HHS Public Access Author manuscript Author Manuscript Discourse Stud. Author manuscript. “Let’s talk about pain and opioids: Low pitch and creak in medical consultations” Peter Joseph Torresa, Stephen Gresham Henry, MDb, Vaidehi Ramanathan, PhDa aDepartment of Linguistics, University of California Davis, CA, USA. bDivision Author Manuscript of General Medicine, Geriatrics, and Bioethics, University of California Davis, Sacramento, CA, USA. Abstract Author Manuscript In recent years, the opioid crisis in the United States has sparked significant discussion on doctorpatient interactions concerning chronic pain treatments, but little to no attention has been given to investigating the vocal aspects of patient talk. This exploratory sociolinguistic study intends to fill this knowledge gap by employing prosodic discourse analysis to examine context-specific linguistic features used by the interlocutors of two distinct medical interactions. We found that patients employed both low pitch and creak as linguistic resources when describing chronic pain, narrating symptoms, and requesting opioids. The situational use of both features informs us about the linguistic ways in which patients frame fraught issues like chronic pain in light of the current opioid crisis. This study expands the breadth of phonetic analysis within the domain of discourse analysis, serving to illuminate discussions surrounding the illocutionary role of the lower vocal tract in expressing emotions. Keywords Doctor-patient interaction; Opioids; Chronic Pain; Creak; Pitch; Prosody; Discourse Analysis; Sociolinguistics 1. Introduction Author Manuscript Despite growing sociolinguistic interest in the study of intraspeaker voice-quality variation and the role that linguistic choices play in shaping identity and creating social meaning (Mendoza-Denton, 2011; Podesva, 2007; Wilce, 1997), there remains little research in the field of clinical sociolinguistics—the application of linguistic or phonetic techniques to the study of communication in the medical space (Ball et al., 2008). This exploratory study attempts to fill this knowledge gap by conducting prosodic discourse analysis (Chafe, 1993) to describe the different vocal features that patients employ when discussing issues concerning chronic pain and opioids with their physician. We found that patients use both low pitch and creaky voice when (1) narrating symptoms or describing pain and (2) requesting opiates. Corresponding Author: Peter Torres, 263 Kerr Hall, Davis, CA, 95616, Tel: 1-310-689-8655, peterjosephtorres@gmail.com. Torres et al. Page 2 Author Manuscript Unlike bruises, the symptoms of chronic pain are not always visible, and so patients use “talk”—among other methods such as illustration and demonstration—to describe their suffering and express their complaints and symptoms to the physician (Heath, 2002). Linguistic practices are especially important in such discussions because there are no objective clinical tests to measure chronic pain (Sullivan and Ferrell, 2005). Consequently, clinical assessments and treatment decisions regarding pain are often based solely on the patients’ verbal manifestations of pain and suffering (Burgess et al., 2008; Turk and Okifuji, 1997; Henry and Eggly, 2013). Author Manuscript The current opioid crisis heightens the relevance of investigations focusing on verbal manifestations of pain. In 2011, the United States Center for Disease Control and Prevention (CDC) declared prescription drug abuse as a national epidemic after deaths from accidental overdose exceeded fatalities from vehicular accidents. Because complaints about chronic pain can be interpreted as drug-seeking behavior (Højsted and Sjøgren, 2007), the manner in which patients communicate their chronic pain to physicians may be different from how they discuss other types of pain. Merrill et al. (2002) found fear and mistrust as dominant themes in discussions of opioids for both physicians and patients. Roberts and Kramer (2014) expressed the need for analyzing linguistic practices in medical appointments where pain medications are discussed because issues related to controlled substances are fraught with questions of ethical and moral nature, as well as the potential for abuse and/or the fabrication of symptoms to justify drug-seeking behavior. Matthias et al. (2013) echo this point, calling for more direct explorations to better understand pain management communication. To meet this need, this study offers a phonetic perspective of how patients discuss chronic pain and opioids at a time of crisis. Moreover, because acoustic or prosodic analysis has mostly been used in clinical sociolinguistics to address pathological speech issues (Cernak et al., 2017; Dudy et al., 2018), this research further broadens the applicability of this method by examining actual doctor-patient dialogue. Author Manuscript 2. Related Literature 2.1. Behavioral Manifestations of Pain Author Manuscript Pain research seeking to improve doctor-patient communication on chronic pain management is necessary; given that both patients and physicians have reported this dialogue to be challenging and frustrating (Henry et al., 2016). Turk and Okifuji (1997) revealed that neither the severity nor duration of pain solely influence physicians to prescribe opioids. Their findings showed that it is, in fact, the behavioral manifestation of pain—which includes audible expression of distress—that impacts such important healthcare decisions. Subsequent studies on pain management (Hughes et al., 2015; Matthias et al., 2013, 2014) agree that there is value in investigating the prosodic nature of displaying pain in doctor-patient dialogue. This study does just that by bridging voice quality issues to their context-specific discourse functions in medical encounters. 2.2. Voice Quality: Modal versus Creak Voice quality, as it is used here, refers to the type of phonation speakers produce in the larynx. A speaker, at will, can strategically assume different voice qualities like high or low Discourse Stud. Author manuscript. Torres et al. Page 3 Author Manuscript pitch, whisper, falsetto, breathy, and creak to portray certain attitudes and emotions (Couper-Kuhlen, 2015; Esling, 2012). Modal phonation refers to the neutral or typical form of speaking, in which speakers talk within the natural pitch range of their voice. The everyday use of the word “pitch” describes the degree of highness or lowness of a tone. Linguistically, pitch is the perception of the rate of vocal fold vibration and that targeting a remarkably low pitch may result in irregularly spaced vocal pulses known as creaky voice or “vocal fry” (Berry, 2001). Creak and, to a certain degree, low pitch1, are both products of epilaryngeal constriction—a process of the lower vocal tract that results in the shortening, bunching, and adduction of the vocal folds, leading to lower pitch frequencies and, consequently, the shifting from modal to non-modal phonations like creak2 (Esling et al., 2019). Author Manuscript 2.3. Interpreting Use of Pitch in Discourse Author Manuscript The introduction of “contextualization cues” in the 1970s—fully interpretable and indexical linguistic signs that possess an embedded context—opened the idea that linguistic styles such as shifting of pitch and use of creak can carry information (Gumperz, 1982; see also Van Dijk 1999, 2011). Later research, including that of Tannen and Wallat (1987: 208), provided evidence that identifiable linguistic cues such as register—”prosodic choices deemed appropriate for the setting and audience”—could be used in the analysis of medical discourse. A speaker’s register can also be described in terms of intonation, the variations in pitch often used in English to express the purpose of utterances as well as understand the attitudes and emotions of other speakers (Couper-Kuhlen, 2015). For example, the use of rising and falling intonation usually helps us understand the difference between statements and questions, among many other ways of regulating discourse. Earlier studies on pitch and intonation elicited the help of actors to portray emotions because of the idea that neutral speech must also be gathered as a baseline for comparison. Using this method, Zetterholm (1998), as well as Bänziger and Scherer (2005), found that both pitch lowering and creak portray sadness. The design of this study expands on Cruttenden’s (1997) work on intonation, which highlights the idea that a speaker’s register is “marked” when the entire pitch configuration of an utterance is transposed towards the higher or lower limits of their vocal range. In this study, we illustrate “markedness” by using the speaker’s average pitch as a baseline to assess instances in which their pitch deviates significantly from the typical. This method allows for the examination of linguistic style and the information that actual spontaneous speech carries without the need of actors or “neutral speech stimuli.” Author Manuscript 1.While epilaryngeal constriction passively lowers pitch, it does not necessarily constitute active pitch lowering. As Moisik (2013) puts it, the constriction synergizes with low pitch. 2.It is important to note that epilaryngeal constriction could result in different types of non-modal phonation, including varying degrees of creak like “harsh” and “pure” (Moisik, 2013). In the present study, we collectively use the term “creak” to refer to all of its variants. Discourse Stud. Author manuscript. Torres et al. 2.4. Page 4 Interpreting Use of Creak in Discourse Author Manuscript Creak is one particular stylistic feature used in creating social meaning that has sparked interest among sociolinguists and discourse analysts. Scholars have studied the relationship between the use of creaky voice and social class (Esling, 1978), gender and sexuality (Henton, 1986; Zimman, 2012, 2013), and stance (Grivicic and Nilep, 2004; Slobe 2018). According to Anderson et al. (2014), the use of creaky voice is becoming increasingly common among young American women and that listeners take the use of creak as untrustworthy and less educated. There are, however, several studies showing that young women’s use of creak is perceived as more dominant and authoritative yet non-aggressive (Borkowska and Pawlowski, 2011; Yuasa, 2010). Author Manuscript While most studies on vocal fry have focused on linguistic attitudes towards young women, there are a few that have examined the contextual application of creak in discourse. More importantly, these studies have looked at the use of voice quality from an intraspeaker perspective —i.e., how an individual employs specific phonation styles in particular discourse contexts—as opposed to the essentializing nature of interspeaker research. Podesva (2007) is among the pioneers in exploring the situational use of creak by taking context, topic, and audience into account. Podesva has demonstrated how a speaker from Long Island named Heath manipulates pitch and strategically employs creak to construct a “diva persona.” Similarly, Mendoza-Denton (2011) analyzed the speech of “Babygirl” and suggested that creak is a discourse-dependent variable employed in the construction of a Chicana hard-of-heart (hardcore) identity. The present study addresses the lack of similar research in realm of medical discourse. Author Manuscript Although creak has been generally understudied in the domain of doctor-patient interactions, there is a key study that provides valuable insight for the present analysis. Notably, Wilce (1997) wrote about the use of creak by Bangladeshi patients to signal weakness, low energy, and misery when interacting with biomedical doctors, herbalists, exorcists, and diviners. Wilce points out that the “markedness” or saliency of creak brings attention to the utterance and lends credibility to the speakers’ reference to their own pain, therefore making vocal fry a learned and internalized social sign that carries particular discourse functions. Similarly, this study pays attention to “marked” speaking turns, in which the speaker’s use of creak and low pitch are perceptually prominent. 3. Methodology Author Manuscript Using prosodic discourse analysis, this study narrows down the different vocal features employed by two patients as they discuss opioids and chronic pain issues with their physicians. 3.1 Participants Both patients and their respective physicians were recruited from a hospital-based primary care resident clinic in California. Both patients are taking opioid for chronic pain. To account for individual sociolinguistic backgrounds, it is important to note that the participants self-identify as females and are native speakers of American English. Interlocutor information are presented in Table1. Discourse Stud. Author manuscript. Torres et al. Page 5 Author Manuscript The primary purpose of patient A’s visit was to request a refill for her opioid prescription after being denied by a different physician due to recent revelations that another pain clinic was already prescribing her with opioids, among other reasons. The physician’s attempt to steer patient A towards alternative means of pain management was met with resistance. On the other hand, Patient B’s reason for scheduling a consultation was to switch to a different type of opioid. Author Manuscript The choice to look at a small number of participants is appropriate in investigations involving language-in-action as it allows for a thorough analysis of the discourse contexts that surround the style shifts in question (Schilling-Estes, 1998). Echoing Podesva’s (2007: 498) perspective on voice quality being a vehicle for social meaning, examining the contextspecific use of linguistic variables lends insight into speakers’ intentions and the functionality of their utterances. In this case, analyzing patterns in which patients employ creak and low pitch informs us about their understanding of the discourse functions both variables serve in the medical space. 3.2. Dividing Discourse into Speaking Turns To adequately and consistently quantify pitch, we separated the data in terms of individual speaking turns, which refers to the entire speech of a specific speaker before another interlocutor mediates and converses. In the following sequence, because of patient A’s quick pause, the physician was able to inject a quick backchannel response. As much as patient A’s second turn is a continuation of the first, we consider them as two separate turns. Author Manuscript 1. P-A: But he wanted me to quit immediately, just like that. 2. DOC: Yeah. 3. P-A: And I told him, “I can’t do that. You can’t just quit -- methadone like that.” It is important to note that “overtalks”—situations in which more than one speaker is talking at once—were excluded from the analysis since they do not provide for an accurate pitch analysis. Brief backchannel responses such as “uh-huh,” “okay,” and “right,” were not considered as well. 3.3. Measuring Pitch and Assessing Creak Author Manuscript Although pitch is the term regularly used in describing the listener’s judgment of what they hear, it is also the perception of the rate of vocal cord vibration in Hertz (cycles per second). In other words, pitch is the subjective attribute of the voice’s fundamental frequency (f0) estimate (Bendor and Wang, 2005). Though certain distinctions exist between the use of both pitch and fundamental frequency (f0), their relationship is established enough to allow us to talk about the speaker’s pitch through f0 measurements (see Gerhard, 2003). This approach enables us to present auditory judgments in a quantitative manner. Pitch or f0 values were obtained using Praat® software (Boersma and Weenik, 2013). Recordings were extracted using Audacity® software (Version 2.1.2, freeware, ©1999–2014. Audacity Team. http://audacity.sourceforge.net/). Background noise was removed through spectral noise gating—a process that works well when the signal in the recording is much louder than the noise3. Discourse Stud. Author manuscript. Torres et al. Page 6 Author Manuscript Author Manuscript Unlike modal utterances, we cannot simply generate accurate pitch values for creak since its f0 is extremely low and the spacing between the pulses are too irregular (Keating et al., 2015). Thus, creaky voice was distinguished from modal voice through perceptual identification followed by examining waveforms and spectrograms to guarantee accuracy. The acoustic analysis in determining creakiness employs the same criteria used by Henton (1986), Gordon and Ladefoged (2001), Podesva (2007), and Mendoza-Denton (2011). Creak is acoustically exemplified by one or more of the following: (1) the irregular spacing of the glottal pulses in wideband spectrographic displays, (2) inconsistencies in f0 values due to slowing of vocal fold vibrations, (3) the abrupt decline in f0, (4) irregularity in the period of each cycle (pitch perturbation or jitter), (5) irregularity in the amplitude of each cycle (amplitude perturbation or shimmer), (6) decreased acoustic intensity relative to modal phonation, and (7) fewer pitch periods per second relative to modal counterpart. For speaking turns that have both modal and creaky segments, only the modal segments were tracked for pitch. This information is summarized in Table 2. Figure 1 summarizes the acoustic analysis section of our methodology. The decision to represent pitch through f0 frequency is defensible because this technique can easily be replicated in the medical setting. For instance, if pitch were to be added to patients’ medical notes in order to take records of how they speak in certain contexts, reporting f0 acoustic data would be viable rather than manually noting impressionistic data. 3.4. Coding of Speaking Turns: Pain-Related vs Non-Pain-Related Author Manuscript To point out the linguistic features both patients use to discuss opioids in light of the current crisis, each speaking turn was coded into different contextual categories using a modified version Chronic Pain Coding System (CPCS) developed by Henry et al. (2016). CPCS focuses on the objective characterization of utterances involving pain and opioids, making it appropriate for the current study. There are three main contextual categories included in the two doctor-patient interactions observed: (a) discussions about chronic pain and the opioid medication used to manage it; (b) discussions related to the other types of pain not involving opioids; and (c) discussions that are not about pain. Tables 3, 4, and 5 present the various subcategories under each of the three primary categories mentioned, together with sample excerpts from both patients. Three coders coded for categories, including a physician and two sociolinguists trained in discourse analysis of medical interactions. The odd number of coders allows for a majority rule in cases of disagreements4. The Fleiss (1971) interrater reliability for these categories by three individual coders is 0.915. Author Manuscript 3.Using a Fourier analysis of the first few seconds of the recording, Audacity creates a noise profile used in filtering out the rest of the recording. Both audio files were filtered using the following setting: a noise reduction of 12 decibels with a sensitivity parameter of 3 and frequency smoothing bands set at 0. Audacity can also generate the “noise residue,” or the noise to be filtered out, which was useful in verifying that the audio was not compromised. 4.In the rare case where all three coders selected distinct codes, each made a case for their decision until an agreement was reached. Discourse Stud. Author manuscript. Torres et al. Author Manuscript 4. 4.1. Page 7 Results and Discussions Acoustic Analysis Table 6 shows Patients A and B’s individual average pitch as well as the mean pitch for the three major categories. Both patients spoke with lower pitch in utterances that focused on pain, more so if the painrelated discussion involved opioids (including requests and positive assessments of opioids, as well as description of pain in relation to the prescribed opioid). The following diagrams provide a visual representation of each patients’ vocal range and the frequency at which each speaking turn’s average pitch occurs. Author Manuscript As shown in Figure 2, the speakers’ pitch in turns that were about pain and opiates were lower. Patient A’s pain-related utterances that were not in the context of opioids have speaking turns that are distributed towards the upper registers of her speaking voice. Towards the second half of her visit, patient A repeatedly raised her voice and overenunciated some words for emphasis. Although pitch is not synonymous to volume, patient A’s pitch went up as she intensely argued. Nonetheless, it is in the context of pain and opioids in which she speaks in the lower range of her register. On the other hand, patient B’s use of low pitch is clearer. Table 7 provides a stratified list of results, specifying the average pitch for each subcategories. After narrowing down the categories, we found that both patients had the lowest pitch when narrating or describing chronic pain or asking for the opioid medication they use to remedy that pain. Figure 3 demonstrates mean pitch distributions for both patients in the following discourse contexts: (i) request for opioid treatment and (ii) description/narration of pain. Author Manuscript It is important to note that the lowering of both patients’ pitch is not restricted to only the concluding segments of a speaking turn. Low pitch also occurs in other parts of the turn and are mostly sustained once initiated. Moreover, there is no correlation between the length of a turn and the potential for pitch lowering. Examples of instances in which both patients lower their pitch when discussing chronic pain and opioids are shown in Figures 4a and 4b below. In what follows, we present our findings on the use of non-modal phonation, creak. 4.2. Entirely Creaky Speaking Turns Author Manuscript Although pitch cannot be measured for creak, we still coded the entirely creaky speaking turns to study the context in which the feature is used by both patients (See Table 8). Of all of the creaky turns, 65 percent of Patient A’s and 85 percent of Patient B’s were about pain. Figures 5 and 6 provide examples of waveforms and spectrograms of words articulated in both creaky and modal phonation. At the bottom of each figure are waveforms zooming into the beginning of the vowel to show the differences in periodicity. Discourse Stud. Author manuscript. Torres et al. Page 8 Author Manuscript The creak and modal examples above adhere to the characteristics outlined in the methodology section. Both creaky examples have aperiodic glottalization, lower acoustic intensity, and significantly lower amplitudes in comparison to their modal counterparts. The spacing between pulses is more regular with the modal examples. Figure 6a also has alternating longer and shorter pulses that is often characteristic of creaking (Keating et al., 2015). Lastly, the pitch values are irregular in Figure 5a and completely undetectable in Figure 6a. 5. Discussion Author Manuscript Our study demonstrates that pitch or f0 values were lower and the use of creak was more apparent when both patients discussed chronic pain and opiates. Because both phonation styles are associated with the lower vocal tract and the use of lower register, it is not unexpected that creaky segments are detected in utterances with low modal pitch, as shown in the examples below. In excerpt (1), patient A informs the doctor about not being prescribed with opioids by another physician. The primary reasons for the denial include patient A’s toxicology results and the fact that she is already taking another opioid from another pain clinic. (The numbers in parenthesis represent the mean pitch of each turn’s modal segment while utterances in bold denote creak). (1) Author Manuscript Author Manuscript 1. Physician: Alright. So -- um -- alright, let’s start with what you would like to talk about. 2. Patient A (139.9 Hz): Okay, and I was very upset with my last visit with Dr. . 3. Physician: M-mm. 4. Patient A (136.1 Hz): He wouldn’t prescribe any an- -- any meds, any pain medication that day, so I’ve been without Norco for almost two months now. 5. Physician: Hm-mm. 6. Patient A (181.4 Hz): I’ve been in extreme pain, with my shoulder that is still hurting. I can’t do physical therapy, because my -- I might even need to have surgery. I wanna see the s-, the surgeon again, because it’s just not healing. 7. Physician: Hm-mm. 8. Patient A (169.5 Hz): It’s not healing at all. And, I mean, he’s with me day and night practically when he’s not at work, and for the past four years we’ve been together, and he can verify -- um -- the pain that I’m going through -- with all this. Lines 6 and 8 are examples of turns in which patient A describes her pain using low pitch and creak. In both turns, the patient’s pitch is seven and thirteen percent lower, respectively, compared to her overall average of 195.298 Hz. In line 4, patient A proves that not all Discourse Stud. Author manuscript. Torres et al. Page 9 Author Manuscript requests have to take the interrogative form. According to Robinson (2001), implicit or indirect requests can take any grammatical form as long as the utterance performs its soliciting function in the context of the medical visit. Line 4 is coded as an opiate request because the purpose of the turn was not only to inform the physician that no opiate has been prescribed but also to suggest that the situation must be addressed. In this turn, the patient’s pitch is 30 percent lower than the baseline. The pairing of low pitch and creak in opiate requests is also discernible in patient B’s speech as shown in excerpt (2) below. In the beginning of this exchange, patient B asks to be prescribed with Cymbalta, Neurontin, and Lyrica, none of which are opiates. She proceeds to talk about the main reason for her visit, which is to get the opiate, Dilaudid, added into her pain contract. Author Manuscript (2) (Utterances in bold denote creaky voice) Author Manuscript 1. Physician: How can I help you today? 2. Patient B (166.0 Hz): Uh, there’s a few things. Uh, the Cymbalta that—I don’t remember his name, but the last doctor I saw—put me on the Cymbalta. 3. Physician: Yeah. 4. Patient B (152.6 Hz): Uh, I don’t have any more and there was [sic] no refills, so -- I’ve still got that. 5. Physician: I can certainly refill that for you if that seems to be helping. 6. Patient B (160.7 Hz): Okay. Yeah, I think it does help a little bit with the anxiety and stuff. I don’t know if it’s helping with the pain, um— 7. Physician: The, the effect on the pain might be sort of subtle... it may, it may be helping to reduce the amount of other pain medicines you require. 8. Patient B (172.2 Hz): Unless—my—cuz [sic] I did pretty good, I think—with the Dilaudid they did give me. And also, uh, I took the Neurontin three times a day. I really have a hard time remembering to do that three times a day all the time. 9. Physician: Uh huh. 10. Patient B (184.5 Hz): And I, I have taken Lyrica before. It was just a two-week trial... Author Manuscript In the example above, patient B lists four medications she wants to be filled, yet it’s worth noting that creak was only employed in the discussion of pain and Dilaudid in lines 6 and 8. In line 6, patient B introduces the idea that Cymbalta does not address her chronic pain, which consequently leads to the reveal that it is the opiate Dilaudid that helps in line 8. The comments on lines 6 and 8 are coded as requests—implicitly delivered through negative evaluation of a non-opiate drug in line 6 and a positive assessment of the opiate in line 8. Starting in line 10 up until opiates are discussed again, patient B goes back to speaking with her regular modal voice, making the use of low pitch on discussions about controlled substances and chronic pain more apparent. Discourse Stud. Author manuscript. Torres et al. Page 10 Author Manuscript Irvine (2001) suggests that we could interpret the motivations behind the use of certain linguistic styles by examining situations in which it is absent. We found that requests for non-opiate medication have no semblance of low pitch and creak, as shown in the example below: 1. Patient B (193.4 Hz): Oh, the Robaxin. Um, could you guys change that from 100 to 120 tablets for a month? Author Manuscript In this excerpt, the request for the muscle relaxant, Robaxin, takes the conventional interrogative form. Such straightforward request contrasts the suggestive opiate requests presented earlier. In fact, both patients started the discussion on opioids by “reporting” its effectiveness while simultaneously highlighting the presence and severity of their chronic pain. According to Robinson (2001) and Gill et al. (2001), reporting—in the form of assessments—is used by patients to implicitly make sensitive requests like asking for addictive medications, without revealing their position towards the request5. Excerpt (3) below illustrates an exchange in which patient B neither uses creak nor lowers her pitch to discuss a condition that is not addressed by an opioid. (3) Author Manuscript 1. Physician: Anything else that uh I can help you with today? 2. Patient B (193.0 Hz): Uh, my asthma. I live in Sunnyville, and it was pretty much fogged in smoke. 3. Physician: Okay. 4. Patient B (190.5 Hz): And it started bothering my asthma right away. 5. Physician: Oh. 6. Patient B (194.7 Hz): My sat was 95, but—I’ve been using my inhaler a lot. Our data show that both patients find value in switching to a distinct register when discussing chronic pain and opiates. Going back to Cruttenden’s (1997) notion about register, the salient shifting in pitch indicates emphasis to what is being said. Considering that both patients overwhelmingly use low pitch and creak in very specific contexts tells us that both linguistic variables are being employed for stylistic work. Specifically, both phonation styles serve as pragmatic resources used to express pain as well as to request the medication that they believe best manages their misery. 5.1. The Nature of Opioid-related Utterances Author Manuscript These findings raise a significant question: what is distinct about the topics of chronic pain and opiates that motivate the change in vocal style? The opioid crisis has brought stigma to the discussion of controlled substances, which has made requesting opioids a fraught process within the medical setting. We already witness such effect in this study, through both patients’ suggestive framing of requests. Roberts and Kramer (2014) found that patients 5.Our data also shows that both patients resort to more straightforward demands only when implicit requests are left unaddressed. Discourse Stud. Author manuscript. Torres et al. Page 11 Author Manuscript orient issues surrounding pain medications as problematic, morally suspect, and easily refusable. Patients have to confront the effects of the opioid crisis by increasing their sensitivity towards the potential concerns that doctors may have about dependency and addiction when prescribing opioids. Frequently, patients find it necessary to defend their moral character and present themselves as credible, responsible, and aware when the topic and requests are sensitive, challenging, have high chances of denial, potentially controversial, and morally fraught. It is evident from our findings that such tasks are accomplished by both patients through the situational use of low pitch and creak. Author Manuscript What sets opioid-related conversations further apart is the fact that it could be a source of disagreement because patients and physicians often do not share the same priorities when it comes to managing chronic pain (Henry et al., 2017). In fact, a post-visit survey given to the participants in this study reveal that both patients ranked “reducing pain intensity” as their most important goal while their respective physicians placed higher emphasis on improving the patient’s overall function. Alerting physicians about the use of specific vocal features in disagreeable discussions like opioids, could alert them to confront the disagreement by returning the conversation to goal setting. Clearly, our findings show that the dynamics involved in medical encounters involving chronic pain and opiates are indeed different from other primary care visits. If the entirety of patient A’s appointment were about her constipation while patient B’s were about her asthma, the linguistic practices they would employ throughout their consultation would less likely involve creak and pitch lowering, as suggested by the way they discussed these same concerns in this study. 5.2. Discourse Functions: Expressiveness of Low Pitch and Creak Author Manuscript Author Manuscript Patients cannot always show visible evidence of chronic pain, therefore their only recourse is to express their symptoms using the primary activity that takes place in medical interactions: talking. This raises the question as to why creak and low pitch are used in conjunction with requests for opiates. What do these variables index? We already know that low pitch/f0 and creak collaboratively portrays sadness, misery, weakness, certainty, and credibility (Bänziger and Scherer, 2005; Borkowska and Pawlowski, 2011; Wilce, 1997; Yuasa, 2010). According to Podesva (2007), the common denominator that describes the polysemic functions carried by a linguistic variable is its “expressiveness”—indexed within the particular discourse contexts in which the variable is repeatedly employed. As such, our analysis of the conversational contexts suggests that the discourse functions low pitch and creak serve include: (a) addressing medical issues that are delicate, important, morally fraught, bound to receive an assessment, and could possibly be refuted; (b) self-reporting chronic pain symptoms, which could be difficult to prove and easily questioned due to lack of methods to properly evaluate it; and (c) requesting addictive painkillers that could easily be refused or interpreted as drug-seeking behavior. 6. Conclusion This exploratory study examined important linguistic features present in two doctor-patient interactions on chronic pain management. Through discourse analysis, we found that the Discourse Stud. Author manuscript. Torres et al. Page 12 Author Manuscript patients lowered their pitch and employed creak when discussing chronic pain and opioids with their physicians. Both features are associated outcomes of epilaryngeal constriction, which in itself carries paralinguistic functions (Moisik, 2013). This study expands the breadth of phonetic analysis within the domain of discourse analysis, leading to the need to explore topics involving the illocutionary role of the lower vocal tract in expressing emotions. Author Manuscript Studying micro-linguistic practices in the medical space may improve the overall efficacy of health communication, which is vital for setting shared pain management goals and reducing inappropriate opioid prescribing (Henry et al., 2017). The situational use of low pitch and creak tells us about how patients frame conversations on opioids in light of the current medical climate. Patients, regardless of intention, are tasked with navigating the challenging and fraught discussion of opioids, knowing well that their requests could be refused or their symptoms questioned. Awareness of the discourse functions of low pitch and creak during discourse may alert physicians about the possible concerns of patients. Recognizing such discernible shift in register could also signal the physician into bringing the discussion back towards their shared goals. Additional investigations would allow us to discover whether men use the same vocal features when interacting with doctors. Lastly, examining the correlation between auditory judgments and prescribing decisions through a perception test could also further this study. Acknowledgements: We would like to express our sincere gratitude to Teun Van Dijk and anonymous reviewers for their expert feedback. Their incisive and thoughtful comments helped reconceptualize several parts of this exploration. Of course, all remaining faults are our own. Author Manuscript Funding: The videos analyzed in this study were collected with funding from the National Institutes of Health (grants KL2TR000134 and UL1TR000002) and the University of California, Davis Department of Internal Medicine. Appendix A:: Coding Scheme R1A Request for Non-Opioid treatment R1B Request for Opioid treatment Request for opioid increase. Request for routine opioid refill. Patient request for opioid switch. Author Manuscript R1C Unclear request regarding Prescription R2 Request for information and logistics Request related to logistics. Patient requests for information. Patient request for non-opioid pain treatments. T2D Opioid related threats/red flags Patient statements about serious opioid side effects and red flags. Statements about less serious opioid side effects. Opioid-related threats. E2 Narration of pain and description of symptoms T1A Positive assessments /satisfaction with non-opioid treatments. T1B Negative assessments / dissatisfaction with non-opioid treatments. Discourse Stud. Author manuscript. Torres et al. Page 13 Author Manuscript T1C Expressions of uncertainty or ambiguity about a non-opioid treatment plan. T2A Positive assessments / satisfaction with opioids. T2B Negative assessments /dissatisfaction with opioids. T2C Expressions of uncertainty or ambiguity about opioids. N1 Disagreement or resistance. N3 Agreement X Non Pain Related Utterances H Other Pain Related Utterances Appendix B:: Pitch Tracking Data for Patients A and B (Summaries at the end) Author Manuscript Patient A Speaking Turn Mean Min Max Range 9 H 164.783 120.73 208.855 88.125 ✔ 12 H 163.907 130.007 204.602 74.595 ✔ H 184.007 128.17 246.504 118.334 ✔ H 184.689 135.189 228.14 92.951 ✔ H 186.901 138.096 214.035 75.939 ✔ H 183.109 125.165 300.135 174.97 ✔ 16 H 179.63 145.518 218.222 72.704 ✔ 22 R1B 136.113 119.113 283.215 164.102 ✔ ✔ 24 E2 177.309 125.172 246.814 121.642 ✔ ✔ 24 E2 185.491 135.693 233.374 97.681 ✔ ✔ 26 E2 169.521 136.813 204.808 67.995 ✔ ✔ 28 H 174.096 129.277 219.569 90.292 ✔ 30 H 173.692 121.563 264.423 142.86 ✔ 32 H 193.155 157.114 262.798 105.684 ✔ 34 H 183.187 158.745 214.013 55.268 ✔ 45 H 197.092 115.095 324.744 209.649 ✔ 47 H 200.927 149.362 335.386 186.024 ✔ 49 H 191.869 115.343 298.943 183.6 ✔ 53 H 192.296 150.457 227.269 76.812 ✔ 67 H 206.175 134.382 303.281 168.899 ✔ 69 H 186.492 147.243 255.286 108.043 ✔ 71 H 195.17 144.291 274.186 129.895 ✔ 73 H 203.499 169.799 249.01 79.211 ✔ 14 Context Opioid Context Pain Context Author Manuscript Author Manuscript 75 H 190.273 144.114 253.458 109.344 ✔ 102 H 190.403 144.018 251.154 107.136 ✔ 104 H 173.239 125.369 246.372 121.003 ✔ Discourse Stud. Author manuscript. Torres et al. Page 14 Author Manuscript Speaking Turn Context Mean Min Max Range Opioid Context Pain Context ✔ Author Manuscript Author Manuscript 106 H 214.232 142.61 266.435 123.825 108 R1B 182.516 139.534 243.977 104.443 ✔ ✔ 110 R1B 157.005 137.904 198.482 60.578 ✔ ✔ 118 R1B 184.221 159.142 280.263 121.121 ✔ ✔ 120 R1B 185.716 155.096 251.27 96.174 ✔ ✔ 122 R1B 187.532 155.309 306.435 151.126 ✔ ✔ 124 H 205.96 134.761 301.223 166.462 ✔ 126 H 173.122 152.454 273.56 121.106 ✔ 128 H 186.943 154.141 239.359 85.218 ✔ 130 H 210.177 142.305 265.149 122.844 ✔ 132 E2 168.11 111.353 246.173 134.82 136 H 210.058 163.922 248.125 84.203 138 E2 149.553 130.15 165.406 35.256 140 H 193.969 143.913 256.311 112.398 ✔ 142 H 189.595 130.251 212.011 81.76 ✔ 144 H 192.092 167.696 228.539 60.843 ✔ 146 H 171.283 121.216 222.346 101.13 ✔ 148 H 192.422 146.019 259.952 113.933 ✔ 158 H 183.827 158.028 216.97 58.942 ✔ 169 H 196.024 157.312 233.51 76.198 ✔ 171 H ✔ 173 H ✔ 186 X 188 X 193 X 208 X 216 ✔ ✔ ✔ ✔ ✔ Author Manuscript 206.509 125.535 221.292 95.757 X 209.451 142.896 256.617 113.721 218 H 189.623 160.729 212.696 51.967 ✔ 221 X 231 X 239 H 203.089 122.546 317.979 195.433 ✔ 247a R2 205.438 140.842 259.43 118.588 ✔ 247b R2 254 H 193.994 167.462 214.392 46.93 ✔ 260 H 203.373 155.662 313.438 157.776 ✔ 262 H 197.593 168.619 232.01 63.391 ✔ 271 H 186.351 115.262 252.639 137.377 ✔ 273 X 201.51 143.765 244.479 100.714 275 X 203.438 136.561 216.375 79.814 ✔ Discourse Stud. Author manuscript. Torres et al. Page 15 Author Manuscript Speaking Turn Context Mean Min Max Range Opioid Context Pain Context Author Manuscript 279 X 303 H ✔ 305 H ✔ 307 H ✔ 311 H ✔ 313 H ✔ 315 H ✔ 317 H ✔ 319 H 325 H 168.923 115.235 220.667 105.432 ✔ 341 H 179.78 129.131 219.357 90.226 ✔ 357 R1C 199.216 154.459 270.33 115.871 ✔ ✔ ✔ ✔ Author Manuscript Author Manuscript 359 T2D 190.259 145.832 238.316 92.484 361 T1B 190.415 157.972 242.936 84.964 ✔ 363 H 192.312 157.468 273.975 116.507 ✔ 365 H 195.694 147.233 274.21 126.977 ✔ 369 H 204.48 126.311 285.751 159.44 ✔ 371 H 215.318 161.889 270.477 108.588 ✔ 373 H 203.487 168.835 253.628 84.793 ✔ 375 H 219.21 141.131 279.262 138.131 ✔ 377 T2D 200.914 143.658 250.758 107.1 379 H 219.21 131.131 289.22 158.089 381 R1B 180.93 146.658 228.98 82.322 ✔ ✔ 383 E2 167.792 127.175 232.103 104.928 ✔ ✔ 387 E2 169.769 132.357 207.39 75.033 ✔ ✔ 389 H 197.88 145.9 289.67 143.77 399 T2A 187.040 142.160 235.904 93.744 401 H 181.878 155.250 252.788 97.538 ✔ 407 H 200.035 124.344 253.802 129.458 ✔ 411 H 199.921 161.501 267.553 106.052 ✔ 425 H 200.073 130.863 284.587 153.724 ✔ 427 H 199.385 156.325 290.202 133.877 ✔ 443 X 200.899 140.533 238.628 98.095 447 X 202.271 143.439 247.937 104.498 451 X 198.665 137.752 287.606 149.854 453 X 199.63 168.179 291.13 122.951 491 H 180.386 159.74 263.798 104.058 ✔ 493 H 169.453 134.69 231.948 97.258 ✔ 495 N1 201.775 155.35 279.71 124.36 ✔ Discourse Stud. Author manuscript. ✔ ✔ ✔ ✔ ✔ ✔ Torres et al. Page 16 Author Manuscript Speaking Turn Context Mean Min Max Range Opioid Context Pain Context 499 H 192.941 145.153 295.153 150 ✔ 503 N1 196.534 149.131 253.357 104.226 ✔ 505 T1B 181.846 155.153 257.07 101.917 ✔ 507 H 180.464 165.953 253.843 87.89 ✔ ✔ Author Manuscript Author Manuscript Author Manuscript 509 H 202.224 156.481 274.004 117.523 511 T2D 187.441 143.887 251.711 107.824 ✔ ✔ 513 T2D 192.885 136.668 255.098 118.43 ✔ ✔ 517 H 203.678 134.614 287.051 152.437 ✔ 521 H 197.946 163.413 286.381 122.968 ✔ 523 H 228.889 154.041 373.66 219.619 ✔ 525 H 207.04 162.16 305.904 143.744 ✔ 527 T2D 199.972 155.253 292.426 137.173 529 T1B 208.439 155.087 296.758 141.671 ✔ 531 N1 203.858 156.22 278.364 122.144 ✔ 533 E2 180.552 128.571 218.881 90.31 535 H 210.425 151.954 302.841 150.887 ✔ 537 H 219.52 138.365 276.878 138.513 ✔ 539 R1C 212.914 156.313 302.98 146.667 ✔ 541 R1C 196.888 134.199 302.186 167.987 ✔ 543 N1 195.955 134.43 288.258 153.828 ✔ 545 H 211.177 174.251 294.834 120.583 547 T2D 204.082 174.684 285.723 111.039 549 H 198.736 173.114 290.267 117.153 551 R1B 188.943 130.45 265.995 135.545 553 N1 201.676 166.885 292.605 125.72 ✔ 555 H 213.585 140.02 294.485 154.465 ✔ 557 H 184.469 132.948 225.704 92.756 ✔ 559 R2 187.834 152.264 221.348 69.084 ✔ 562 R1C 225.412 177.142 285.305 108.163 ✔ 564 N1 203.757 150.412 292.27 141.858 ✔ 566 H 208.579 138.25 288.655 150.405 ✔ 568 T1B 203.338 135.995 300.237 164.242 ✔ 570 R2 213.39 154.772 284.089 129.317 ✔ 572 H 235.596 156.295 310.296 154.001 ✔ 574 H 214.883 169.737 313.636 143.899 ✔ 576 H 221.902 152.495 279.186 126.691 ✔ 578 H 220.564 169.27 292.588 123.318 596 E2 179.961 119.002 215.887 96.885 ✔ ✔ 598A E2 170.218 138.593 204.603 66.01 ✔ ✔ Discourse Stud. Author manuscript. ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ Torres et al. Page 17 Author Manuscript Speaking Turn 598B Context T2D Mean Min Max Range Opioid Context ✔ Pain Context ✔ Author Manuscript Author Manuscript Author Manuscript 244.784 158.553 397.457 238.904 600 H 202.84 171.158 298.37 127.212 ✔ 602 H 214.079 153.633 290.977 137.344 ✔ 604 T2A 213.443 148.843 282.315 133.472 ✔ ✔ 606 T2D 200.694 181.216 282.63 101.414 ✔ ✔ 608 H 190.604 155.014 287.312 132.298 ✔ 610 H 191.14 163.84 261.937 98.097 ✔ 614 N1 180.394 152.649 267.89 115.241 ✔ 618 N1 198.061 177.285 302.704 125.419 ✔ 620 N1 214.741 174.543 292.981 118.438 ✔ 622 H 199.76 171.271 302.349 131.078 ✔ 626 T1B 206.748 162.679 290.65 127.971 ✔ 628 E2 187.54 142.816 260.246 117.43 630 N1 199.482 158.285 289.327 131.042 ✔ 632 H 206.054 178.051 308.142 130.091 ✔ 634 H 204.211 176.155 311.083 134.928 ✔ 636 H 209.86 158.6 311.024 152.424 ✔ 638 H 207.745 166.702 314.258 147.556 ✔ 642 H 192.382 154.829 292.289 137.46 ✔ 644 H 183.437 173.388 281.969 108.581 ✔ 646 R2 199.755 163.888 267.17 103.282 ✔ 654 H 212.745 166.702 314.258 147.556 ✔ 658 X 202.745 146.702 242.258 95.556 660 R2 202.366 145.511 304.686 159.175 ✔ 664 H 182.366 145.511 244.686 99.175 ✔ 674 N1 219.293 157.325 282.693 125.368 ✔ 676 H 205.766 174.508 310.662 136.154 ✔ 678 H 203.675 143.088 316.62 173.532 ✔ 680 T2D 218.837 151.444 306.27 154.826 686 H 207.516 137.65 271.281 133.631 ✔ 688 H 219.621 197.805 279.329 81.524 ✔ 690 H 189.258 168.801 297.935 129.134 ✔ 692 H 180.787 160.833 198.839 38.006 ✔ 696 H 190.905 180.207 273.403 93.196 ✔ 698 H 198.72 156.461 290.714 134.253 ✔ 700 H 190.164 169.417 282.201 112.784 ✔ 702 R2 187.912 149.705 226.951 77.246 ✔ 704 R2 184.456 133.748 229.723 95.975 ✔ 706 R2 183.906 143.671 205.45 61.779 ✔ Discourse Stud. Author manuscript. ✔ ✔ ✔ ✔ Torres et al. Page 18 Author Manuscript Speaking Turn 708 Context T2D Mean Min Max Range Opioid Context Pain Context 183.333 170.926 283.102 112.176 ✔ ✔ Shaded rows denote entirely creaky speaking turn.* Patient B Speaking Turn Context Mean Min Max Range Opioid Context Pain Context Author Manuscript 8 X 166.005 140.428 235.628 95.2 10 H 152.624 125.386 173.289 47.903 ✔ 12 R1A 138.099 121.188 159.123 37.935 ✔ 16 H 199.853 100.725 305.317 204.592 ✔ 18 T1C 160.675 134.325 186.21 51.885 ✔ 26 T2A 177.498 157.108 198.591 41.483 ✔ ✔ 28 T2A 172.222 141.605 300.588 158.983 ✔ ✔ Author Manuscript Author Manuscript 30 H 179.623 138.108 241.546 103.438 ✔ 32.1 H 184.535 100.304 287.728 187.424 ✔ 32.2 H 34 H 145.018 114.272 162.64 48.368 36 R2 176.542 135.526 197.539 62.013 38 R1B 162.562 122.836 194.049 71.213 40 H 42 H 44 H 54 H 56 H 58 H 177.983 122.46 195.193 72.733 ✔ 60.1 H 161.995 101.918 261.306 159.388 ✔ 60.2 H 129.865 106.853 152.301 45.448 ✔ 64 T1C 178.737 119.766 285.424 165.658 ✔ 66 H 139.418 123.954 154.07 30.116 ✔ 76 H 156.025 111.376 187.743 76.367 ✔ 78 H 154.524 115.797 368.434 252.637 ✔ 80 H 147.852 134.949 177.544 42.595 ✔ 82.1 T1A 179.491 104.647 288.368 183.721 ✔ 82.2 T1A 151.381 109.638 261.261 151.623 ✔ 88 T1B 155.842 121.583 181.912 60.329 ✔ 90 X 167.448 103.391 170.093 66.702 94 X 176.489 108.333 183.876 75.543 98 X 168.945 132.691 185.592 52.901 102 X 214.178 147.974 308.762 160.788 ✔ ✔ ✔ ✔ ✔ ✔ 158.633 119.406 230.721 111.315 141.135 118.557 156.776 38.219 ✔ ✔ ✔ ✔ Discourse Stud. Author manuscript. Torres et al. Page 19 Author Manuscript Speaking Turn Context Mean Min Max Range Opioid Context Pain Context Author Manuscript Author Manuscript 104 X 194.182 101.334 326.326 224.992 106 X 153.429 100.17 199.934 99.764 108 X 162.764 114.468 163.471 49.003 110 X 160.113 114.763 170.869 56.106 122 N3 ✔ 126 T1A ✔ 128 R1B 130 R1B 132 T2A 168.296 127.908 208.535 80.627 142 H 214.225 117.833 302.108 184.275 144 R1B 165.902 106.659 298.712 192.053 148 H 166.445 111.287 243.513 132.226 150 R1B 156 T2A 185.18 143.841 278.464 134.623 158 H ✔ 160 H ✔ 162 H 182.424 140.498 210.107 69.609 ✔ 168 H 195.336 133.528 277.355 143.827 ✔ 170 H 172 E2 154.31 128.947 174 E2 139.91 182 H 148.48 184 H 186 H 188 H 190 146.012 121.302 195.268 73.966 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 184.763 55.816 ✔ ✔ 129.548 176.8 47.252 ✔ ✔ 117.604 195.388 77.784 ✔ 136.775 113.395 163.366 49.971 ✔ 142.943 116.584 181.127 64.543 ✔ H 180.28 115.232 208.892 93.66 ✔ 192 H 156.778 121.067 184.443 63.376 ✔ 196 X 166.715 137.05 183.725 46.675 198 X 161.427 137.1 223.511 86.411 202 X 177.771 135.833 196.231 60.398 206 X 178.27 140.421 205.967 65.546 212 X 183.598 141.789 217.686 75.897 ✔ Author Manuscript 218 T2C 195.333 130.635 316.825 186.19 220 H 170.216 120.906 317.006 196.1 222 H 226 T2D 188.009 228 T2D 230 H 232 T1A ✔ ✔ ✔ ✔ 150.631 229.475 78.844 ✔ ✔ 133.578 118.2 152.534 34.334 ✔ ✔ 142.084 110.959 179.786 68.827 ✔ 141.32 115.79 155.252 39.462 ✔ Discourse Stud. Author manuscript. Torres et al. Page 20 Author Manuscript Speaking Turn Context Mean Min Max Range Opioid Context Pain Context 234 H ✔ 236 T1A ✔ 238 T1A 146.74 120.588 163.096 42.508 ✔ 242 H 166.32 143.79 202.252 58.462 ✔ 147.206 115.181 197.838 82.657 ✔ Author Manuscript 244 H 246 E2 248 T2C 250 E2 254 T2C 184.612 148.452 226.275 77.823 256 H 149.581 131.797 181.258 49.461 ✔ 258 H 163.581 135.16 227.925 92.765 ✔ 262 X 187.25 153.025 216.713 63.688 146.941 128.79 167.802 39.012 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ Author Manuscript 264 H 266 X 184.591 145.828 200.406 54.578 268 X 211.534 189.824 293.409 103.585 270 X 176.493 134.692 182.49 47.798 274 H 165.194 152.004 197.879 45.875 ✔ 276 H 207.691 154.513 268.775 114.262 ✔ 280 H 196.265 146.468 236.122 89.654 ✔ 282 H 284 H 172.484 124.512 191.048 66.536 ✔ 288 T1B 174.823 120.912 240.868 119.956 ✔ 294 T1B 296 T1C 306 308 310 H ✔ 312 H ✔ 314 X 185.551 142.377 194.58 52.203 316 X 192.964 125.025 237.809 112.784 320 X 190.5 119.506 270.648 151.142 322 H ✔ ✔ ✔ 172.266 136.523 249.397 112.874 R1B 135.04 111.836 175.426 63.59 ✔ ✔ R1B 145.448 130.644 196.753 66.109 ✔ ✔ ✔ Author Manuscript 324 X 194.728 144.047 281.585 137.538 328 X 175.73 149.311 326.779 177.468 336 H 147.309 110.679 223.24 112.561 338 X 180.55 128.311 191.98 63.669 342 X 344 X 191.532 122.608 205.782 83.174 346 X 170.863 145.521 173.785 28.264 Discourse Stud. Author manuscript. ✔ Torres et al. Page 21 Author Manuscript Speaking Turn Context Mean Min Max Range 350 X 178.061 128.689 196.973 68.284 352 X 183.844 133.152 203.237 70.085 354 X 358 X Opioid Context Pain Context Author Manuscript 360 X 184.729 117.03 190.83 73.8 362 X 191.642 121.77 289.364 167.594 370 X 190.976 122.748 296.334 173.586 372 H 169.734 113.919 212.009 191.09 374 T2A ✔ ✔ 376 T2C ✔ ✔ 380 X 193.931 118.783 216.974 98.191 382 X 187.229 129.394 221.764 92.37 384 X 192.489 133.18 197.468 64.288 386 X 187.489 123.18 207.468 84.288 390 X 180.862 120.746 214.232 93.486 392 X 191.623 126.211 205.629 79.418 ✔ Author Manuscript Author Manuscript 406 X 176.623 123.211 225.629 102.418 412 R1A 193.426 120.306 281.224 160.918 ✔ 414 T1B 182.169 113.654 227.654 114 ✔ 416 R1A 170.456 112.05 219.238 107.188 ✔ 420 X 422 X 426 X 176.332 144.982 200.638 55.656 428 X 178.467 112.391 208.232 95.841 432 X 172.776 119.618 192.262 72.644 442 H 190.682 145.073 287.774 142.701 446 E2 448 X 209.309 155.192 297.991 142.799 450 X 199.723 125.112 310.076 184.964 464 H 160.604 119.066 193.752 74.686 466 E2 470 H 143.25 116.832 472 E2 163.308 474 H 163.476 476 T1B 208.878 478 T1B 481 H 189.654 122.573 290.198 167.625 ✔ 485 T1C 194.172 161.523 296.41 134.887 ✔ 487 H 165.668 116.098 299.208 183.11 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 166.954 50.122 119.957 187.37 67.413 121.175 171.586 50.411 ✔ 144.578 295.197 150.619 ✔ ✔ ✔ ✔ Discourse Stud. Author manuscript. Torres et al. Page 22 Author Manuscript Speaking Turn Context Mean Min Max Range Opioid Context Pain Context ✔ 491 N3 493 T1C 196.118 118.226 283.964 165.738 ✔ 495 H 172.827 160.222 200.849 40.627 ✔ 507 X Author Manuscript Author Manuscript 526 H 181.551 123.824 219.34 95.516 ✔ 528 R2 177.551 113.924 199.324 85.4 ✔ 532 H 155.571 137.435 193.596 56.161 ✔ 534 H 536 E2 538 H 162.539 115.185 195.999 80.814 ✔ 540 H 178.268 142.035 280.391 138.356 ✔ 544 H 172.569 116.159 203.385 87.226 ✔ 548 H 554 H 198.014 112.315 205.929 93.614 556 T2D 134.136 110.134 164.631 54.497 558 E2 560 H 564 N3 574 H 578 H 145.543 114.816 185.662 70.846 ✔ 582 H 145.782 112.054 188.418 76.364 ✔ 584 H 171.92 118.073 198.94 80.867 ✔ 586 H 188.267 120.89 192.344 71.454 ✔ 588 H 142.464 127.835 163.073 35.238 ✔ 592 R2 198.572 150.694 308.15 157.456 ✔ 598 H 130.993 115.929 163.524 47.595 ✔ 598 H 150.993 120.855 203.446 82.591 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 164.302 135.567 187.228 ✔ 51.661 ✔ Shaded rows denote entirely creaky speaking turn.* Appendix C Patient A Summary Author Manuscript Mean Min Max Range Overall Mean : 195.298 148.785 266.402 117.617 Pain 194.868 149.127 267.366 118.238 Pain: Opioid 184.919 142.710 253.420 110.709 Pain Non-Opioid 197.316 150.706 270.797 120.090 Discourse Stud. Author manuscript. Torres et al. Page 23 Author Manuscript Mean Min Max Range Non-Pain 202.791 142.818 249.591 106.773 Pain: Opioid Mean Min Max Range Narration of pain and description of symptoms 173.256 129.790 221.426 91.635 Request for Opioid treatment 175.372 142.901 257.327 114.426 Positive assessments / satisfaction with opioids. 200.242 145.502 259.110 113.608 202.320 156.212 284.349 128.137 Unclear request regarding Prescription 208.608 155.528 290.200 134.672 Request for information and logistics 195.632 148.050 249.856 101.806 Negative assessments / dissatisfaction with non-opioid treatments. 198.157 153.377 277.530 124.153 Disagreement or resistance 201.411 157.501 283.651 126.149 Other Pain Related Utterances 196.490 149.827 269.928 120.101 Opioid related threats/red flags Pain: Non-Opioid Author Manuscript Patient B Summary Author Manuscript Mean Min Max Range Overall Mean : 171.108 126.549 220.609 94.744 Pain 165.816 124.725 219.141 95.416 Pain: Opioid 161.017 129.391 214.048 84.657 Pain Non-Opioid 166.968 123.606 220.363 97.998 Non-Pain 182.552 130.493 223.785 93.291 Pain: Opioid Mean Min Max Range Narration of pain and description of symptoms 152.509 126.151 182.978 56.827 Request for Opioid treatment. 150.993 118.655 212.042 93.386 Positive assessments / satisfaction with opioids. 175.799 142.616 246.545 103.929 175.629 135.959 236.967 101.008 151.908 126.322 182.213 55.892 Request for Non-Opioid treatment. 167.327 122.899 219.862 102.014 Request for information and logistics. 184.222 133.381 235.004 101.623 Expressions of uncertainty or ambiguity about opioids. Opioid related threats/red flags Pain: Non-Opioid Author Manuscript Positive assessments /satisfaction with non-opioid treatments. 154.733 112.666 216.994 104.329 Negative assessments / dissatisfaction with non-opioid treatments. 180.428 125.182 236.408 111.226 Expressions of uncertainty or ambiguity about a non-opioid treatment plan. 180.394 134.073 260.281 126.208 Agreement 164.302 135.567 187.228 51.661 Other Pain Related Utterances 164.747 122.899 215.644 94.436 Discourse Stud. Author manuscript. Torres et al. Page 24 Author Manuscript References Author Manuscript Author Manuscript Author Manuscript Anderson RC, Klofstad CA, Mayew WJ, et al. (2014) Vocal fry may undermine the success of young women in the labor market. PloS one 9(5): e97506. [PubMed: 24870387] Ball MJ, Perkins MR, Müller N, et al. (2008) The Handbook of Clinical Linguistics. Oxford: Blackwell Publishing Ltd. Bänziger T and Scherer KR (2005) The role of intonation in emotional expressions. Speech communication 46(3–4): 252–267. Bendor D and Wang X (2005) The neuronal representation of pitch in primate auditory cortex. Nature 436(7054): 1161. [PubMed: 16121182] Berry DA (2001) Mechanisms of modal and nonmodal phonation. Journal of Phonetics 29(4): 431– 450. Boersma P and Weenink D (2013) Praat: doing phonetics by computer [Computer program]. Version 5.3.51. http://www.praat.org/ Borkowska B and Pawlowski B (2011) Female voice frequency in the context of dominance and attractiveness perception. Animal Behaviour 82(1): 55–59. Burgess DJ, Crowley-Matoka M, Phelan S, et al. (2008) Patient race and physicians’ decisions to prescribe opioids for chronic low back pain. Social science & medicine 67(11): 1852–1860. Cernak M, Orozco-Arroyave JR, Rudzicz F, et al. (2017) Characterisation of voice quality of Parkinson’s disease using differential phonological posterior features. Computer Speech & Language 46: 196–208. Chafe WL (1993) Prosodic and functional units of Language. In: Lampert MD and Edwards JA (eds) Talking Data: Transcription and Coding in Discourse Research. Hillsdale, N.J.: Lawrence Erlbaum Associates., pp. 24–43. Couper-Kuhlen E (2015) Intonation and discourse. In: Tannen D, Hamilton HE, and Schiffrin D (eds) The Handbook of Discourse Analysis. John Wiley & Sons, Inc. Hoboken, NJ, USA, pp. 82–104. Cruttenden A (1997) Intonation. Cambridge: Cambridge University Press. Dudy S, Bedrick S, Asgari M, et al. (2018) Automatic analysis of pronunciations for children with speech sound disorders. Computer speech & language 50: 62–84. Esling JH (1978) The identification of features of voice quality in social groups. Journal of the International Phonetic Association 8(1–2): 18–23. Esling JH (2012) Voice Quality. In: Chapelle CA (ed.) The Encyclopedia of Applied Linguistics. Oxford, UK: Blackwell Publishing Ltd. DOI: 10.1002/9781405198431.wbeal1271. Esling JH, Moisik SR, Benner A, et al. (2019) Voice Quality: The Laryngeal Articulator Model. Cambridge Studies in Linguistics ; 162. Cambridge, United Kingdom ; Cambridge University Press. Fleiss JL (1971) Measuring nominal scale agreement among many raters. Psychological bulletin 76(5): 378–382. Gerhard D (2003) Pitch Extraction and Fundamental Frequency: History and Current Techniques. Department of Computer Science, University of Regina, Regina. Gill VT, Halkowski T and Roberts F (2001) Accomplishing a request without making one: A single case analysis of a primary care visit. Text–Interdisciplinary Journal for the Study of Discourse 21(1–2): 55–81. Gordon M and Ladefoged P (2001) Phonation types: a cross-linguistic overview. Journal of phonetics 29(4): 383–406. Grivicic T and Nilep C (2004) When phonation matters: The use and function of yeah and creaky voice.: 1–11. Gumperz JJ (1982) Discourse Strategies. Cambridge: Cambridge University Press. Heath C (2002) Demonstrative suffering: the gestural (re) embodiment of symptoms. Journal of Communication 52(3): 597–616. Henry SG and Eggly S (2013) The effect of discussing pain on patient-physician communication in a low-income, black, primary care patient population. The Journal of Pain 14(7): 759–766. [PubMed: 23623573] Discourse Stud. Author manuscript. Torres et al. Page 25 Author Manuscript Author Manuscript Author Manuscript Author Manuscript Henry SG, Chen M, Matthias MS, et al. (2016) Development of the Chronic Pain Coding System (CPCS) for characterizing patient-clinician discussions about chronic pain and opioids. Pain Medicine 17(10): 1892–1905. [PubMed: 26936453] Henry SG, Bell RA, Fenton JJ, et al. (2017) Goals of Chronic Pain Management. The Clinical journal of pain 33(11): 955–961. [PubMed: 28244944] Henton CG (1986) Creak as a sociophonetic marker. The Journal of the Acoustical Society of America 80(S1): S50–S50. Højsted J and Sjøgren P (2007) Addiction to opioids in chronic pain patients: a literature review. European Journal of Pain 11(5): 490–518. [PubMed: 17070082] Hughes HK, Korthuis PT, Saha S, et al. (2015) A mixed methods study of patient–provider communication about opioid analgesics. Patient education and counseling 98(4): 453–461. [PubMed: 25601279] Irvine J (2001) ‘Style’ as distinctiveness: The culture and ideology of linguistic differentiation. In: Eckert P and Rickford JR (eds) Style and Sociolinguistic Variation. Cambridge, U.K.: Cambridge University Press, pp. 21–43. Keating P, Garellek M and Kreiman J (2015) Acoustic properties of different kinds of creaky voice. In: Proceedings of the 18th International Congress of Phonetic Sciences, 2015, pp. 0821–1. Matthias MS, Krebs EE, Collins LA, et al. (2013) “I’m Not Abusing or Anything”: Patient–physician communication about opioid treatment in chronic pain. Patient education and counseling 93(2): 197–202. [PubMed: 23916677] Matthias MS, Krebs EE, Bergman AA, et al. (2014) Communicating about opioids for chronic pain: A qualitative study of patient attributions and the influence of the patient–physician relationship. European journal of pain 18(6): 835–843. [PubMed: 24921073] Mendoza-Denton N (2011) The semiotic hitchhiker’s guide to creaky voice: circulation and gendered hardcore in a Chicana/o gang persona. Journal of Linguistic Anthropology 21(2): 261–280. Merrill JO, Rhodes LA, Deyo RA, et al. (2002) Mutual mistrust in the medical care of drug users. Journal of General Internal Medicine 17(5): 327–333. [PubMed: 12047728] Moisik S (2013) The epilarynx in speech. Doctoral Dissertation. University of Victoria, Victoria, BC, Canada. Podesva RJ (2007) Phonation type as a stylistic variable: The use of falsetto in constructing a persona 1. Journal of sociolinguistics 11(4): 478–504. Roberts F and Kramer JS (2014) Medication and morality: Analysis of medical visits to address chronic pain. In: Hamilton H and Chou WS (eds) The Routledge Handbook of Language and Health Communication. London: Rouletge, pp. 477–489. Robinson JD (2001) Asymmetry in action: Sequential resources in the negotiation of a prescription request. Text–Interdisciplinary Journal for the Study of Discourse 21(1–2): 19–54. Schilling-Estes N (1998) Investigating “self-conscious” speech: The performance register in Ocracoke English. Language in society 27(1): 53–83. Slobe T (2018) Style, stance, and social meaning in mock white girl. Language in Society 47(4): 541– 567. Sullivan M and Ferrell B (2005) Ethical challenges in the management of chronic nonmalignant pain: negotiating through the cloud of doubt. The Journal of Pain 6(1): 2–9. [PubMed: 15629412] Tannen D and Wallat C (1987) Interactive frames and knowledge schemas in interaction: Examples from a medical examination/interview. Social psychology quarterly 50(2): 205–216. Turk DC and Okifuji A (1997) What factors affect physicians’ decisions to prescribe opioids for chronic noncancer pain patients? The Clinical journal of pain 13(4): 330–336. [PubMed: 9430814] Van Dijk TA (1999) Critical discourse analysis and conversation analysis. Discourse & Society 10(4): 459–460. Van Dijk TA (ed.) (2011) Discourse as Structure and Process. Reprinted. Discourse studies a multidisciplinary introduction / ed. by van Dijk Teun A.; Vol. 1. London: SAGE. Wilce JM (1997) Discourse, power, and the diagnosis of weakness: encountering practitioners in Bangladesh. Medical Anthropology Quarterly 11(3): 352–374. [PubMed: 9292862] Discourse Stud. Author manuscript. Torres et al. Page 26 Author Manuscript Yuasa IP (2010) Creaky voice: A new feminine voice quality for young urban-oriented upwardly mobile American women? American Speech 85(3): 315–337. Zetterholm E (1998) Prosody and voice quality in the expression of emotions. In: Fifth International Conference on Spoken Language Processing, 1998. Zimman L (2012) Voices in Transition: Testosterone, Transmasculinity, and the Gendered Voice among Female-to-MaleTransgender People. Linguistics Graduate Theses & Dissertations. University of Colorado at Boulder, Colorado. Zimman L (2013) Hegemonic masculinity and the variability of gay-sounding speech: The perceived sexuality of transgender men. Journal of Language and Sexuality 2(1): 1–39. Author Manuscript Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 27 Author Manuscript Author Manuscript Figure 1. Diagram showing how the data were analyzed acoustically. Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 28 Author Manuscript Author Manuscript Figure 2. Mean pitch distributions for each of the three coding categories. The width of each violin plot indicates the distribution of the mean pitch located at that point, while the bar shows the overall average for that category. Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 29 Author Manuscript Author Manuscript Figure 3. Mean pitch distributions of opioid requests and pain narration separated from the rest of the speaking turns. The width of each violin plot represents the pitch distribution located at that point while the bar represents the average pitch for that category. Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 30 Author Manuscript Author Manuscript Figure 4a. Example of pain narration from Patient A Waveform and spectrogram of a turn in which patient A discusses pain. The dense line on the spectrogram represents pitch. The mean pitch of the present turn is below her overall average and covers only the lower 22 percent of her vocal range. Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 31 Author Manuscript Author Manuscript Figure 4b. Example of pain narration from Patient B Waveform and spectrogram of a turn in which patient B discusses pain. The dense line on the spectrogram represents pitch. The mean pitch of the present turn is below her overall average and covers only the lower 20 percent of her vocal range. Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 32 Author Manuscript Author Manuscript Author Manuscript Figures 5 and 6. Creaky and modal versions of the same word Waveforms and spectrograms illustrating the creak and modal versions of the word “Pain” and “Bad” spoken by Patients A and B, respectively. The line on the spectrogram refers to pitch. The waveform for the beginning of the vowel is also presented. Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 33 Table 1. Author Manuscript Patient-physician information taken from their pre-consultation questionnaire. Patient Resident Physician Pseudonym Gender Age Range Pseudonym Gender Age Range Patient A Female Late 50s Physician A Female Mid 20s Patient B Female Mid 40s Physician B Female Early 30s Author Manuscript Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 34 Table 2. Author Manuscript Distribution of speaking turns collected and analyzed. Speaking Turn Patient A Patient B Modal turns (analyzed for pitch) 166 136 Entirely creaky (not analyzed for pitch) 17 41 183 177 Total number of speaking turns Author Manuscript Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 35 Table 3. (a) Chronic pain (Opioid context) Author Manuscript Discussions about chronic pain and the opioid medication used to manage it. Patient A Patient B Author Manuscript Request for opioids If you could give me methadone pills. I would be happy with that. I just feel like it [Dilaudid] should be put on there [Pain Contract]. Narration/description of chronic pain (Treated with opioids) I woke up crying last night. I--it’s just still all up here in the shoulder. I’ve been miserable the last couple days from my neck and shoulders. Positive assessment of opioid treatment I know what is safe for me. I’ve been prescribed with Norco, and it has been working. I’m still alive. I haven’t been needing Zofran as much since they switched me to that Dilaudid. Negative assessment of opioid treatment None None Ambiguous assessment of opioid None (On effectiveness of opioid) It depends on the day, it depends on what I’m doing. Opioid-related red flags and threats. I’ll self-medicate if I have to. I’ll go on the streets. I’m not gonna go through withdrawals. I probably take more (opioid) overall than I would normally, when I’m in a lot of pain. Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 36 Table 4. (b) Other pain (Non-opioid context) Author Manuscript Discussions about other types of pain that does not involve opioids Patient A Patient B Author Manuscript Request for non-opioid medication I think I need a refill on Dulcolax. The Robaxin. Um, could you guys change that to 120 tablets a month? Unclear requests So are you gonna prescribe me anything? None Request for information How many refills do I have on the ibuprofen? Would it be okay if my daughter picks the prescription up? Positive assessment of non-opioid treatment None Neurontin does help with some of the um, like my skin or my hair hurting, Negative assessment of non-opioid treatment Ibuprofen is not good for my stomach. I’ve tried it all. I’ve tried a party bag of ice until I can’t even feel it anymore. Ambiguous assessments of nonopioid treatment None Robaxin doesn’t really help, but it does General agreements None Yeah, especially right in here. I would definitely try trigger point injections General disagreements No, I’m not gonna do Tylenol, because it’s not good. None Other pain-related utterances I’ve had constipation. I’m kinda worried about asthma. Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 37 Table 5. (c) Non-pain Author Manuscript Discussions that are not about pain Non-pain Patient A Patient B We can do Rite-Aid, yeah. Honestly, it’s just to go to Taco Bell— Author Manuscript Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 38 Table 6. Author Manuscript Average pitch of all modal speaking turns per category. Patient A Patient B Overall average pitch 195.298 Hz 171.108 Hz All pain-related utterances 194.868 Hz 165.816 Hz (a) Chronic pain (Opioid context) 184.919 Hz 161.017 Hz (b) Other Pain (Non opioid-context) 197.316 Hz 166.968 Hz (c) Non-pain 202.791 Hz 182.552 Hz Author Manuscript Author Manuscript Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 39 Table 7. Author Manuscript Average pitch for each subcategory Average Pitch (in Hz) Patient A Patient B 195.298 171.108 Request for opioids 175.372* 150.993* Narration/description of chronic pain 173.256* 152.509* Positive assessment of opioids 200.242 175.799 Baseline: overall average pitch (a) Chronic pain (Opioid context) Negative assessment of opioids -- Ambiguous assessment of opioids -- Opioid-related red flags and threats. 202.320 -175.629 151.908* Author Manuscript (b) Other pain (Non-opioid context) Request for non-opioid medication -- Unclear requests 208.608 Request for information 195.632 Positive assessment of non-opioid treatment Negative assessment of non-opioid treatment -198.157 167.327* -184.222 154.733* 180.428 Ambiguous assessment of non-opioid treatment -- 180.394 General agreements -- 164.302* General disagreements 201.411 -- Other pain-related utterances 196.490 164.747* 202.791 182.552 (c) Non-pain Author Manuscript Non-pain (*) Asterisk indicates that the average pitch of that category is lower than the speaker’s overall average pitch. Author Manuscript Discourse Stud. Author manuscript. Torres et al. Page 40 Table 8. Author Manuscript Distribution of completely creaky speaking turns Number of creaky turns Patient A Patient B Context: Pain 11 35 Context: Non-pain 6 6 None of the completely creaky turns were measured for pitch. Author Manuscript Author Manuscript Author Manuscript Discourse Stud. Author manuscript.