Hello, and welcome, to the ratcliffe institute, for advanced, study at harvard. My name is claude rizzini, i'm the executive, director of the fellowship, program. The radcliffe, institute, brings together, students, scholars, and practitioners. From across disciplines, and professions, to create and share transformative. Ideas. You can participate, in this laboratory, of ideas by attending public programs such as this one, visiting virtual, exhibitions. Or pursuing, the special collection. Held at the schlesinger, library. To learn more about radcliffe, please feel free to sign up to receive information, on news, and events, at. Radcliffe.harvard.edu. We'll begin the program with the presentation, by tina tallon. After the presentation, the speaker will respond to questions from the audience. Please use the q a feature on zoom to submit your questions, at any time throughout the program. We ask that you keep your questions, brief, to allow. Us to address, as many as possible, in the time that we have together. Tina tallon, is a composer. A computer, musician. Vocalist, and scientist. She learned to play the piano and the violin, as a child. She went on to study biological, engineering, at mit. And spend the early years of her career, researching, the biomedical. The biomolecular. Basis, of pancreatic, cancer, and endometriosis. In 2010. She received a commission, from the league of imaginary. Scientists. She was asked to write the sound for a science, and art exhibit, relating, the surface of bacteria, on planet, earth. To that. And that of a close-up, views of planet, mars. The exhibit, went on. View, at powerhouse, museum, in sydney, australia. And later at the museum of contemporary. Arts in los angeles. Music has since been widely performed, internationally. By many ensembles. And musicians. Her first string quartet, selected, defrosting. Won grand prize, in the parma, student composer, competition. And her dissertation. Lucinia. Won an ascap, morton, gold, young composer, award. She was also the recipient, of a palo, and diamond, general, commission. To support the composition. Of new work for violinist, composer, and composer, kurt road. For viola, and live music viola, and live, electronics. Talon, is active, not only as a composer, and passionate, advocate for new music, but also as a vocalist. Educator. And art documentarian. She frequently, performed, as a soloist. With the mit. Chamber, chorus, and concert, choir. And other groups. Additionally. She served, as assistant, to the, artistic, director, of the boston modern orchestra, project. And an assistant producer, for the grammy-winning. Bmop. Sound label. She also served, general manager, of the san diego's, fresh, sound, music series. Academically. Her research, interests, include, technological. Mediation. Of the human voice. Virtual, tactility. The relationship, between summer static, and music cognition. Computational. Modeling of energetic, relationship, between various, musical parameters. Based upon newtonian, mechanics. Development, of software, for spectral, analysis, and composition. Algorithmetric. Composition. And computational. Approaches to musical. Musicological. Inquiry. Tina holds an sb, degree. In biomedical, engineering, and music from mit. And then mfa. In composition. And music theory from brandeis, university. She is currently, completing, her phd, in composition. At the university, of california. San diego. A radcliffe, tina is working on shrill. An electro-acoustic. Chamber, opera, that will examine the ways that gender bias, in the development, and regulation, of voice technology, has shaped its history. Everything, from microphones. To modes of transmission. Has been optimized, for lower voices. And the gender invective, directed, at women, in the media, has changed very little since the dawn of the broadcast, era. Taught from the perspective, of four female, identified. Vocal laborers. A switchboard, operator. A radio announcer. A modern presidential, candidate. And a virtual digital, assistant. It will grapple, with questions, of embodiment. Virtuality. Agency, and identity, construction. The set will include, novice cultural, electronic, instruments. And interactive. Elements. Ultimately. Shrill, will entry. Will entreat listeners. To examine, their own biases. And inspire, advocacy, for voices. That are silenced, in our society. By, structural, inequality. And with that, it is my pleasure to give the floor to tina talan. Thank you so much claudia, for that very warm introduction. And to the radcliffe, institute, for advanced study for its support, of my research, and creative, practice. Today, we're going to be discussing. A little bit of the research, that is underpinning. What will eventually become the opera, that claudia, told you about.
Um. And it's also going to, be, manifesting. In a book, project. So i want to thank you all for joining me this morning. I'm coming to you from cambridge, massachusetts. Where i sit on the native land of the massachusetts. People. So for the past decade, there's been a lot of discourse. On technological. Bias, particularly, surrounding, algorithmic, justice. And of course these biases, aren't new. We've also seen a lot of recent research on racial bias. And the technological, development, of camera sensors, film and lighting equipment and i have a number of fantastic, colleagues who are engaged in that work. However while there's been significant. Attention, given to bias in the way that we discuss, vocal labor and its relationship, to power structures. There's actually been much less discussion. And research, on the deep-seated, technological. Underpinnings, of these biases. First i'd like to mention that while much of the discourse, reinforces, a gender binary, this is not just a problem that affects people who identify, as female. And of course our understanding of gender identity, and expression, has become much more nuanced, over the past century. So i desperately, hope that we can begin, incorporating, these more sophisticated. Understandings, of gender in our studies and analyses. Moving forward. So my interest in the subject, began from an early, age. I actually grew up in a pretty conservative, household. In which my father regularly, listened to rush limbaugh. So for me hearing the word shrill in reference to female politicians. Was, unfortunately, pretty commonplace. By the time i turned 18. In 2008. Yes i'm a 90s, kid. And i proudly registered to vote for my first presidential, election. I was pretty sick of shrill. But unfortunately, 2008. Also inc, coincided. With a resurgence. In the word shrill. With hillary clinton's, first run for president. An article, in the columbia, journalism, review caught my attention around the same time entitled, shrillery. And one of the, subtitles. Is. Is clinton's, problem as basic as her voice, and this of course, led me to wonder is it as basic as, her voice, and how much more is there to this argument. Um. Now of course we all know by now that there is but the focus on shrillness, hasn't exactly, subsided, in the past decade, uh here's a clip of democratic, strategist, julie roginski, in an exchange with tucker carlson. Back in 2016. During clinton's second, run. You have a bunch of men. Both running against her and also in the pundit class. Who are saying things about her that she's shrill that she's shouting maybe she's true maybe she's shouting what does that have to do with because, when was the last time you had a man be accused of being shrill, that is such a code word, a lot of women. I don't think i've ever heard a man be called. No man on television, has called hillary clinton shrill. Because that is a felony offense in. Everybody. Nobody uses the word shrill. And i love i think at this point all of us can think of at least one or two thousand. Examples. Now regardless, of your politics, that's a much longer conversation, that we will surely be grappling with in the, particularly, on the left uh over the next four years, uh we can't deny that the voice is the site of so much when it comes to identity, construction, both individually. And societally. And as we gain a better understanding, of the history of words such as shrill, are we simply replacing, them with different words such as likability. And electability. So before we dig into this history, i'd first like to, uh, begin by asking the question. What is, sound. Um. There though we all experience, it in different forms on a daily basis, i don't think it's safe to assume that we all, share the same understanding. Of how it works or even what it is, while there's certainly philosophical. Considerations. Here at play, hence the tree having fallen in a forest. For the sake of time i'm going to begin by focusing on just the technical ones that'll be relevant to our discussion, later. So sound in essence. Is the vibration. Of particles, in a medium. So here we have a nice little, uh. Illustration. Of, some of the waves. That govern the movement of these particles, in various media and of course these media can be solid, they can be liquid, and they could be gaseous, such as the air that we breathe, every day. Now there are two words that kind of are relevant, in this discussion, one of which is frequency, which is related to how fast these particle, moves, these particles, move, this is related, to what we would perceive, as pitch of a sound. An amplitude, which is, related, to how far the particles, move and that is loosely related.
To. Volume, well there's an asterisk there and we'll talk about some of those loose relations. In a minute. One of the ways that we represent. Sound, as audio engineers, and researchers. Is through the use of waveforms. And these graph, amplitude, of the sound, over time and if you zoom in on one of these i'm sure you've all seen a waveform, at some point. You can actually see these tiny vibrations, being graphed, and if you imagine one of those, little red particles. That were in our previous illustration, moving back and forth. If we turn that on its side you now can see these vertical. These vertical traces, of its motion. Now. No, talk about the history of voice technology, would be complete. Without, listening to one of the first ever recordings. Of the human voice. So this, is, a, um, what you're about to hear is, what's called a phone, autograph. Or a phone autograph, a graphic, transcription. Uh which came from an inventor named edward leon scott's. Invention called the phonatograph. Which, comes from 19, 1857. Excuse me. What this was was essentially, a giant. Uh. Diaphragm. Or this thing that looks almost like a balloon that collected, these vibrations, in the air, and then, via a needle. Would actually trace them onto, a piece of sooted paper and that's what you see here on the right. And both of these i've gotten from. A website that does an amazing, job of studying and archiving, these early sonic, transcription. Devices, called firstsounds.org. So this is a, um. A phone autograph, of, edward leon scott himself. Singing, uh the french, song. Uh eau claire de la. Luna. So this one dates back from, april of 1860. And you can hear that while we can't really quite make out, any sort of uh consonants. There are no fricatives, or sibilants, we mainly do hear just vowels. Um, you can make out some semblance, of the human voice, and you can see on the right, that some of these vibrations, are actually traced, into the soot on the paper. Another, way that we. Look at sound and that we represent, it is through the use of what is called a spectrogram. Now the spectrogram, takes into account not only amplitude, but also, frequency. So it looks at all of these different, speeds of vibration, of particles, in the air over time. So the spectrogram, that you see before, you is one, of, this. Audiophile. Which is uh, the unfortunate, recording, of me singing, three different vowel sounds, now what you'll notice is that if you were to go try to play what i just sang on the piano you would probably, press one key for all three of them that would be middle c or thereabouts. Um. And that key, that appraisal, of the quote unquote pitch of the sound, is what. Is. Resultant, from the fundamental. What we call the fundamental, frequency. And that's of course this lowest, line, on the spectrogram. And, this is also the segment of sound that we often associate. Um. With vowels, in human speech and singing, now there's obviously, a lot more information. In this graph this is everything above which is known as the overtones. And if these are in. Integer multiples, of each other they are known as harmonics, so for instance if you take your fundamental, if it's, 200, hertz and you multiply, it by two three four five and six you end up with these even distributions. Of horizontal, lines above the sound. Now you can also see that there's, significant. Variation. Between these three sounds even though they seem to be the same pitch. And that's why the overtones, are so important so it's all of this higher frequency, information.
That Allows us to distinguish, between different types of vowel sounds in human speech, or in the case of consonants. It's essentially. Pretty much conditioned, on everything, other than the fundamental. So, without. Uh, leaving behind our discussion. Of the types of technology. And representations. Of, uh, sound, we also have to take into account, the way in which humans, perceive, sound. And here i also want to say that this is what i'm about to describe is just one way that humans perceive, sound. People perceive sound in a lot of different ways that don't necessarily. Involve. The human hearing apparatus. As pictured, so of course, the, vibrations, in the air travel through your auditory, canal. They. Slam, into your tympanic, membrane or your eardrum, and then a series of organic, transducers. Then, allow. Those vibrations, to be transmitted, and transduced, into electronic, signals that our brain. Then, interprets, as sound, in this specific, paradigm. Of course there are many other ways. That, we experience, sound and people who are deaf, may not necessarily, experience, sound in the same way. And i want to make sure we give voice to that. Now. In terms of that specific paradigm that i just mentioned. I also want to be clear that we don't always perceive, frequencies. The same way. So this is what's known as the fletcher monsoon, equal loudness, contour, or a set of the fletcher munch than equal loudness, contours. Um. And these are the result, of. Research, that came from. Two researchers, in the 1930s. Named harvey fletcher and wilden munson. Um, and basically they wanted to gain a more detailed, understanding, of how the human hearing apparatus, processes, different frequencies. Um i prefer to look at this graph. This way so, upside down. Because here we can kind of talk about. The hearing range in terms of our ear sensitivity. So above the top on the horizontal, axis we have frequency. And then along, the side. On the vertical axis we have, uh sound, pressure level, uh, or intensity, in decibels. And so essentially you can see that we have more sensitivity. In sounds between, 1000. 1000, and approximately, 7000, hertz. And this sensitivity. Falls off. Below that and above it. So, what does this mean. Um, this means that we're essentially, more sensitive, to higher frequency, sounds. And if you don't believe me we can do a nice little example. That was 200, hertz, and here is 2 000 hertz. And i can pretty much guarantee, you, that, you found one of those to be a little bit more piercing, and most likely it was the 2000, hertz, tone. Now here's a spectrogram, of both of those sounds you can see that uh first of all these are sine tones so there's not a lot of other, uh spectral, or frequency, information, in the sound, and also they're clearly the same volume as shown by, the the, um. Same coloration, here. And so. Um. What this means why is this, you know, why is this relevant. Well, um. Evolutionary, biologists, have hypothesized. That of course we're a little bit more sensitive, to higher frequency, sounds because that's where a lot of warning, calls. Uh exist, in terms of frequency, space. Um. There's selective, pressure. For. Organisms, that can more quickly respond, to, these types of sounds. Also acoustically, we're better able to localize, higher frequencies, because the wavelength, of those waves, of those sound waves are shorter. So again, more, accurate. Uh. Appraisal, of these sounds. More quickly allows for better snap decisions, when survival's, on the line. So. Now that we've got a little bit of a basis. In sound. Here i'd like to, discuss some of the history of the technology, that's involved. Many people peg the beginning, of voice technology. Or at least electrically, mediated, voice technology. In 1876. With the, invention. Or rather the patenting, of the telephone, that's a much longer story for the another time, but i think it's actually really important to begin this story about two decades earlier in 1853. And the reason for this has to do with the intersection, of voice technology, infrastructure. Surveillance, capitalism. And policing. Some of you may have recently, seen. Sarah cooper's netflix, special, everything's, fine. Which if you haven't seen it yet i highly recommend, it it captures. The current time that we're living in like nothing else that i've seen before.
And One of the sketches. That it begins, with, is, a faux kenburn, style documentary. About the history, of voice technology, and policing, called karen's. Narrated by whoopi goldberg. And of course it talks about the invention of the telephone, as a way for white women, to more efficiently. Report the presence of african americans, in their community. In boston. And while it may seem like sarcasm. Uh and a nice little bit of comedy, it's actually not too far from the truth. So in, 1858. Sorry, 1853. Uh, reverend augustus, r pope just up the road for me in somerville, massachusetts. Patented the electric burglar alarm, unfortunately. He was a reverend and not a businessman. So he didn't exactly make bank with it however. Uh edwin, holmes. Came along and bought the patent off of him in 1857.. So. Um. Augustus, r pope unfortunately. Passed away from typhoid, that year, um. And, and edwin, holmes went and launched on a career of trying to sell. Uh burglar alarms. Now there wasn't a huge market for them in boston because unfortunately there wasn't a lot of crime, uh and so no one really wanted to go through and invest, in this infrastructure. Uh also people didn't really trust electricity. Yet, um electric light bulbs, uh didn't really come, into, prominence, until the early 20th century, so. People were really hesitant, to kind of electrify. Their, uh window, sashes, and door frames. Um. So in 1859. Holmes moved his operation to new york city. And he called new york city in in his writings where he talks about this move he said. Um. It is the city where every thief in the nation made his home. Which. Unfortunately. Has some white nationalist, and anti-immigrant. Undertones. Actually. Um. And one of my colleagues, dr leslie harris has a fantastic, book entitled in the shadow of slavery. African-americans. In new york city, between, 1626. And 1863. And i'd actually really encourage you to read, uh some of her research. On. What life was like. During that era. So throughout, the, 1960s. The popularity, of the electric burglar alarms spread and soon, holmes began setting up central monitoring, stations in order to ensure prompt service. Um, and he even actually set up his own, private security, force. And of course, when we look at the reconstruction. Era there are absolutely, racial undertones, to this as well. Um, so in 1877. Uh he said hey things are going great in new york city, how about i send my son edwin t holmes back to boston to copy this system, there, now in, 1860. Uh sorry 1876. We of course had, the invention, of the telephone, or the patenting, of the telephone. And the establishment. Of the american bell telephone, company, in boston. And so, when, edwin. T arrived, in boston. There was already infrastructure, starting to be laid for a telephone, service throughout the boston metropolitan. Area, and so, what edwin, t's, innovation, was, is that he said hey, people aren't using telephone, lines at night so how about we, kind of do a little bit of uh, of a service share on the telephone, lines. At night to monitor, our burglar alarms. Uh and so we kind of have this. This weaving together of the infrastructure. Of surveillance, capitalism. And, telephony. So. Unfortunately. Um the infrastructure. Was, not great. In 1877. In 1878. Um, and so, there were a lot of switchboard, operators, that were needed to be able to monitor, all of these different exchanges.
Um. In. 1877. We see here a picture on the left of what a switchboard, operation, room may have looked like and it was staffed by primarily. Young white boys. This led to a lot of issues particularly, given some of the bugs in the system. Apparently. Young boys would frequently, swear at the customers, who were often irate because of delays, or denials, of service, apparently this even led to some physical altercations. Uh and this was really bad for business. Um, and so, in 1877. Edwin, t said you know what i think i'm just going to start hiring young women because of course they're much more docile. And their customers, are going to be less likely to swear at them. Um, of course, he only hired primarily. U.s born upper to middle class white women for this work. And so, white women became a critical, human component, of the bell system's, infrastructure, in terms of ensuring. Both customer, satisfaction. And profit. Profitability. So. Scholar, venus, green, uh who in her book race on the line, uh. Details, an enormous, amount, of uh, extremely, detailed research, on the bell system and she actually used to work for the bell system before becoming a professor, i highly recommend reading that book if you're interested. Um, she outlines, a lot of, the, uh parameters, involved, in the bell systems hiring practices for operators, so, in 1910. 93.7. Of the bell system, uh percent of the bell system operators were us-born, white women and they were mostly. Young women. Some reasons, for exclusion. From hiring, included. Race. A refusal, or inability, to work weekends and holidays, which was often tied to religion. Jews were excluded. And then the apparently the most common reason was quote unquote unacceptable. Accents, and here we get into. The policing, of female voices. So, this is from a switchboard. Operation, conference in 1906. In which one of the trainers. Notes that the training of the voice to become soft. Low, melodious. And to carry well is the most difficult, lesson an operator. Has to learn. Now something else very interesting happened in 1906. As well and this is where we get kind of the dawn of broadcast, radio in the united, states. So this is in december 21st, i believe is when one of the first radio broadcasts, was actually transmitted, from brant rock in massachusetts. And by the time that we start, getting, popular, adoption of broadcast, radio is one of the primary, means of information, dissemination. We also start getting a lot of criticism. Of female voices, on air. Um. So some of the criticisms. Include their voices, are flat or they are shrill, and they're usually pitched far too high to be modulated, correctly. And these, types of criticism, go on and on and on and on, and on and on and i don't have time to get into all of them but you can absolutely, see the tenor of a lot of these. What's interesting, is that of course people would have been used to female voices, uh, coming, through various types of speakers, for almost 50 years now. And suddenly, this has become problematic. So why is that, this article published in 1928. In scientific, american, sheds a little bit more light on the issue. So, the particularly, relevant, here, uh thing here is comes from dr j.c steinberg, of bell labs which is one of the preeminent. Actually one of the only. Audio technology, research institutions, in the united states. We don't have time to talk about monopolization. At a t. And trust busting, but that also plays an enormous part in the story. So one of the things that was published here is that secondly, the speech characteristics. Of women, when changed to electrical, impulses, do not blend with the electrical, characteristics. Of our present day radio equipment. So that's quite curious so i decided to start digging into this so what, what specifically, were the electrical, characteristics. That didn't blend, and, how do they how did they have such far-reaching, implications. So here's a brief from the, bell laboratory, record in 1927. Uh by the aforementioned. Dr, j.c, steinberg. And here's one of the bits of information that's particularly. Salient, so he notes that the elimination. Of all frequencies. Above 5000, cycles, affects only slightly the interpretation, of masculine, speech. For women's voices such elimination, would produce considerable. Degradation, and interpretation. And it is estimated, that frequencies, as high as 7000, must be transmitted, to give possibilities, of interpretation, corresponding, to a man's voice. Okay. Uh. That's, sure that's a great statement to be able to keep. Tabs on, but he takes it a step further and he thus asserts that it it appears that nature has so designed woman's speech that it is always most effective, when it is of soft and well modulated, tone. So unfortunately. Regardless, of the fact that there was a clearly, articulated.
Uh Technical, problem that they knew about. Rather than fixing it they just blame, uh, female voices. For causing the problems as opposed to the technology. For the types of limitations, that it imposed. So why 5 000 hertz why couldn't they just fix it you know if they knew that it wasn't working. Why didn't they do anything to counteract. That. Well the radioactive, 1927. Played, a big role in this so as radio stations. Um. As radio stations popped up all across the united states during the 1920s. Unfortunately, we ended up with a lot of signal interference, particularly, in urban areas there really was. It was the wild wild west there was kind of no regulation, in terms of who could start broadcasting, on what frequencies. So the radioact, allotted, 10 000 hertz of bandwidth, to each station. Keep that in mind because. Amplitude, modulation, radio, or am radio as many of you are most familiar with, actually, ends up doubling the bandwidth, of the signal so we have our baseband, signal which is, situated, here at the microphone, it goes through a bunch of equipment, that modulates, it and then sends it out of the radio tower so, if we have a 10 000 hertz limit on the bandwidth of our modulated, signal it means that we need to have a 5 000 hertz limit on the baseband, signal. So here's where some of, these limitations. Have been applied to the equipment. And of course. You know once you have a limitation, like that people start. Uh designing, all their equipment to cater to it and it becomes set in stone. Now there's another issue at play here that, that involves, measurement. In terms of what may have informed, the decision, to go with 10 000 hertz there are a lot of different. Types of factors that were involved in that decision but one of them has to deal with faulty research. And i'd like to invoke, something that you might be familiar with. If you've seen the hbo, docu-series, chernobyl, or have ever read any of the books. And that is the concept of 3.6, rondon, which uh one of the supervisors. Said oh that's not great but that's not terrible. As you see in the picture on the left it turns out that 3.6, runkin was actually. The limit of detection, for the equipment that they were using, and when they brought in a. More industrial, strength dosimeter. They found out that the levels were actually greater than 15. 000. Runkin, which is actually, really really terrible. Um. And this of course was, an enormous, tragedy, that led to. Thousands, of deaths, and, decades, of, uh health problems for people in the area and it's one of probably the largest environmental, disasters. In human history. So the importance of good measurement, devices. And clear communication. Is extremely, important. So let's talk about some of the measurement devices, being used. In, the early, history of voice technology. So this is a picture of one of the early condenser, microphones, that may have been used for auditory, research. And unfortunately, we find out that this couldn't actually. Accurately, measure anything, above 5000, hertz. So here's a paper from the physical review in which they actually, outline, calibration. Procedures. And take early data with this. And we find out that, beyond, 5000. Cycles per second the energy, is so low, as to be impossible, of measurement with the apparatus. Used. So, here's some of the data that they got from this microphone. Which is graphed over here on the right side of the slide, um. Also one thing that's funny about this is they only used. Six different subjects. To test their voices and take data from, four of which were male two of which were, female and then they later, average these results, together. This is something that frequently, happens, where they just. Test. Significantly. Fewer. Women in these studies. And so unfortunately, we're not represented, well in the data either. And of course you can see that these graphs, uh. They stop at 5000. Hertz, so there's just simply no measurement being taken, of all the frequency, information. Above that, um. So essentially they knew that they couldn't detect it. They didn't care they, may have known that you know as equipment got better it was important but unfortunately, things had kind of become. What we call path dependent, or you know a standard gets set and unfortunately, that sticks around the qwerty keyboard, is actually another great example, of path dependency. Um, especially, when it shows up on your phone because of course thumbs, are very different from typing with 10 fingers.
So. In response to this early consumer, radio, in 1920, of course couldn't accurately, really receive, anything above 5000, hertz, uh and this led to a lot of issues, so what does this actually sound like. Um. I have a number of examples, prepared, for you, on the left we have the high voice. And on the right we have the low voice. Both saying the same sentence which i'm not going to tell you because i don't want to prime you for identification. So here is our first sound sample these are band limited, to represent, what. These voices may have sounded like on radios, in the 1920s. Thirteen. Thick thistles, start, five fires. Thirteen. Six thistles. Start, five, fires. And what you'll notice is that you can most likely pick out significantly. More. Consonants. In the lower voiced example, here, and if we zoom in on these spectrograms. Um. By the way i, have been focusing on 5000, hertz, we're not going to have time to get into it but spoiler, alert. Voice band frequencies, actually became limited, between, 300, and 30 400, hertz so it actually was even more reduced, from 5 000 hertz so that's what i've. Exemplified, here. And here we see, that, uh there's just simply less information. Now because, of the way that things have been limited if you look at the spacing, of those bands in the high voice vertically. You can see that there's much more space, in between them which means that we're just getting less information, for our brain to be able to go through all of these pattern matching algorithms, that it does. To identify, different vowels and consonances. Whereas there's much more information. In the lower voice. Here's what those files sounded like. Before. I process, them with kind of the full fidelity, recording. 13. Thick thistles, start, 5, fires. 13. Thick thistles. Start, 5, fires. And presumably, it's much easier, to understand, what's being said, um, you know it's, hypothesized. That as long as we have, up to about by the way the human hearing range extends from about 20 hertz at the bottom to 20 000 hertz at the top that's why this scale, is uh. So much more expanded. And. With that extra information. You know at that point we're doing pretty good in terms of uh intelligibility. And articulation. Um, so we would think okay, we're probably, fine now right. Uh unfortunately. We're not because we still have to contend with a lot of things including, compression. Algorithms, which i'm sure you're all very familiar with after having spent the last year on zoom. Uh so. Here's a pictorial, representation. Of of compression. Uh basically we're trying to reduce the amount of data that's being transmitted, because, of course we. Need to be judicious, with our use of bandwidth. Um. And so this results in all kinds of artifacts. And the same thing happens with audio so i'm going to play you. Just the artifacts. From, a type of compression known as mp3, compression which i'm sure you're all familiar, with. What i've done is i've subtracted, out the original signal so you just hear, the things that result, from. Compression. So it might sound like a very garbled, uh, ghastly. Uh. Rendition. On those sound files that i just played you uh we started with the lower voice and then, ended with a higher voice, and what you may notice is that it's actually easier to make out what the sentence was in the second example, and you might think oh that's great you know that's not so bad right, that's actually really really bad because that means that the compression, artifacts, are, situated, in a frequency, range that interferes, with speech cognition. The more that we can hear the original signal that means it's more in the space where it's actually, interfering.
And I'm sure all of you, now at this point are very familiar with those types of artifacts, and of course, as we take into account, the types of, microphones. Uh the types of, earbuds. The types of bluetooth, compression, that are further applied to the signal. We end up with worse and worse results, a great example of this was the great yanni laurel debate of 2018. If any of you remember that. In which compression. On the internet was on full display and its effect on speech, cognition. Um, and of course, as i mentioned before there are a lot of other types of biases, that are technological. In nature that influence. How people are seen and heard, um, this is a what's called a shirley, card from kodak, in the 1970s. That of course. Showed how, their. Processing, and production methodologies. Were biased, towards, white skin. And of course, other researchers. Have been looking into, algorithmic. Bias. Joy bolemweeny, is an amazing researcher, at mit, who focuses. A lot of her work on computer vision and biases, and computer vision although her work has actually extended beyond that and she's formed something called the algorithmic, justice league, which now tries to. Hold different types of corporations. And institutions, to account in terms of the way that they are using algorithms. So what are some of the systemic, outcomes, of this well first of all i'd like to say that we did not have the first african amon uh, first african-american. Owned and operated, radio station until, 1949.. Um. African-americans. Were, predominantly. Only allowed on the radio, not as announcers, but as performers. Um, and so it was actually. It took a very very long time, for, um. For, not only markets but. For people to allow. African americans, to occupy, positions, where they could control the narratives that were being presented. On, air. Some of this also has to do with education. And access, in that. Many, white schools, were, investing in radio equipment and unfortunately, those same investments, were not made. In segregated. Schools. Also, this contributed. To, something, that a, scholar named cien. Ngai, has termed animatedness. Particularly when we look at racial exploitation. In performance. Um and she notes in her book ugly feelings that to be animated, in american culture is to be racialized. In some way. And so there's inherent, exploitation. Of african americans, and people of color in particular, in the media and this has stuck around since the dawn of the broadcast, era. Um. A few last points here this was from the democratic, debate in. 2019. Cnn interviewed a democratic, debate, coach on how the many women candidates, that we had this year could be, not be perceived as shrill, and she of course says to very purposely, slow your pace, and lower the tone a bit, because that will add meaning or gravitas. To whatever it is you're talking about. Um. So we're getting the same advice, that we have had, since, 1906. Right nothing has really changed, there are many studies. That have cited. Um the systemic, effects, on. Female voices that have occurred throughout history this one notes that there was an average drop of about 23, hertz, in, the fundamental. Of speech in young australian, women, over about 50 years. There are perceptions. Of. Trustworthiness. That are influenced, by voice pitch in gendered context, and of course we all are familiar with vocal fry. Right and the study in 2014. Uh, examined, how vocal fry may undermine the success of young women in the labor market. Now uh in 2019. The women's media center, did a report, on, racial and gender biases, in the media, and we see that these are still persisting. Right there's still. Vast disparities. Um, between women and men. In the media landscape. And then of course. Even. Even greater disparities. In terms of the incorporation. Of people in color, people of color in news rooms. And so we have a lot of work to do. So what, what do we do at this point right where do we go from here if it seems like we haven't been able to overcome, a lot of these technological. Problems, and it seems like maybe technology, is not catching up. What do we do. One important thing that we can do is support, women. And people with minoritized. Gender identities. Particularly, of women. Of color in audio production, roles. This begins by hiring. More women and particularly women of color in audio engineering, um.
A. Report from the audio engineering society in 2008, found that women make up only about five to seven percent of professional, audio engineers. And then a report in 2019. From ufc, usc, annenberg, found that women make up, 21.7. Of artists. 12.3. Of songwriters. And only 2.1. Of producers. And then within that, the 871. Producers, that they included in the study only, four of them were women of color that's, 0.46. Percent. So we have a lot of work to do, one thing that we can also do if we're working in the music or entertainment, industries. Uh is, insist, on inclusion writer some of you may have remembered, um. Francis mcdormand's, speech from the oscars in 2018. Which she ended with the two words inclusion, writers and so you can make demands that your production, team. Represents, an equitable, society. Right. Even if you're not in charge of casting. You can absolutely make sure that people incorporated, in that process, you can also get involved with education, and training organizations. Such as sound, girls, eql. Techny, women's audio mission. And audio girl africa, these all do amazing, work in providing, opportunities. For women and people with minoritized. Gendered identities. In terms of, you know placing, them with mentors, and providing, funding, and access to equipment. And of course if you're a politician. You can have an audio advocate, or an audio savvy staffer. This is someone who might know how to communicate, with audio engineers, and ask for specific, modifications. In clear, concise, terms, of course. Oftentimes, things are very cramped in terms of time. And so you want to be able to, quickly and effectively, make the modifications. That you need, and of course in general if you lead a research, or a design, team, think about the demographic, make of your of your team, does it reflect what you're trying to study. Or what you're trying to create. If not, maybe you need to go back and consider, who you're actually, involving. Uh in your research. Um, think about where you're getting your data and how it's collected, what are the types of biases, that are incorporated, at those stages. And then lastly what types of tools are you using for measurement, and what are the social histories, of those tools.
Um. Ultimately, we all need to be engaged, in critically, analyzing. All the things that contribute. To the systemic, disparities. And inequities, that we see not only in the media but in society. In general. So i'd entreat everyone, to question your biases. And question, never stop questioning the histories. Of the tools that you're using. So that we can all work towards. Trying to correct, some of the things that have been in place for a very long, time. So. Thank you, very much. I'm now going to turn it back over to claudia, for some q a. Thank you tina. Uh this was, uh, fantastic. Really really, uh thank you. We do have questions not. Surprisingly. And the first one is what are your thoughts, on elizabeth, holm's artificial. Voice lori. And how it shaped her public, image. Yeah i, imagined, that that was probably, going to come up, um. I had actually been asked about that in in some interviews, before. So the first thing i would like to say, is, that. When it came to the lawsuits. A lot of people actually cited. This voice that many people. Claim is performative. Um. As, evidence, of her untrustworthiness. And i would like to say that i. Resist that somewhat, because i think that again. That's engaging, in in this privileging, of the voice. As, a portion of someone's identity. Um. That, can be criticized. In these inequitable, ways, and so. Um i do hesitate. To kind of grab onto those things however i will say. That there. I've done some analyses. Um, i have some raw data, uh and actually raw audio from interviews that she did uh with npr i believe back in. 2005. Um. In which, the theory is that she had let her guard down. And had kind of adopted a more natural, tone and there was a disparity. Of, i think almost 100, hertz i'd have to go back and look at my data, i'm happy to pass that along if, whoever, asked that question if you want to send me an email i can show you my analysis. Um. And so you know kind of if you take averages over the course of someone speaking, you can look at where, the kind of confidence, intervals, are. Between, what a, seemingly, normative. Mode of speech is and what a. Potentially, performative, mode of speech is. So, um. Yeah, there are definitely. Some disparities. Between what people say, is likely. Her natural voice and what is likely a performative, one. But i do hesitate. To enter into that mode of criticism. Just because it is. Highly gendered. And there's a very problematic, history of it. That's great now i must read this it says, what, a fantastic. Capital, letters, talk, thanks, exclamation. Mark. Thank you. Capital, letters. Can digital, processing, of female, higher. Frequency, voices. Modulate, them into a more high. Fidelity, transmission, frequency, distribution. Assuming, yes, is this a good path forward. Um. I wish i wish i could ask the questioner. Uh. A few questions about their question i realize that's not necessarily. Uh something that we have the ability to do right now. So. In terms, of. Being, able to secure. Higher fidelity, representations. Of a voice in terms of having, a, greater, amount of frequency, components, involved. Right. We're dealing with a lot of different, layers of mediation, here first of all we're dealing with a microphone. We're dealing, with, uh the sampling, technology, that we're using which is a digital, um. You know digital to analog converters, which, is. Something we didn't even get to talk about. We're dealing with software. Algorithms. Then we're dealing with kind of going back out, over the air we're dealing with streaming, we're dealing with uh, data compression.
Algorithms. Multiple layers of that right because especially if you're listening, on a bluetooth, device, bluetooth, adds an additional, layer of compression or it often adds an additional layer, of bluetooth, compression. On the already compressed signal, so. Even if you have an extremely high fidelity, recording. As kind of like your baseband, signal, there are still. Multiple, modes. Of, um. Signal, loss, or, signal interference. I'm not. Those maybe aren't the, best terms to use, um. But there's, multiple ways, in which, we have no control over what actually happens. And this is where and i think a lot of the problems, start. Um. Because again we talk about path dependency, right we talk about the the qwerty keyboard, um we talk about the ways that all of these things are set in stone, in some cases. Um. And it seems as if there's nothing that we can really do so you can start, always start with the best quality audio that you can possibly, get. But of course. It feels, in some ways as if there's not really anything we can do because we don't we can't control what happens to it after we send it out to whatever transmission. Modalities. It needs to go through, so i i'm sorry that that's not a particularly. Rosy. Outlook. On the situation. Um. That's what it is. Uh, the next question asks how is your operation. Going to engage. Uh the issues that you've covered, in your talk. That's a fantastic, question, and you're absolutely correct. Questioner, in that i did not talk about my own creative practice, at all in this discussion. Um. So i'm still in the early stages, of kind of working on the libretto, for the opera or the text that i'm going to be setting. Um, and so that's kind of why i'm in this research modality. But i my creative practice deals a lot with technical mediation, of the voice i'm, particularly. Interested. In ways that we can manipulate. Intelligibility. Uh using a lot of these same frequency, characteristics, that we've talked about i, use, uh, some. Uh algorithms called convolutional, synthesis, that allows me to look at the spectral profiles. Of things such as instrument, sounds or sounds in nature or whatever else and look at the similarities.
Between Human vocal sounds and figure out ways of interpolating, between them. So. This idea of kind of. Looking at the liminal spaces. Between human speech and identity. And other types of sounds that we're familiar, with. To try to draw connections. And, create context, that allow us to kind of, um find new ways of looking at things so that's one of the ways that i i deal with that, um. Also the opera, is going to include some sculptural, electronic, instruments. So there's a lot of these, i think that are pretty obvious, right switchboards. Uh modular, synthesizers. Use a lot of things that i think are pretty reminiscent, of switchboards, in terms of plugging cables in and out of different terminals. Um. And this i think is, by creating sculptural, instruments. We can physicalize. A lot of these things right we respond, as humans to things such as resistance. Right physical resistance. That i think, um. On a conceptual, level, we also experience, every day but maybe haven't physicalized, and so i feel like this idea of embodying. Some of these resistances, can be really cathartic, for people. A lot of my dissertation. Actually is focused, on. Electronic, music as a mode to de-personalize. In response to trauma and so this idea of taking something that's virtual. And turning it into something that's physical. Can i think be, really really instructive. And kind of shed new light on on some of these things. So i realize that's not entirely, specific. But. Somewhat outline some of the approaches that i hope to be able to take. Okay. Um. The next question, uh in a new york times, uh article, this past sunday, on cnn, reporter, abby phillips. So the writer, play, praised, abby philippe's, voice, and her speaking, style, slow, deliberate, currents, distinct, form. This thing from the ratatat. Verbal, spray, that has characteristic. News for a decade. Comment on that. Yeah, so as many. Of us who watched, uh i believe abby phillips was a contributor on cnn, uh over what was that two weeks ago now um. I. Made the mistake of having cable news on in the background, for what seven days straight, um, and i also found abby phillips to be an amazingly. Calming, presence, um, because, of her. Uh. Well i mean. Her, poison, everything about the way that she approached the material, but um. You know i do think that uh. She, does have a very calming, uh tone of voice and it's interesting. To note this right, um, particularly, when we see a lot of male anchors who often.
Kind Of yell, uh. This goes a little bit back to that clip that we looked at from fox news where, tugger carlson, is literally. Yelling. Literally, yelling. Uh, and uh julie roginsky. Is of course, maintaining, a very level tone of voice. Um. I've also seen uh kamala harris be praised, for her calm demeanor. And tone of voice and when we look at the way that. Uh, particularly. Female vocality. And particularly. Female vocality, of african-american. Women is analyzed, in the media there are definitely. Gendered and racialized. Notions, of what authority, looks like that we have to be very. Um. Very, aware of, in terms of what the discourse, is like i think this is also similar. To, times in which people have praised, um, african-american. Men for being articulate, right we see this with barack obama and it's actually extremely, problematic. To praise someone for that because basically what you're saying is okay you've defied, this stereotyped. Notion of how i think you should speak. Um. And so, while yes i do think that on the surface, right um. It's fine to note that of course. Abby phillips demeanor was very calm i do think we need to be aware, of some of the underlying, biases, that might be conditioning. The way that we react to someone's, voice. Yeah. The next question. Different languages, have different voices. In my native norwegian. My voice has a higher frequency, than when i speak english can you comment on this. Yes thank you so much for for bringing that up there are so many other things that i wanted to cover in this talk that i didn't get to, uh and language. And the role that language, different languages, play in terms of the spectral, profiles. Um, is absolutely. Something that i think, i definitely, need to. Discuss. More. In my work and research more in my work. There are some researchers. Out there who are looking at this and who are definitely, taking these more comprehensive. Approaches. To looking at the role of language, and the spectral profiles of different languages. On, different types of equipment. For instance, you know, audio technology, didn't just develop in the u.s, right germany, had a huge, uh hand, russia had a huge hand in developing, early audio technology, that in the united states we didn't get access to until after world war ii, um. And so, language, of course plays, very distinct, roles, on. How voices are perceived and they're also societal, notions that might be different, uh in different cultures about how the voice functions, in terms of, power authority, identity construction. Um. So, i can say i'm not, i'm obviously, not fluent in norwegian, and don't know i'm not an expert. And i'm not a linguist so i can't necessarily, comment in that specific, case, but i do think that it's important. To realize, that different languages, just have different spectral, profiles. And that different technologies. Will will. Will act differently, on these different spectral, pro, profiles. It's also interesting to note that in tonal languages, such as mandarin. We. Need to focus on different types of parameters, when we're doing this analysis, right because the actual, gestures, the frequency, gestures. Influence, meaning, so not only are we just looking at kind of a semantic context but the actual sonic. Um. Gestures. Do, also impart meaning and so, as something that is meaning bearing we need to give special. Um. Uh special attention to that in our analyses. Um, and yeah i think there's a lot more work that needs to be done on that for sure, so thank you for bringing up that point that's an excellent question. Um. Going back to the very last part of your talk. How do we reform, and democratize. Listening, and mediation, technologies. In practices. When resources, such as high-speed, internet connections, and pro-quality. Audio, gear. Are often prohibitively. Expensive. Yeah this is a really fantastic. Question and i think, especially, now in 2020, after we've all gone through almost a year. Uh or nine months now i guess, of, um. Having to deal with an intensely. Mediated, existence. Kind of. Entirely. Conditioned, upon our internet connections. Um. You know the. The importance. Of.
Access, Comprehensive. Universal, access, to high-speed, internet connections, has become more and more important not only for education, but for health care, um for information. I mean everything, now right, it's it's, pretty clear, that, one's internet connection. Um. Influences, not only how they come across to others but what type of resources, they have access to. And so i think that we absolutely, need to take care in making, sure, that. This infrastructure. Is. Is democratized. This infrastructure. Is regulated. I have hopes that we'll be able to. Roll back some of the things that happened with net neutrality, in the previous, administration. Although, that doesn't solve the problem right we still need better access. When it comes to public utilities. For, high-speed, internet connections. There's a lot of research being done there in terms of how we can, ensure. High-speed connections. Especially to rural populations. Similar to for instance how the united states postal service. Is required, to provide service, to people in rural communities whereas if we privatize, that system. That access. Could be compromised, and so if we kind of think about high-speed, internet access in the same way. We might potentially, find ways forward that could be more equitable. And then in terms of audio gear yeah this is absolutely, a problem. Um. You know there there are just certain costs, that are involved. Um, and, we certainly, can't expect, everybody, to be able to outlay those costs right i'm, using an external. I mean it costs 99. But for many people that's that's actually cost prohibitive, right, so i think in educational, context, i think schools should be providing. Good audio equipment, uh to students. Um. Employers, could potentially, do this as well, i think that could go a long way in helping so that private citizens don't necessarily have to bear the burden of that and i will also say that. There's often an element of toxic masculinity. Involved when we get into conversations. About gear. That's maybe a subject that i don't want to broach with two minutes left. Um. But yeah i would also like to see the culture, change around discussions, about gear, uh so it's less. Gendered, and those discussions, about kind of good solutions, for people. Are a little bit. Less discriminatory. Because that also becomes problematic. Well. Thank you. Thank you for these. Fascinating. Presentation. And uh definitely, thoughtful. Uh. Answered, and and your perspective, thank you for your perspective. Um, i want to thank you the audience for the terrific, questions. I hope uh you'll be able to join us for other radley, virtual, uh programs. You can find out about future programs. And watch videos, of parts of past events, at. Radcliffe.harvard.edu. And with that thank you again for joining today, and i'll see you soon. Bye. You.
2020-11-29