Listening to Interviews: Attending to Aurality, Emotions, and Atmospheres in Qualitative Analysis

Abstract: Sociologists commonly record interviews; however, most coding and analysis is done from transcriptions rather than from the audio recording itself. Transcribing sound into text is an often unquestioned and almost hallowed step in qualitative research. Yet, the question of what is lost in transcription is worth revisiting given advances in artificial intelligence, ethics, and new conceptual concerns that challenge processes of knowledge production. In this article, I aim to offer inspiration for listening to interviews—data that can be heard—in qualitative research. I discuss practical, ethical, and conceptual considerations that may come with analyzing interview recordings, tracking the reasons why scholars have come to rely on transcripts and the possible shortcomings of doing so. I argue that by adhering to a tradition of working with only the transcript, researchers miss rich layers of sensory, emotional, and embodied data. Instead, I suggest the use of listening as one of the key methods in analyzing the constitutive role of emotions and atmospheres in qualitative interview research.

The impetus for this paper came from my experience as a sociologist attending an oral history conference. I was struck to hear most presenters playing excerpts from interviews in their papers. They offered analysis of the pauses, hesitations, emphases, excitability, laughter, and so on, in the interview dialogue. It felt like participants, their character and sentiment, were in the room. The interview itself seemed more transparent, with the exchange between interviewer, interviewee, and the interview space often included, and aspects of the dynamic and setting audible. Most starkly, the audio-clips conveyed and sparked discussion about emotions that can be heard in the tone, exhalations, and stammers. I contrasted this to papers, including my own, presented at sociology conferences where most commonly, snippets of interviews are shared via quoted text from transcripts, with only a few aspects of intonation carried over, e.g., pauses denoted by ellipsis and emphasis by italics. Sociologists commonly record interviews; however, most coding and analysis is done from transcriptions of interviews rather than from the audio recording itself. Researchers working with specific methods, such as conversation analysis, routinely analyze audio (ASHMORE & REED, 2000; AYAß, 2015). However, in sociology more widely, scholars have noted that transcription of sound into text is an often unquestioned and almost hallowed step in qualitative research (DAVIDSON, 2009; NASCIMENTO & STEINBRUCH, 2019; POLAND, 1995). I wondered, why do sociologists commonly record interviews only to use the written transcript? And what is lost in this practice? As I discovered, this is not a new question, but it is one worth revisiting, especially given advances in automated technology, ethics, and new conceptual challenges for qualitative inquiry. [1]

Overall, in this paper I offer inspiration for listening to interviews—data we can hear—in qualitative research. I outline practical (Section 2), ethical (Section 3), and conceptual (Section 4) considerations that may come with analyzing interview recordings, tracking the reasons researchers often rely on transcripts and the shortcomings of doing this. I argue that by adhering to a tradition of working with only the transcript, researchers unnecessarily miss rich layers of sensory data like emotions and atmospheres. I therefore offer a more specific provocation for qualitative research on emotions (Section 5). The possibility of listening to interviews is of special relevance for research on emotions because they are widely cited as one of the main elements of data transcription sacrifices (HAMMERSLEY, 2010; MERO-JAFFE, 2011; TESSIER, 2012; WAINWRIGHT & RUSSELL, 2010). Emotions are mercurial, subjective, and often evade expression. Thus, scholars have found them hard to interpret, and sometimes off bounds (BROWNLIE, 2011; WETTERGREN, 2015). Aurality does not fully alleviate this challenge. But it can offer ways to hear the feelings that simmer in and around words as it might be possible to hear the emphasis in speech—sarcasm, trepidation, pride; or the embodied aspects of emotions—gasps, sighs, sobs; or at the edges of expression—mumbles, titters, pauses, fishing for words. To highlight the value of analyzing interview audio for research on emotions, I conclude the paper by exploring how listening to interviews can open insights into feelings and atmospheres (Section 6). To do this, I draw on sociological projects in which researchers have experimented with audio-coding and listening methods, as well as lessons from adjacent disciplines, such as oral history, where the emotional and aural elements of recorded interviews are established as central to both research analyses and outputs. In this, my paper contributes to the "growing awareness of the significance of the sonic world as open to qualitative inquiry" (BULL & BACK, 2016, p.16; see also HALL, LASHUA & COFFEY, 2008, p.1037), as I reconsider sound as thick with emotions and atmospheres. [2]

Working from transcriptions has become a commonsense process for qualitative sociologists, an essential part of one's methodological training and research practice thereafter. WAINRIGHT and RUSSELL argued that "transcripts, the verbatim write-up of audio-recorded interviews, are a respected and in many ways fetishized form of data" (2010, p.1). This is supported by STONEHOUSE's review of method guides for qualitative researchers, who found that almost all "assume whole-interview transcription as a necessary component of interview analysis" (2019, p.3). In literature, scholars have given several reasons for the necessity of transcription in qualitative research. Foremost among these have been immersion, accuracy checking, and searchability (GREEN, FRANQUIZ & DIXON, 1997). Others have noted its role in anonymizing participants (WERNER, MEYER & BISCHOF, 2023). In transcribing a recording, the researcher is immersed in each word of the interview, an invaluable means of getting close to the data. The investigators can check the interview against the transcript, ensuring that what was said is documented and analyzed, rather than only their recollections of the interview. These reasons are important. However, there is a general assumption that researchers are doing their own transcriptions, and indeed, conducting their own interviews. [3]

It is not always the case that the interviewer, transcriber, analyst and researcher are the same person (TILLEY & POWICK, 2002). Scholars commonly outsource transcription to research assistants or third-party transcription services, and in larger funded projects, may even outsource some of the interviewing. In these cases, the researcher is perhaps neither immersed in nor checking the audio as part of the transcription process. With the advent of AI and automated transcription functions, such as the transcripts produced from Zoom recordings, it is likely that soon, in many cases, no human will listen back to the recorded interview at all. While automated transcripts are rough and currently require editing, using automated transcripts will become more common (BOKHOVE & DOWNEY, 2018) as accuracy improves. Technology, in this case, has made the traditional rationale for full reliance on transcription less relevant. Advances in software also make audio-coding a valid addition or alternative. Software that makes audio searchable and enables audio-coding has been available since early 1991 (STONEHOUSE, 2019, p.3). However, this advance has clearly not taken off in sociology yet, perhaps in part because of the assumed process of transcribing interviews for analysis. [4]

This practice of transcription also means that there has been little open discussion about its value to specific projects. It is not common for published studies to detail the process of transcription, even though it involves choices and shapes analysis (STONEHOUSE, 2019; WAINRIGHT & RUSSELL, 2010). As noted, transcripts can be produced by a range of people, the interviewer, a research assistant, a third-party service, or a machine.¹⁾ There are also distinct types of transcription, from exhaustively annotated to "just the gist." [5]

Scholars working with sociolinguistics, conversation analysis and reconstructive biographical research often transcribe aural aspects of speech, including annotating sonic elements of speech such as prosody, or adding memos about embodied aspects of the research encounter (ASHMORE & REED, 2000; AYAß, 2015; BAMBERG & GEORGAKOPOULOU, 2008; BECKER, POHN-LAUGGAS & SANTOS, 2023). However, this is not the norm in qualitative research. CRAIG, McINROY, GOULDEN and EATON (2021, p.4) observed that in "pragmatic transcripts," the current standard, "the interview dialogue is transcribed verbatim from the recording and no attempts are made to neutralize the loss of multidimensional elements of the interview, such as the participant's speed, pace, intonation, song, hesitation, interview context, or background noise." This loss is underscored in ANDERSON and JACK's study of rural women's lives (2016 [1998]), where several of the researchers "found discrepancies between our memories of interviews and the transcripts because the meaning [they] remembered hearing had been expressed through intense vocal quality and body language, not through words alone" (p.158). Had the researchers not also been the interviewers, this disjuncture may not have been noticed, and the untranscribed data lost to the study. As the traditional benefits of transcription—immersion, accuracy checking, and searchability—become less relevant, it is worthwhile reevaluating whether such trade-offs maintain their value. However, the issue of anonymity remains an important one, particularly in sensitive research, where it could be too risky to reveal a person's voice. In such cases, it may be possible to listen to and analyze the audio carefully but not include it in outputs where identification may create risk. [6]

COLLINS, LEONARD-CLARKE and O'MAHONEY (2016) argued that the difference between hearing a spoken utterance and reading a transcribed text is not so much a loss as an altogether different outcome—listening to a recording is not the same sensory experience as reading a text, nor do written and spoken words necessarily create the same meanings. Such considerations are valuable as they dislodge the assumed requirement of transcription, showing that it is not merely a question of the quality of a transcript but its form. COLLINS et al. reminded us that the transcript, as with the audio recording, is not the interview. It is worth recounting this study in more detail to understand the authors' argument about the differences between listening and reading, a difference that opens up productive questions about how qualitative data are analyzed. [7]

COLLINS et al. focused specifically on disfluencies—sounds like "um," "ah," "er"—that pepper everyday speech. In some cases, disfluencies are included in verbatim transcriptions and analyzed. Offering a case study, they harked back to an interview the lead author had conducted for a project on physicists 20 years earlier, where, after presenting the transcript to his interviewee for approval, the scientist requested the removal of disfluencies as they made him sound unsure on the topic he discussed. COLLINS was reluctant to edit the transcript, given his training's emphasis on verbatim transcription as best practice. However, he came to see that what the interviewee said was true. When listening to someone speak, people filter out disfluencies, but in print they are made more significant. In the recording, the scientist sounded careful and expert, yet on the page he read as unsure and hesitant. Taking this experience as a cue, the authors revisited the interview and tested how people listen to and read the same utterance differently. Overall, they found that the transcript as a text can redirect the researcher away from pursuing the overall meaning, which is more immediate when listening to speech. It can attune the reader to disfluencies in text, where they stand out starkly as a variation from the narrative flow and come to be attributed meaning. There will certainly be cases where reading will attune the researcher to something important missed in the act of listening, but this is an argument for using recordings and transcripts in concert. Overall, the study showed that listening to interviews offers different kinds of data than reading a transcript and provides a valuable means to consider how modes of analysis shape the analysis itself, what is included or excluded. [8]

However, all this is not to presume that transcripts are simply flat and uninspiring. Transcripts can be evocative, for example, in stimulating a sensory memory of the interview. WAINWRIGHT and RUSSELL (2010) referred to this phenomenon as "subvocalization—when we hear in our minds the words we read" (p.3). They argued that by stimulating memory (assuming the reader is the same person who conducted the interview) the transcript "might play a part in bringing us back to that interview, that house, that day, etc." (ibid.). However, they contended that while transcripts can transport the reader back, this "may not be as effective as hearing the recorded voices all over again while engaged in analytical reflection" (ibid.). This contention is supported by researchers in oral history where listening back to recorded interviews, even when done by the interviewer, can throw up inflections and emotions that were not recalled without audio stimulus. This can be especially important when time has elapsed between interview and analysis or re-analysis. For example, BORNAT (2016) reflected on the value of relistening as what HOLMES (2010) called "emotionalization of reflexivity," that is, for the interviewer to also hear their own contribution to an emotional dynamic. When BORNAT (2016) listened back to an oral history she had recorded with a mill worker early in her career, she heard it anew—not only the interviewee's emotions but her own as a junior researcher, and thus the emotion of the interview, evident in both pace and redirection. BORNAT wrote,

"I was unprepared for her emotion in recalling this time in her life. [...] It wasn't just that she had veered away from the topics I wanted to hear her talk about; I simply didn't know what to do with the emotion she expressed in the interview situation" (p.43). [9]

Re-listening to the recording summoned reflection on the emotions expressed, but also how they were shaped by the interview. What could be remembered and interpreted from an interview recording changed over time, with the audio data crucial in allowing a return to the interview to hear details potentially not heard and noted in an earlier transcription. Listening to an interview can engage senses, memories, and emotions in ways that bring new insights to data and their collection, enhancing the transcript's role as stimulus in analysis. [10]

As noted, many authors, including those who assumed scholars should work from transcripts, have acknowledged that emotion is one of the most important things lost in transcription (e.g., MERO-JAFFE, 2011; WAINWRIGHT & RUSSELL, 2010). Of course, much important research has and can be done about emotions by analyzing text, including novels, lyrics, poetry, letters, diaries, ethnographic fieldnotes, qualitative survey data, and transcripts of interviews. But in the case of qualitative interviews, it is usually noted that some levels of emotion that are discernible in a recorded interview are lost when it is transcribed and analyzed as text. Such qualifications are often made in asides, for example, in a study by SALDAÑA (2015), where it was noted of a quotation from a transcript:

"Not included in this particular interview excerpt are the emotions the participant may have experienced or talked about. [...] A reader may not have inferred that specific emotion from seeing the line in print. But the interviewer, present during the event and listening carefully to the audio recording during transcription, noted that feeling in his tone of voice" (p.595). [11]

In such asides, researchers can be seen to be working around the losses in the transcription process as best they can, to carry over the emotions experienced in the interview itself. Many investigators keep detailed notes to supplement transcripts, for example writing memos (GLASER & STRAUSS, 1968 [1967], p.108). However, such gaps could be remedied if audio were included more thoroughly in analysis and dissemination. Researchers working with audio cannot magically solve issues of subjectivity in the reconstruction of the emotions expressed, but they can provide another potential point of access to the co-creation of meaning in an interview. An additional practical concern is that analyzing text pairs well with the dominance of text-based modes of dissemination, such as books and journal articles; this is a point I will address later. For now, I continue with a provocation to loosen the rigid reliance on analyzing transcripts as the best practice. [12]

Listening to interviews also provides opportunities for scholars to engage creatively with pressing and complex ethical concerns about listening and voice, a discussion that has remained strangely bereft of aural dimensions. There is a huge amount of research on the role of listening within an interview—the importance of open-ended questions, active listening, and so on (e.g., ABRAHAMS, 2017; GUSCHKE, 2023; LAVEE & ITZCHAKOV, 2023). Rarely mentioned, though, is listening at the analysis stage of research. Work on listening has also proposed alternatives to dominant traditions—the need to listen into silences, to polyphonous voices, or to listen against the grain (GUSCHKE, 2023; 't HART, 2023). Similarly, scholars have done a great deal of research on the ethical importance of "voice" in qualitative research (e.g., JACKSON & MAZZEI, 2009). However, as CHANDLER, ANSTEY and ROSS observed, while there is an "emphasis and honoring of the participants' own words as generative of meaning and knowledge," "it is rare to hear the actual voices of the research participants in a presentation or in text" (2015, p.1). [13]

Researchers working on voice have primarily focused on the concept of voice. They have used it to elevate stories and experiences that are otherwise marginalized. Such work is methodological as scholars have tried to decenter the researcher as the main storyteller and to establish methods for centering interviewees as narrators of their own lives. While there is no need to be reductive in claiming that voice, in this conceptual sense, must be equated with people's actual speaking voices in research, it remains a little surprising that even in writings on the ethics of voice, methods for listening to voices and including them at stages of the research process have been rarely discussed. For example, in a collection dedicated to voice in qualitative inquiry, JACKSON and MAZZEI (2009) offered a series of valuable and provocative essays that each dealt with voice in a critical and conceptual fashion, but only a few of the chapters mentioned the issue of audio and transcripts, and only in passing. For example, in MAZZEI's own chapter she stated:

"Instead of the one-dimensional way in which we often treat data on the page in the form of field notes and transcripts, our data may speak to us differently with a multiplicity of voices in the context in which it was presented [...] If we are bound to the text, be it the text of the transcript or the script that is spoken by our participants, then we are bound by the text" (2009, p.53). [14]

While MAZZEI raised this question, she went on to analyze excerpts from transcripts without reference to aural aspects of the participants' story as spoken in the field. This is not a criticism, as the chapter is valuable in extending our conceptual understanding of voice. Rather, the example showed that even in contexts where the limits of transcripts are acknowledged and the ethical aspects of listening to voice are central, the analysis of transcribed text remained the default. [15]

In the small amount of research in which scholars have explored audio-coding, they have frequently raised the issue of ethics. They have argued that through listening to audio in the coding, analysis, and dissemination phases of research, it is possible to achieve some objectives of elevating marginalized voices and/or conducting sensitive research. In their work on coding audio and video from interviews with sexual and gender minority youth, CRAIG et al. (2021) noted that the methodological literature held little information about options beyond analyzing written transcripts. They remarked that, "while transcript analysis is widely used across disciplines, it may have limitations—particularly for research involving marginalized populations" (p.1) specifically. CRAIG et al. found that elements only deduced by analyzing the audio, such as pace and intonation, were particularly "essential to cross-cultural research and research with marginalized populations" (p.5). CRICHTON and CHILDS (2005) similarly argued for researchers to work with the technological innovations at hand to "clip and code" audio, explicitly to "enable participant voice" (p.40). They explained that "the tools and methods used have evolved in such a way that the voice of the participants under study has the potential to be stronger when digital audio recordings are clipped and coded [...] rather than directly transcribed" (ibid.). Speaking to the particulars of what audio adds to analysis, they explained that coding audio allows researchers to attend to not only what is said, but how it is said, and in what context (pp.47-48). [16]

As I will discuss in the second half of this paper, audio is valuable for understanding emotion and atmosphere in ways that extend the analysis achieved with a transcript. The point made in the audio-coding literature is that embodied listening to actual voices, in coding and analysis, can generate ethical benefits for the conceptual inquiries about listening and voice that underpin post-structural literature on qualitative research. Far from leading to a "too literal" interpretation of these ethics, the actual inclusion of voice via audio-coding allows for a robust and critical analysis of representation and listening. [17]

It is worth adding that in two of the audio-coding studies scholars also advocated using audio in dissemination to further center the voices of participants. CHANDLER et al. developed a "hypermodal" method of dissemination, using text, audio, and video, to, as they explained, "more fully honor the voices of our research participants by enabling our audiences to experience the power of the data through listening to it with their own ears" (2015, p.3). Similarly, CRAIG et al. incorporated audio into their research outputs as "audio-enhanced dissemination" which, they argued, emphasized the "importance of the participant's voice in qualitative research" (2021, p.6). [18]

There are obvious ethical issues with including audio (and especially video) excerpts of participants in studies where anonymity is needed, as people can be more readily identified by voice and other audio clues, e.g., about location. But increasingly, as with the assumption that transcription is the gold standard, the idea that anonymizing participants is best practice has also been questioned. Some scholars have not anonymized participants, others have allowed participants to choose (MOORE, 2012; TILLEY & WOODTHORPE, 2011). Using audio in dissemination of research will not always be appropriate. However, where it is appropriate, the data evident in a recording can diversify the voices heard in research spaces. "Audio-enhanced dissemination" can also open valuable sociological information, including that which may or may not be noted by the researchers themselves, and invite audiences into sociological analysis. To name just one example, accents, in some cases, may convey the class and ethnicity differences or similarities between interviewer and interviewee, giving insight into the relationality that shapes the research encounter. In this vein, sharing audio can be an important "check" on the interpretation of a researcher, as was the aim of transcription, making the dynamics of data collection audible. [19]

Among scholars who have critiqued the assumed good of transcription, two schools of thought have primarily influenced academics rethinking a divide between text and audio, namely anthropological critiques of Western ocularcentrism and post-structural criticisms of a speech/writing hierarchy. I think it is important to consider these influences. Researchers from these fields showed that limiting our analyses to either seeing or hearing is problematic and recuperates questionable assumptions. Furthermore, they demonstrated that a turn to the aural will change both our methods and the ideas that underpin them. WAINWRIGHT and RUSSELL (2010) argued that the wider epistemological assumptions of social scientific research are part of why transcription still dominates, and audio-coding remains obscure in practice. As they noted, "deeper challenges arise because of epistemological and sensorial differences between thinking critically and analytically about what we read compared to what we hear" (p.3). They cited a western tradition of ocularcentrism—privileging vision as the sense of scientific observation and verification (see also HOWES, 1991). [20]

The fundamental conflation between seeing and knowing has been called a "despotism of the eye" and scholars have pointed out the profusion of "sight" metaphors in discussions of research (IKONEN, 2022, p.2020). DAZA and GERSHON, for example, called for methods that go "beyond ocular inquiry," to take up "multisensory" and "aural, embodied" inquiry (2015, p.641). Seeing a sound-driven sociology as a tool for ethical research, they cited hearing as a modality where reverberations between people, places, and objects are included and "the impossibility of closing an earlid is a possibility for more socially just, ecological methodologies" (p.639). However, BACK, a sociologist who has also promoted listening as a method, was careful to qualify that hearing is not a superior sense but rather one whose full repertoire—including its dark side—has been undervalued and underexplored (2023, n.p.). For BACK, a turn away from the ocularcentric, or the spectacle, has not offered an easy remedy, but access to a different scale, in that "thinking with sound and thinking through sound, invites us to pay close attention to those things that are also going on that might be in the background—that might not be the big spectacles of life" (ibid.). In these arguments, scholars have sought to question the assumed centrality of vision as the scientific sense. They have explored the dynamics of how research might be opened up to be multisensory in ways that can enhance analyses, attending to new layers and scales of social life. WAINWRIGHT and RUSSELL (2010) argued that it is only with more "take-up" of methods such as audio-coding that we can begin "challenging the necessity of transcripts for the systematic analysis of interviews, [and] think critically about the way in which interviews only gain their legitimacy as 'data' once they are transformed into text" (p.3). [21]

The second conceptual influence that has guided researchers towards reading and listening as modes of analysis is post-structural, grounded in DERRIDA's troubling of representation. Particularly pertinent to the discussion here was DERRIDA's (1978 [1967]) analysis of the binary between speech and writing, and his critique of the idea that speech is more present and immediate, with writing a lesser form that comes after the fact. Such thinkers were particularly influential in anthropology where the concept of "writing up" was deeply problematized, with the act of writing recognized as productive of culture rather than descriptive (CLIFFORD & MARCUS, 1986). In light of this, scholars have been preoccupied with notions of "writing" and "text," and those turning to sound have been wary not to resurrect the oral as a "truer" source of data, or to cast the tape-recorder a "quasi-magic time-machine" (RICHARDSON, HAWORTH & DEAMER, 2022, p.683). Considering this turn, MacLURE (2009) argued that qualitative analysis must begin by forsaking ideas about the intact unmediated voice and the possibility of "capturing" it altogether. Instead, she embraced "the insufficiency of voice—its abject propensity to be too much and never enough" (p.97). Taking in shades of emotion, she had in mind:

"a kind of 'voice research' that would attend to such features as laughter, mimicry, mockery, silence, stuttering, tears, slyness, shyness, shouts, jokes, lies, irrelevance, partiality, inconsistency, self-doubt, masks, false starts, false 'fronts' and faulty memories—not as impediments or lapses to be corrected, mastered, read 'through' or written off, but as perplexing resources for the achievement of a dissembling, 'authentic' voice" (p.98). [22]

While MacLURE was primarily focused on "voice" in the conceptual sense, she located audio-based analysis as one avenue for putting this line of inquiry into practice, given that "the voices that are 'heard' in research texts carry so few of the qualities associated with the spoken voice" (ibid.). Taking a post-structural stance, she outlined the problems associated with transcribing recordings into text but was careful to qualify that she did not find audio to be "truer," as something is always lost. She was interested rather in what is lost—what disappears, MacLURE argued, is not only the "phonic substance of speech—its sound and cadence," but also the "double-dealing, mischief-making qualities [...] that lie on the boundaries of language itself, such as laughter, gasps, tears, sneers, and silences" (ibid.) Thus, while she did not seek to privilege speech over writing, MacLURE saw those aural modes of analysis as ones that "may offer useful and powerful resources for research" (p.111). In such analyses, as with critiques of ocularcentrism, there is no lazy antidote, such as privileging sound over sight, but rather the endorsement of an exploratory pluralism where the opening of senses and attention can provide new potentials for the generation of vibrant data and analyses. [23]

In the following sections I will focus on examples of inquiry in which researchers have listened to interview audio, and in doing so, opened paths into analyzing emotion. Such coding can be like the coding of transcripts, in terms of centering "emotion words," for example, when participants talk about feeling worried or happy. However, as I will demonstrate, analyzing interview audio can also register aspects of emotion that are diffuse, ecological, and complexly social. Listening to interviews can elevate emotions heard in expression, tone, sounds, or even silence, as well as atmospheres and emotional resonances, highlighting the relationality between people, environments, and social worlds. In this last section of the paper, I aim to draw together inspiration for scholars who might, having taken in the argument so far, be contemplating how they might listen to interviews in emotions research. [24]

The researchers who promote the necessity of transcription and those who question it have agreed that the process of rendering audio into text sacrifices emotion. In interdisciplinary qualitative research, there have been efforts to remedy this and to engage with the aurality of interviews, analyzing emotion as expressed in tone, pace, emphasis, laughter, and tears; as well as emotion that is withheld, in pauses, shifts in tone, silence (e.g., GALLAGHER, 2020; KINGSTON, 2024). [25]

In her work on emotions and reflexivity, BROWNLIE (2011) argued that we cannot simply rely on "emotions talk" but must find ways to approach the non-narratable and the non-verbal aspects of emotional expression—perhaps especially for people whose experiences fall out of middle-class modes of therapeutic dialogue. On a similar note, WETTERGREN (2015) asked "how do we know what they [participants] feel?", and called attention to the role of unsaid or withheld emotion in interviews. How something is said, when speech changes tone, breaks off, or breaks down, can be heard in interview audio and may provide alternate ways to listen for the meaning that is made in interviews. I consider this approach valuable to analyses that explore emotions specifically; however, given that emotions are central to how people express all manner of experiences, listening for emotion could add to studies well beyond an emotion focus. [26]

An inspiring example of how audio can become central to analyzing emotions in qualitative research was outlined in THOMSON's Australian Generations oral history project. In a paper on "indexing emotions," THOMSON (2019) explored how "hearing the voice and sound of personal testimony can expand and enrich our understanding of its emotional register" (p.2). In the interview, emotion was conveyed in bodily and vocal cues. A question may have sparked recall of an emotional episode, and as that "remembering generates a recurrence of emotion, we can see our narrator hunching over, perhaps unable to speak, or we can hear the breath squeezed out of her voice, a change in pitch or pace, or a lingering sigh" (p.2). [27]

THOMSON argued that even though oral historians have worked with aurality and emotions for a long time and are "alert to the fact that a transcribed text of the words of the interview is a partial and inadequate rendering of the event" (ibid.), locating and analyzing emotions remains difficult. The process of listening back in real time is laborious, and it can be difficult to determine what feeling is being expressed. To address this, Thomson and his team worked with the National Library of Australia to pioneer a method for coding emotion. Open-access, time-coded transcripts are embedded with the audio for targeted listening.²⁾ People accessing the oral history recordings can search for emotions in the time-stamped summaries that accompany the audio and then play excerpts from interviews, directly accessing the sound. In this way, researchers can use the text summaries as an index for the audio which becomes the main source for the interview. With the support of the Studies in Oral History journal, the article about this method is also embedded with links to the audio for each interview quotation analyzed in the paper. The reader therefore listens to the interview while reading analysis that attends to the aural aspects of the data. [28]

The study highlighted the practical possibilities for audio-coding and dissemination that could be used much more widely than oral history. THOMSON claimed that by listening to emotion in interviews we can begin to understand relationships between emotions at the time people are recalling, emotions at the time of retelling, and the emotional dynamic of the interview (ibid.; see also HOLMES, 2017). Oral history can serve as a precedent for the social sciences in moving toward listening to interviews in analysis, with its recognition that "much of the emotional meaning in interviews is available through the sound of the voice" (THOMSON, 2019, p.2). [29]

To demonstrate the gravity of what insights can be garnered by listening to the sounds and emotions that can be heard in audio, I suggest it is useful to consider oral historian GAMMERL's (2015) work on the changing emotional trajectories of gays and lesbians in rural west Germany between 1960-1990. In close analysis of a recorded interview, GAMMERL showed that even subtle sounds like breath can be emotionally significant, as much as overt expressions such as crying, which are more likely to be noted in a transcript. GAMMERL was interviewing a man who had tested positive for HIV and had concealed his homosexuality and diagnosis from his wife and child. The interviewee began to sob at a point in the interview where he described a wonderful experience of solidarity, after having been composed throughout his account of several grief-stricken experiences. When GAMMERL listened back to the interview recording, he noted that at the beginning of the account of the grief-stricken story, the interviewee "took a deep breath and cleared his throat with a characteristic noise before saying: 'This was the worst week of my life'" (p.157). With this intake of breath, the interviewee had held back emotion and tears to get through the story. It was only later, once getting to "firmer ground" in the conversation, that he was able to let go, and then sobbed. [30]

GAMMERL made the point that it is not always the most overt expressions of emotion, here sobbing, that indicate the most emotionally difficult period in the person's life and indeed, in the interview itself. Much subtler sonic cues can speak to the gravity of what an interviewee felt and is feeling. The emotional significance of the data can be more fully realized when researchers analyzing interviews can listen very closely to both the said and the unsaid. [31]

Another emotive aspect of interviews that can get lost in transcription is the emotional atmosphere, or sonic elements of what ANDERSON (2009) described as mood, feeling, ambience, tone, and other ways of naming collective affects (p.78). Some scholars have noted that often in guides on how to conduct interviews for research, atmosphere is the enemy. Surveying the methods literature, HALL et al. declared: "The priorities could not be any clearer: a time and place free of distraction, secured against interruption; no noise allowed" (2008, p.1023). Any noise, or atmospheric sound, risks distracting from the interview, and makes play-back and transcription difficult. [32]

Every researcher has come across a place in their transcripts that says "[inaudible]" or "[overtalking]." But increasingly, taking inspiration from long traditions of ethnography, interview methods are veering toward place-based approaches, such as walking interviews. Theoretical concerns about ecology and human relations with animals and things also invite scholars to think more deeply about decentering human subjects and attending to material relationalities. The inclusion of atmosphere and environment via audio recordings of interviews has a rich potential to generate important data that can extend sociological knowledge through an ecological lens. It is not only the interviewer and interviewee who co-create meaning. [33]

The anti-noise ethos of interview etiquette runs deep. HALL et al. found that while atmospheric sound, or "noise" is so much a part of interview recordings that researchers have harnessed technologies to edit it out, using sound filters and transcription to "exclude [them] from subsequent analysis" (p.1025). HALL et al. considered what qualitative data might look like if there was no muting of noise in audio or tidying of a transcript. They asked: "In what ways might background noises and incidental sounds, audible context, and interference, lend character to the interview and aid analysis? What would an interview receptive to noise and the everyday look and sound like?" (p.1026) [34]

To explore these questions, the researchers conducted walking interviews in urban spaces. They drew their conceptual framing from LEFEBVRE's rhythmanalysis (2004) and focused on what sounds did in the interview. Listening back to their audio recordings HALL et al. were struck by the agency of noise in the interview, directing the very route the walking interview took, as the interviewer and interviewee responded to "sirens, traffic control, music, ringtones, announcements, advertising, and the continual audible notice, before anything is seen, of what lies ahead or round the corner" (2008, p.1034). The noises also directed the topics of conversation, opening ways to speak about experiences of urban space. HALL et al. found that the noisy space also afforded a different power dynamic in the interview. Questions and answers did not fall on the silence of the "ideal" interview room but were drawn into the everyday hubbub of place. As they pithily summarized, "we want to emphasize the leveling effect that noise can have on relations of inquiry and the 'space' that it makes for everyday speech" (p.1035). In this study the shift away from an assumed good in the research process—here "quiet"—allowed for a deeper questioning of rote research practices. The agentic capacities of the environment, not just the interviewer and interviewee, had particularly rich outcomes for researchers who sought to understand inequalities in urban space; an insight that could carry over to diverse studies of situated experiences. [35]

Along with aural atmospheres, it is possible to think about the intervention of specific sounds as socially significant. Focusing on the intrusion of one (notorious) noise in particular—the leafblower—FLINT explored the potential of listening to interview audio as a "more-than-human methodology" (2022, p.521). FLINT said,

"The buzz of the leafblower echoed across the space of the college campus where I was conducting research, yet as an 'inaudible' sound, it is brushed over, cleaned up, erased in the transcripts, and from write-ups of the data and research" (p.522). [36]

So intrusive was the leafblower that it forced FLINT to rip her headphones off during transcription and demanded further critical attention. FLINT raised the issues of transcription, how the leafblower exceeded it and was "made inaudible" by transcribing sound into text. However, she was interested more in how the leafblower, as a more-than-human sound, when "heard, listened to, and encountered, offers a critical entry point for analysis [...] attending to the sound of the leaf-blower makes possible particular questions about the indexicality of sound and place" (p.523). [37]

What FLINT discovered by tuning into rather than out of the leafblower's vibrations was that they resonated with her inquiry into white supremacy on campuses in the US South. She concluded that the leafblowers were sonic embodiments of the gated community ethic of the campus, and often intervened into the interview, setting the interviewee into topics about the manufacture of image and brand. Listening to the leafblower became political, a way into the microaggressions and control of space, but also into conventions of how people listen, what gets heard, and what gets drowned out (p.537). The study offered a rich example of how (the sources of) specific sounds can be tracked for their meaning-making capacities within the interview. As with THOMSON, FLINT developed a method for disseminating sound. She brought the sonic onto the page and included links to audio files as well as visualizations of the sound wavelengths where the leafblower came into the interview. [38]

Wavelength visualization has also been used by researchers to significant effect in a study of nurses' emotions, demonstrating how aural analysis of emotions and atmosphere can be incorporated into text-based publishing formats. Deviating from the standard interview, COTTINGHAM and ERICKSON (2020) invited nurses to keep audio-diaries to document their emotions, especially "more private, unpopular, and spontaneous emotions" (p.549). They found that the waveform visualizations could not

"fully capture the sighs, deep inhales, uses of sarcasm, or many of the other subtle features of spoken diary entries that convey emotional meaning. They do, however, offer some insight into how speed, pitch, and pauses correspond to [...] levels of emotional energy that vary across time" (p.560). [39]

Brought together, the audio and the waveform visualizations also conveyed the sonic atmosphere of where and when the audio diaries were recorded, for example, "the background features of beeping medical equipment, music, traffic noises, sounds of participants walking, sighing, laughing, and crying" (p.561). The authors noted that such aspects usually frustrate transcription and are therefore cut, but in the audio, and in visual representations that seek to foreground aurality, these data became crucial. [40]

Nurses made audio diaries which spoke to, for example, feelings of stress and tiredness, evoked by the clamor of hospital noise, a reminder of the constant vigilance needed to monitor the edge between life and death, amidst the busyness of keeping a ward running at all hours. Had the interviewer arranged to meet in a quiet room for a conversation that could be seamlessly transcribed, these data would not have been registered. More of the emotional experience of the nurses, especially the atmospherics and environment in which their emotions were generated, came forward because audio was a primary site for qualitative analysis. [41]

Scholars working with such methods, attuned to the sonic reverberations that intermingle within environments, sounding relations between humans, animals, machines, changes in rhythms of times and seasons, movements between spaces, have the capacity to further attune us to ecological relations in social analysis (WHITEHOUSE, 2017). Such methods help researchers to bring further depth to the interview method by acknowledging a wider set of sensory influences in the co-creation of meaning. [42]

Just as transcribing interviews for coding and analysis is common practice in the social sciences, so too the usual modes for presenting interview findings are text-based, read with the eyes. In this paper, I have argued that there is great value—especially for emotions research—in shifting this practice to consider the aurality of the recorded interview itself, heard with the ears, as a primary site for analysis and dissemination. With shifts in the scale of projects, the distribution of labor, and the outsourcing of transcription, as well as the advent of AI and automated transcripts, it is an apt moment to revisit the "fetishized" or assumed status of transcripts in qualitative research. I have demonstrated that some of the initial rationale for working only from transcripts is no longer sacrosanct. It is not reason enough to foreclose the rich layers of data and ethical interventions that can register with listening to interviews. [43]

Theoretical shifts in qualitative research have sought to decenter the authority of the social scientist to generate a more ethical and reflexive practice. This decentering has become more crucial with a need to focus on the relational and ecological aspects of social life, and the implications of human-centered thinking and action. This approach might entail finding new ways to explore aurality within interviews, using different interview techniques, or analyzing the audio of interviews alongside or instead of transcripts (POMERANTZ & FEHR, 1997, pp.70-71). [44]

As I have explained, there are some methods that already attune to sound—conversation analysis, sociolinguistics, reconstructive biographical research, and oral history—and help scholars create detailed transcriptions to include information that will elicit aural, emotional, or embodied aspects of the interview. These approaches present a helpful toolkit for how sociologists might analyze recordings and the emotions and atmospheres that can be heard within them. I have argued that qualitative inquiry could also learn from interdisciplinary sound studies, particularly in thinking critically about methodologies for analyzing sensory data and the act of listening in research. Timbre, tone, pause, silence, and other aspects of sound might become rich foci for the analysis of emotion in recordings. [45]

Forging these links could inspire new developments in a qualitative inquiry geared toward the senses. Furthermore, by focusing on the act of listening in analysis the researchers could make the sensory, emotional and atmospheric dimension of the interview, and indeed the research process, more readily felt. It would not only increase the diversity of data, including emotion and environmental happenings, and thus show different dimensions of its production, but would also pose further questions about the role of listening as a relational and positioned mode of attunement and analysis. [46]

Auralities in research, as some of the examples I have drawn together illustrate, have the potential to include dimensions of emotional experience that are not easily put into words, as well as the agency of the non-human world, be it atmospheres or relations with environments too often quietened and excluded as "just background noise" in interviews. While the technology to search and code audio has been available to researchers for decades, they have stuck to ingrained habits and epistemologies and have kept the take up of aural methods at a lag speed. However, it is possible to shift such habits and experiment with listening to interviews, perhaps alongside reading them, as researchers build new repertoires of methods and methodologies to train hearing as a primary tool of analysis. [47]

It is with such experimentation that scholars can further a robust critical discussion about the challenges and benefits of aural methods. Oral history, sociolinguistics, reconstructive biographical research (ASHMORE & REED, 2000; AYAß, 2015; BAMBERG & GEORGAKOPOULOU, 2008; BECKER et al., 2023) and sound studies (FLINT, 2022; GAMMERL, 2015; THOMSON, 2019) can provide models for inspiration, critique, and adaptation. Researchers might begin to dust off old tapes, or call up long forgotten wav files, and listen to interviews resonant with emotions and atmosphere, in a move toward a more aural qualitative analysis. The next steps might then also include further mixing of methods, enlisting video and live analysis to more fully encompass the visual, the embodied, and the atmospheric, all of which are active in the entangled scene of qualitative knowledge production. [48]

My sincere thanks to the editors of this special issue and to the peer reviewers for their supportive and rich critique. Thanks also to Amy VANDERHARST for helpful reading suggestions on listening.

1) For an entertaining exception to this see HOLMES (2010), who documented how the annotations of a "cheeky transcriber" (p.152) intervened into her reading of the transcript, injecting humor, but also co-creating meaning by noting the action or irony in and around the interview. Here the transcriber becomes an unintended co-analyst, highlighting the interpretive and reflexive aspects of the transcription process itself (such wry asides are not yet offered by AI). <back>

Abrahams, Hilary (2017). Listen to me: A reflection on practice in qualitative interviewing. Journal of Gender-Based Violence, 1(2), 253-259.

Anderson, Ben (2009). Affective atmospheres. Emotions, Space, and Society, 2(2), 77-81.

Anderson, Kathryn & Jack, Dana C. (2016 [1998]). Learning to listen. Interview techniques and analyses. In Robert Perks & Alistair Thomson (Eds.), The oral history reader (3rd ed., pp.11-26). New York, NY: Routledge.

Ashmore, Malcolm & Reed, Darren (2000). Innocence and nostalgia in conversation analysis: The dynamic relations of tape and transcript. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 1(3), Art. 3, https://doi.org/10.17169/fqs-1.3.1020 [Accessed: July 25, 2024].

Ayaß, Ruth (2015). Doing data: The status of transcripts in conversation analysis. Discourse Studies, 17(5), 505-528.

Bamberg, Michael & Georgakopoulou, Alexandra (2008). Small stories as a new perspective in narrative and identity analysis. Text & Talk, 28(3), 377-396.

Becker, Johannes; Pohn-Lauggas, Maria & Santos, Hermilio (2023). Introduction: Reconstructive biographical research. Current Sociology, 71(4), 552-566.

Bokhove, Christian & Downey, Christoper (2018). Automated generation of "good enough" transcripts as a first step to transcription of audio-recorded data. Methodological Innovations, 11(2),1-14, https://journals.sagepub.com/doi/full/10.1177/2059799118790743 [Accessed: September 22, 2024].

Bornat, Joanna (2010). Remembering and reworking emotions: The reanalysis of emotion in an interview. Oral History, 38(2), 43-52.

Brownlie, Julie (2011). Not "going there": Limits to the professionalisation of our emotional lives. Sociology of Health & Illness, 33(1), 130-144.

Bull, Michael & Back, Les (2016) Introduction. In Michael Bull & Les Back (Eds.) (2016). The auditory culture reader (pp.1-28). London: Bloomsbury Academic,.

Chandler, Rasheeta; Anstey, Erica & Ross, Henry (2015). Listening to voices and visualizing data in qualitative research: Hypermodal dissemination possibilities. SAGE Open, 5(2), https://journals.sagepub.com/doi/full/10.1177/2158244015592166 [Accessed: September 22, 2024].

Clifford, James & Marcus, George (1986). Writing culture: The poetics and politics of ethnography. Oakland, CA: University of California Press.

Collins, Harry, Leonard-Clarke, Willow & O'Mahoney, Hannah (2016). "Um, er": How meaning varies between speech and its typed transcript. Qualitative Research, 19(6), 653-668.

Cottingham, Marci D. & Erickson, Rebecca J. (2020). Capturing emotion with audio diaries. Qualitative Research, 20(5), 549-564.

Craig, Shelley L.; McInroy, Lauren B.; Goulden, Ami & Eaton, Andrew D. (2021). Engaging the senses in qualitative research via multimodal coding: Triangulating transcript, audio, and video data in a study with sexual and gender minority youth. International Journal of Qualitative Methods, 20, 1-12, https://journals.sagepub.com/doi/full/10.1177/16094069211013659 [Accessed: September 22, 2024].

Crichton, Susan & Childs, Elizabeth (2005). Clipping and coding audio files: A research method to enable participant voice. International Journal of Qualitative Methods, 4(3), 40-49, https://journals.sagepub.com/doi/full/10.1177/160940690500400303 [Accessed: September 22, 2024].

Davidson, Christina (2009). Transcription: Imperatives for qualitative research. International Journal of Qualitative Methods, 8(2), 35-52, https://doi.org/10.1177/160940690900800206 [Accessed: September 22, 2024].

Daza, Stephanie & Gershon, Walter S. (2015). Beyond ocular inquiry: Sound, silence, and sonification. Qualitative Inquiry, 21(7), 639-644.

Derrida, Jacques (1978 [1967]). Writing and difference (transl. by A. Bass). Chicago, IL: University of Chicago Press.

Flint, Maureen A. (2022). More-than-human methodologies in qualitative research: Listening to the leafblower. Qualitative Research, 22(4), 521-541.

Gallagher, Michael (2020). Voice audio methods. Qualitative Research, 20(4), 449-464.

Gammerl, Benno (2015). Can you feel your research results? How to deal with and gain insights from emotions generated in oral history interviews. In Helena Flam & Jochen Kleres (Eds.), Methods of exploring emotion (pp.153-162). London: Routledge.

Glaser, Barney G. & Strauss, Anselm L. (1968 [1967]). The discovery of grounded theory: Strategies for qualitative research. London: Weidenfeld and Nicolson.

Green, Judith; Franquiz, Maria & Dixon, Carol (1997). Myth of the objective transcript. TESOL Quarterly, 31(1), 172-176.

Guschke, Bontu Lucie (2023). Fire inside me—Exploring the possibilities of embodied queer listening. Culture and Organization, 29(6), 564-581.

Hall, Tom; Lashua, Brett & Coffey, Amanda (2008). Sound and the everyday in qualitative research. Qualitative Inquiry, 14(6), 1019-1040.

Hammersley, Martin (2010). Reproducing or constructing? Some questions about transcription in social research. Qualitative Research, 10(5), 553-569.

Holmes, Katie (2017). Does it matter if she cried? Recording emotion and the Australian Generations Oral History Project. The Oral History Review, 44(1), 56-76,

Holmes, Mary (2010). The emotionalization of reflexivity. Sociology, 44(1), 139-154.

Howes, David (1991). The varieties of sensory experience: A sourcebook in the anthropology of the senses. Toronto: University of Toronto Press.

Ikonen, Essi (2022). The sonic meanings of the life and death of my disabled brother: Listening as a method for qualitative research. Qualitative Health Research, 32(13), 2019-2029.

Jackson, Alecia Y. & Mazzei, Lisa A. (2009). Voice in qualitative inquiry: Challenging conventional, interpretive, and critical conceptions in qualitative research. London: Routledge.

Kingston, Sarah (2024). Audio research methods, attitudes, and accessibility theory: Using audio vignettes to elicit attitudes towards sex work. Qualitative Research, 24(3), 647-668.

Lavee, Einat & Itzchakov, Guy (2023). Good listening: A key element in establishing quality in qualitative research. Qualitative Research, 23(3), 614-631.

Lefebvre, Henri (2004). Rhythmanalysis: Space, time, and everyday life. London: Continuum.

MacLure, Maggie (2009). Broken voices, dirty words: On the productive insufficiency of voice. In Alecia Y. Jackson & Lisa A. Mazzei (Eds.), Voice in qualitative inquiry: Challenging conventional, interpretive, and critical conceptions in qualitative research (pp.97-113). London: Routledge.

Mazzei, Lisa A. (2009). An impossibility of full voice. In Alecia Y. Jackson & Lisa A. Mazzei (2009). Voice in qualitative inquiry: Challenging conventional, interpretive, and critical conceptions in qualitative research (pp.45-62). London: Routledge.

Mero-Jaffe, Irit (2011). "Is that what I said?" Interview transcript approval by participants: An aspect of ethics in qualitative research. International Journal of Qualitative Methods, 10(3), 231-247, https://doi.org/10.1177/160940691101000304 [Accessed: September 22, 2024].

Moore, Niamh (2012). The politics and ethics of naming: Questioning anonymisation in (archival) research. International Journal of Social Research Methodology, 15(4), 331-340.

Nascimento, Liandro da Silva & Steinbruch, Fernanda Kalil (2019). "The interviews were transcribed", but how? Reflections on management research. RAUSP Management Journal, 54(4), 413-429.

Poland, Blake D. (1995). Transcription quality as an aspect of rigor in qualitative research. Qualitative Inquiry, 1(3), 290-310.

Pomerantz, Anita & Fehr, Barbara J. (1997). Conversation analysis: An approach to the study of social action as sense making practices. In Teun A. van Dijk (Ed.), Discourse as social interaction (pp.64-91). London: Sage.

Richardson, Emma; Haworth, Kate & Deamer, Felicity (2022). For the record: Questioning transcription processes in legal contexts. Applied Linguistics, 43(4), 677-697.

Saldaña, Johnny (2015). Coding and analysis strategies. In Patricia Leavy (Ed.), The Oxford handbook of qualitative research (pp.581-605). New York, NY: Oxford University Press.

Stonehouse, Paul (2019). The unnecessary prescription of transcription: The promise of audio-coding in interview research. Research in Outdoor Education, 17(3), 1-19, https://digitalcommons.cortland.edu/reseoutded/vol17/iss1/3/ [Accessed: September 22, 2024].

't Hart, Dorinda (2023). COVID times make "deep listening" explicit: Changing the space between interviewer and participant. Qualitative Research, 23(2), 306-322.

Tessier, Sophie (2012). From field notes, to transcripts, to tape recordings: Evolution or combination?. International Journal of Qualitative Methods, 11(4), 446-460, https://doi.org/10.1177/160940691201100410 [Accessed: September 22, 2024].

Tilley, Liz & Woodthorpe, Kate (2011). Is it the end for anonymity as we know it? A critical examination of the ethical principle of anonymity in the context of 21st century demands on the qualitative researcher. Qualitative Research, 11(2), 197-212.

Tilley, Susan A. & Powick, Kelly D. (2002). Distanced data: Transcribing other people's research tapes. Canadian Journal of Education, 27(2 & 3), 291-310.

Wainwright, Megan & Russell, Andrew (2010). Using NVivo audio-coding: Practical, sensorial and epistemological considerations. Social Research Update, 60, 1-5.

Werner, Cosima; Meyer, Frank & Bischof, Susann (2023). Foundations, strategies, and techniques of anonymizing transcripts in qualitative research: A practice-oriented introduction. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 24(3), Art. 12, https://doi.org/10.17169/fqs-24.3.4067 [Accessed: September 22, 2024].

Wettergren, Åsa (2015). How do we know what they feel?. In Helena Flam & Jochen Kleres (Eds.), Methods of exploring emotions (pp.115-124). London: Routledge.

Whitehouse, Andrew (2017). Loudly sing cuckoo: More-than-human seasonalities in Britain. The Sociological Review, 65(1), 171-187.

Barnwell, Ashley (2025). Listening to interviews: Attending to aurality, emotions, and atmospheres in qualitative analysis [48 paragraphs]. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 26(1), Art. 6, https://doi.org/10.17169/fqs-26.1.4345.

Forum Qualitative Sozialforschung / Forum: Qualitative Social Research (FQS)

ISSN 1438-5627

Funded by the KOALA project

Creative Commons Attribution 4.0 International License