference for future studies.

References

Benney, M., Riesman, D., & Star, S. A. (1956). Age and Sex in the Interview. American Journal of Sociology, 62(2), 143–152. http://www.jstor.org/stable/2773344

Brownlow, S., Rosamond, J.A. & Parker, J.A. (2003). Gender-Linked Linguistic Behavior in Television Interviews. Sex Roles, 49, 121-132.  https://doi.org/10.1023/A:1024404812972

ExtraTV. (2023, March 12). Angela Bassett’s Kids Say Mom DESERVES Oscar Win (Exclusive). Youtube [Video]. https://www.youtube.com/watch?v=iM_bBHk-1HM

ExtraTV. (2023, March 12) Oscars 2023: Brendan Fraser Brings His Sons to the Red Carpet (Exclusive). Youtube [Video]. https://www.youtube.com/watch?v=dad48BVyR2A&ab_channel=extratv

ExtraTV. (2023, March 12) Oscars 2023: Colin Farrell’s Son Henry Says He’s a NICE Dad (Exclusive). Youtube [Video]. https://www.youtube.com/watch?v=WZRN4i0mIoo&ab_channel=extratv

ExtraTV. (2023, March 12) Oscars 2023: Miles Teller on Possibility of MORE ‘Top Gun’ Movies (Exclusive). Youtube [Video]. https://www.youtube.com/watch?v=BJMNzyb5Jtk&ab_channel=extratv

E! News. (2023, March 12). Oscars 2023: MUST-SEE Read Carpet Moments | E! News. Youtube [Video]. https://www.youtube.com/watch?v=-bu4OPD1PCM

Lawson, C. E., & Draper, J. (2021). Working the red carpet: a framework for analysing celebrities’ red carpet labour. Celebrity Studies, 12(4), 635-648. https://doi.org/10.1080/19392397.2020.1750969

Lundén, E. C. (2021). Twilight of the Idols. In Fashion on the Red Carpet: A History of the Oscars, Fashion and Globalization, 207–226. Edinburgh University Press. http://www.jstor.org/stable/10.3366/j.ctv1vtz84g.15

Mohammed, M. M., & Kadum, S. A. (2016). A Discoursal Analysis of Gender Differences in Selected English TV Interviews. Journal of the College of Languages (JCL), (34), 1-14. https://jcolang.uobaghdad.edu.iq/index.php/JCL/article/view/19

Nathanson, E. (2021). Aging on and off the red carpet: Joan Rivers, celebrity culture and postfeminist television. Celebrity Studies, 12(1), 51-66. https://doi.org/10.1080/19392397.2019.1608839

Vanity Fair. (2023, March 12). After the Awards with Vanity Fair. Youtube [Video]. https://www.youtube.com/watch?v=ncn7h59Ko-U

Wang, H. C. (2009). Language and ideology: gender stereotypes of female and male artists in Taiwanese tabloids. Discourse & Society 20 (6), 747-774. https://www.jstor.org/stable/42889296

celebrities, gender bias, gender differences, interviews, Media, tones

How Polite is Your Professor? A Gendered Analysis of Hedging as a Tool for Student Engagement at UCLA

Sociolinguistics

Layla Hernandez, Yasleen Robinson, Charlotte Norris

Throughout their lectures, professors typically engage with their students. This process often requires the professors to implement certain linguistic devices in their speech that allow for them to sound less aggressive and threatening. These linguistic features include forms of hedging. Both male and female professors rely on hedges to further display politeness when interacting with their students. In this study, we focused on how professors utilize these hedge words in ways that promote themselves in ways that are more approachable and less authoritative. We hypothesized that female professors would utilize hedges more than their male professor counterparts. Specifically focusing on the frequency of the usage of hedge words, we analyzed four sociology professors from UCLA through recordings after attending their lectures. We carefully listened to each audio and transcribed them through Conversational Analysis (CA) to further allocate the number of hedges they used when speaking with their students. A detailed analysis revealed that male and female professors do not yield significant patterns in their uses of hedges. In fact, they used them very similarly in terms of frequency and style.

Introduction

For our project, we investigated the hedges that are used by both women and men professors during their interactions with their students during their lectures. More specifically, we aimed to highlight how the usage of these linguistic markers help to ease the declarative statements they share with their students. We hypothesize that the women professors will hedge more frequently throughout their discussions with their students because hedging is unequivocally involved and ingrained in women’s speech (Lakoff, 1972). We also hypothesize that both male and female professors will utilize hedging as a means to soften their assertive/strong statements to further promote interpersonal interaction and more inclusivity within the classroom.

Background

Professors and their approaches hold great influence over the success of their students. In this way, they consciously work to provide interpersonal interactions with their students while also actively working to supply a learning environment that is inclusive and welcoming (Van Petegem et al., 2005). This process typically requires the professors to employ linguistic strategies that portray them to be less threatening and aggressive through their choice of words (hedging). Hedges are words or utterances that serve to “reduce the force of a statement” (Wright & Hosman, 2009, p. 143). The professor’s reliance on hedge words prevents them from sounding or behaving like they know everything, rather it serves as a buffer that allows for a response or objection. In this way, professors create an open forum in which students feel comfortable probing or questioning the content they are teaching without fear of judgment. Our focus will also turn to the gendered patterns of hedging among men and women professors, specifically focusing on who uses hedges more frequently.

Methods

Our main focus included examining instances of hedging and their role in yielding the authority of professors in student-teacher interactions. We focused specifically on UCLA professors in the Sociology department to avoid linguistic variation that could occur due to variations in lecture content. We collected recordings from 4 different Sociology professors at UCLA which will include three males and 1 female. We collected our data through audio recordings after attending 4 different lectures, where we keyed in on student and professor interactions. We extracted our data from two interactions from each lecture. Where we then transcribed each exchange through Conversational Analysis (CA). Through this form of analysis, we highlighted the different hedges that both female and male professors use to soften their strong statements when speaking to their students. To achieve this, we looked at each transcript individually to identify patterns specific to each professor followed by a comparison between professors of the same gender and then across genders to look for any similarities or differences that are connected to larger gendered-language patterns. We also wanted to measure the frequency of how many times each hedge word was used by each professor. Thus, after transcribing we allocated the number of hedge words they included in their interactions.

Results and Analysis

Interestingly, our results indicated that there was not a significant pattern when comparing hedge use between female and male professors. Our initial hypothesis that women would produce more instances of hedging when interacting when students in comparison to male professors was completely in misalignment with our results. In contrast, we found that both genders utilize hedges in similar ways that would further elevate their statements from sounding too forceful or impolite. We also found that in terms of frequency, male and female professors depend on the practice of hedging similarly. There wasn’t a significant pattern that demonstrated that one gender utilized the linguistic device more than the other. If anything, it dispels the common notion that women rely on hedging more than men because we see that is not the case here, both genders rely on hedges almost equally.

Figure 1: Professors’ Usage of Hedges During Student Interactions

Our first participant, professor 1, was unique because he was the only male as well as the only professor to demonstrate a wide and dependent use of hedging when interacting with his students. The bar graph above illustrates how professor 1 drastically differs from the other professors when examining their reliance on hedges during their interactions with their students. After completing professor 1’s (CA), we found that he accounted for a total number of 15 different instances of hedging. Each professor yielded some form of hedging, however, it seems that professor 1 superseded every other professor in terms of hedge use. After comparing him to the other professors that did not align with this pattern, we connected his use of hedging to his linguistic style and postulated that hedging may be a central linguistic feature in his everyday speech.

Figure 2: Transcription excerpt of Professor 1’s use of hedging

In the (CA) extracted from Professor 1’s (male) lecture, you can see a handful of the hedge words he used while engaging with his students. We classified his frequent use of the word “right” as a hedge word because we noticed that he would insert them right after finishing his strong and declarative statements. It was apparent that he depended on “right” as a means to establish an open forum that allowed for the student to respond or interject. We also interpreted his usage of “right” as a way to provoke silent agreement. This hedge word was especially important when interacting with other students as the professor relied on it with each interaction he made with his students. We also highlighted that the professor used the hedging phrase “I don’t think” several times as well. It was very clear to us that he depended on this phrase as a means of expressing uncertainty. This is an important practice because professors work to avoid behaving or sounding like they know everything because it leaves room for students to object, gauge, or comment on their statements.

In contrast, Professor 2, 3, and 4 all displayed significantly less instances of hedge use during their interactions in comparison to Professor 1. In contrast, professor 2 (female) and professor 3 (male) included hedge words 6 times whereas professor 3 (female) only displayed 5 instances of hedges when engaging with her student. Our findings indicated that both genders relied on hedges similarly, in terms of numbers and surrounding context. Each professor integrated some sort of hedge word in ways that lessened the force of their statement, signaled agreement, or expressed cautious uncertainty.

Figure 3: Transcription excerpt of Professor 3’s use of hedging

For instance, as depicted in the conversational analysis above professor 3 (female) sporadically included hedges like “sort of” and “you know” to soften her claims and statements. We concluded that she turned to this linguistic variable to display politeness and avoid authoritative brute force that would make her sound less approachable and more aggressive. This approach allowed for a fluid and authentic interaction when replying to an inquiry made by her student, one that didn’t involve the professor imploding information on the student but rather one that left room for doubt and allowed room for responses, objections, and even interjections.

Figure 4: Transcription excerpt of Professor 4’s use of hedging

Similarly, Professor 4 (male) employs the linguistic device to lessen the magnitude of his assertions. The (CA) above shows that he utilized the phrase “it’s like” on three different occasions to display some sort of indirectness. We deemed the phrase as an instance of hedging because he used it to ease his claims so that he didn’t sound overly forceful or assertive. We also derived the conclusion that he could be employing this device as a way to create an inviting forum for further commentary or inquiries. After gathering our observations, we found that each professor depended on hedges in ways that further display themselves along with their statements as more polite to steer clear from sounding rude or unapproachable. Some obviously more than others but regardless this linguistic variable enabled them to engage with their students in an authentic and less authoritative manner.

Discussion and Conclusion

To conclude, our results revealed that both male and female professors utilize hedging to soften their statements. We found that professors exhibited hedging as a means to soften their statements to some degree in their speech. Professors 2, 3, and 4 used hedging in similar amounts and Professor 1 stood out as an outlier in his much more frequent use of hedging. This finding contradicts our hypothesis that female professors would hedge more frequently than men. It also contradicts Lakoff’s (1972) findings and his claim that hedging is an integral part of women’s speech, so much so that women hedge more frequently than men. If our findings had been in alignment with Lakoff (1972), we would have seen female professors hedge more frequently than male professors but that was not the case.

While we did not find noticeable differences in use of hedging on the basis of gender, further analysis of our data with a less specific focus on a particular linguistic variable could reveal patterns that are not visible when the scope is narrowed on one variable. Our study was faced with multiple limitations that may have affected the outcome. With a larger sample of professors and more time to analyze the data, patterns that couldn’t be found or generalized with such a small sample could emerge. Our teaching population was also limited to one subject (sociology) and one educational setting (large university). An analysis across subjects or in different educational settings that have different goals for students might have led to different results.

There are a few other explanations for our findings and directions that our study could be taken in. Perhaps there were no noticeable patterns across male and female professors because professors have different styles of engaging students during interactions. Professor 1 (male), who hedged noticeably more than his colleagues, may choose to engage his students in direct interaction through hedging while the other professors may choose other linguistic tools to achieve the same goal. It is also possible that these professors are not focused on creating an inclusive lecture through student interaction and instead attempt to foster engagement through other parts of their lecture, resulting in a lack of hedging in their speech. Maybe their focus is not on creating an inclusive environment and rather each professor has a unique goal in the classroom that further influences their style of communication.

Furthermore, our findings raise some important questions concerning gender dynamics in teaching and in other social contexts. The Lakoff (1972) study demonstrated that gendered linguistic patterns do exist in some social contexts, but our research did not find this pattern in one particular context of university teaching. Perhaps, our findings could influence or lead to further research that could investigate how gendered speech patterns that have been documented in some contexts (e.g. hedging) appear in other social contexts such as other professional environments, in conversations between established friends, or conversations between strangers.

References

Lakoff, Robin. (1972). Language and Woman’s Place. Language in Society, 2(1), pp. 45-80. http://www.jstor.org/stable/4166707?origin=JSTOR-pdf

Oliveira, A.W., Sadler, T.D., Suslak, D.F. (2007). The linguistic construction of expert identity in professor–student discussions of science. Cultural Studies of Science Education, 2, pp. 119-150. https://doi.org/10.1007/s11422-006-9039-4

Van Petegem, K., Creemers, B. P. M., Rossel, Y., & Aelterman, A. (2005). Relationships Between Teacher Characteristics, Interpersonal Teacher Behaviour and Teacher Wellbeing. The Journal of Classroom Interaction, 40(2), pp. 34–43. http://www.jstor.org/stable/23870662

John W. Wright II & Lawrence A. Hosman (1983). Language style and sex bias in the courtroom: The effects of male and female use of hedges and intensifiers on impression information. Southern Journal of Communication, 48(2), pp. 137-152. https://doi.org/10.1080/10417948309372559

Zakia, Anisa (2018) Pragmatic study on hedging as politeness strategy in online newspaper. Universitas Islam Negeri, pp. 13. https://repository.uinjkt.ac.id/dspace/handle/123456789/45815

classroom, hedges, professors

アンチ vs. Haters: How do Individualist Americans and Collectivist Japanese Net 民 Express Hate Online?

Sociolinguistics

Kayenat Barak, Emily Moreira, Sae Tsunawaki, Karin Yamaoka

While social media has been a revolutionary tool for facilitating access to resources and information and connecting people globally, the power to hide behind anonymous platforms has also equipped many with the ability to spread hate online. Our project analyzes such hate comments written by Japanese and American audiences to gain insights into the sociocultural factors that shape the nature of online hostility. We chose four celebrities: one American female, one American male, one Japanese female, and one Japanese male, and used multiple social media platforms to collect a total of 120 comments. Upon categorizing these comments by type, tone, and directness, we found that there are no significant differences between comments targeted towards Japanese celebrities and American celebrities. This conclusion is fascinating, as it shows that values that characterize a certain culture, such as politeness and collectivism, and the linguistic barriers they may pose were not scientifically sound when it comes to the online world.

Introduction and Background

In today’s highly technological and globalized society, social media has quickly made itself a necessity in social relationships around the world, but the power to hide behind anonymous platforms has also equipped many with the ability to spread hate online. Ranging from cancel culture to haters, online hate culture has become so integrated to our society that it has manifested itself as an interesting sociolinguistic phenomenon. It is this precise overlap between sociolinguistics and online speech around that world that we wanted to research.

Our project compares hate comments targeted towards Japanese and American celebrities as a way to gain insight into the distinct sociocultural contexts and linguistic patterns that shape the expression of online hostility. Understanding the complexities of hate speech requires an exploration of its linguistic, cultural, and sociological dimensions. By comparing hate comments written by Japanese and American audiences, we can begin to understand the variations and similarities in the ever complex issue of online hostility. Given this context, our research question then is: How are written negative comments online directed towards popular entertainment public figures similar or different between Japanese and American speakers?

Defining hate comments becomes more intricate when we consider the linguistic, cultural, and sociological differences between Japanese and Standard American English. Negative comments are extreme forms of “online incivility”, and they can be recognized and understood differently depending on various normative concepts and cultural backgrounds (Schmid et al., 2022). To establish a comprehensive framework of what is considered a “hate comment”, we will adopt the definition proposed by Alzarouni who writes, “those (online hate) directed to a person or a particular group to express intolerance for the individual or their ideology” (2022, p. 5). Japan’s culture emphasizes hierarchy, group harmony, and politeness which translates into a high language positionality and indirect speaking style. However, American culture promotes individualism which is expressed in the way Americans speak, such as the lack of structured high language and freedom to discuss social taboos. Additionally, resistance towards progressive ideas may be observed due to Japan’s more conservative socio-political structure (Feldman, 2023). Given the contrast in these two societies, we hypothesized that Japanese hate comments would center around comments that came as a result of maintaining the national status quo, such as promoting standardized beauty norms and criticizing mannerisms, whereas American hate comments would directly criticize celebrities based on the hater’s individual moral code rather than as a result of social constraints.

Methods

Establishing a way to test this hypothesis meant that we needed to test the frequency of our linguistic variables within a controlled environment. That being said, we collected 120 hate comments on Twitter, TikTok, and Instagram directed towards Japanese and American celebrities across 4 categories. These four categories test the correlation and frequency of hate between gender and race: Japanese male, Japanese female, American male, and American female. In order to find hate comments for these 4 categories, we collected online hate directed to our representative celebrities, Kuro-chan, Ryuchell, Timothée Chalamet, and Demi Lovato.

From left to right: Timothée Chalamet, Demi Lovato (American celebrities), Kuro-chan, and Ryuchell (Japanese celebrities)

We collected this data by choosing the 30 most interacted hate comments made in 2023 under social media posts made in the same year from the celebrities’ personal social accounts. Only comments written in Japanese and made by Japanese speakers were analyzed for our Japanese celebrities; conversely, only comments written in American English were analyzed for our American celebrities. From these 120 comments, we then analyzed frequency across three linguistic variables: type of identifiable hate, tonality, and toxicity. By identifiable hate, we wanted to categorize the hate comment in regards to what it directly criticized which we classified into 6 groups: appearance, gender identity/sexuality, private affairs, personal hatred, mockery, and violence. It is important to note that hate comments were not restricted to one category, rather they could be categorized under multiple if they fit the requirements. An example of this would be a comment made to Demi Lovato, “It’s the chin for me. 🤪” which was classified as both mockery and an appearance hating comment. In order to analyze tonality, comments could also be categorized into 4 major tonal groups: serious, comedic, pitiful, or neutral. Lastly, the variable of ‘toxicity’ stems from existing previous research conducted by Won Ik Cho and Jihyun Moon on Korean online hate comments. Cho and Moon define toxicity as 1) dependent on how hateful the speech is, and 2) its ability to influence either the victim or other commenters (2021, p. 4-5). In order to gauge toxicity, we adapted this concept to instead analyze the “directness/indirectness” of the hate comments. An example of an indirect comment would be, “lowk basic but ok” which was directed to Timothée Chalamet vs. a direct comment “気持ち悪いから目覚めんな🤢🤢🤢” (“I hope you never wake up because you are disgusting”) directed towards Kuro-chan. Through this methodology, our main goal was to count and compare the frequency of these different linguistic markers and see if there were any differences, similarities, or correlations across Japanese and American hate comments, and subsequently, global hate culture.

As a disclaimer, we wanted to acknowledge some expected fluctuations in types of online hate for each of our categories. At first, we aimed to pick noncontroversial, nationally, but not internationally, acclaimed celebrities of each country with approximately similar fame in each nation in order to minimize data alteration of any kind. However, this proved to be difficult as we were unable to find hate comments directed to popular Japanese male and female celebrities. The only hate comments we could find were made to already “controversial” Japanese celebrities. Kuro-chan is problematic for his online comedic persona which is perverted, misogynistic, and violent in nature. Ryuchell is scrutinized by the Japanese public for being a transgender woman. Given these factors, we assumed our data would be skewed and expected higher rates of violent comments directed towards Kuro-chan and more hate comments regarding gender identity directed towards Ryuchell. We also expected a higher rate of gender and sexuality hate comments directed towards Demi Lovato as well, as they are bisexual and nonbinary (she/they pronouns). Our predictions were proved partly true as our data reflects that only Ryuchell and Demi received comments in regards to their sexuality or gender identity, while neither Kuro-chan and Timothée received comments of that nature. However, the hate comments were still dispersed across several different categories.

Findings

Context

Figure 1: Aggregate counts of corresponding content categories for each comment

The data obtained from the context categories revealed both cross-cultural and cross-gender differences among the four celebrities. There were a few distinct findings we did not anticipate prior to the data collection, such as the high count of personal hatred comments toward the Japanese female celebrity as seen in Figure 1. This may come as a result of a widely disputed scandal Ryuchell was involved in last year. Yet, the discrepancy between Ryuchell and the Japanese male celebrity is still extremely wide. Timothée similarly was involved in controversy due to his dating rumors^[1], but the comparable count of personal hatred comments to the American female celebrity may indicate that Japanese commenters are more harsh than Americans to controversial celebrities.

Other data that stands out in Figure 1 is the large count of insult/mockery comments targeted at the Japanese male celebrity. Again, we see a similar trend where there is a significant discrepancy between genders, rather than cross-culturally, in Japan and America. This can be explained by the general social media culture in Japan where trolling is considered a common and entertaining form of online conversation (Kaigo, 2017). Although internet trolling is seen in American online platforms as well, Japanese entertainment and comedy often involve dark humour. Another aspect of Japanese humor we see in comments and in many social conversations is “ohghiri” (大喜利), which are unique jokes that involved comedic takes on seemingly neutral or one-dimensional topics. Therefore, the combination of dark humor and ohghiri are commonly seen as a tactic of online trolling in Japan; especially when taking into account that the chosen Japanese male celebrity is a comedian, his haters may be influenced to convey their sentiments in a similar manner, perhaps as a way to mock his occupation.

Contrary to our initial hypothesis, our data showed that American celebrities received more appearance-related comments compared to their Japanese counterparts, with the aggregate count being 18 and 5, respectively. Additionally, appearance-related comments were more common for male celebrities compared to their female counterparts – almost 4 times as much for Japan, and twice as much for America. The discrepancy between genders may be attributed to the notion that commenting on a man’s appearance is less of a social taboo than commenting on a woman’s appearance. Since oftentimes women are subjects of strict beauty standards, perhaps this discourages individuals from criticizing women based on their appearance. Particularly for Timothée Chalamet, his occupation as an actor increases the frequency of hate comments regarding this appearance as attractiveness plays a significant role in his promotional activities.

Another finding that defied our hypothesis was that there were an equal number of hate comments targeting sexuality and gender identity for both Japanese celebrities and American celebrities. As we anticipated, such hate comments were exclusively directed at the female celebrities who were both members of the LGBTQ+ community. However, we assumed that a conservative society like Japan would avoid discussing topics relating to sexuality, while American progressivism meant that we expected Demi Lovato’s hate comments to be primarily about her sexuality. Surprisingly, we saw almost equal amounts of hate comments about sexuality and gender identity, where a Japanese female celebrity received 6 and the American counterpart received 7. This may be because social media permits anonymity which in turn allows commenters to code-switch and not adhere to their expected linguistic social restrictions. Japanese commenters, therefore, may feel less obligated to uphold social harmony online and express underlying opinions they would not say otherwise. Additionally, the nature of online platforms transcends cultural boundaries. The “international” online community lacks rigid cultural norms or taboos that regulate how certain topics should be treated or discussed. Therefore, Japanese commenters may be code-switching into the ‘online language’ or culture that allows them to engage in conversations that might be considered taboo in Japan.

Tone

When looking at tonality, our results show that Japanese female celebrities received the most ‘serious’ tone comments. There were no drastic differences between Japan and America when looking at tonality, excluding the pity tonality. We expected for there to be a higher rate of pity-toned comments for Japanese celebrities than for American celebrities due to Japanese culture which emphasizes politeness and humility. However, contrary to our predictions, there were more pity-toned comments for Americans overall. This may be due to the ability to be vulnerable in America, where influencers and celebrities are often encouraged to be transparent to their audience. As American celebrities tend to appeal to the masses by relating to their humanity, when celebrities face setbacks they are in turn treated with the same humanity and pity.

Another explanation for the pity tonality could be the presence of negative politeness in American online culture, as many of the pity comments we counted displayed negative politeness. Commenters at times used indirect speech to display their discontent. By framing it as a suggestion rather than a comment, the hater aimed to preserve the other person’s autonomy and avoid sounding too forceful. We think individuals in America rely on negative politeness strategies to express their criticism and avoid confrontation due to “cancel culture”.

Toxicity/Directness

Figure 3: Comments for each celebrity separated into direct or indirect

In regards to toxicity, or directness as we refer to it in our study, we originally expected Japanese comments to be primarily indirect whereas American comments to be direct. This is based upon Japan’s high language and how it tends to place a greater emphasis on respect, politeness, and conformity, which in turn discourages individuals from expressing explicit or controversial opinions openly. In contrast, American culture generally emphasizes freedom of expression and individualism, which may lead to more direct comments being shared online. However, our results show that there was no particular pattern cross-culturally, but that a correlation did exist between genders. Overall, male celebrities tended to receive more direct comments than their female counterparts as displayed by our data.

We think that due to gender roles in both countries and views on women in Japan as submissive, they had more implicit comments. Also, it could be considered taboo to comment on a woman’s appearance directly since it can contribute to this objectification and reinforce harmful gender stereotypes. Society holds women to different standards than men when it comes to appearance and behavior and implicit comments could be used to imply those notions. Regarding the more explicit comments for Japanese male celebrities, Japanese fan culture could cause a certain devotion and obsession that can sometimes lead to more explicit or possessive comments being directed towards male celebrities.

Future Research

Due to this study’s time constraints, we cannot be certain of this idea, but a proposal for future research to test this would be to increase the sample size by collecting thousands of hate comments made to multiple celebrities within our four categories. Frequency of hate comments and their types should then be categorized and compared between not only Japan and America, but other countries around the world as well. Additionally, our findings showed that there was a stronger correlation between hate comments and gender vs. hate comments and race; thus, analyzing hate comments in accordance to both the gender of the commenter and the gender of the celebrity would also be interesting. Applying the concept of toxicity will provide insight as to whether or not speech tends to be harsher or softer when criticizing individuals of the same gender identity.

Conclusion

Ultimately, our findings showed that when categorized into types of identifiable hate, tonality, and toxicity, hate comments made by Japanese and American English speakers did not differ significantly from each other. This conclusion in particular is fascinating as it shows that the preconceived notions we had about polite vs. impolite society and linguistic barriers they may pose were not scientifically sound. In fact, it instead points that perhaps the online world may be a liminal space where culture, language, age, gender, and other sociolinguistic factors are broken down, creating instead a universal shared language and culture.

References

Alzarouni, E. (2022). Detection of Hateful Comments on Social Media Detection of Hateful Comments on Social Media. Rochester Institute of Technology. https://scholarworks.rit.edu/cgi/viewcontent.cgi?article=12308&context=theses

Cho, W. I., & Moon, J. (2021). How does the hate speech corpus concern sociolinguistic discussions? A case study on Korean online news comments. ACL Anthology. https://aclanthology.org/2021.nlp4dh-1.3/

Feldman, O. (2023). Challenging Etiquette: Insults, Sarcasm, and Irony in Japanese Politicians’ Discourse. 93–116. https://doi.org/10.1007/978-981-99-0467-9_5

Kaigo, M. (2017). The Japanese Internet Environment. Social Media and Civil Society in Japan, 1–35. https://doi.org/10.1007/978-981-10-5095-4_1

Schmid, U. K., Kümpel, A. S., & Rieger, D. (2022). How social media users perceive different forms of online hate speech: A qualitative multi-method study. New Media & Society, 146144482210911. https://doi.org/10.1177/14614448221091185

celebrities, cultural differences, gender differences, hate comments, social media, tones

“American Slang” in Global Pop: The Adoption of AAVE by L2 English Speakers

Sociolinguistics

Ashley Ghodsian, Maddie Kostant, Kat Escobar, Maxime Guerra

Much of the previous work that has studied African American Vernacular English (AAVE) has focused on either native speakers of AAVE or native speakers of Standard American English (SAE) who adopt certain language features of AAVE into their speech (a phenomenon known as “language crossing”). This study investigates the adoption of AAVE features into the language of individuals who speak English as a second language (hereafter, “L2”). We hypothesized that our L2 speakers would exhibit language crossing into AAVE in a manner similar to that of native SAE speakers’ crossing, but may have different (likely unconscious) motivations for doing so. Specifically, we expected that any language crossing into AAVE by our L2ers would not only be motivated by an attempt to index proximity to Blackness (as with non-Black, native SAE speakers) but also by a desired proximity to an international conception of “American-ness,” and that this indexicality would differ for men and women as has been observed for native AAVE speakers. We analyzed the English-language interviews of two fluent, L2 English-speaking hip-hop artists who sing in Spanish in order to understand both the rates at which and the act sequences in which they adopt features of AAVE, and found evidence in favor of our hypothesis.

Introduction and Background

The origin of hip-hop and rap stems from the oral practices of enslaved Africans that were brought over to the Americas. Its style of deliverance is meant to embody the roles of the storyteller and culture historian that resemble those in traditional African society (Álvarez-Mosquera, 2015). One of the most integral aspects of hip-hop is its use of AAVE to index African-American culture.

Given the level of prominence that hip-hop has attained in global popular culture, music, and language and its considerable exposure in global media (Tamasi & Antieau, 2014), non-Black hip-hop artists whose native variety of English is Standard American English often end up adopting and appropriating aspects of AAVE (Chun 2001). Rampton (2020) has coined this phenomenon “language crossing.” Language crossing occurs when speakers borrow features from dialects that do not “belong” to them and is particularly problematic when the language being “crossed” into is socially marginalized. Because hip-hop is so closely associated with African-American culture, non-Black artists are left to differentiate themselves from the social expectations of their own culture and reinforce their authenticity as hip-hop artists through the use of crossing into AAVE (Álvarez-Mosquera, 2015).

This study builds on all of the aforementioned previous research into language crossing by investigating the non-native use of AAVE in specific language communities with whom it had yet to be explored. Also notably, this study builds on the work done by Britt & Weldon (2015), who found that African-American women who natively speak AAVE use linguistic features more similar to those of SAE compared to their male counterparts. Exploring these gendered differences was also relevant to our project design.

As such, this project aimed to answer the following research questions: (1) In what speech contexts and with what goals—conscious or otherwise—do non-Black hip-hop artists who are L2 English speakers adopt features of AAVE into their speech? And (2) How does said feature adoption vary depending on the speakers’ genders? As previously mentioned, L2 English speakers are a speech community whose potential appropriation of AAVE has yet to be investigated. We hypothesized that like non-Black native SAE speakers, our L2 speakers would exhibit language crossing into AAVE, but that they may have had different (again, likely unconscious) motivations for doing so. Namely, we expected that any language crossing into AAVE by our L2ers would not only be motivated by an attempt to index proximity to Blackness (as with non-Black, native SAE speakers) but also by a desired proximity to an international conception of “American-ness” altogether. Following Britt & Weldon’s findings, we also anticipated that this indexicality would differ for men and women, as it does for those who are native AAVE speakers.

Methods

As previously mentioned, we analyzed the English-language interviews of two fluent, L2 English-speaking hip-hop artists who sing in Spanish to investigate our claims: J Balvin, a Colombian reggaeton artist, and Rosalía, a Spanish pop artist. Two interviews were analyzed for each artist, one in which they primarily discussed topics such as their global success and interactions with famous American celebrities, and one in which they focused on their cultural and/or family backgrounds. For both J Balvin and Rosalía, their “success” interviews were chosen to be The Tonight Show Starring Jimmy Fallon, a major American late-night talk show that both artists had attended, as a method of controlling for the interviewer. The “personal” interviews were harder to control for in the same way. J Balvin’s was hosted by the YouTube channel HardKnockTV and Rosalía’s was hosted by Billboard Music’s YouTube channel. Crucially, neither of the “personal” interviews was as nationally televised in the U.S. as the “success” interviews were, and they were both centered more around personal content.

For each interview, we noted each instance in which the artist adopted at least one phonological or morphosyntactic AAVE feature and observed the frequency with which they were adopted in relation to the speech context (e.g., establishing comradery with the interviewer, discussing their childhoods, etc.). Many of the features of AAVE that this study investigates have already been widely studied by linguists. As such, we primarily followed a list of features presented by Walters (1992), found in Figures 1 and 2 below, which includes AAVE phonological features as well as AAVE morphological and lexical features. In order to minimize possible confounds, any features of AAVE that overlapped with aspects of the speakers’ L2 English dialect coming from their L1 Spanish (e.g., dropping of a final ‘g’ in verbs ending in ‘-ing’) were not considered in our analysis.

Figure 1. AAVE Style (Walters 1992): AAVE Phonological Features

Figure 2. AAVE Style (Walters 1992): AAVE Morphological and Lexical Features

The results in the following section are analyzed both quantitatively and qualitatively. The quantitative analysis considers the features of AAVE used per minute by each artist (measuring by the rate of feature adoption as opposed to considering a net total measurement is motivated by the fact that each artist’s respective videos were not equal in duration), and the qualitative analysis explores select examples of AAVE usage and aims to explain them within the lens of the conversational contexts they were used in in order to relate them back to the research question and hypotheses. We counted each syntactic, phonological, and lexical feature as one (1) instance of AAVE. For example, in his “success” interview, J Balvin produced the sentence “Yo, give it up for his Spanish!” We counted the lexical item [yo] as one feature, and the phrase [give it up for] as a second.

Results

We found that both J Balvin and Rosalía adopted features of AAVE more frequently during their “success”-oriented videos than during their “personal” interviews. Specifically, J Balvin adopted AAVE features at a rate of 3.42 features per minute during the “success”-oriented interview, compared to 0.75 features per minute during his “personal” interview. Rosalía adopted AAVE at a rate of 0.78 features per minute during her “success” interview, compared to 0 features per minute for the “personal” interview.

J Balvin utilized AAVE features throughout his “success”-oriented interview. However, he only used AAVE a few times during his “personal” interview: once when establishing comradery with the interviewer, and once when he described moving to New York and the inspiration he drew from the city’s hip-hop culture. Examples from J Balvin’s “personal” and “success” interviews can be seen below. AAVE features are highlighted using bold font:

Personal

Talking about moving to New York

5:08: And that was the vibe that I was looking for.

Success

Talking about style for Guess magazine

4:09: You know, sort of Miami vibe

4:13: Shout-out to Miami.

Rosalía did not use AAVE features as frequently as J Balvin for both interview types. However, there was still a disparity between the results for her “personal” and “success” interviews: while she did adopt AAVE features in the “success” interviews when talking about her new album, “Motomami,” Rosalía did not adopt any AAVE features at all during her “personal” interview. An example from her “success” interview is shown below, again with AAVE features highlighted in bold:

Success

Talking about success of new album, “Motomami”

1:54: ‘Cause I feel like Motomami’s an energy.

Our quantitative results are illustrated in Figure 3 below:

Figure 3. Adoption of AAVE Features per Minute by Artist: J Balvin adopted features more frequently than Rosalía for both interview types, and both artists adopted features more frequently during the “success” oriented interviews.

Additionally, we did not count AAVE lexical items that have entered the Standard American English (SAE) lexicon such as “gonna,” “wanna,” and “you know” toward our tallied per-minute totals for each artist. This was a subjective decision that we made based on our prediction that the use of these terms is not generally used by SAE speakers to index social identity in the same way that strictly-AAVE terms are.

Discussion

As mentioned, J Balvin used AAVE features more frequently in both of his interview types than Rosalía did, and both of them adopted AAVE more frequently in their “success” interviews than in their “personal” interviews. These findings are in line with our hypothesis, which predicted that both artists would use aspects of AAVE in contexts for which they have an incentive to communicate their connection to an “American” identity. Additionally, the overall difference between J Balvin’s and Rosalía’s utilization of this variety is in line with Britt & Weldon’s (2015) observations of native AAVE speakers’ gendered differences. It is important to note that these are only numerical comparisons, and we cannot assert any statistical significance.

Fortunately, the “success”-oriented interviews for J Balvin and Rosalía covered similar topics, including friendship (with both the interviewer and U.S. celebrities), major performances and releases, fashion, and the artists’ success with English-speaking audiences. The “personal” interviews generally explored the artists’ childhoods and family backgrounds, the artistic inspirations they drew from their home cultures, and their immigration stories. The thematic similarities between the two “personal” interviews and the two “success” interviews, respectively, allowed us to analyze them together.

The fact that J Balvin used AAVE throughout his “success” interview, which only covered topics related to his success in the U.S., while he only used AAVE in the “personal” interview when discussing moving to New York provides evidence that he crosses into AAVE when trying to communicate his proximity to “American-ness.” Additionally, our findings that Rosalía did not use AAVE features as frequently as J Balvin for both interview types is in line with the aforementioned gendered differences observed among native AAVE speakers. However, this could also be due to the fact that Rosalía’s English is not as strong as J Balvin’s, and thus she may not be as confident in switching between English varieties. At the same time, there was still a disparity in the results for her “personal” and “success” interviews: while she did adopt AAVE features in the “success” interviews when talking about her new album, “Motomami,” Rosalía did not adopt any AAVE features during her “personal” interview. This may be because her “personal” interview did not cover any topics related to the U.S., and focused mainly on her inspiration from Spanish culture. Therefore, unlike in J Balvin’s interview, in which he talked about moving to the U.S. and New York’s hip-hop culture, Rosalía did not have any incentive to index proximity to “American-ness” based on the thematic components of her interview.

Conclusion

In conclusion, we found that both J Balvin and Rosalía adopt features of AAVE more frequently when they are talking about their international success and topics related to the United States. These results provide evidence that L2 English speakers adopt features of AAVE in order to index a connection to a global conception of “American-ness.” However, despite our study corroborating our hypothesis, there were several limitations: (a) Rosalía’s English is not as fluent as J Balvin’s; (b) we only worked with a small sample size; and (c) the categorization of act sequences and features is inevitably subjective (e.g., the way we differentiated topic types into “success” versus “personal”; ambiguous lexical items that we encountered such as “gonna,” “wanna,” and “you know” that are now a part of the Standard English lexicon). Additionally, it is worth noting that Rosalía and J Balvin are celebrities; their results are not necessarily representative of all L2 English speakers since they have a particularly strong incentive to connect to Americanness (it is essentially their job to do so if they want to break into the American market).

Finally, although we did observe a gendered difference in the frequency of adoption by J Balvin and Rosalía, the Britt & Weldon study that found gendered differences for native AAVE speakers measured this difference in terms of manner (primarily, phonetic), not in terms of frequency. However, since our study was forced to disregard much of the phonetic information from J Balvin’s and Rosalía’s interviews in order to minimize confounds with their L1 accents, we were unable to conduct a similar analysis. As such, we cannot definitively state that our results align with those of Britt & Weldon (2015).

In the end, our research suggests that L2 English speakers use AAVE features to communicate a connection to the “American identity,” and calls for a future, more robust experiment.

References

Álvarez-Mosquera, P. (2015). Underlining authenticity through the recreolization process in rap music: A case of an in-group answer to an identity threat. Sociolinguistic Studies, 9(1), 51.

Britt, E., & Weldon, T. L. (2015). African American English in the middle class.

Chun, E. W. (2001). The construction of white, black, and Korean American identities through African American Vernacular English. Journal of Linguistic Anthropology, 11(1), 52-64.

Rampton, B. (2020). Crossing. The International Encyclopedia of Linguistic Anthropology, 1-5.

Tamasi, S., & Antieau, L. (2014). Language and linguistic diversity in the US: An introduction. Routledge.

Walters, K. (1992). Supplementary materials for AFR 320/LIN 325: Black English. Master’s thesis, University of Texas at Austin.

AAVE, cultural identity, L2 speakers, music

Pitch Level of Female Characters in East Asian Media

Sociolinguistics

Hannah Shin, Emily Matsuda, Cindy Xiaoxuan Wang

The idea of femininity is often grounded to common elements such as being tender, sweet, and obedient (Lee et al., 2002). This study aimed to test the relationship between one’s level of pitch and the aforementioned characteristics– specifically the role of East Asian media in promoting gender stereotypes through the implementation of various pitch levels. In order to address this question, we conducted a pitch analysis of female fictional characters in popular East Asian shows by obtaining the average fundamental frequency of a speech string through Praat (Boersma & Weenink, 2023). Unlike the hypothesis that higher pitch would correlate with the character’s degree of femininity, we found no significant difference in the average F0 value of stereotypically “feminine” and stereotypically “masculine” female characters. This finding suggests that pitch level alone does not override other non-linguistic and linguistic factors that altogether contribute to the perception of a “feminine” persona.

Introduction and Background

One’s style of speech serves as a unique indicator of their personality, gender, mood, age, and perhaps even their occupation. Even if we were to talk to someone over the phone, we would be able to learn a lot about the speaker’s identity due to this tight connection between speech and persona. This made us ask the question of: to what extent does one’s pitch level correspond to the speaker’s personality and perceived femininity? By analyzing fictional female characters in East Asian media, we attempted to identify the role that pitch level plays in reinforcing or rejecting gender stereotypes.

It is important to note that the average pitch level of East Asian females tends to differ from Western populations— hence the reason why we made within-group speech comparisons. For instance, Japanese women have higher pitches than Dutch women due to “the association of high pitch with attributes of physical and psychological powerlessness in the Dutch and Japanese cultures” (Van, 1995, p. 253). Additionally, Van (1995) reported that women with higher pitch are idealized and preferred by the general public in Japan, as high pitch level is correlated with many favorable social characteristics (Klofstad et al., 2012).

In East Asian culture, femininity is often associated with characteristics such as: being tender, sweet, obedient, and careful, while masculinity is described as having leadership, being confident, brave, ambitious, independent, and physically strong (Lee et al., 2002). For instance, characters designed to be “traditionally feminine” will exhibit submissive qualities as mentioned above, and pitch level may be adjusted to highlight their identities (Collins, 2011). We therefore hypothesized that female characters with a stereotypic “feminine” personality would produce higher pitch speech sounds on average than female characters who are depicted to be less “feminine.”

Although previous studies have revealed that sociocultural factors lead to a discrepancy between the speech of East Asian women and Western women, and a preference for high pitch in general, not many studies specifically investigate the role of East Asian media —specifically with regards to the use of specific linguistic styles in character portrayal— in perpetuating corresponding gender stereotypes. That is why our study aimed to address the significance of a female character’s speech style in East Asian media.

Methodology

To examine the variations of pitch levels among female characters in East Asian media, we conducted a quantitative analysis on the data obtained from audio clips, with the help of Praat. Specifically, two polarizing female characters were chosen from each of the following shows from different East Asian countries: iPartment (Chinese), Shitsuren Chocolatier (Japanese), and Secret Garden (Korean), leading to a total sample of 6 characters. Comparatively “feminine” and “masculine” characters were selected based on the aforementioned characteristics of masculinity and femininity (Lee et al., 2002). For each character, we obtained a 30-second continuous speech sample during a neutral conversation to avoid any highly emotional conversations, such as arguments and crying scenes. We utilized Praat (Boersma & Weenink, 2023) to measure the average, maximum, and minimum F0 levels of each character’s speech samples (Figure 1). We then compared our data and conducted an independent t-test (Table 2) to examine if the results of the comparison of average pitch levels for “feminine” vs “masculine” female characters were significantly different– indicated by a p-value of 0.05 or lower.

Figure 1: Example of character spectrogram through Praat (Gil Ra Im)

Results and Analysis

With respect to within-group comparisons, the Korean TV show Secret Garden showed results consistent with our hypothesis. The traditionally “feminine” character Ah-Young had an average F0 of 309.2 hertz, while the “masculine” character had an average F0 value of 271.2 hertz. Thus, the average F0 values obtained from this TV show provided positive evidence for our initial hypothesis that high pitch largely contributes to formulating a traditionally feminine persona. It is also important to note that the maximum and minimum F0 values possessed by the Korean feminine characters were highly similar, with the numerical difference between their maximum F0 values only being 19.8 Hz, and their minimum F0 values only differing by 4.5 Hz.

For the Japanese TV show, the stereotypically “feminine” character also had a higher average pitch than the “masculine” counterpart. Saeko, the traditionally “girly” female lead, had a mean F0 of 260.9 Hz while Kaoruko had an average pitch of 236.9 Hz. The two characters’ minimum and maximum pitch values highly aligned with one another as well, with the difference between their maximum values being 32.95 Hz and the difference between their minimum values being only 1.42 Hz. Such closely overlapping values in the minimum and maximum pitch range throughout the speech sample indicate that despite differences in average pitch level, people utilize their whole vocal range during regular day-to-day speech.

Although the data from the Korean and Japanese shows were in favor of our hypothesis, the Chinese show, iPartment, demonstrates results that suggested otherwise. The masculine character surprisingly exhibited a higher average pitch throughout her speech; specifically, the mean F0 value for the masculine character YiFei was 315.4 Hz, which is considered to be conventionally high. In comparison, the feminine character, Nuolan, had a lower average F0 of 236.9 Hz, which was significantly lower than the F0 of YiFei.

Table 1: Comparison of Average Pitch Levels

Lastly, in order to test for the statistical significance for any observed pitch differences, we conducted an independent t-test of the average value of the feminine and masculine characters’ pitch levels. The average pitch level of all stereotypically “feminine” characters was 269 Hz, while the average pitch level for the stereotypically “masculine” characters came out to 274.5 Hz. The t-test suggested that there was no statistically significant difference between these two values, as the p-value came out to p= 0.45. These findings, therefore, were insufficient in leading us to accept the initial hypothesis that the average pitch level of female characters plays a critical role in portraying a stereotypically feminine, girly persona in East Asian media.

Table 2: Independent T-Test (Feminine vs. Masculine)

Discussion and Conclusion

The sample as a whole did not support the hypothesis, as the average pitch level did not significantly differ between feminine and masculine female characters. Although we were able to observe within-group differences for the South Korean and Japanese media, we were unable to identify the predicted pattern of higher pitch in feminine characters within the Chinese drama. This suggests that pitch isn’t the only factor that contributes to one’s feminine persona– there are possible non-linguistic factors, such as appearance (ex. short hair, fashion), occupation, etc. in play that interact to achieve specific characteristics in the media. For instance, other non-linguistic similarities found within the feminine characters was their similar sense of fashion. They dressed themselves in softer colors, more feminine accessories, and wore more skirts and dresses compared to the masculine characters who had a more gender-neutral haircut, dressed in darker clothing, and largely wore pants. It is also important to note that the “masculine” female character in Secret Garden had a nontraditional occupation as a stuntwoman. Her role as a stuntwoman displayed characteristics that were previously identified as being primarily masculine: brave, ambitious, independent, and physically strong (Lee et al., 2002), which could have been a more salient factor in the portrayal of gender norms.

Additionally, it is highly likely that the impressions of masculinity and femininity differ between the three East Asian countries. Although East Asian countries may share common values and practices, there are several invariants: one of them is the link between gender norms and education level. In particular, Chinese culture often correlates masculinity with the possession of a PhD degree (Shanghai Star, 2005). In China, even a third gender type besides men and women has been proposed specifically for females with a PhD degree– this is due to the traditional belief that women are physically and mentally weaker than men, which results in unequal perceptions behind the cognitive capabilities of men and women. In comparison, such prejudiced thoughts regarding gender stereotypes and education are not as prominent in Japan and Korea. This finding above suggests that there are subtle differences in gender beliefs within East Asian culture, which might contribute to why we were not able to observe a static trend in the linguistic style of feminine female characters in media.

Figure 2: Characters from *Shitsuren Chocolatier* and their corresponding pitch analysis

Other linguistic factors that contribute to one’s persona could be the speaker’s word-choice, pitch contour, etc.– specifically in the Korean drama Secret Garden, the feminine female lead Ah-Young would often include a word-final nasal sound “ㅇ” that is associated with a playful, cute tone in Korean culture. For instance, she would add the “ㅇ” sound to the end of a neutral phrase “그랬어?” [kɨlɛs͈ʌ], creating an ungrammatical yet stylistic production “그랬엉?” [kɨlɛs͈ʌŋ]. Comparatively, the masculine female lead Ra-Im spoke in a more strict, direct, blunt tone throughout the drama, with the absence of a word-final nasal sound.

Some limitations of this study include the possibility of interference from background noise and thus an inconsistency in speech sample quality. Also, the speech samples were rather short, which might be insufficient in catering to all the variations in the stories’ settings and changes in characters’ personalities and corresponding degrees of femininity (if any occurred). In the future, the research could be improved by including an analysis of several media sources within one culture, instead of one representative film. Additionally, a longer speech sample that encodes the pitch variance throughout the entirety of the drama would lead to a more accurate analysis. Lastly, the speech sample could be cleaned up to minimize “noise.” This could be achieved by feeding the audio clip through a software that isolates linguistic sounds (aka speech sounds) with nonlinguistic ones, or by adjusting the level of the background noise on a higher-quality audio file.

Throughout this research, we heavily emphasized how the use of pitch levels is relevant in Chinese, Japanese, and Korean media to portray both feminine and masculine characteristics. However, it is important to note that a particular language is spoken differently depending on which linguistic community the speaker belongs to within a single country as well. For instance, an existing article regarding the pitch levels of female speech in two different Chinese villages — Jiuying Village and Taoyuan Village — explores how a specific language can be utilized and spoken differently depending on which linguistic community the speaker belongs to (Deutsch et al., 2009). Moreover, it was concluded by the authors that “the overall pitch level of a speaker’s voice is influenced by a mental representation that is acquired through exposure to the speech of others” (Deutsch et al., 2009), indicating that the speaker’s experience and lifestyle in a specific linguistic community affects their tones and pitches in the long run. For future research, focusing on a specific language and comparing how that language is spoken differently in various linguistic communities (hence the emergence and progression of regional dialects) may be beneficial for in depth analysis of a specific language.

It is also plausible that women’s voices are gradually getting deeper overall. The article published by BBC explores how social transformation is mirrored to our speech style, “women today speak at a deeper pitch than their mothers or grandmothers would have done, thanks to changing power dynamics between men and women.” (Robson, 2022). Since the expectations and social norms are changing over time, it is reasonable that our speech style is adjusting to them. To extend upon this research, we could investigate recent shows (past 5 years) from East Asian countries, as our focused media was relatively old and failed to capture the social norms in today’s society.

Altogether, this study revealed that our speech style and persona potentially have a bi-directional relationship, especially with the strong presence of the media in our day to day lives. It allowed us to consider the broad questions of: “What makes us perceive a certain character as feminine versus masculine? Could this factor potentially be linguistic in nature?” Despite not obtaining significant cross-cultural findings, the research uncovered a possibility of Korean and Japanese media implementing high pitch to portray femininity, and the potential for future research that reflect each country’s distinct position surrounding the notion of gender and in-depth analysis of the role of regional dialects in shaping persona.

References

Boersma, P., & Weenink, D. (2023). Praat: doing phonetics by computer [Computer program]. Version 6.3.10, retrieved 3 May 2023 from http://www.praat.org/

Collins, R. L. (2011). Content analysis of gender roles in media: Where are we now and where should we go? Sex Roles, 64(3-4), 290–298. https://doi.org/10.1007/s11199-010-9929-5

Deutsch, D., Le, J., Shen, J., & Henthorn, T. (2009). The pitch levels of female speech in two Chinese villages. The Journal of the Acoustical Society of America, 125(5). https://doi.org/10.1121/1.3113892

Klofstad, C. A., Anderson, R. C., & Peters, S. (2012). Sounds like a winner: Voice pitch influences perception of leadership capacity in both men and women. Proceedings of the Royal Society B: Biological Sciences, 279(1738), 2698–2704. https://doi.org/10.1098/rspb.2012.0311

Krahé, B., & Papakonstantinou, L. (2019). Speaking like a man: Women’s pitch as a cue for gender stereotyping. Sex Roles, 82(1-2), 94–101. https://doi.org/10.1007/s11199-019-01041-z

Lee, B. S., Kim, M. A., & Koh, H. J. (2002). Development of Korean gender role identity inventory. Journal of Korean Academy of Nursing, 32(3), 373-383. https://doi.org/10.4040/jkan.2002.32.3.373

Robson, D. (2022, February 25). The reasons why women’s voices are deeper today. BBC Worklife. https://www.bbc.com/worklife/article/20180612-the-reasons-why-womens-voices-are-deeper-today

Shanghai Star. (2005, March 4). Women PhDs the 3rd type of people besides men, women? China Daily. https://www.chinadaily.com.cn/english/doc/2005-03/04/content_421834.htm

van Bezooijen, R. (1995). Sociocultural aspects of pitch differences between Japanese and Dutch women. Language and Speech, 38(3), 253–265. https://doi.org/10.1177/002383099503800303

East Asian media, fundamental frequency, Gender Stereotypes, Media, vocal pitch

Is Instagram the New Duolingo? Investigating How Social Media Usage Affects Heritage Language Competence

Bilingualism

Sarah Bassiry (Sky), Anna Harutyunyan, Akina Nishi, Jasmine Shao

Diaspora communities and heritage language speakers are a very unique population when it comes to language and bilingualism. Heritage speakers vary greatly in their language skills, language background, and environment. As heritage speakers are generally exposed to their heritage language only at home or in other limited contexts such as a cultural community group, this study investigated to see if social media may also be a context of heritage language exposure for some heritage speakers. If so, this study investigates the role social media might play in language competence. Four Eastern Armenian and four Mandarin heritage speakers attending UCLA and one native speaker in each language, were participants in this study. The participants were given a language background survey, a grammaticality judgment test, and an elicitation task judged by a dominant native speaker using a Likert scale. Initially, we expected to see a positive correlation between social media usage in the heritage language and the participants’ heritage language skills. However, the results did not provide sufficient evidence to support this hypothesis. Thus, further testing with a larger sample size is recommended to further investigate whether or not there might be a correlation.

Introduction and Background

The concept of bilingualism is interesting to discuss in the context of heritage speaker populations, as these groups include speakers of a vast variety of language competence. Heritage language speakers ‘may or may not be frequent users of the language or exceptionally proficient in the language” (Mahootian, 2000). Many heritage speakers are also receptive bilinguals, which Nakamura (2019) notes, “does not imply zero production of the weaker language but highly limited language use in interaction.” These speakers usually have strong listening and comprehension skills but weaker production skills in their heritage language.

This study sets out to determine if there is a correlation between heritage speakers’ social media consumption in their heritage language and their competence in that language. A prior study on bilinguals and newspaper consumption conducted by Moring (2011) et al., found media to be an instrumental resource in maintaining language competence. Flewitt and Zhao (2020) also conducted a study investigating the relationship between heritage speakers and social media usage. In their study, they observed two adolescent Mandarin Chinese heritage speakers and their usage of WeChat to connect with relatives in the heritage language. They found that WeChat acted as a “springboard to include multilingual practices” and facilitated a motivating environment of supportive feedback from interlocutors for their language to grow.

Given the previous research in the field, we hypothesized that our study would show that more frequent social media consumption in the heritage language will be correlated with a higher degree of language competence. Previous studies have noted that in the case of young children, engagement with social media often involves “several types of learning and socialization occurring all at once” (Zhao, 2019). In this study, we aimed to observe the role social media activity plays in the heritage language competence of young adults at UCLA.

Methods

The population this study investigated was heritage speakers of either Mandarin or Eastern Armenian who attend UCLA, testing 4 participants for each language. This project utilized survey methods, grammaticality judgment tasks, and elicitation tasks to investigate the research questions.

Our general survey asked participants basic questions regarding demographics, social media usage in the heritage language, and information regarding details of heritage language background and exposure. Two separate grammaticality judgment tests with 20 questions each were created by bilingual speakers of each language from the research team. The Eastern Armenian task tested plural suffixation and word order, the morphosyntactic structure of the language. The Mandarin task tested participants’ knowledge of aspect markers and measured words. We chose to test structures that differ from English, which are commonly lost by heritage speakers. They also included romanization and audio recordings of the tasks to account for the varying degrees of literacy.

The elicitation tasks consisted of a story-retelling exercise where the participants were virtually shown an animated short story from Youtube in the heritage language and then asked to retell the story in the heritage language. A Likert scale was then used by a native speaker to judge each participant’s fluency levels as they viewed the recording of the participant’s elicitation.

Results

The participants in the study varied in the production and comprehension levels of their respective heritage languages, and their use of social media in the heritage language.

Eastern Armenian heritage speakers varied in their production and comprehension levels of the language. Based on the results of the elicitation and grammaticality tasks, it appears that receptive knowledge is not necessarily related to expressive knowledge. For instance, the native speaker ranked participants B2 and B3 as the lowest proficient speakers of the heritage language, but these same participants had two of the highest scores for correctly identifying grammaticality judgments in their heritage language.

Figure 2. Grammaticality & Elicitation Tasks Results (Eastern Armenian).
Figure above shows that participants had varying results on the grammaticality judgment tasks and the elicitation task. Only B4 has high scores on both tasks.

Additionally, the Eastern Armenian heritage speakers who score the highest on the elicitation and grammaticality judgment tasks consumed social media in various amounts, ranging from rarely or never to as frequent as daily consumption. Participants B2 and B3 scored the highest on the grammaticality judgment tasks, but B3 said that they rarely or never consume social media in their heritage language, while participant B2 did so on a daily basis, in formats of video, audio, and text. Participant B4 scored highest on the elicitation task and the second highest on the grammaticality judgment tasks, and used social media in the form of Instagram video multiple times a month. Lastly, participant B1 scored the lowest on the elicitation task and had a somewhat low score on the grammaticality judgment task. They consume social media in their heritage language on a weekly basis in the form of audio and text.

All in all, the data from our Eastern Armenian speakers does not represent any correlation between social media usage and heritage language competence.

Figure 3. Self-Evaluated Language Interaction Survey Results (Mandarin).
Figure above shows that more than half of the participants received formal education in Mandarin and only a quarter of the participants use social media in Mandarin.

Our Mandarin heritage speaker participants also varied in their levels of fluency. Diverging from the lack of relationship we observed in Armenian, those who performed highly on one test usually performed highly on the other. Only one of our four participants for Mandarin, participant A2, expressed consumption of social media in Mandarin, and this participant also had the highest score on the grammaticality judgment tasks. This speaker could serve as one data point to support our hypothesis that heritage language social media consumption is correlated to higher levels of retention in the heritage language. However, participant A4 scored second highest on both tasks, but they do not consume social media in Mandarin. Participants A1 and A3 also do not consume social media in their heritage language and received lower scores on both tests. Although the rough possibility of a correlation exists in the Mandarin data set, there is not enough evidence to definitively support our hypothesis.

Figure 4. Grammaticality & Elicitation Tasks Results (Mandarin).
Figure above shows that participants had varying results on the grammaticality judgment tasks and the elicitation task. Only A2 has high scores on both tasks.

Discussion and Conclusion

As a result of our investigation, we were unable to find supporting evidence for our initial hypothesis that higher social media consumption in the heritage language corresponds with higher competency in the heritage language. Some factors that led to this conclusion is the lack of concrete and frequent social media usage among the participants. For instance, in the Mandarin heritage speakers cohort, participant A2 claimed they use social media and received the highest scores on both tasks as seen in Figure 4, however, participant A4 who claimed they don’t use social media got the second-highest scores in the task. Similarly, in the Armenian heritage speaker cohort, participants B2 and B3 received similar results in both tasks (Figure 2) and the two highest on the grammaticality judgment task. However, participant B3 claimed they use social media never/rarely, while participant B2 uses social media on a daily basis. The contrast can be thought of to be due to formal instruction in the heritage language.

Our new finding, formal education, correlated with higher scores on both tasks. For instance, participants B2, B3, and B4 received formal instruction in learning Eastern Armenian. The Mandarin heritage speaker who performed the highest on both tasks, A4, also received formal education in Mandarin. This correlation was of interest for other scholars who claim that the lack of formal education has negative effects on the acquisition of the heritage language; children who don’t receive formal education do not learn the formal registers, vocabulary, and complex structures in written language (Montrul, 2010). Our data reflects the aforementioned claim as the participants who did not receive formal instruction in Eastern Armenian and Mandarin performed the lowest or second lowest on the grammaticality judgment task which tests their knowledge of the language grammar. While formal education can be a factor that influences higher competency in the heritage language, this correlation has to be interpreted in terms of production or grammatical knowledge. This is because the participants who received formal education performed well on the grammaticality judgment task, but not as well on the elicitation task. These results show that analysis between formal education and heritage language competency should be further investigated.

Another factor that can influence the participant’s competence in the heritage language is the age of exposure. There is contrasting data on how the age of exposure relates to language competence. Proponents of the Critical Period Hypothesis claim that children should be exposed to a language at a certain age to be able to acquire it native-like (Mahootian, 2020). Others claim that other factors such as ‘the child’s temperament or motivation’ can affect their learning of a language (Yuldasheva, 2021). In our research, we found that 7 out of the 8 participants in our study claimed that they were exposed to their heritage language from 0 to 3 years old, with one exception, A4, who claimed they were exposed to their heritage language from 4 to 8 years old. In addition, more than half of the participants, 5 out of 8, claimed that they were exposed to both English and the heritage language at the age of 0-3, simultaneously. The participant who performed the best on both tasks in Mandarin claimed to have been exposed to English and Mandarin simultaneously from the age of 0-3. The participant who performed the best on both tasks in Eastern Armenian stated that they were exposed to English from the age of 4-8 and Armenian, from the age of 0-3. This shows that age of exposure may not play a major role in the heritage speaker’s language competency. Perhaps, the difference lies in simultaneous and sequential bilingualism. Age of exposure to the heritage language may be a sufficient factor in the heritage language competency, however, this factor has to be accompanied by other factors, perhaps, social media consumption and formal education in the heritage language.

Another finding is that there may be a correlation between longer attendance to formal instruction and higher social media usage. Social media in the heritage language provides an extension to the traditional classroom setting. Going back to Zhao & Flewitt’s case study, heritage language speakers’ language practices on WeChat extended their “Chinese-speaking social world and provided them with opportunities to engage in contextualized, transnational heritage language and literacy practices beyond their home and Saturday Chinese school environment.” (Zhao & Flewitt, 2020). We could make the claim that formal instruction and social media in the heritage language complement one another to learn, and a longer history of instruction may motivate the bilingual speaker to seek more natural and conversation-based learning environments, such as social media.

Overall, claiming that the heritage speakers’ competency in the heritage language correlates with their social media consumption would be invalid based on the findings of our research. Other variables that can account for this contrast are formal education in the heritage language or age of exposure to the language. Perhaps, formal education in the heritage language could account for these three heritage speakers’ higher performance on the grammaticality judgment and in one of the cases, the elicitation task. In a similar manner, the age of exposure to the language can be a variable that influences the results. A detailed analysis above shows that these are sufficient factors of heritage language competency and the limitations of our study prevent further research into these variables. Some of the limitations of our study include a smaller sample size. With a larger sample size, the findings of our research could be more reliable and generalizable.

However, it was interesting to see a small sample size showcase such a wide variety of language proficiency. Bilingual heritage speakers have unique backgrounds in their heritage language because of many internal and external factors, such as “the role of the school system and other institutions, the historical experiences of particular language communities, the unique circumstances involved in the adoption by some communities or individuals of proxy HLs as part of the complex multiple identities of contemporary life, and the specifiable impact of a language ecological pattern over the life cycle of individuals and families” (Lo Blanco, 2001). Our participants had different characteristics, such as high proficiency in listening but lower proficiency in writing, or vice versa, as well as the amount of formal education, ranging from a couple of years to over 10 years.

For future research, if these methods are adapted, we suggest that elicitation tasks are performed in an in-person setting rather than virtual, where the native speaker listens to the heritage speaker’s elicitation and rates them in person. Additionally, the elicitation tasks should be reviewed by different native speakers. We believe that these changes will allow the results to be more reliable and generalizable.

References

Lo Bianco, J. (2001). Language Policies: State Texts For Silencing and Giving Voice. Difference, Silence and Textual Practice: Studies in Critical Literacy, 31-71.

Mahootian, S. (2019). Bilingualism. Routledge.

Montrul, S. (2010). Current issues in heritage language acquisition. Annual Review of Applied Linguistics, 30, 3-23.

Moring, T., Husband, C., Lojander-Visapää, C., Vincze, L., Fomina, J., & Mänty, N. N. (2011). Media use and Ethnolinguistic Vitality in bilingual communities. Journal of Multilingual and Multicultural Development, 32(2), 169–186. https://doi.org/10.1080/01434632.2010.541918

Nakamura, J. (2019). Receptive bilingual children’s use of language in interaction. Studies in Language Sciences, 18, 46-66.

Yuldasheva, D. (2021). Age and The Second Language Acquisition. Research Jet Journal of Analysis and Inventions, 2(04), 124-130.

Zhao, S. (2019). Social media, video data and heritage language learning. In G. Falloon, J. Rowsell, & N. Kucirkova (Eds.), The Routledge International Handbook of Learning with Technology in Early Childhood. Routledge.

Zhao, S., & Flewitt, R. (2020). Young Chinese immigrant children’s language and literacy practices on social media: a translanguaging perspective. Language and Education, 34(3), 267–285. https://doi.org/10.1080/09500782.2019.1656738.

heritage speakers, language proficiency, social media

Impassioned Speech: ¿Lo has visto?

Bilingualism

Nikki Lee, Kiara Mares, Jamie Seals, Ally Shirman

In everyday conversations, bilingual speakers frequently code-switch between their languages. With our modern society, we can see this becoming prevalent in media during emotional scenes with bilingual characters; specifically in movies and TV shows. In this study, we investigated three different shows and movies with bilingual actors who use their heritage language as a part of their character to see whether there is a correlation with code-switching in highly emotional situations. We chose three sources to gather this data that have at least one bilingual main character; Jane the Virgin (2014), Everything Everywhere All At Once (2022), and Modern Family (2009). For each source, we analyzed around 120 minutes to have an equal amount of data. Analysis of this data found that code-switching during emotional scenes did occur, but was more prevalent in more recent media. The study helps to show that code-switching is becoming more representative for bilingual speakers in modern media.

Introduction and Background

Entertainment media heavily relies on creating a connection between the viewers and the characters and themes presented. The characters in media are often the biggest attraction because of their relatability. Thus, the representation of characters with different backgrounds, including gender, ethnicity, age, socioeconomic status, etc. is very powerful in shaping viewer and societal perceptions. Characters rely heavily on language and emotion to give away key details and shape who they are and what their role is. This is especially evident in American entertainment media with actors who play ethnic characters, with cultures other than dominant U.S. culture. Bilingualism and code-switching are some of the biggest tools used to portray identity within these characters’ roles. This makes their heritage language use very strategic and intentional for character depiction. Recognizing the power of media and how it aims to reflect real-life ideas, our project focuses on where entertainment media uses second languages, and how it intersects with emotion to reflect ideas about a language or a character.

Our project focuses specifically on bilingual actors within entertainment media who use their heritage language as part of their character. We want to investigate when this heritage language is used and if it is associated with highly emotional scenes or situations. Research on language and emotion in real-life scenarios has indicated that language use and choice can be influenced by emotions that we correlate with that language. Looking specifically at American media, our goal is to observe whether or not heritage language use occurs in more emotional situations than non-emotional scenes, and identify if there are emotions that occur more often in a heritage or dominant language. This project will be focusing on language contexts, both positive and negative, and then observing what language it occurs in. Based on our own experiences with code-mixing, media depictions, and language perceptions, we hypothesize that in highly emotional situations, characters will code-switch to their heritage language, other than English. We predict that the characters are written to represent real-life bilingual speakers to deepen their character and the media. Many viewers may find this relatable or can connect with the cultural identity of both languages being presented – which could contribute to further media representations.

Gaps in Research

Though bilingual speakers often use one language in specific settings, and the other in other settings (e.g. heritage language at home and in informal contexts, and dominant language in formal settings like work, especially if that second language is the standard language of their country), they also often use both languages at once (Ferguson, 1959). This most commonly happens when speaking to other bilingual people, but it does happen elsewhere as well. This use of both languages at once is referred to most commonly as code-switching, where interlocutors will use both languages at their disposal within the same utterances, often without consciously making the decision to do so. Though there are many ways and uses of communicating via multiple languages at once, a prominent trigger for code-switching comes from emotionally arousing situations. One study found that, through direct observation of participants in a high-stress problem-solving task, people code-switched more frequently when expressing negative emotions. Specifically, participants switched to their heritage language to express these negative feelings rather than using the dominant language (Williams, Srinivasan, Liu, Lee, Zhou, 2020). We anticipate that our study will recognize some of the same characteristics of code-switching and plan to focus on a character, then identify the number of situations in which this occurs.

Additionally, in these highly arousing emotional situations, it is more likely that speakers use their first language to react, rather than their second language. We see this in the study conducted by Ariana Mohammadi titled, “Swearing in a second language: the role of emotions and perceptions.” Additionally, a study conducted by Shahrzad Mahootian covers these emotional situations in their study and they state, …the list of evocative switches was “more meaningful” in Spanish, that they carried “more emotional power,” (“emotional statements will be said in Spanish. English is not sufficient”) that they were used to “create solidarity,” that they were “much stronger” when said in Spanish (Mahootian 2005). This study helps to show how speakers would be more motivated to code-switch in order to convey their words more effectively between themselves and their listeners. As we have shown, there has been a good amount of research done on the connections between code-switching and heritage language use with emotion but this research has not been carried over to the media. There has not been research done specifically on character use of code-switching and heritage language use and the emotion of their character or scene. We hope to fill this gap with our research in this project and we can anticipate that the media we have chosen for our study will be able to reflect these previous findings.

Methods

To understand whether code-switching occurs in half or more emotional instances in media, we used three sources that each include at least one bilingual main character and are from the 21st century. Specifically, Jane the Virgin (2014), Everything Everywhere All At Once (2022), and Modern Family (2009).

From left to right: *Jane the Virgin (2014), Everything Everywhere All At Once (2022), Modern Family (2009)*

Additionally, we made sure that all three of these sources’ bilingual characters’ heritage languages were relevant to the characters themselves, but not a dominant focus for the storylines. All bilingual characters were interacting in a space where their heritage language is not the dominant language (which is English in all three sources).

We analyzed roughly 120 minutes of each source to ensure that the amount of data we collected from each would be equal, based on the length of Everything Everywhere All at Once. (We looked at three Jane the Virgin episodes (each ~40 minutes), and 6 Modern Family episodes (each ~20 minutes)).

We noted (coded) each instance of an emotional scene in each source, as well as instances of code-switching within these emotional scenes. This would allow us to calculate a ratio (number of code-switching instances in emotional scenes divided by the total number of emotional scenes), and thus a percentage, to understand how accurate our hypothesis is.

Before coding each source, we defined what “emotional” meant to ensure that each researcher was coding via the same guidelines. Highly emotional instances could come in the form of:

Voice: raised voice, yelling, screaming, whimper
Exaggerated body language
Clearly emotional facial expressions that last (e.g. no microexpressions)
Crying

Results

Everything Everywhere All At Once

4 code-switching instances in 8 total emotional scenes → 50% code-switching

Jane the Virgin

15 code-switching instances in 35 total emotional scenes → 43% code-switching

Modern Family

8 code-switching instances in 47 total emotional scenes → 17% code-switching

Combined: 27/90 → 30% of all emotional scenes across the three sources included code-switching.

Everything Everywhere All At Once was the only source to meet the hypothesis that at least half of the emotional scenes would include code-switching. However, it is important to note that there was quite a bit of diversity in how many instances of emotional scenes there were per source. Everything Everywhere All At Once only included eight, likely because it has deep themes compared to the sitcom-like shows, and the scenes are generally longer. Modern Family, on the other hand, is fast-paced (only 20 minutes per episode) and includes generally more exaggerated characters, leading to their behaviors being more exaggerated and emotional as well. This is reflected in the 47 total emotional scenes, compared to the eight from the movie.

However, even though Modern Family had the most emotional scenes, it had the lowest ratio (17%) – this is likely because there is only one bilingual character (Gloria), giving her fewer opportunities and reasons to code-switch (e.g. code-switching out of instinct when stepping on a Lego, versus code-switching to communicate an idea to someone who also speaks English and Spanish that she wouldn’t otherwise be able to explain).

Visual Media

Patterns and Analysis

Results of our study did not exactly align with our data but did highlight some interesting patterns. There were three particular patterns that were recurring in the data and had a similar basis to previous research in the realm of bilingualism in the media.

The first pattern observed from our data was the use of bilingual code-switching for reference terms. In all three of the media forms, characters code-switched when referring to another character they were close to or comfortable with. In Jane the Virgin, this was observed with the consistent referral to Jane’s grandmother as “abuela,” even when the rest of the sentence was in English. ‘Gloria’ in Modern Family also did this frequently when referring to her husband with terms of endearment such as “mi amor” (season 9, episode 3) and when referring to Haley’s children as “mis niños” (season 11, episode 6). This pattern touched on previous research that found that heritage language is used when speakers are more comfortable, which would make sense for using with family (Cho, 2015).

Modern Family Phil makes Gloria cry

Another pattern observed was use of bilingual code-switching as a form of addressing certain audiences. During scenes where characters were in the presence of other bilingual characters, or when they wanted to speak to someone directly, sometimes to the extent of using language to be private. This example was seen in Everything Everywhere All At Once when ‘Evelyn’ is standing up to ‘Gong Gong’ and switches between languages to focus on different scenarios, despite him understanding each of the languages. Evelyn chooses to code switch from Cantonese, English, and finally Mandarin to allow the listeners of that scene, her daughter and husband, to better understand her feelings. Other scenes like in Jane the Virgin used code-switching to Spanish to make sure the police didn’t understand their plan of action while he was in the room. This pattern is reflective of previous research that concluded that negative languages are more typically expressed in the heritage language, which could explain why it was used to exclude someone from a conversation (Williams, Srinivasan, Liu, Lee, Zhou, 2020).

Everything Everywhere All At Once (2022) – Evelyn Stands Up To Gong Gong | HD

Jane the Virgin used code-switching to Spanish

Finally, we observed bilingual code-switching as an indicator of urgency. Characters who were experiencing a sense of urgency and trying to communicate this hurry used code-switching to do so. This example can be seen in Everything Everywhere All At Once when Evelyn is dealing with multiple stressful situations while her family members are adding to that stress, which makes her frantic. The audience can see her emotions overflow as she code switches from Mandarin to English in order to fully express her frustration in that situation. Given that urgency could be related to other feelings, such as stress and desperation, this observation came closest with our hypothesis.

Despite the shows having different content matter, being different genres and having different amounts of bilingual characters (and thus different instances of code-switching), these were patterns that overlapped in at least two of the three media examples used. The patterns occurred in various examples throughout each of the shows and seemed to be consistent parts of the character and their role in the plot. Each of the patterns identified was relevant in various instances throughout the shows and highlighted some real-life uses for code–switching and bilingualism. Although it wasn’t what we were expecting, it definitely taught us a lot about what to expect for bilingualism in both media and everyday life.

Limitations

Within our study, there were some limitations that potentially contributed to the results being different than our predictions, but definitely impacted the process. In relation to content, our project primarily focused on three media examples. Had we used more than two shows and one film, our results may have had more range in code-switching and examples of bilingualism. Essentially, a wider range in media examples would have been more beneficial for getting a wider scope of media examples. This connects with the second limitation: variation within the media content examples. Modern Family definitely had more “feel-good,” positive moments which is different from Jane the Virgin which dealt with heavier themes and was more of a drama. While this variation was definitely better for maximizing examples of bilingual code-switching, the lack of consistent storylines and genres didn’t allow for the same amount of opportunities for code-switching during impassioned speech. Instead, the variation in content and storylines would have made code-switching more of a chance occurrence than a predictable feature.

Finally, perhaps the most impactful limitation in our research project was that most characters were the only bilingual characters in the show. This was specifically the case for Modern Family, where ‘Gloria’ was the only character who spoke Spanish and didn’t have any other main character to speak it with. In Jane the Virgin, even though ‘Jane’ and ‘Xiomara’ were Spanish-English bilinguals, they would respond in English and ‘Abuela’ was the only one who spoke Spanish the entire time. Each of these limitations could have contributed to the results and for future studies would be important to avoid.

Future Study

Considering the limitations previously mentioned and in collaboration with the patterns mentioned, a refined, future study would certainly involve a wider range of media for observation. For standardization purposes, it would be important for shows to be produced and based in the US, be from the same range of years (as in our example, the last two decades/the 21st century), and have the same amount of minutes watched, but this new study would focus on shows with more bilingual characters. Considering these elements and our previous interest in bilingualism and code-switching in the media, further research conducted could be based upon media that covers different backgrounds to be more inclusive towards the American audience. We can see a trend of this from big companies, like Disney, who have produced movies like Encanto and Turning Red. With this trend of inclusivity, there is a high possibility of representation for bilingual speakers in future media. In turn, this would allow for more possible studies to look into the instances of code-switching during emotional scenes.

Conclusion

This analysis of bilingualism in media is a good first step for better understanding how bilingual and multicultural people are represented through TV, specifically to American, English-dominant audiences. However, no broad conclusions can be drawn from this study alone; without a large sample size, we may have missed important themes and patterns on the representation of code-switching from real life, or the patterns that we did find may not appear to the same degree on a larger scale.

A larger sample size, both in terms of the amount of minutes analyzed for each source, and the amount of diverse sources chosen, should be used to create a clearer picture of how writers choose to represent bilingual characters’ language use in media.

Despite the limitations and potential to make this study broader, this introductory analysis still has shown how some patterns of code-switching (proven to be common via research studies mentioned earlier) in real life are accurately represented in media as well. As the representation of multicultural and multilingual people in the United States is being more and more encouraged in American media, the more we are actually seeing it. As in Everything Everywhere All At Once – the most recent, and most diverse, media source used in this study, it was the one source that did meet our hypothesis (50% code-switching in emotional scenes).

As screenwriters continue to portray a more accurate representation of the diversity in the United States, the more we should expect to see multi-language dialogue, filled with code-switching in these shows and movies. In future studies, a specific focus on the function of code-switching by bilingual characters, both in and out of emotional contexts, could give even more insight into how multilingual and multicultural people are represented in older media versus current media. Some examples of code-switching (and code-mixing) functions include keeping conversations private from others, raising the status of the agent who is speaking, making side comments or humorous remarks, emphasizing what is being said, or reiterating something (Gumperz, 1982). Taking this approach and categorizing the linguistic reasons for code-switching would be a great potential first step for understanding where United States media companies are at in terms of true representation of the diversity in this country.

References

Cho, G. (2015). “Perspectives vs. Reality of Heritage Language Development Voices from Second-Generation Korean-American High School Students.” https://files.eric.ed.gov/fulltext/EJ1065385.pdf

Ferguson, C. A. (1959). “Diglossia” WORD, vol. 15, no. 2, pp 325-340., https://doi.org/10.1080/00437956.1959.11659702

Gumperz. (1982). Conversational code switching. Discourse Strategies, 59–99.

https://doi.org/10.1017/cbo9780511611834.006

Lee, J. S. (2010, April 23). The Korean Language in America: The Role of Cultural Identity in Heritage Language Learning. Taylor & Francis Online. Retrieved February 5, 2023, from https://www.tandfonline.com/doi/abs/10.1080/07908310208666638

Mahootian, S. (2005). Linguistic change and social meaning: Codeswitching in the media. International Journal of Bilingualism, 9(3-4), 361-375.

Mohammadi, A. N. (2020). Swearing in a second language: the role of emotions and perceptions. Taylor and Francis Online. https://www.tandfonline.com/doi/full/10.1080/01434632.2020.1755293

Williams A, Srinivasan M, Liu C, Lee P, Zhou Q. Why do bilinguals code-switch when emotional? Insights from immigrant parent-child interactions. Emotion. 2020 Aug;20(5):830-841. doi: 10.1037/emo0000568. Epub 2019 Mar 14. PMID: 30869940; PMCID: PMC6745004

bilingualism, Code-Switching, heritage speakers, Media

El tema que hablamos de: Spanish Heritage Speakers and Language Maintenance in the U.S. and France

Bilingualism

Ashley Ghodsian, Madeleine Kostant, Maxime Guerra, E Perez

There are various classifications for bilingual speakers in the formal study of bilingualism. Among these are heritage speakers, who have acquired a minority language in the home prior to gaining dominance in their majority language of the society upon being exposed to it in the community and in school. This results in unbalanced bilingualism in detriment to the minority (or heritage) language, a process of grammatical loss or weakening known as attrition. The emphasis of this study is whether this attrition in the heritage language can be attributed to transfer from the majority language, or if it is due to general loss of input or other extralinguistic factors. In order to accomplish this task, this study analyzed a specific syntactic (or sentence structure) construction, prepositional relative clauses, in populations of English-dominant and French-dominant heritage speakers (“HSs”) of Spanish. We hypothesized that previously reported attrition in the grammars of English-dominant Spanish HSs was due to dominant language transfer from English and that French-dominant HSs would not exhibit this same attrition. We conducted both receptive judegment tasks and oral production tasks in order to test this phenomenon, and found evidence that seems to largely be in favor of our hypothesis. We conclude by commenting on the broader implications of our research on bilingualism and pedagogy, especially as it relates to heritage language maintenance.

Introduction / Background

This study aims to examine heritage speakers’ attrition through the lens of dominant language transfer by probing this question with two groups of HSs: English-dominant Spanish HSs and French-dominant Spanish HSs. To do so, we focus on the syntactic structures of prepositional relative clauses. Relative clauses (“RCs”) are clauses used to modify a noun (e.g., In The dress that Mary wore was very old, the clause that Mary wore is an RC that modifies the noun phrase the dress.).

Prepositional relative clauses, which are notable because they contain a prepositional object, have two possible realizations in a sentence. The first is preposition-stranding (“P-stranding”), which occurs when a preposition is “stranded” at the end of a relative clause (e.g., The track which they train on _ is slippery and dangerous). The second, pied-piping, occurs when the preposition and the wh-element move together to the front of the clause to form a relative clause (e.g., The track on which they train _ is slippery and dangerous). All three languages explored in this study—Spanish, French, and English—accept pied-piping constructions. Our central motivation behind choosing these three languages, however, stems from the fact that English accepts P-stranding, but it is ungrammatical in both Spanish and French. We will return to this cross-language split of grammaticality with P-stranding shortly.

Several previous studies on this topic have inspired our investigation and experimental design. Kim (2007) found that Korean HSs in the U.S. exhibited more attrition of a specific syntactic construction than their counterparts in China. Although this had to do with a different construction than P-stranding, the findings demonstrate language transfer in action, specifically because Korean and Chinese share the same grammatical setting with respect to the construction being tested, whereas English holds a different setting from the other two. This study has been the largest source of motivation for us to investigate dominant language transfer for other types of syntactic constructions.

Further research on P-stranding and Spanish heritage speakers inspired our experiment design, notably Depiante and Thompson (2013) and Pascual y Cabo and Soler (2015). Depiante and Thompson (2013) conducted acceptability judgement tasks with heritage Spanish-English bilinguals, asking their participants to judge pied-piping and P-stranding in various constructions, including relative clauses. They determined that Spanish HSs generally accepted P-stranding more than their Spanish-dominant counterparts. A follow-up study by Pascual y Cabo and Soler (2015) replicated the experiment and further found that the timing of English acquisition also had a notable effect on the rate at which the speakers accepted P-stranding. Both of these studies relied on receptive judgement tasks in which participants read sentences and evaluated their grammatical acceptability. Our study seeks to contribute to this conversation by combining oral production tasks alongside receptive tasks.

The other gap in research that we hope to fill has to do with testing the two different groups of heritage speakers that we are studying, as previously mentioned. Our research question is as follows: is the apparent attrition of Spanish heritage speakers’ grammar (i.e., the acceptance of preposition-stranding) due to a general loss of input or to transfer from the grammar of the dominant language? We are able to investigate this question largely because the three languages we are exploring have the same pattern as the three languages in the Kim study: P-stranding is grammatical in English but ungrammatical in Spanish and French. Therefore, if we find no difference between our English-dominant and French-dominant speakers, we would not be able to attribute any potential attrition to transfer from the dominant language. However, if we find that our English-dominant HSs accept and produce P-stranding at a higher rate than the French-dominant HSs—which is what we hypothesize for this study—we can make the claim that this is due to the fact that English accepts P-stranding and French does not, and therefore that attrition is due to dominant language transfer in the case of prepositional RCs.

Methods

We recruited six participants total; five Spanish-English bilinguals from the U.S., and one Spanish-French bilingual from France. All of the speakers were heritage speakers of Spanish, with ages ranging between 19-23 years old. After recruitment, participants completed our experiment in several stages. First, they completed the production task over Zoom, during which they were shown a presentation with clipart images and accompanying sentences, some of which were test sentences and some of which were fillers. Each test sentence contained a prepositional object, and participants were prompted to rephrase that sentence with a relative clause so that we could see where they placed the preposition. An example of one of these slides is shown in Figure 1 below.

Figure 1. Example from the elicitation task in which participants are prompted with the first sentence (left), “The children talk about a classic movie,” and are then required to complete the second one (right), which asks “What is this? This is the classic film…” and results in the response “… about which the children talk” if participants use pied-piping or “…which the children talk about” if they use preposition-stranding.

After the production task, participants completed a “pre-survey” called the Bilingual Language Profile (BLP) (https://sites.la.utexas.edu/bilingual/), a tool that linguistics researchers use to determine the language exposure and dominance of bilingual speakers. The BLP contains 19 main questions, each with their own sub-questions, which probe participants’ language exposure, linguistic history, daily use, confidence in fluency, and linguistic identity. All of this data is then used to automatically generate a dominance score, which determines which of the two languages the participant is dominant in. We used the dominance score and our participants’ self-reported ages of exposure in order to ensure that all participants were heritage speakers.

After the BLP, participants completed a main survey (the receptive task) in which they ranked sentences with relative clauses with prepositions, in addition to filler sentences, on a scale of 1 to 5 based on how natural or unnatural they sounded according to their intuitions. Our test conditions were evenly divided between pied-piping and P-stranding sentences. An example of one of our test sentences is shown below in Figure 2.

Figure 2. Example from the receptive task where participants were asked to rate sentences on a scale of 1-5 based on how natural or unnatural they sounded. This sentence is a pied-piping example, and it reads: “The film about which they speak has won many awards.”

Results

We found that for both the elicitation and receptive tasks, the Spanish heritage speakers from the U.S. produced or accepted P-stranding at a higher average rate than the participant from France. It is important to note that we ended up having four (4) Spanish heritage speakers from the U.S. and one (1) from France. Initially, we had five (5) participants from the U.S., but one of them rated the aforementioned “good” (i.e., grammatically correct) control sentences in the acceptability judgment task too low; as such, we disregarded their data.

Figure 3 below presents the production task results. The U.S.-based Spanish heritage participants produced P-stranding constructions 50% of the time, compared to 0% of the time for the French participant. For example, each participant was given the prompt: “Los meseros trabajaban con una cocinera famosa. Esta es la cocinera famosa…” (‘The waiters worked with a famous cook. This is the famous cook…’) and asked to complete the sentence. Two out of four (50%) of the U.S. participants responded with “…que los meseros trabajaban con” (‘…whom the waiters worked with’), stranding the preposition “con” at the end of the phrase. In contrast, the French participant responded “…con que los meseros trabajaban” (‘…with whom the waiters worked’). Given that P-stranding is not grammatical in French or Spanish, these results are in line with our hypothesis, which predicts that the English-Spanish bilinguals would produce more P-stranding than their French-Spanish counterparts due to transfer from English grammar.

Figure 3. Average Production Results Across Speaker Groups: out of all opportunities to produce P-stranding (every relative clause with a preposition), the 4 U.S.-based speakers produced it 50% of the time whereas the participant from France did not produce it at all and produced pied-piping 100% of the time.

Figure 4 below presents the acceptability task results. As explained above, the participants were given a mix of P-stranding, pied piping, and grammatical and ungrammatical filler sentences in Spanish, and asked to rate their grammatical acceptability on a scale of 1 to 5 (5 being the most acceptable). The U.S.-based Spanish heritage participants rated P-stranding constructions to be an average of 3.33, compared to 1.67 for the French participant. For example, when given the sentence “La serie que hablan de tiene críticas muy buenas” (‘The show which they talked about has really good reviews’), the Spanish-English bilingual speakers gave it an average acceptability rating of 4 compared to 2 for the Spanish-French bilingual speaker. These results align with our hypothesis, which predicts that the Spanish-English bilingual speakers would accept P-stranding more than the Spanish-French bilingual speaker due to transfer from English grammar.

Figure 4. Average Acceptability Ratings Across Speaker Groups: Spanish-English bilinguals judged P-stranding with an average rating of 3.33 out of 5 on an acceptability judgement scale and the Spanish-French bilingual judged P-stranding with an average rating of 1.67 out of 5.

Discussion

Our data from both the elicitation and acceptability judgement tasks provides evidence in favor of our hypothesis that attrition experienced by Spanish heritage speakers with respect to the production and judgement of prepositional relative clauses is due to grammar transfer from their dominant language. It is important to note that due to our small sample size, these results are just numerical comparisons of averages; we cannot assert statistical significance.

The asymmetry between our speaker groups (four (4) from the U.S. and only one (1) from France) is another significant limitation for this experiment. It was difficult to recruit Spanish heritage speakers from France given that we are based in the U.S. and considering the time constraints on our experiment. For example, one of our U.S.-based participants never produced P-stranding constructions in her elicitation task. If she had been our only participant from the U.S., our results would have shown no attrition in the production task for either speaker group. However, we are attributing this to her dominance score from the Bilingual Language Profile (see “Methods”), because she was the only participant whose score was slightly dominant in Spanish. The other three (3) Spanish-English bilingual speakers were dominant in English, and the Spanish-French bilingual speaker was dominant in French. Our hypothesis allows for these results because it attributes attrition to grammar transfer from one’s dominant language; therefore, a heritage speaker dominant in Spanish may not exhibit attrition to the same degree as one dominant in the majority language of their society.

Another potential limitation could be the fact that our French participant had been learning Spanish formally in school for most of his life, while none of the U.S.-based participants reported taking Spanish for more than a couple of years. Foreign language educator Dr. Kathleen Stein-Smith highlights the discrepancies in second-language education between the U.S. and European countries in her 2013 Ted Talk, “The U.S. foreign language deficit” (https://www.youtube.com/watch?v=8CZ7zu5Aeu0). She argues that educational institutions, businesses, and government-funded programs can help to promote foreign-language teaching at the societal level. While our French participant was a heritage speaker of Spanish (and therefore not learning Spanish as a “second” language), the fact that France has such a comparatively robust Spanish language curriculum was unavoidable and may have skewed our results. If a speaker is explicitly taught prescriptive grammatical rules—for example, not to strand prepositions—they are less likely to violate these rules than speakers with no formal education in a language they only acquired in the home.

Conclusion

This experiment allowed us to explore the source of grammatical attrition in Spanish heritage speakers in terms of producing and accepting prepositional relative clauses. We conducted an elicitation task and an acceptability judgement task for five (5) participants: four (4) Spanish heritage speakers from the U.S. and one (1) from France. We found that on average, the participants from the U.S. produced and accepted P-stranding constructions more than the participant from France. Given that English allows P-stranding and French and Spanish do not, these results provide evidence in favour of our hypothesis that grammatical attrition in heritage speakers is caused by transfer from the grammar of speakers’ dominant languages. In other words, our study suggests that the participants from the U.S. are allowing and producing P-stranding in Spanish because they are dominant in English, which allows P-stranding. These findings shed light on how and why heritage speakers experience grammatical attrition, and have pedagogical implications for reinforcing heritage speakers’ language maintenance in the classroom. Bilingual education advocate Aminah Ghanem sheds light on the importance of reforming the U.S. language education system in order to promote multilingualism in her 2019 Ted Talk, “Reforming Our Bilingual Education System” (https://www.youtube.com/watch?v=vvCUk9qJQmE). She highlights the positive effects of multilingualism in terms of celebrating diverse cultures, preventing language discrimination, and the marginalisation of heritage speakers. We hope that our study will contribute to this conversation and promote future research and educational policy decisions as they affect heritage bilinguals.

attrition, heritage speakers, multilingualism, P-stranding