How Music Works

by David Byrne

  Throughout the history of recorded music, we have tended to value convenience over quality every time. Edison cylinders didn’t really sound as good as live performers, but you could carry them around and play them whenever you wanted. LPs, revolving slower, didn’t sound as rich as 45s or 78s, but you didn’t have to attend to them as much. And cassettes? Are you kidding? We were told that CDs would last forever and sound squeaky clean, but they really don’t sound as good as LPs, and the jury is out regarding their durability. The spectrum of sound on analog mediums has an infinite number of gradations, whereas in the digital world everything is sliced into a finite number of slivers. Slivers and bits might fool the ear into believing that they represent a continuous audio spectrum (psychoacoustics at work), but by nature they are still ones and zeros; steps rather than a smooth slope. MP3s? They may be the most convenient medium so far, but I can’t help thinking that the psychoacoustic trickery used to develop them—the ability to cause the mind to think and feel that all the musical information is there when in reality a huge percentage has been removed—is a continuation of this trend in which we are seduced by convenience. It’s music in pill form, it delivers vitamins, it does the job, but something is missing. We are often offered, and gladly accept, convenient mediums that are “good enough,” rather than ones that are actually better.

  Where does this road of compromise end, and does it really matter if we lose a little quality along the way? Isn’t the quality or accuracy of a recording somewhat irrelevant to music’s use and enjoyment? We laugh out loud at antics on fuzzy, grainy, and atrociously low-resolution YouTube postings, and we talk to our loved ones on mobile-phone networks with voice quality that would make Alexander Graham Bell roll over in his grave. Information theory tells us that the amount of bits needed to communicate certain kinds of content—what someone is saying, or the antics of a cat, for example—can really be much lower than we think. If we only need to understand the verbal content of someone on the telephone, then the quality can be surprisingly bad and we’ll still know what our friends and family are saying. It doesn’t seem to matter that so much is missing. Maybe “good enough” is okay.

  Or maybe not. Reacting to this tendency, some musicians have decided to go back to analog recording, and some have perversely gone out of their way to make their recordings sound as lo-fi as possible—as bad as they can get away with. They want to get as far from digital cleanliness as possible. Why would bad quality, fuzziness, and distortion imply that the music is more authentic? The idea is that if one accepts that crisp and clean recordings are inherently soulless, then the opposite, dirty and rough, must therefore be straight from the heart. That might not sound logical, but that’s the way we think. It’s all part of the recurring belief that conflates new technologies with being inauthentic. Bad—even fake bad, in this way of thinking—means good. It’s confusing, because most digital music does not sound “bad.” If anything, it sounds conventionally good—clean, spotless, with a full range of frequencies. Though it is actually less rich sounding than previous technologies, it fools the ear into believing that it sounds better. It’s this shiny, glossy quality that is considered suspect by many music fans. In response, they overvalue the easily audible drawbacks of a previous era—the hiss, crackle, and distortion. In my opinion, realness and soul lie in the music itself, not in the scratches and pops of old records. So, while the cleanliness and “perfection” of much current music is not a guarantee of a moving musical experience, neither is their opposite.

  If, following the lead of the phone company, we find ourselves talking about communication and information transmission when we talk about music, then maybe some of the sonic richness of LPs is indeed superfluous and can be eliminated with no serious loss. Could this work with speech as well? Yes and no. Music has more going on simultaneously than speech, for starters. Looking at a reproduction of a painting is certainly not the same as standing in front of the real thing, but an awful lot of the emotion, intent, ideas, and sensibility can indeed be communicated—even via a cheap reproduction. Similarly, I can be moved to tears by a truly awful recording or a bad copy of a good recording. Would I be moved even more if the quality were higher? I doubt it. So why bother?

  There does come a time, however, when the richness of the retinal or aural experience is so diminished that the communication—in this case the enjoyment of the music—becomes unintelligible. But how can we define that? I first heard rock, pop, and soul songs on a crappy-sounding transistor radio, and they changed my life completely. The sound quality was atrocious, but that tinny sound was communicating a wealth of information. Though it was an audio transmission that carried the news, it was the social and cultural message embedded in the music that electrified me as much as the sound did. Those extra-musical components that got carried along with the music didn’t demand a high-resolution signal—good enough was good enough. I’m not saying that tinny sound should be considered satisfying or desirable, or that we should never strive for more than “good enough,” but it’s amazing how much lo-fi or lo-res information can communicate. Live concerts don’t generally have perfect sound either, but they can move us deeply.

  Now I begin to ask myself if the fuzziness and ambiguity inherent in lowquality signals and reproductions might actually be a factor that gives the viewer or listener a way in. I know from writing lyrics that some details—names, places, locations—are desirable; they anchor the piece in the real world. But so are ambiguities. By letting the listener or viewer fill in the blanks, complete the picture (or piece of music), the work becomes personalized and the audience can adapt it to their own lives and situations. They become more involved with the work, and an intimacy and involvement becomes possible that perfection might have kept at bay. Maybe the lo-fi music crowd has a point?

  MUSIC SOFTWARE AND SAMPLE-BASED COMPOSITION

  Music composition changed a lot with the advent of digital recording. As we’ve seen, the first digital samples were short and primarily used by the phone company. People used them for gimmicks and special effects, but these early developments did not have a wide musical influence. Soon enough, though, it became possible to grab or sample a whole bar of music, and though the samples were not of super-high quality, it was enough. Looping beats became ubiquitous, and rhythm tracks made of sampled measures (or shorter intervals) now function as the rhythm bed in many songs. You can “hear” the use of Akai, Pro Tools, Logic, and other digital recording and sample-based composition in most pop music written in the last twenty years. If you compare recent recordings to many made in previous eras, you might not know what makes them sound different, but you would certainly be able to hear it.

  The effects the software has affected not just the sonic quality but also the composition process. Maybe you can also hear the effects of the ones and zeros that make up digital recording, though that may be less true as time goes by and technology passes the limits of the differentiating aspects of our hearing. What you do hear, though, is a shift in musical structure that computer-aided composition has encouraged. Though software is promoted as being an unbiased tool that helps us do anything we want, all software has inherent biases that make working one way easier than another. With the Microsoft presentation software PowerPoint, for example, you have to simplify your presentations so much that subtle nuances in the subjects being discussed often get edited out. These nuances are not forbidden, they’re not blocked, but including them tends to make for a less successful presentation. Likewise, that which is easy to bullet-point and simply visualize works better. That doesn’t mean it actually is better; it means working in certain ways is simply easier than working in others. Music software is no different. Taking another avenue would make music composition somewhat more tedious and complex.

  An obvious example is quantizing. Since the mid-nineties, most popular music recorded on computers has had tempos and rhythms that have been quantized. This means that the tempo never varies, not even a little bit, and the rhythm
ic parts tend toward metronomic perfection. In the past, the tempos of recordings would always vary slightly, imperceptibly speeding up or maybe slowing down just a little, or a drum fill might hesitate in order to signal the beginning of a new section. You’d feel a slight push and pull, a tug and then a release, as ensembles of whatever type responded to each other and lurched, ever so slightly, ahead of and behind an imaginary metronomic beat. No more. Now almost all pop recordings are played to a strict tempo, which makes these compositions fit more easily into the confines of the recording and editing software. An eight-bar section recorded on a “grid” of this type is exactly twice as long as a four-bar section, and every eight-bar section is always exactly the same length. This makes for a nice visual array on the computer screen, and facilitates easy editing, arranging, and repairing as well. Music has come to accommodate software, and I have to admit a lot has been gained as a result. I can sketch out an idea for a song very quickly, for example, and I can cut and paste sections to create an arrangement almost instantly. Severe or “amateurish” unsteadiness or poorly played tempo changes can be avoided. My own playing isn’t always rock steady, so I like that those distracting flubs and rhythmic hiccups can be edited out. The software facilitates all of this. But, admittedly, something gets lost in the process. I’m just now learning how to listen, value, and accommodate some of my musical instincts that don’t always adhere to the grid. It makes things a little more complicated, as I still use the software, but I sense that the music breathes a little more as a result of me not always bending to what the software makes easiest.B

  Courtesy of David Byrne

  Sometimes, after having begun a songwriting process using computer software, I find that I have to sing or play it apart from that recording to free myself from its straightening tendencies. In singing a song freely, for example, I might find that the note at the high point or at the end of a melodic arc “wants” to hang on just a little bit longer than the grid of measures on the computer might indicate as normal. The result is that a verse might end up being nine measures long, rather than the traditional eight. Alternately, a half measure might feel like a nice emotionally led extension, as well as giving a short, natural feeling breathing space, so then I’ll add that half measure to the grid on the software. In a recent collaboration with Fatboy Slim, I discovered that he often added “extra” measures to accommodate drum fills. It felt natural, like what a band would do. Shifting off the grid is sometimes beneficial in other ways, too. If a listener can predict where a piece of music is going, he begins to tune out. Shifting off an established pattern keeps things interesting and engaging for everyone, though it sometimes means you have to avoid the path of least resistance that digital recording software often offers.

  Quantizing, composing, and recording on a grid are just some of the effects of software on music. Other effects are created by the use of MIDI, which stands for musical instrument digital interface. MIDI is a software/ hardware interface by which notes (usually played on a keyboard) are encoded as a series of instructions rather than recorded as sounds. If you strike a middle A on a keyboard, then the MIDI code remembers when that note was played in the sequence of notes in the composition, how hard or quickly it was struck, and how long it was sustained. What’s recorded is this information and these parameters (a bit like a piano roll or the tines on a music box), but not the actual sound—so if that sequence of instructions is played back, it will tell the keyboard instrument to play that note and all the rest exactly when and how you played it previously. This method of “recording” takes up much less computer memory, and it also means the instructional information recorded is independent of the instrument played. Another MIDI-equipped instrument with a completely different sound can be told to play the same note, at the same time. What began as a piano sound might be changed later to synthesizer strings or a marimba. With MIDI recording, you can make arrangement decisions easily and at any time.

  The way MIDI remembers how hard or fast you hit a note is by dividing the speed of the note strike into 127 increments. The speed of your hit will be rounded off so that it will fall somewhere in that predetermined range. Naturally, if you strike a key faster, slower, or with more subtlety than the MIDI software and its associated sensors can measure, then your “expression” will not be accurately captured or encoded; it will be assigned the nearest value. As with digital recording, music gets rounded off to the nearest whole number, under the assumption that finer detail would not be discernible to the ear and brain.

  There are instruments that can be used to trigger MIDI fairly well: keyboards, some percussion pads, and anything that can be easily turned into switches and triggers. But some instruments elude capture. Guitars aren’t easily quantifiable in this way, nor are wind, brass, or most bowed string instruments. So far, the nuances of those instruments have been just too tricky to capture. Using MIDI therefore tends to entice people away from using those instruments and the kinds of expression they are uniquely good at. A lot of MIDI-based recordings tend to use arrays of sounds generated or at least triggered by keyboards, so, for example, it’s easiest to play chord inversions that are keyboard friendly. The same chords on guitars tend to have a different order of the same notes. Those keyboard chords, then, in turn, incline composers to vocal melodies and harmonies that fit nicely with those specific versions of the chords, so the whole shape, melody, and arc of the song are being influenced, not just the MIDI parts and instruments. As soon as technology makes one thing easier, it leaves a host of alternatives in the dust.

  The uncanny perfection that these recording and compositional technologies make possible can be pleasing. But metronomic accuracy can also be too easy to achieve this way, and the facile perfection is often obvious, ubiquitous, and ultimately boring. Making repetitive tracks used to be laborious and time-consuming, and the slight human variations that inevitably snuck in as a musician vainly attempted to be a machine were subliminally perceptible, if not consciously audible. A James Brown or Serge Gainsbourg track often consisted of a riff played over and over, but it didn’t sound like a loop. Somehow you could sense that it had been played over and over, not cloned. Imagine a row of dancers moving in unison—something that has a huge visceral impact. Besides implying hard work, skill, and precision, it also works as a powerful metaphor. Now imagine that same row created by a series of mirror reflections or CGI technology. Not so powerful.

  For many years, DJs, mixers, and hip-hop artists constructed tracks from digital samples of riffs and beats taken from existing recordings. Some artists lifted entire hooks and choruses from pop songs and used them like a knowing reference or quote (P. Diddy does this a lot, as does Kanye West), the way you might quote a familiar refrain to a friend or lover to express your feelings. How many times has “put a ring on it” been used in conversation? Song references are like emotional shortcuts and social acronyms.

  In much contemporary pop, if you think you’re hearing a guitar or piano, most likely you’re hearing a sample of those instruments from someone else’s record. What you hear in such compositions are lots of musical quotations piled on top of one another. Like a painting by Robert Rauschenberg, Richard Prince, or Kurt Schwitters made of appropriated images, ticket stubs, and bits of newspaper, this music is a species of sonic collage. In some ways it’s metamusic; music about other music.

  However, many artists found out relatively quickly that when songs were created in this way, they had to share the rights and profits with the original record companies and the original composers of those fragments and hooks. Half of the money from a song often goes to the source of a hook or chorus. A few of my songs have been sampled in this way, and it’s flattering and fun to hear someone add a completely new narrative to something you wrote years ago. The singer Crystal Waters sampled a Talking Heads song (“Seen and Not Seen”—whoa, what a weird choice!) in her hit “Gypsy Woman (She’s Homeless),” and that wonderful song of hers (which I even covered at one point) bears no rela
tion to the original. As far as I know, she or her producers were not intentionally referencing the original song; they simply found something about it, the way the groove felt or its sonic texture, useful.

  Besides the fact that being sampled pays well, I don’t feel that it compromises the original. Anyone can tell it’s a quote, a sample, right? Well, that depends. Trick Daddy, Cee Lo, and Ludacris used another Talking Heads song, the demo “Sugar on My Tongue,” in their hit song “Sugar (Gimme Some).” In that case the reference to the original was obvious, to me at least. They used our chorus hook for their own. But since that was never a popular Talking Heads song (it was a demo included in a boxed set), that “quotation” aspect was probably lost on most listeners. (For those keeping track, I did get paid and credited as a songwriter on that track, not a bad second life for what was originally a demo!) Since the song was already a bit of silly sexual innuendo, those guys taking it one step further was no big deal. But, if someone hypothetically proposed repurposing the hook of a song I’d written as a new song about killing Mexicans, blowing up Arabs, or slashing women, I would say no.

  Many artists soon began to find that the ubiquitous use of samples could severely limit the income from “their” songs, which led many of them to eventually stop or curtail their use of sampling technology. Sometimes even what is easiest (grabbing a beat from a drum break on a CD, which only takes minutes) becomes a thing to be avoided. Bands like Beastie Boys picked up instruments they’d put down years earlier, and hip-hop artists either disguised their samples better, found more obscure ones, or, more often, began to create or buy tracks made from scratch (often on synthesizers and drum machines). Technology, or rather the aspect of technology that enabled the use of copyrighted material, had, thanks to the efforts of rights enforcers and guardians, sent some musicians back to the drawing board. A group of hot young programmers soon emerged whose skills were in constructing, for use by others, these grooves made from scratch. With a relatively inexpensive piece of gear or software, you could make contributions to major songs from your bedroom. In contemporary hip-hop, there is now often no relationship between a composition’s backing track and a simulation of a live performance by musicians in the traditional sense. In the early days, there were live DJs using vinyl to loop drum breaks, but now everything—every instrument—is sampled, processed, or in some way shamelessly and boldly artificial.