Perception vs. Acoustics in Voice Training の全文書き起こし

higher harmonic content, but perhaps missing those slowest components, the fundamental and and you know, even higher harmonics than the fundamental, is that the brain extracts the mathematical difference in hertz in frequency between those higher harmonics. 00:05:27.920 --> 00:05:34.240 Periodic sound being what it is, we would represent it with equally spaced harmonics in a frequency space. 00:05:34.399 --> 00:05:48.319 Um, I I think it's really interesting to look at that same question through a different lens, which is essentially, you know, that that Fourier transform is looking at a time domain phenomena that's you know, it's pressure changes over time, essentially. 00:05:48.480 --> 00:05:50.399 And um it's really interesting. 00:05:50.639 --> 00:05:57.839 I think in voice pedagogy, especially voice science, much less so, but in voice pedagogy, we don't spend a ton of time looking at waveforms. 00:05:58.000 --> 00:06:05.439 Um, Ingo, you you you introduce voice acoustics through the waveform in principles of voice production, which I think is amazing. 00:06:05.680 --> 00:06:13.759 And um, and then you rightly point out that waveforms are highly redundant, and so it makes sense to move into a sort of a summary mode to understand it. 00:06:13.920 --> 00:06:15.120 And that is absolutely true. 00:06:15.199 --> 00:06:29.199 But if one removes that fundamental frequency, you know, do a Fourier transform, zero it out in the spectral data, and do an inverse filtered uh transform back to a waveform, the the periodicity is still there. 00:06:29.360 --> 00:06:48.160 Like I think that's the interesting story with this question, is that you have to remove a lot of harmonics before you get to the point where the sound itself is either so chaotic or the sound itself is low enough in amplitude that we stop hearing that experience of the pitch. 00:06:48.399 --> 00:06:53.040 One of the things that I think is fascinating, and this falls apart at the edges. 00:06:53.120 --> 00:06:56.800 So if you really drive a cochlea hard, this is not true. 00:06:57.120 --> 00:07:10.560 But the pitch that we hear tends to be related to the frequency, sorry, not the pitch, the the tone color of the pitch that we hear tends to be related to the frequency range of the harmonics that we are presented with. 00:07:10.800 --> 00:07:15.839 And um, and so I think that that is the thing that tells me it's a phantom experience. 00:07:16.079 --> 00:07:25.759 Because if we were actually being presented with energy oscillating at that slower fundamental frequency, we would hear like a tonal color world that aligned with that frequency. 00:07:26.000 --> 00:07:27.360 So that's how I think about it. 00:07:27.439 --> 00:07:32.240 I and there's more than one way to even teach people this basic concept. 00:07:32.399 --> 00:07:34.000 Certainly just play them sounds. 00:07:34.079 --> 00:07:38.560 Like that's the way to get the idea across, and it freaks people out every single time you do it. 00:07:38.720 --> 00:07:46.560 Um, but I I personally, even though most people present it through a frequency domain lens, I think the fascinating one is a time domain lens. 00:07:46.879 --> 00:07:52.079 There's a a term that hasn't made it into our community very much. 00:07:52.160 --> 00:07:58.160 It's called the greatest common divisor between the frequencies. 00:07:58.319 --> 00:07:59.360 So let's give an example. 00:07:59.519 --> 00:08:11.600 If you had two frequencies, one was uh 400 hertz and the other one was a thousand hertz, and those are the only two frequencies you had, the common divisor would be 200 hertz. 00:08:12.000 --> 00:08:20.800 Okay, you can divide a thousand by two hundred and get uh an integer number, and you can take four hundred and divide by uh two hundred and get an integer number. 00:08:21.040 --> 00:08:26.639 So that means that the common divisor between 400 and 1,000 is 200. 00:08:26.879 --> 00:08:29.120 And that is what we hear. 00:08:29.519 --> 00:08:36.720 And uh and that is also what the pitch extraction algorithms that are available now everywhere uh try to do. 00:08:36.879 --> 00:08:52.000 They try to find the greatest common divisor, not just what you see in the spectrum, but going into that concept of uh the brain actually does that division, as you have said earlier, and produces that fundamental frequency. 00:08:52.159 --> 00:08:59.600 And this is why we still hear the fundamental frequency over the telephone, uh, where often frequencies below 300 Hertz are not even present. 00:08:59.679 --> 00:09:11.600 They're they're filtered out, but we still hear the fundamental or a baritone singing in front of an orchestra where probably very little of that lowest frequency information is even making it past the proscenium. 00:09:12.080 --> 00:09:12.480 Yep. 00:09:13.039 --> 00:09:14.639 Very brief uh addition. 00:09:14.720 --> 00:09:21.200 I I I think of it like Inga was just describing, that the common denominator, which would be the greatest. 00:09:21.279 --> 00:09:40.159 Uh and if you if you took a bunch of sine waves, if you're thinking in that version, that are there with the missing fundamental, and you do them additively, they will create a pressure wave that repeats at the rate of that missing fundamental frequency because it's the lowest or the greatest common denominator. 00:09:40.399 --> 00:09:48.000 So we perceive the periodicity of that sound wave still stays at the fundamental frequency rate, which we perceive then as the pitch. 00:09:48.159 --> 00:09:49.919 So same, same thing. 00:09:50.320 --> 00:09:55.120 Um and and as as been pointed out, our hearing is actually very weak lower. 00:09:55.360 --> 00:09:58.799 We you know, we perceive this actually comes up a little bit a later question. 00:09:59.039 --> 00:10:04.080 Our our we perceive vibration with two different senses, touch and hearing. 00:10:04.320 --> 00:10:17.840 And hearing is most sensitive from a musician's point of view in the top octave

of the piano, you know, two to four thousand hertz up in that area, and feeling is more sensitive around middle C, around 250 Hertz. 00:10:18.159 --> 00:10:31.759 So um we if you played, if you played that that fundamental frequency for most pitches that that many of us sing, it's barely audible, barely audible to our sense of hearing. 00:10:32.000 --> 00:10:40.960 So we're imposing the higher frequency content to that fundamental frequency in terms of timbre and other things in anyway. 00:10:41.120 --> 00:10:43.840 That's yeah, with another sense, yeah. 00:10:45.200 --> 00:10:46.960 Nice description, yeah. 00:10:47.840 --> 00:10:48.159 Great. 00:10:48.240 --> 00:10:49.759 Well, let's let's move to the next question. 00:10:49.840 --> 00:10:52.559 Uh uh Ingo, I'm gonna start with you if that's okay. 00:10:52.720 --> 00:11:00.320 Um the next question we've got is why do scientists use the term pitch extraction when determining fundamental frequency? 00:11:00.799 --> 00:11:12.080 I was actually angry about that for many years, because we all know that pitch belongs to perception and fundamental frequency belongs to measurement, acoustic measurement. 00:11:12.320 --> 00:11:15.840 But why do so many scientists maintain the word pitch? 00:11:16.000 --> 00:11:33.759 It's not only because it's a uh one-syllable word and easy to write, but it's also because they're trying to actually uh get at what we were just talking about, the fundamental frequency that we perceive as individuals, even though it may not be there. 00:11:34.000 --> 00:11:53.039 And so we have looked at all the algorithms uh that are currently available, whether it's uh uh Pratt or Vocevista or many others, and they all end uh do that uh uh measurement of the greatest common divisor that we've just been talking about. 00:11:53.200 --> 00:12:06.399 Um and so the the scientific community has gotten uh in tune with us, so to speak, to get the best they can with instrumentation from what we actually hear. 00:12:06.559 --> 00:12:14.399 And that's why I'm not so upset now that they use the word pitch, because they're trying to get to what we perceive. 00:12:15.039 --> 00:12:24.159 I I think there's some there's some edge cases that that make it tricky and really kind of poke at at what you're saying that pitch is a perceptual phenomenon. 00:12:24.559 --> 00:12:40.399 Because if you if you listen to a, you know, as as close to a sine wave as can exist in the natural world, like if you listen to a really simple oscillation um and change its intensity, the intensity of the output, our perception of what the pitch is is gonna change. 00:12:40.639 --> 00:12:44.960 And it'll bend up in one direction and bend down in the other direction. 00:12:45.279 --> 00:12:51.840 But as a physical thing, the way that a pitch tracking algorithm would measure it would measure no change. 00:12:52.080 --> 00:13:12.559 And so, like there are like interesting edge cases like that that, or think about a phenomenon like subharmonics, when those pop into a speaking voice, like there are liminal thresholds where you'll hear the higher pitch and then you'll hear the lower pitch, and like at what point do you move from one to the other? 00:13:12.639 --> 00:13:15.840 Is like I'm not even sure how one would answer that question. 00:13:16.159 --> 00:13:31.039 Well, we we just published a study on that, and uh we found that at least informally we hear it uh when uh the subharmonic is greater than 30 percent of the fundamental of the harmonic series. 00:13:31.279 --> 00:13:36.000 Well, uh if I knew that's when the lower pitch starts to come in and starts to dominate. 00:13:36.399 --> 00:13:40.000 I would have asked for more if I knew I was gonna get my questions answered. 00:13:40.320 --> 00:13:41.200 That's great. 00:13:43.759 --> 00:13:44.399 Some of your thoughts. 00:13:44.559 --> 00:13:45.200 Oh, no, sorry. 00:13:46.000 --> 00:14:01.840 I was I was just gonna say, Ingo, I um I I have a decent understanding of pitch tracking algorithms, and I I work really actively in Pratt, and I know, for example, the difference between like a raw cross-correlation approach versus a filtered autocorrelation approach. 00:14:01.919 --> 00:14:11.200 And and I understand that the the autocorrelation approach transforms things into a Fourier transform domain, right? 00:14:11.360 --> 00:14:24.399 And then it I think looks for the alignments of the harmonics like we're talking about, and then it inverse filters it, but then a raw cross-correlation is really sliding the waveform against itself, looking for alignment. 00:14:24.559 --> 00:14:25.759 Is that is that right? 00:14:26.080 --> 00:14:28.799 Yeah, I think that's that's exactly what's going on. 00:14:28.879 --> 00:14:29.519 Yeah. 00:14:30.080 --> 00:14:30.480 Cool. 00:14:30.639 --> 00:14:31.440 Thank you. 00:14:32.240 --> 00:14:34.559 Um, Ken, what are some of your thoughts? 00:14:35.279 --> 00:14:37.360 I'm I'm good with what we said already. 00:14:37.840 --> 00:14:42.000 A lot of this is above my pay grade, so let's move ahead. 00:14:42.720 --> 00:14:43.519 Fair enough. 00:14:43.679 --> 00:14:44.559 Fair enough. 00:14:44.799 --> 00:14:52.399 Um well uh let's uh uh um uh let's come back to uh to this next question here. 00:14:52.720 --> 00:14:55.200 Um and Ken, maybe I'll start with you here. 00:14:55.519 --> 00:15:02.639 Why are low frequency vocalizations perceived as rougher than high frequency vocal vocalizations? 00:15:02.960 --> 00:15:16.960 So if you're singing, if I use the word pitch, if you're singing a low pitch, a low sung pitch, and you have good chord closure, good rapid chord closure, you have a lot of high frequency spectral content in the signal that's going on. 00:15:17.200 --> 00:15:30.799 And high frequency spectral content, if it has higher numbered harmonics that are close enough together, they start beating against each other to our perception and introduce a buzziness, an auditory roughness. 00:15:31.039 -->

00:15:41.120 But it's it's a periodic roughness, not the roughness of asymmetry or pathology, which is a noise kind of roughness. 00:15:41.360 --> 00:15:46.639 So it's a it's a it's a we in the pedagogy world, we consider it a good roughness. 00:15:47.039 --> 00:16:00.639 When you sing in a high pitch, the spacing of the potential lower harmonics of that pitch are wide enough uh relative to our hearing sensitivity that they don't beat against each other. 00:16:00.799 --> 00:16:17.200 And those high, which you need to have auditory roughness from harmonic content, it has to be high, closely clustered and resonated harmonics that trend, you know, put in the signal these closely clustered beating against each other harmonics. 00:16:17.360 --> 00:16:31.679 When you sing a really high pitch, those that level of the uh spectrum is way beyond our hearing sensitivity, well off the keyboard, and it may also not be being resonated by any higher resonances of the vocal tract. 00:16:31.919 --> 00:16:41.919 So you just hear the more or less smoother lower end of the spectrum where the harmonics are well spaced, and you just get this pure sort of round, warm color. 00:16:42.080 --> 00:16:44.080 Thank you, Ian Howell, for that information. 00:16:44.639 --> 00:16:54.000 Whereas with the uh lower, so bases are really buzzy because they're singing low pitches with a lot of resonated, high clustered harmonic content. 00:16:54.240 --> 00:17:07.359 Sopranos are really smooth, and every voice type migrates from relatively buzzier low pitches to relatively smoother high pitches, and basses get from really buzzy to sort of buzzy. 00:17:07.519 --> 00:17:14.799 Sopranos go from sort of buzzy to really smooth, and everybody else is somewhat in between in terms of where their range lands with these things. 00:17:15.039 --> 00:17:17.440 So perceptually, that's that's how that works. 00:17:17.599 --> 00:17:23.920 And you can use that information pedagogically to actually indirectly train laryngeal function. 00:17:24.559 --> 00:17:32.799 If the output target goes from buzzier to smoother, that tends to invite the better function from the larynx. 00:17:32.960 --> 00:17:35.119 We get appropriate levels of closure. 00:17:35.440 --> 00:17:46.079 Uh the other piece to it is that shorter, thicker cords that close with more contact and for longer each cycle generate more high frequency content. 00:17:46.480 --> 00:18:06.799 Longer, thinner cords that have a shorter uh close quotient uh tended to generate a steeper spectral slope with fewer high spectral contact or less high spectral content, which also contributes to this buzzier to smoother uh across range uh percept that we that we have. 00:18:08.079 --> 00:18:10.720 Uh Ingo, uh what are some thoughts you had on that? 00:18:11.359 --> 00:18:13.599 Uh just to add to that, that's very, very good. 00:18:13.759 --> 00:18:16.559 Um, Ken, you you summarized it very well. 00:18:16.720 --> 00:18:24.000 Uh in the whole uh area of sound perception by humans, there's something known as a critical band. 00:18:24.480 --> 00:18:35.359 A critical band is a band of frequencies in which the cochlear uh mechanics can no longer isolate one pitch from another. 00:18:35.519 --> 00:18:51.279 Um and so if two harmonics or two any two frequency end up in one critical band, then what we perceive basically is how these two frequencies fight each other as opposed to uh hearing them individually. 00:18:51.599 --> 00:18:57.920 And uh critical bands can be well uh defined uh over a whole range of frequencies. 00:18:58.079 --> 00:19:09.440 Um I love uh the book that we no longer use, but uh Harvey Fletcher uh was uh a pioneer in our field. 00:19:09.519 --> 00:19:18.000 Uh he actually was the first one to establish an acoustics laboratory at uh Bell Telephone Laboratories after World War II. 00:19:18.640 --> 00:19:26.319 And uh his book is called Speech and Hearing in Communication, and he always puts the two together. 00:19:26.480 --> 00:19:33.119 He always has what we perceive with our auditory system together with what we produce with our vocal system. 00:19:33.440 --> 00:19:40.079 And um, anyway, critical band theory is uh important factor in in all of this discussion. 00:19:40.400 --> 00:19:52.880 I I've read that book cover to cover several times, and and beyond the information being not only interesting and good, and also being part of a really fascinating historical sweep, he is a hell of a writer. 00:19:53.200 --> 00:19:53.599 Yeah. 00:19:53.920 --> 00:19:57.920 My god, what like what lucid prose about complex ideas? 00:19:58.160 --> 00:19:59.440 Just beautiful. 00:19:59.759 --> 00:20:00.400 Yeah. 00:20:00.799 --> 00:20:01.200 True. 00:20:01.680 --> 00:20:12.720 Um, I as a as a performing artist, just to throw it out there too, that there's always the interesting question about like what why would we perceive auditory roughness in a voice at this pitch versus this pitch? 00:20:12.960 --> 00:20:30.799 Um and you know, we tend to use this shorthand, like Sundberg uses this in Science of the Singing Voice, where we say any two harmonics that would fall within, you know, basically a little bit narrower than an equal-tempered major third will will trigger auditory roughness. 00:20:30.880 --> 00:20:34.000 And that's that's a simplification, that's a good simplification. 00:20:34.079 --> 00:20:41.680 And we could understand that as the harmonics go lower in frequency, that critical band actually gets larger, and as they go higher, it gets narrower. 00:20:41.759 --> 00:20:48.079 So, like there, anybody who wants to jump off the plank into that world, there's more interesting stuff to look at. 00:20:48.400 --> 00:21:00.400 But singing with other people, like that opens up some real options because all of a sudden you can load your your own cochlea with frequency information that you couldn't make yourself. 00:21:00.640 --> 00:21:12.079 And anybody who has sung close harmony understands that like

there are these intervals that you will hit where it is as though the sound is a solid thing in the room. 00:21:12.480 --> 00:21:25.440 And it's be I think it's because you know the the human brain's response to the cochlea being stimulated within critical bands is like it's it's that's how we experience it. 00:21:25.519 --> 00:21:28.480 It would be like sweet versus sour, right? 00:21:28.559 --> 00:21:34.160 It's just like a a sense a sensory experience of that physical phenomenon. 00:21:34.319 --> 00:21:40.480 Um, but any anybody who has sung close harmony go sing a whole step with another human being in a stairwell. 00:21:40.640 --> 00:21:45.039 And like I feel like that gets the idea across almost better than anything else. 00:21:46.720 --> 00:21:47.440 That's great. 00:21:47.599 --> 00:21:48.160 That's great. 00:21:48.240 --> 00:21:49.599 I'll have to remember to do that. 00:21:50.079 --> 00:21:53.359 Um all right, let's let's go to our next question here. 00:21:53.519 --> 00:21:56.240 Um, uh because we're talking a little bit about buzziness. 00:21:56.319 --> 00:21:59.440 So I think this next question really kind of uh uh helps with this. 00:21:59.519 --> 00:22:03.440 Um And um Ken, I might come back to you to start here. 00:22:03.680 --> 00:22:12.000 How much of our pedagogical language conflates the sensory experiences that we informally describe as fuzziness? 00:22:19.039 --> 00:22:24.799 Conflation in I think involves conflating two different things at least. 00:22:25.359 --> 00:22:27.200 You can't conflate something with itself. 00:22:27.440 --> 00:22:34.000 So I think we're conflating our hearing sense of frequency with our feeling sense. 00:22:34.400 --> 00:22:38.319 So our vibratory senses peak around middle sea. 00:22:38.720 --> 00:22:47.839 And our our colleague that we co-teach with Chadley Ballantine has really brought this information to the fore from hearing impaired research. 00:22:48.079 --> 00:22:52.319 How do people with no hearing or very low hearing experience sound? 00:22:52.480 --> 00:22:57.839 They experience it as vibration and it peaks around 250 Hertz and then it tapers off. 00:22:57.920 --> 00:23:00.880 And above a thousand hertz, we cannot feel vibration. 00:23:01.359 --> 00:23:03.440 We can only feel vibration up to a thousand hertz. 00:23:03.599 --> 00:23:05.759 That's roughly soprano high C. 00:23:06.079 --> 00:23:11.119 We're just beginning to get into the strength of our hearing perception at a thousand hertz. 00:23:11.200 --> 00:23:16.559 It really peaks somewhere probably near around 3,000 hertz in the top of the piano. 00:23:16.720 --> 00:23:22.960 So the difference between those two perceptually causes us to conflate those experiences. 00:23:23.200 --> 00:23:37.359 So for example, singers generally, particularly non-treble singers, basses, baritones, and tenors, and even the lower octave of treble voices, need high frequency content in their spectrum to carry over orchestras. 00:23:37.920 --> 00:23:38.960 Why do they need it? 00:23:39.119 --> 00:23:43.759 Because high spectrum content in the top occur of the piano is where our hearing is sensitive. 00:23:44.000 --> 00:23:49.920 And if we have that there, it'll cut over the spectrum of the orchestral background and so forth. 00:23:51.440 --> 00:24:05.200 When that is present strongly in, say a tenor's high note or a bearitone's high note, we feel vibration that we think we're feeling the content of that high frequency content. 00:24:05.680 --> 00:24:06.480 We are not. 00:24:06.960 --> 00:24:10.799 We are feeling strengthened low frequency content. 00:24:11.440 --> 00:24:15.759 But we associate it because it's correlated with that high frequency content. 00:24:15.920 --> 00:24:20.720 Because if you strengthen any part of the signal, my understanding, please correct me if I'm wrong on this. 00:24:20.880 --> 00:24:24.319 If you strengthen any part of the signal, it's like the whole signal is stronger. 00:24:25.039 --> 00:24:36.960 If you strengthen the high part, so we we feel the low frequency content buzzing away when that high frequency content is present, and we conflate and think we're feeling the high frequency content. 00:24:37.440 --> 00:24:44.640 We're not, but it's still actually useful biofeedback because it's there when we get a strong signal. 00:24:44.960 --> 00:24:47.119 And so we conflate those things. 00:24:47.359 --> 00:24:50.480 Um that's at least a start into this equation. 00:24:50.559 --> 00:25:07.519 Well, we're we feel low frequency content as feelable vibration, we hear high frequency content as a ringing, you know, a high buzzy timbre that Italians call squillo, and we have various terms in the history of pedagogy for it. 00:25:07.759 --> 00:25:09.920 At least that's a a stab at this subject. 00:25:11.599 --> 00:25:13.359 I I will make the next stab. 00:25:13.599 --> 00:25:17.759 Um, I I think I'll say a controversial thing. 00:25:18.000 --> 00:25:23.839 I think that a lot of voice teaching is really challenging to falsify. 00:25:24.240 --> 00:25:31.599 And uh and this is one of the things that I think is kind of a home run, is what Ken is talking about. 00:25:31.680 --> 00:25:35.359 And so, for example, I there's no way to gather evidence about this. 00:25:35.440 --> 00:25:38.400 This is just my lived experience as a singer and a voice teacher. 00:25:38.720 --> 00:25:58.160 But any voice teacher I've encountered who has tried to get a singer to experience some sort of forward vibratory sensation, and you know, we can certainly talk about whether that's a good pedagogical approach or not, but they always use bright vowels for that. 00:25:58.319 --> 00:26:05.119 So they'll use an an A or an E kinds of sounds, like snarly sounding vowels. 00:26:05.359 --> 00:26:17.279 And I think people have this association that the feeling is caused by the brightness that is present in those vowels that is not present in like an ooh, for example. 00:26:17.519 --> 00:26:20.079 And

um, and it's it's exactly backwards. 00:26:20.240 --> 00:26:34.400 It's actually the the monumental magnitude of the first formants of those vowels that is almost certainly generating the sensory experience that is like a vibrotactile experience. 00:26:34.559 --> 00:26:44.559 Um, bridging these things is bone conduction hearing, like that it's a it's a fascinating bridge between uh airborne hearing and our sense of vibration. 00:26:44.799 --> 00:26:53.039 And part of me does wonder how much conflation there is in people's minds that they they think they're feeling, but actually they are hearing. 00:26:53.200 --> 00:26:56.720 Um, Chadley and I actually just finished up a study which looks at this. 00:26:56.799 --> 00:26:58.960 We were basically doing like audio audiograms. 00:26:59.039 --> 00:27:11.680 Like if you were to go to an audiologist and they would check your hearing, except we did it with um playing people's own voices back to them in various parts of their body with contact like vibration speakers. 00:27:11.839 --> 00:27:13.599 And um and it is really interesting. 00:27:13.680 --> 00:27:27.119 Like the threat, the threshold for feeling vibration as a sensation, like Ken said, is shockingly low in the frequency domain, given what I think most singers think is causing that feeling of vibration. 00:27:28.079 --> 00:27:50.240 And even lower than that is the actual resonance frequency of the tissues that are vibrating, uh not just the receptors, which you say are tuned to around 250, but the resonance frequency of buckle tissue and other tissue uh in our um system is more like 120 Hertz or so. 00:27:50.400 --> 00:27:59.279 Um and so when we do uh resonant voice therapy or anything with semi-occluded vocal tract, we always say it has to be easy and has to be buzzy. 00:27:59.440 --> 00:28:02.160 But it's good for speech because that's where we are. 00:28:02.240 --> 00:28:04.799 We're in 100 and 200 hertz. 00:28:04.960 --> 00:28:10.799 By the time you get up to 500 hertz, or as you mentioned, a thousand hertz, none of those sensations are there anymore. 00:28:11.200 --> 00:28:15.200 You've got to be really loud to have those sensations. 00:28:17.839 --> 00:28:18.480 That's great. 00:28:18.640 --> 00:28:20.079 I appreciate the discussion here. 00:28:20.160 --> 00:28:21.759 Uh let's uh let's go to our next question. 00:28:21.839 --> 00:28:24.160 And uh Ingo, I'd like to start with you. 00:28:24.319 --> 00:28:31.039 Um, why is vocal fry more of a perceptual phenomenon than a mechanistic one? 00:28:33.039 --> 00:28:59.920 Well, uh from what I have learned, again, from perception of low frequencies, is that when frequencies get somewhere 70 or 50 hertz or lower, then um those frequencies aren't perceived by our brain anymore as places on the cochlea, but rather as individual events that are sent uh to the auditory nerve and finally to the brain. 00:29:00.079 --> 00:29:11.200 In other words, what you hear is just pup, pup, pup, pup sounds, individual

sounds that aren't necessarily connected in a continuum the way normal sound is. 00:29:11.359 --> 00:29:18.160 And and that the the breakover uh region there is around 70 hertz or so. 00:29:18.559 --> 00:29:44.400 And but that means that you can have many, many mechanisms that produce the same thing, as long as there is the perception of sound, gap, sound, gap, you can use all kinds of different um mechanisms inside in the vocal folds and still produce that same effect, that same feeling of gap, sound, no gap, or sound, no sound, the sound, no sound. 00:29:44.720 --> 00:29:47.920 And that's why I don't think it's a mechanism. 00:29:48.160 --> 00:30:06.000 And uh the people who put M's in front of mechanisms like M1 and M2 and so forth use M0 for Fry, and I have not yet been convinced that there's any one mechanism that produces this, but there are many, many mechanisms that give you that perception. 00:30:06.400 --> 00:30:20.720 I think there's um in the in the metal screamo distorted singing practice world, there there's also really interesting use of the term fry. 00:30:20.960 --> 00:30:35.279 Like it it nests as a compound term, so there's something called a fry scream, for example, that um would definitely be considered to be pitched, that has nothing to do, I think, with with what most of us would call vocal fries. 00:30:35.440 --> 00:30:42.559 So, like I would say even uptake of the term itself is is fracturing at this point in a really interesting way. 00:30:42.720 --> 00:30:47.519 Interesting because it points us towards new new sounds maybe we didn't know the human voice could make. 00:30:47.680 --> 00:30:49.519 Um, I want to ask you a question though, Ingo. 00:30:49.759 --> 00:31:13.359 I I always wonder about this like how much of that 70 hertz-ish cutoff is because of the periodicity, and how much of it is because the nature of the vocal tract when excited is to dampen rapidly and especially dampen higher frequency energy, probably more rapidly than lower frequency energy. 00:31:13.599 --> 00:31:15.359 How much of it is the gap? 00:31:15.599 --> 00:31:30.559 Which is to say, if my vocal tract were like a wine glass and I pulsed at 70 hertz and the resonance just kept going, you know, for several seconds, would we still perceive it in the way that we perceive it now? 00:31:31.359 --> 00:31:48.000 Well, in terms of the limited experiments that I did, and that was many years ago with uh Anat Kedar, um what we did is we simply changed the length of the gap where there was no sound uh and left everything the same. 00:31:48.160 --> 00:32:02.240 Uh in other words, the pulse that uh that was created by collision of the vocal fold and then the resonance stopped and then we changed the gap, and that's how we determined 70 hertz was the transition frequency. 00:32:02.480 --> 00:32:20.240 But you're right, if uh if you figure out a way of not dampening uh the energy in the vocal tract easily, even at 70 hertz, you could still have a continuous sound, and that probably then would not end up being fry-like. 00:32:20.400 --> 00:32:30.400 Uh so you could have normal uh vibration um in whatever register you want to call it, maybe down to 50 hertz or even lower. 00:32:30.799 --> 00:32:31.359 Very cool. 00:32:31.920 --> 00:32:35.680 So I I I think your your question is right on target. 00:32:37.519 --> 00:32:39.599 And I'd like to get your thoughts on some of this. 00:32:40.160 --> 00:32:41.039 Very, very few. 00:32:41.200 --> 00:32:47.440 First of all, ignore if you're right, if there's no mechanism there, then maybe the zero is an appropriate mechanism zero. 00:32:47.599 --> 00:32:48.640 There is no mechanism. 00:32:49.920 --> 00:32:51.039 Okay, I get it. 00:32:51.279 --> 00:32:52.799 Zero means no mechanism. 00:32:52.960 --> 00:32:53.200 Okay. 00:32:53.599 --> 00:32:53.759 Right. 00:32:54.000 --> 00:32:54.559 I got it. 00:32:54.880 --> 00:32:55.359 Right. 00:32:55.680 --> 00:33:01.359 So the other one was uh you can take a click, a spaced click. 00:33:01.440 --> 00:33:03.359 I I I make this comment. 00:33:03.519 --> 00:33:07.759 I say pressure variations are experienced by humans on a on a continuum. 00:33:07.920 --> 00:33:12.000 If they're really slow, we experience that as weather. 00:33:12.240 --> 00:33:12.559 Yeah. 00:33:13.119 --> 00:33:18.880 If they are if they are abrupt but infrequent enough, we perceive that as rhythm. 00:33:19.759 --> 00:33:24.880 But if you get them close enough together, then we start perceiving it as pitch. 00:33:25.440 --> 00:33:45.519 So it may be, I'd be very curious to see, because you we we've got a I've got a little app that does this where that Christopher Besh put together for us, actually, as a colleague, where he starts with clicks and he speeds up the frequency of the clicks, and you hear them as individual rhythmic clicks, and Benji gets fast enough, all of a sudden you start hearing them as pitch. 00:33:45.680 --> 00:33:56.480 And if you run that through a Fourier transform, you'll start getting frequency, frequencies appearing, but you don't get the frequencies from the individual clicks, even though every sound has frequency content. 00:33:56.640 --> 00:34:01.359 You just get this vertical striation, you know, and then anyway, it's very, very fun. 00:34:01.440 --> 00:34:06.720 So then, and then if you get them fast enough, fast enough, then they're supersonic and we can't hear them at all. 00:34:07.279 --> 00:34:11.840 So we go from weather to rhythm to pitch to supersonic. 00:34:12.320 --> 00:34:12.960 All right. 00:34:13.440 --> 00:34:13.760 Okay. 00:34:15.519 --> 00:34:18.079 I uh I'm I'm I'm really enjoying the discussion. 00:34:18.159 --> 00:34:21.920 So I uh we've got a couple more questions here that I'd like to get to here. 00:34:22.079 --> 00:34:28.400 So um let's uh maybe move away from some of the mechanistic discussion and talk about sound pressure. 00:34:28.559 --> 00:34:30.800 Um and Ingo, I'll I'll start with you again here.

00:34:30.960 --> 00:34:37.920 Why does sound pressure level not always correspond to perceived loudness? 00:34:38.639 --> 00:34:41.760 Well, I'd bring us right back to Harvey Fletcher. 00:34:41.840 --> 00:34:49.280 There is something known as the Fletcher-Munson curves, and they uh equal loudness curves. 00:34:49.760 --> 00:34:53.840 Um everybody in our field should know that diagram. 00:34:54.079 --> 00:35:02.639 Uh that should be the number one uh along with maybe uh uh formant uh frequencies or whatever of vowels. 00:35:02.880 --> 00:35:12.159 But it it shows that uh our ears are most sensitive to frequencies um around a thousand to three thousand Hertz. 00:35:12.480 --> 00:35:26.880 And um if you want to get the same loudness perception uh at lower frequencies and higher frequencies, then you have to give more sound pressure level to get the same uh effective loudness. 00:35:27.280 --> 00:35:34.719 Um so um but this is not how a sound level meter is calibrated. 00:35:34.800 --> 00:35:41.840 We it can be calibrated to have that same uh difference uh that we have from our auditory system. 00:35:41.920 --> 00:35:50.000 But most of the time um the sound level meters just gives us the amount of sound physically that is there. 00:35:50.559 --> 00:35:57.519 But our auditory system gives us basically what is known as the Fletcher-Munson curves. 00:35:57.760 --> 00:36:01.599 Um yeah, maybe Ian you can add to that or everybody else. 00:36:02.159 --> 00:36:17.760 I think it's a little bit like asking a question as to what color a flower is, and then it depends on whether you're a bird or a human, because you know, we we can't see ultraviolet, for example, and some colors radiate in beautiful ultraviolet color. 00:36:18.079 --> 00:36:41.679 And so the I think the entire world that we know is mediated through our senses, like like we everything is beyond a veil that we can't really pierce, and um, and so you know, the the Fletcher Munson Equal Loudness Contours is like a great example of this, um, because we shouldn't anticipate being able to hear certain frequencies. 00:36:41.840 --> 00:36:45.119 Well, this is so true, it's in the law, right? 00:36:45.199 --> 00:36:56.000 Like OSHA and NIOSH requirements for work sites, they can have much higher intensity sounds as long as they fall outside of the range of hearing sensitivity. 00:36:56.159 --> 00:37:04.400 Um, and there's there's a low frequency world going on outside right now that is cacophonous that we're just not we're not aware of and we'll never be aware of. 00:37:04.639 --> 00:37:16.559 Um, I I think I think this is another thing that voice teachers really need to know about, and not even from a mathematical model point of view, but just from a practical application point of view. 00:37:16.880 --> 00:37:31.519 Because if if you're a voice teacher and you teach in an eight by eight studio, you are significantly closer to your singer than you will be if they're an acoustic singer than you will be when they are on a stage. 00:37:31.760 --> 00:37:35.840 And proximity to a sound source, because the inverse square law, right? 00:37:36.000 --> 00:37:40.400 Proximity to a sound source, it's going to be more and more intense the closer you are. 00:37:40.559 --> 00:37:46.239 And one of the interesting things about those contours is that the more intense the sound is, the flatter they are. 00:37:46.480 --> 00:38:12.400 And so the implication of this is if you are trying to train someone to just like sing the bejees out of music on a stage and put out a lot of perceptible sound from a distance, and you are not overwhelmed by the warmth and brightness in the studio that will then be attenuated in the concert hall, then you're actually not training them to make the sounds that they need to make. 00:38:12.559 --> 00:38:30.079 And and so this is this is a strong argument, I think, and this is not a podcast about voice teachers, but but like it's a strong argument for teaching some of your lessons in a recital hall and sitting 40 feet away from people because those curves kick in and it makes a big difference than the sound a singer makes. 00:38:31.199 --> 00:38:37.199 And the and directivity of the sound is something very important, uh, particularly in the animal world. 00:38:37.360 --> 00:38:59.440 We just studied uh the vocalization of the P-haw bird, which is a uh small bird uh in South America, um lives high in the canopy uh area, and uh that bird um they have recorded of producing 110 decibels, presumably at one meter distance. 00:38:59.599 --> 00:39:02.079 I say presumably, but that's impossible. 00:39:02.239 --> 00:39:13.679 I went through and I computed the caloric intake of that bird, and and the it would the amount of wattage that you get out of the mouth is more than what that bird takes in during the thing. 00:39:15.519 --> 00:39:31.920 And so what's happened is people couldn't get close enough to the bird to really put a sound level meter right near its uh beak, but they uh uh uh uh extract it backwards, thinking that it was perfectly uh, you know, uh spherical sound. 00:39:32.159 --> 00:39:32.800 But it isn't. 00:39:32.960 --> 00:39:36.719 It was very directed, and there, therefore, they got that high number. 00:39:37.039 --> 00:39:39.920 But that that that also happens in in the in the hall. 00:39:40.079 --> 00:39:59.039 I mean, uh men say that why do the women always outdo me uh, you know, with the high frequencies uh back deep into the into the uh concert hall is because the high frequencies are more direct, they can be beamed, the low frequencies spread locally and you know don't reach very far. 00:39:59.440 --> 00:40:11.280 That's why if you're if you're ever staged to sing upstage, you better be like seven inches from the back wall and it better be made of plaster so that your sound will actually bounce out into the hole. 00:40:13.119 --> 00:40:13.840 And go ahead. 00:40:14.480 --> 00:40:15.519 A couple of comments. 00:40:15.840 -->

00:40:24.960 In addition to the threshold of hearing in that curve, which is a wonderful curve, the top of the curve is the threshold of pain. 00:40:25.119 --> 00:40:25.360 Yeah. 00:40:25.599 --> 00:40:32.400 So just right where the where the threshold of hearing is the most sensitive, in that sort of, I I say two to four thousand. 00:40:32.480 --> 00:40:39.199 I see that little, there's a little dip right in the pretty much the top octave of the piano, whether it's one to three or two to four. 00:40:39.440 --> 00:40:47.840 There's a dip in where it starts hurting if it's that intense, uh, which is actually how toddlers get their parents' attention. 00:40:47.920 --> 00:41:07.920 They really have a lot of high frequency content right in the area, and that has a lot to do with the dimensions of the uh ear canal, which is also a resonator, which boosts that frequency region from the outside world before it hits the eardrum and even raises it higher than it was in the air before it hits the eardrum. 00:41:08.159 --> 00:41:10.960 So, yeah, there's that as well. 00:41:11.119 --> 00:41:25.280 And as you mentioned, the high frequency content is much more linear, which is why the tenors and the the you know non-treble voices need a really strong high frequency component to zing out over everything in the hole. 00:41:25.360 --> 00:41:39.840 And I have heard reports of people saying as the tenor fans the audience like this and tries to hose them down with the sound, you can hear the loudness, the loudness per se swing by you with almost like a Doppler effect. 00:41:40.079 --> 00:41:40.719 Yeah. 00:41:41.679 --> 00:41:42.320 Yep. 00:41:43.519 --> 00:41:44.159 That's great. 00:41:44.320 --> 00:41:45.039 That's great. 00:41:45.199 --> 00:41:45.679 That's right. 00:41:46.000 --> 00:41:46.320 All right. 00:41:46.400 --> 00:41:48.159 Uh, a couple more questions here. 00:41:48.320 --> 00:41:50.800 Um uh Ian, I'll start with you on this one. 00:41:50.960 --> 00:41:51.280 Yeah. 00:41:51.519 --> 00:41:58.880 Um, what is acoustically weak may be perceptually strong, and what is acoustically strong may be perceptually weak. 00:41:59.119 --> 00:42:01.360 So, how does that affect voice training? 00:42:01.679 --> 00:42:03.440 Yeah, this is one of my favorite things. 00:42:03.599 --> 00:42:12.000 Um, we've brought it up a little bit when we were discussing uh uh vibrotactile sensation versus hearing sensation. 00:42:12.559 --> 00:42:39.039 Um, but you know, the the the more that I've learned about acoustics and dug into the scientific literature around it and you know considered what what is the transfer between the flow world and the pressure world, and what rates of change are actually generated in in the vocal tract and above the glottis as we phonate. 00:42:39.440 --> 00:42:50.559 Um, one, it becomes transparently clear that every acoustic recording that anybody has ever listened to has a high pass filter, right? 00:42:50.639 --> 00:42:52.159 So we should say that out

of the gate. 00:42:52.239 --> 00:42:56.079 And so we immediately lose pressure changes that are slower than that. 00:42:56.320 --> 00:43:12.239 But two, what becomes really clear is that the the slowest oscillating resonance of the vocal tract, what we'd call the first formant in a radiated sound, probably, as a physical thing is just a it's a high amplitude oscillation. 00:43:12.639 --> 00:43:41.599 And to the point that when we sing in a way where the second formant dominates and has a higher amplitude oscillation than the first formant does, like those are those are the edge cases, those are money notes, those are high belt sounds, those are you know Pavarotti or Corelli going up to their high C and doing a specific tuning thing so that they really have a standing wave that's generated by their second format. 00:43:41.840 --> 00:43:56.960 And so, um, as a physical thing, like I I try to understand how our equal loudness contours distorts our sense of the physical nature of the pressure pattern that the human voice creates. 00:43:57.199 --> 00:43:58.960 And so, like a E. 00:44:00.719 --> 00:44:16.239 Or even like a oh, I don't know if this will come through my zoom filter, but like any sort of overtone singing type thing where we hear a really piercing high frequency overtone, it is really present in our percept of the sound. 00:44:16.480 --> 00:44:33.840 But if you record that and then break out the different components and don't look in a spectrum, which has probably a bias of some sort that is boosting higher frequency components, but you just look in a waveform, the amplitude of that thing is physically small. 00:44:34.079 --> 00:44:52.000 And it's fascinating to me because if we're looking to leverage nonlinearities and we're looking for interactive qualities between what the vocal folds do in voice production and the response and the behavior of the vocal tract, like the stronger those interactions, the stronger the interaction is. 00:44:52.239 --> 00:45:08.800 And so I think a lot of the sounds that we think of as being strong because our hearing is so sensitive, are actually weak and probably don't have as much to do with sort of the ongoing contributions to self-sustaining oscillation as we think. 00:45:08.880 --> 00:45:16.639 And because our ears are insensitive to the slower components, we don't like we have to train ourselves to notice them to an extent. 00:45:22.960 --> 00:45:32.559 But I'm as a teacher, I'm on a kick at this point to just get everybody to notice the warm part of their sound, because the warm part of their sound is the physically strong part of the sound. 00:45:32.719 --> 00:45:40.639 And and if we want to build acoustic energy and we're looking to make power, like that's where joy is found, is in that part of the spectrum. 00:45:42.239 --> 00:45:44.000 I can chime in on that. 00:45:46.239 --> 00:45:55.599 The warm part of the sound, the lower part of the spectrum uh plays typically, in some sense, the first format, right? 00:45:55.679 --> 00:45:57.440 It is the first format content. 00:45:57.760 --> 00:46:06.320 And vowels are made of two different colors of sound as since uh individual frequencies have a tone color. 00:46:06.559 --> 00:46:16.719 You have a lower, warmer color and lower frequency sound, ooh and o like literally if you isolate them and play them, and higher frequencies have other vowel-like colors, high enough it just sounds like E. 00:46:17.280 --> 00:46:19.599 So you have this range of other colors. 00:46:20.320 --> 00:46:29.039 We most of our vowel targets in language come from the higher spectrum content from the second format contribution. 00:46:29.199 --> 00:46:29.360 Yeah. 00:46:29.519 --> 00:46:36.800 And I do a I do a childoscope whisper uh exploration that that is helps train resonance tuning. 00:46:36.880 --> 00:46:41.840 Because if you introduce a whisper, it has uh theoretically every frequency in it. 00:46:42.000 --> 00:46:43.760 It's like a sweep of frequencies. 00:46:43.920 --> 00:46:48.719 So it completely populates the transfer function of the vocal tract. 00:46:48.960 --> 00:47:02.639 And if you and you can hear, then in that case, your resonance structure and the formats you hear on the outside are the same thing because you've completely populated the resonance structure with spectral content and it radiates, and you can hear those. 00:47:05.360 --> 00:47:11.360 We hear the second form of contribution way stronger because our hearing is stronger at the higher end. 00:47:11.440 --> 00:47:27.599 So if I do an example, you can possibly hear the second form of content of the that series of vowels. 00:47:28.480 --> 00:47:31.519 But you can hear the lower one if I do ingressive clicks. 00:47:38.000 --> 00:47:39.679 And they're they're both there. 00:47:43.679 --> 00:47:46.719 And you can train your ear to begin to hear them both. 00:47:46.960 --> 00:47:52.239 Well, hearing that lower one, like Ian was saying, we tend to ignore it for two reasons. 00:47:52.480 --> 00:47:59.119 It's not the language sound we're going for, it's the complementary sound that completes the vowel and makes it sound nice. 00:47:59.360 --> 00:48:08.400 But the the the identifying color of the vowel is most for most vowels, except for ooh and o comes from that second format contribution. 00:48:14.800 --> 00:48:21.360 Not from the I call that I call the higher one the over vowel. 00:48:21.920 --> 00:48:24.960 I just for the studio I use terms that are friendly. 00:48:26.639 --> 00:48:27.679 Let's call it the overvowel. 00:48:27.760 --> 00:48:29.039 I could have called it the higher vowel. 00:48:29.280 --> 00:48:32.400 And the under the the one, the the other one I call the under vowel. 00:48:32.800 --> 00:48:49.360 And for the warmth, we need sufficient undervowel in the radiated signal to have a nice balanced kaudoscoro tamer, which is pedagogy term, historic pedagogy term for a balance of high and low frequency content, warmth and ring, warmth

and clarity. 00:48:49.760 --> 00:48:53.599 But but it's a trick to start hearing that that undervowel color. 00:48:53.679 --> 00:49:12.639 So normally if I do a most people with with good hearing that haven't lost hearing off the top, hear the and they don't hear the so Ken, why don't you just call it the uh higher color and the lower color? 00:49:13.039 --> 00:49:14.000 Yep, you could. 00:49:15.199 --> 00:49:19.199 But but it has a bowel-like color that they identify with. 00:49:21.920 --> 00:49:24.079 But you're right, it's the higher color and the lower color. 00:49:24.159 --> 00:49:25.280 I could have called it anything. 00:49:25.360 --> 00:49:26.320 It's too late in. 00:49:26.559 --> 00:49:41.920 I've already I'm already in print calling it the overval and the undervalued, and now that's spread out, they're spread out through the pedagogy world, but identify it as the second form of contribution and the first form of contribution so they know, you know, more scientifically what's creating it, you know. 00:49:42.159 --> 00:49:47.199 But in the studio, I don't tell my students, can you give me a little more second harmonica on your first format? 00:49:47.440 --> 00:49:50.480 I don't do that kind of thing to the poor student in the studio, you know. 00:49:51.119 --> 00:49:54.559 Can you give me a little bit more of that warmer vowel color, more than higher, you know, the brighter? 00:49:54.800 --> 00:49:59.280 I actually in the studio I try to talk about the brighter and the darker war uh colors. 00:49:59.440 --> 00:50:02.480 I give it we need more of the high, brighter and the low darker. 00:50:02.559 --> 00:50:04.880 So I use a variety of terms for that. 00:50:05.440 --> 00:50:06.880 But fair enough. 00:50:07.119 --> 00:50:09.039 I just, you know, I made up another term. 00:50:09.199 --> 00:50:12.079 I know you didn't care for that when I first presented that. 00:50:14.480 --> 00:50:15.599 Um all right. 00:50:15.679 --> 00:50:19.519 Let's uh Ingo, do you have any thoughts on uh on this topic that you'd like to share? 00:50:19.920 --> 00:50:30.239 Uh no, I think it was well well explained by uh our great, great uh teachers of singing of all styles. 00:50:30.320 --> 00:50:31.920 So much appreciated. 00:50:32.880 --> 00:50:33.440 All right. 00:50:33.599 --> 00:50:36.480 Um I'm gonna come to the next question here. 00:50:36.639 --> 00:50:42.400 Um uh and uh um uh uh Ingo, we'll we'll we'll start with you. 00:50:42.639 --> 00:50:54.480 Can we still conceptualize voice production through the linear non-interactive source filter model and then simply tack on ideas related to non-linearities? 00:50:55.519 --> 00:50:56.800 Generally, no. 00:50:57.840 --> 00:50:58.079 Amen. 00:51:01.280 --> 00:51:44.880 We can start there, but we don't end there, particularly now with the uh new uh sound productions that everybody wants to make with multiple sound sources, uh, for example, the two folds and the ventricular folds vibrating at the same time, and maybe even the aerial epiglottic fold, then you have interactions between these sources, uh, and you have the interaction of all the sources with the vocal tract, and that creates actually new frequencies because every time, uh, if you for a moment uh allow me to say a sinusoid is one of the frequencies, every time you multiply a sinusoid by another sinusoid of a different frequency, you get what are known as difference frequencies. 00:51:44.960 --> 00:51:48.079 There are literally new frequencies that are produced. 00:51:48.320 --> 00:51:50.880 And the linear theory will never give you that. 00:51:51.039 --> 00:51:57.760 I don't you can work all your life on it, it will never give you uh that difference frequency. 00:51:58.400 --> 00:52:06.320 And that's what a lot of the uh heavy metal singers now are counting on, is to have this enormous spectrum. 00:52:06.559 --> 00:52:19.360 For example, if you have two frequencies, um uh say one of them is 20 hertz and the other one, no, let's say one is 120 hertz and the other one is 100 hertz. 00:52:19.519 --> 00:52:22.000 The difference between the two is 20 hertz. 00:52:22.559 --> 00:52:30.719 In a linear system, you would never have that frequency produced, but in a nonlinear system that will actually show up in a spectrogram. 00:52:30.880 --> 00:52:32.480 It's physically there. 00:52:32.880 --> 00:52:39.920 And so uh we have to learn how to deal with the interactions and the nonlinear concepts. 00:52:40.000 --> 00:52:46.559 Uh it's uh sometimes frustrating, but uh it uh it it belongs to our world. 00:52:47.199 --> 00:52:49.039 I um is it okay if I jump in? 00:52:49.280 --> 00:52:54.960 Yeah, on this um I'm gonna say things that I think will make people mad. 00:52:55.280 --> 00:52:57.679 So I'll just preface with that. 00:52:58.000 --> 00:53:08.320 Um I I come at I come at this question from from like a voice pedagogy classroom teacher point of view. 00:53:08.559 --> 00:53:15.920 And you know, voice acoustic, one of the things I love about how Ken presents voice acoustics is it is at its core phenomenological. 00:53:16.079 --> 00:53:24.320 Like it's about sounds and it's about registration things that occur, which can then be explained through acoustical models. 00:53:24.480 --> 00:53:35.599 Um, but I think a lot of every year there's a group of graduate students in voice who go through voice pedagogy classes and sit through voice acoustics units. 00:53:35.760 --> 00:54:09.519 And I think come out the other side confused because I think that there are a visual vocabulary that arises out of the linear source filter model, which makes people imagine that the the discretized spectrum from the vocal folds and the spectral domain vocal tract transfer function, and then the radiated combination of those two things, that that is actually describing a characteristic of voice production. 00:54:10.000 --> 00:54:18.639 That the vocal folds are actually generating a series of independent sign tones, which are then

individually acted upon in the vocal tract. 00:54:18.880 --> 00:54:21.840 And we we see it all over our teaching models. 00:54:21.920 --> 00:54:37.039 People will say, oh, your pharynx is a Coke bottle and your mouth is a Coke bottle, and they're different pitched coke bottles, and that's why that's why when the pure tones go through those Coke bottles, you know, you get a first formant and a second formant, which I think as a model means well. 00:54:37.119 --> 00:54:41.199 I don't think that's really grounded in reality in a physical sense necessarily. 00:54:41.440 --> 00:54:49.840 And um, and I'm not really convinced that the voice pedagogy field has grappled with two quotations, which I would love to share. 00:54:50.000 --> 00:54:51.440 I'm gonna paraphrase them. 00:54:51.760 --> 00:55:17.039 One is Gunnar Font, if if you go to his book where he lays out the source filter model as as he amalgamated it, um, one of the things he says is uh paraphrasing, he says it's a quasi-scientific misunderstanding to imagine that the contribution of the vocal folds is a literal set of harmonic overtones. 00:55:17.519 --> 00:55:23.920 Um, and then he goes on to describe it in this mathematically elegant manner, as though it were, right? 00:55:24.079 --> 00:55:27.679 And that's the beauty of understanding both sides of that at the same time. 00:55:27.920 --> 00:55:40.239 Uh, if you go back to Helmholtz's work, his English translator, uh Alexander Ellis, uh, it's it's in the margins, unfortunately, like it's in his editor's notes. 00:55:40.559 --> 00:55:54.639 But at one point he he writes that um Helmholtz did not actually think that the vocal folds produced harmonic partials, which were individual sign tones. 00:55:55.440 --> 00:56:06.639 But as his theory of hearing presumed that that's how the ear processed things, it just makes sense to imagine it as though that is what the nature of voice production is. 00:56:06.800 --> 00:56:31.360 And like the deeper I've gotten into into voice acoustics and really trying to understand nonlinearities as a as a way to understand voice production itself rather than post hoc voice analysis, uh, I have just it has gotten harder and harder to literally imagine that the linear source filter model is accurate. 00:56:31.599 --> 00:56:37.199 And and so if that is true, what fills that space? 00:56:37.440 --> 00:56:45.280 Like what model could we use to simply introduce you know voice acoustics to voice pedagogy communities? 00:56:45.519 --> 00:56:48.000 I kind of think is the model you use in your book, Ingo. 00:56:48.159 --> 00:57:03.760 I mean, that's that's what I keep coming back to is understanding it as flow pressure in a time domain, and then very having a clear understanding of what it is that a Fourier transform does to that signal so that we don't reify the output of it. 00:57:04.400 --> 00:57:15.360 Uh, because otherwise we're we're thinking that harmonics are bouncing back when really it's just a resonance that's bouncing back in time or out of time with the oscillation of the vocal folds. 00:57:15.920 --> 00:57:20.000 So maybe I'll give it my voice scientist card now if that's entirely incorrect. 00:57:20.079 --> 00:57:22.800 But but I am I am curious about your thoughts. 00:57:25.039 --> 00:57:28.079 You want me to dive in before Ingo straighten this all out and gets your vocal? 00:57:28.320 --> 00:57:28.880 Yeah, dive in. 00:57:28.960 --> 00:57:29.360 Go ahead. 00:57:29.840 --> 00:57:39.599 So um, well, first of all, I don't think I I taught pedagogy for a number of years, 26 years when I was teaching at Lawrence. 00:57:39.760 --> 00:57:49.679 And I don't think I ever taught the linear source filter theory, because we've known about nonlinearity for quite a while. 00:57:49.920 --> 00:58:01.360 So I always, whether or not, you know, and again, some of this is above my pay grade because I I like to tell people uh very honestly, I'm not a scientist. 00:58:01.599 --> 00:58:04.800 I'm a voice teacher that tries to be science informed. 00:58:05.360 --> 00:58:10.320 So I will always yield to the scientist to straighten me out if I'm if I'm misrepresenting it. 00:58:10.559 --> 00:58:14.000 Second thing I want to say is these are models of explanation. 00:58:14.159 --> 00:58:22.159 And as we know, all the way from the statistician George Box, all models are wrong because you're modeling something that's too complex. 00:58:22.480 --> 00:58:28.800 The only thing that isn't that isn't that isn't an inaccurate model is the thing itself. 00:58:29.360 --> 00:58:31.119 And you're trying to explain the thing itself. 00:58:31.199 --> 00:58:36.800 So you have to do a model which deliberately simplifies and you know eliminates some information. 00:58:36.960 --> 00:58:39.039 So therefore, we teach models. 00:58:39.119 --> 00:58:50.320 So I always thought the nonlinear source filter theory as the way of explaining these phenomena, even though phenomena, even though realizing that's also got problems. 00:58:50.639 --> 00:59:08.559 Um and and uh to pick up on what Ian was saying, the problem with the terminology, one of the problems with the terminology, and I still use this model because I think it's very useful in certain ways of the source filter, is that implies there's an implicit chronology. 00:59:09.199 --> 00:59:10.960 Source filter. 00:59:11.599 --> 00:59:18.800 So you have this chronology, and we all know that the only thing the vocal folds do is open and close at the fundamental frequency. 00:59:18.960 --> 00:59:30.559 Now they can do that opening and closing with some variety, thicker or thinner, longer or shorter uh open or closed phase, uh uh greater or lesser contact quotient. 00:59:30.719 --> 00:59:37.199 So we have the closed quotient, the contact quotient, and the rapidity of the closure versus the opening. 00:59:37.360 --> 00:59:46.400 All of those affect the nature of that single fundamental frequency oscillation, which has an impact on the potential

spectrum, right? 00:59:46.639 --> 00:59:55.280 If you have more closure, longer, longer closure per cycle, and quicker closure, you're gonna have more high frequency content in the output. 00:59:55.440 --> 00:59:57.199 But what is that high frequency content? 00:59:57.280 --> 01:00:01.199 It isn't the harmonic series in a single pulse. 01:00:01.440 --> 01:00:04.320 In a single pulse, it's a noise. 01:00:05.679 --> 01:00:07.519 It's a noise sweep. 01:00:07.679 --> 01:00:15.760 And if you even in the Fourier transform, if you choose the right bandwidth, you will see an individual pulse is a resonance-shaped noise. 01:00:15.840 --> 01:00:29.119 It has the incomplete performance structure of the vocal tract because it's basically exciting the air in the vocal tract, which which gets excited relative to its own transfer function, its own, you know, resonance structure. 01:00:29.199 --> 01:00:32.719 So you'll get a series of pulses that have a resonance structure. 01:00:32.960 --> 01:00:44.880 What then is being periodically re-it doesn't introduce a set of harmonic frequencies, though they're they're included in the noise frequencies, they're not special at that point. 01:00:45.119 --> 01:00:58.559 And the only ones that are featured are the ones that land within the bandwidths of those resonances, the critical band that Ingo was talking about earlier, those get particularly strengthened, and the ones in between don't get much much help. 01:00:58.960 --> 01:01:06.960 Then that noise, that shaped, that resonance-shaped noise is reiterated at the fundamental frequency. 01:01:07.199 --> 01:01:26.800 And the periodicity of that fundamental frequency then enables only harmonic frequencies to survive in the radiated spectrum, but only those harmonic frequencies that previously landed in each individual pulse within the bandwidth of resonances of the vocal tract. 01:01:27.360 --> 01:01:37.039 So you have it's interactive from the start, which leads me to say this, which so far I can't get any scientists to agree with me on. 01:01:37.119 --> 01:01:38.719 So you want some controversy. 01:01:38.800 --> 01:01:40.239 So here's the controversy. 01:01:40.480 --> 01:02:07.679 What we normally think of as the source, the vocal fold oscillation, and the filter is the transfer function of the vocal tract, is it turns out in an on the on an individual pulse, impulse of the vocal fold closure, which we all know if we've known since at least the 90s in cook, you could probably straighten me out since probably much earlier, that it's this rapid closure that is the strongest pressure change of the airflow. 01:02:07.760 --> 01:02:10.400 So that's where the real power is, is in that closure. 01:02:11.519 --> 01:02:13.440 It excites the vocal tract. 01:02:13.760 --> 01:02:15.360 I lost my train of thought there. 01:02:15.519 --> 01:02:21.519 The oh, so the the thing that's being reiterated is that transfer function of the vocal tract. 01:02:21.599 --> 01:02:21.920 Yeah. 01:02:22.480 --> 01:02:35.360 And the only harmonics that that get even a chance to be strengthened by the periodicity of the vocal fold oscillation are the ones that landed within that the bandwidths, the the critical bandwidth of the resonances. 01:02:35.599 --> 01:02:42.159 Then those get filtered further than by the periodicity of the vocal fold opening and closing. 01:02:42.320 --> 01:02:53.679 So I say, this is the part that's controversial, that the so-called filter, which is the transfer function of the vocal tract, is actually crucial in the source itself. 01:02:53.920 --> 01:03:01.920 Each individual pulse is the source sound, and each individual pulse is a noise, is a resonance-shaped noise. 01:03:02.239 --> 01:03:19.440 So then where do we get harmonic content by the periodicity, the ongoing periodicity of that sound only reinforces that subset of harmonics that landed in resonances, the periodicity essentially equals harmonicity. 01:03:20.639 --> 01:03:32.159 So then the source that we call the source, the periodic oscillation of the vocal folds, is actually filtering those periodic noise-shaped resonances for harmonic frequencies. 01:03:32.239 --> 01:03:36.800 So the source is the filter and the filter is the source, and they're all so interactive that you can't sort them out. 01:03:36.880 --> 01:03:39.840 So there's my Cliff Notes version of the R. 01:03:40.159 --> 01:03:52.559 Based on that, maybe our introductory textbooks in the future should simply start to say that sound is produced in humans and animals with a source airway system. 01:03:54.000 --> 01:03:56.000 So we never separate the two. 01:03:56.159 --> 01:04:00.480 We always say that the sound is produced with a source airway system. 01:04:00.880 --> 01:04:02.719 Yeah, I'm good with that. 01:04:03.440 --> 01:04:04.159 Absolutely. 01:04:10.320 --> 01:04:10.960 Yep. 01:04:12.239 --> 01:04:27.760 I I think one of the things that I appreciate about your working, that I I feel like the vocabulary of the linear source filter doesn't really accommodate, certainly doesn't invite people into, is just like the notion of the flow pulse skewing. 01:04:27.920 --> 01:04:36.880 Like the interaction of the vocal tract air mass at that time scale seems really important for the sound that the voice ends up making. 01:04:37.119 --> 01:04:48.880 And and again, if we if we if we separate the vocal folds phenomenologically, like it's actually really difficult to understand that interactive property of the vocal tract. 01:04:49.519 --> 01:05:08.000 And we should say that you mentioned Gunnar Fant, you know, he and his uh colleague Ananta Padmanaba uh back in the 70s already agreed that this was not a linear combination between two systems, but that there was always feedback between them and they interact with each other. 01:05:08.159 --> 01:05:13.760 Even though, as you say, in his book, he then goes on and explains everything with a linear concept. 01:05:13.920 --> 01:05:18.800 But that's

because, you know, our brains uh just don't follow these things that quickly. 01:05:18.960 --> 01:05:22.960 We have to understand it in simple tune uh terms first. 01:05:23.199 --> 01:05:27.599 And then then we uh add the uh the complexities. 01:05:27.760 --> 01:05:27.920 Yeah. 01:05:28.800 --> 01:05:38.800 I think um I think this is a good example of the very first thing that I said, which is that there are frequently multiple lenses to try and encounter a phenomenon through. 01:05:39.039 --> 01:05:49.920 And um, and so the the condition that I think is really interesting is like how how do we understand a resonance that falls between two harmonics? 01:05:50.239 --> 01:05:54.320 Because if let's say you have a resonance at 2.5 FO, right? 01:05:54.480 --> 01:05:56.400 Just imagine that interaction. 01:05:56.639 --> 01:05:59.760 Um the spectrum, if you have a long wind. 01:06:00.239 --> 01:06:18.800 Fourier transform the mathematics of that are going to prioritize representing that acoustic energy in terms of multiples of the fundamental frequency, just because the periodicity itself is interacting with every one of those sine cosine like phase shifted pairs to do the testing of the complex signal. 01:06:18.960 --> 01:06:25.280 Um, and so I think it's really easy to imagine, oh, there's energy at 2 FO and energy at 3 FO. 01:06:25.679 --> 01:06:31.920 But if you look within a single pulse response, actually there is energy at 2.5 FO. 01:06:32.239 --> 01:06:34.159 It just keeps getting reinterrupted. 01:06:34.320 --> 01:06:37.119 It just keeps getting restarted and restarted and restarted. 01:06:37.199 --> 01:06:46.480 So it never has the ability to build the energy that we would normally see in a standing wave in the vocal tract if a more virtuous alignment took place. 01:06:46.880 --> 01:06:56.480 And um it's just really fascinating to me, like looking at that as the phenomenon that a waveform represents, it seems fairly obvious there isn't energy at 2FO and 3FO. 01:06:56.559 --> 01:06:57.280 There is energy at 3. 01:06:57.920 --> 01:07:12.480 But even the linear theory tells us that the there the resonance has a bandwidth, and therefore it's not just at one frequency, but it's a spreading of an entire uh collection of frequencies in that bandwidth. 01:07:12.800 --> 01:07:17.760 So if a harmonic lands anywhere in the bandwidth, it's still uh energized. 01:07:19.519 --> 01:07:25.119 Yeah, I just think it's an interesting um it's an interesting mind puzzle, I guess I would say that. 01:07:25.199 --> 01:07:26.559 It's an interesting thought experiment. 01:07:26.800 --> 01:07:27.440 Yeah. 01:07:27.760 --> 01:07:28.239 Right. 01:07:28.880 --> 01:07:29.199 Okay. 01:07:29.519 --> 01:07:37.360 Well, I'd like to start with uh, or I'd like to end maybe with a broader question for all three of you, if and if I can ask if you uh keep your answers brief. 01:07:37.519 --> 01:07:51.920 Um but uh the uh uh you know for singers, teachers, and researchers, uh what's the single most important distinction between perception and acoustics that you wish more people understood? 01:07:52.000 --> 01:07:53.840 And and Ken, let's start with you. 01:07:56.400 --> 01:08:00.400 Um Well, things are rarely as they seem. 01:08:00.719 --> 01:08:03.599 It's useful to know the difference. 01:08:03.760 --> 01:08:07.599 And the singer sings with what is called procedural knowledge. 01:08:08.079 --> 01:08:10.880 Declarative knowledge is the science stuff, the measurable part. 01:08:11.039 --> 01:08:15.199 Procedural knowledge is what it sounds and feels like to the singer to do it. 01:08:15.519 --> 01:08:32.720 Realizing that difference and then being guided by a knowledgeable helper, teacher, as to what is what are these interactions going to feel like and seem like to you as the producer of it is actually the knowledge the singer needs to know. 01:08:33.119 --> 01:08:47.840 Uh I think the teachers should learn the declarative knowledge because it's a conversation between those two that guides their instruction, a very informed conversation between what's actually going on and what does it seem like to you, right? 01:08:48.079 --> 01:08:51.119 Uh, but but the singer sings with the what it seems like. 01:08:52.079 --> 01:08:52.960 That's that's the thing. 01:08:53.439 --> 01:09:00.640 Once they know the feel it feelable path and the hearable path of sounds, that's and that's what they've sung with for forever. 01:09:01.199 --> 01:09:07.920 Uh, but the the declarative knowledge is really helping us refine how to guide them to those those sensations. 01:09:08.000 --> 01:09:14.159 So it's a it's a a a wonderful conversation that um um it needs to be ongoing. 01:09:14.319 --> 01:09:18.000 So I'm that's why we need to be science-informed voice teachers. 01:09:18.640 --> 01:09:18.960 Gotcha. 01:09:20.960 --> 01:09:22.239 Yeah, that's a cool question. 01:09:22.319 --> 01:09:24.640 Uh we could talk about for another hour, probably. 01:09:24.720 --> 01:09:37.199 Um, if I were to give a just one response, I I don't want voice teachers to be of afraid of science and afraid of learning more about this, sort of first and foremost. 01:09:37.359 --> 01:09:52.640 And so this is something I've I've heard Ken say frequently, and I I really believe it too, that um I I don't know a single scientific study that has set out to study what really terrible feeling singing feels like. 01:09:52.880 --> 01:09:59.119 And or or singing where when you go through a register transition, it is just completely awkward and feels awful. 01:09:59.279 --> 01:10:10.640 Like, I think most of these phenomena that we can identify, harmonics crossing formants or tuning specific acoustic things in certain ways, or setting up standing waves in certain ways. 01:10:10.800 --> 01:10:19.279 Like, from the singer's point of view, I I think we are just analyzing after the fact what feels amazing to them. 01:10:19.600 --> 01:10:27.279 And so I think the information is

really useful for teachers as sort of predictive guardrails. 01:10:27.439 --> 01:10:34.560 Like you really can learn a lot about how a voice, uh, all other things being equal, like should be able to behave. 01:10:34.640 --> 01:10:40.159 And so you can make predictions about what sounds you should ask of the people that you're working with. 01:10:40.399 --> 01:11:07.920 But at the end of the day, I I think what I think what all of the knowledge points to is that you know, singing is just a magnificent wild yop that the human body can make, and it feels great, and we should never attempt to um reduce it or sterilize it by imposing language on it that is just is just trying to capture how great singing is. 01:11:12.319 --> 01:11:24.479 Uh well, I've gotten very interested in uh not only human production, but also the production of uh birds and mammals and uh in general the animal world around us. 01:11:24.640 --> 01:11:35.680 And I'm so amazed by the fact that all species are tuned so sharply to a very, very small section of the total sound that's available to us. 01:11:36.319 --> 01:11:59.760 And uh and so for me the beauty is to study the sound in its general term um and uh and then and then figure out what is one individual in in a given species or a cross species um really doing with their sensory information, uh with that little sliver of sound that they can receive. 01:12:00.239 --> 01:12:19.680 Um and uh yeah, to me uh therefore acoustics itself is not any more important uh than perception, but it perception is every bit as important as acoustics, but we have to understand it across individuals and across species. 01:12:22.079 --> 01:12:22.880 That's great. 01:12:23.119 --> 01:12:23.760 Well, wonderful. 01:12:23.920 --> 01:12:25.600 Well, thanks uh uh everybody. 01:12:25.680 --> 01:12:28.800 I I uh I really appreciate the the discussion. 01:12:29.039 --> 01:12:35.039 Um uh and uh I'm gonna be thinking about that wild, magnificent yacht line for a little while there, Ian. 01:12:35.359 --> 01:12:35.760 Thank you very much. 01:12:36.159 --> 01:12:38.319 And then we can measure it and understand it. 01:12:38.640 --> 01:12:40.000 Right, right, right, right, exactly. 01:12:40.079 --> 01:12:41.279 That's that that's the point, right? 01:12:41.439 --> 01:12:44.800 So um well, I'll go ahead and uh just close us out here. 01:12:44.880 --> 01:12:51.279 Um so thank you again, Ingo, uh, Ian, uh Ken for really thoughtful, really engaging in the conversation. 01:12:51.439 --> 01:12:57.680 Um, this has been Toe-to-Toe with Ingo, a podcast from the National Center for Voice and Speech, where evidence matters. 01:12:57.840 --> 01:13:00.319 Questions are welcome and disagreement drives discovery. 01:13:00.479 --> 01:13:02.880 Until next time, I'm Andrew Parker. 01:13:02.960 --> 01:13:05.920 Keep asking good questions and follow where the evidence leads. 01:13:06.399 --> 01:13:07.520 We'll see you next time. 01:13:08.560 --> 01:13:13.760 Must have a code that you can live by. 01:13:15.439 --> 01:13:24.960 And so become yourself because the past is just a good.